Parvez Mahbub Thesis

Impact of Urban Traffic and
Climate Change on Water Quality

from Road Runoff
S. M. Parvez Bin Mahbub

Master of Engineering
Bachelor of Science in Civil Engineering
A Thesis Submitted in Partial Fulfilment of the Requirements

for the Award of the Degree of Doctor of Philosophy
School of Urban Development

Queensland University of Technology
March 2011
KEYWORDS
Climate change, heavy metals, hydrocarbons, multivariate data analyses, pollutant

build-up, pollutant wash-off, rainfall simulation, urban water quality, water quality
mitigation.
i
ii
ABSTRACT
Urban traffic and climate change are two phenomena that have the potential to
degrade urban water quality by influencing the build-up and wash-off of pollutants,
respectively. However, limited knowledge has made it difficult to establish any link
between pollutant buildup and wash-off under such dynamic conditions. In order to
safeguard urban water quality, adaptive water quality mitigation measures are
required. In this research, pollutant build-up and wash-off have been investigated
from a dynamic point of view which incorporated the impacts of changed urban
traffic as well as changes in the rainfall characteristics induced by climate change.
The study has developed a dynamic object classification system and thereby,
conceptualised the study of pollutant build-up and wash-off under future changes in
urban traffic and rainfall characteristics. This study has also characterised the build-
up and wash-off processes of traffic generated heavy metals, volatile, semi-volatile
and non-volatile hydrocarbons under dynamic conditions which enables the
development of adaptive mitigation measures for water quality. Additionally,
predictive frameworks for the build-up and wash-off of some pollutants have also
been developed.
iii
iv
TABLE OF CONTENTS
Keywords ................................................................................................... i
Abstract....................................................................................................iii
Table of Contents ..................................................................................... v
List of Tables ........................................................................................... xi
List of Figures ........................................................................................ xv
List of Appendices ................................................................................. xix
Statement of Original Authorship ........................................................ xxi
Acknowledgements ............................................................................ xxiii
Dedication ............................................................................................ xxv
List of Publications from this Research ........................................... xxvii
Chapter 1 Introduction ........................................................................... 1
1.1. Background .......................................................................................................... 1
1.2. Research Problem ................................................................................................ 1
1.3. Aims and Objectives ............................................................................................ 2
1.4. Research Hypothesis ............................................................................................ 3
1.5. Scope of the Research .......................................................................................... 3
1.6. Justification for the Research ............................................................................. 5
1.7. Innovation and Contribution to the Knowledge ............................................... 6
1.8. Outline of the Thesis ............................................................................................ 7
Chapter 2 Water pollution: From the perspective of changes in urban
traffic and climate change ....................................................................... 9
2.1. Introduction .......................................................................................................... 9
2.2. Urban Water Pollution ........................................................................................ 9
2.3. Primary Water Pollutants ................................................................................. 10
2.3.1. Gross pollutants ........................................................................................................... 11
2.3.2. Solids ........................................................................................................................... 11
2.3.3. Nutrients ...................................................................................................................... 12
2.3.4. Oxygen Demanding Materials ..................................................................................... 14
2.3.5. Toxicants ..................................................................................................................... 16
A. Heavy Metals ............................................................................................................... 16
B. Hydrocarbons............................................................................................................... 19
2.4. Pollutant Processes............................................................................................. 21
2.4.1. Build-up of Pollutants .................................................................................................. 21
2.4.2. Wash-off of Pollutants ................................................................................................. 25
2.5. Urban Traffic and its Impact on Water Quality ............................................. 28
2.5.1. Road Traffic: Australia Wide Perspective ................................................................... 28
v
2.5.2. Impact of Road Traffic on Water Quality .................................................................... 30
2.6. Climate Change and its Impact on Water Quality ......................................... 33
2.6.1. Climate Change: Global and Australian Perspective ................................................... 33
2.6.2. Impact of Climate Change on Water Quality............................................................... 38
2.7. Conclusions ........................................................................................................ 41
Chapter 3 Research design and methods ............................................ 45
3.1. Background ........................................................................................................ 45
3.2. Research Methodology ...................................................................................... 45
3.3. Site Selection ...................................................................................................... 46
3.4. Sample Collection .............................................................................................. 50
3.4.1. Build-up Sample Collection ......................................................................................... 50
3.4.2. The Wet and Dry Vacuum System .............................................................................. 50
A. Equipment .................................................................................................................... 50
B. Definitions ................................................................................................................... 51
A. Standardisation of the Build-up Sample Collection Procedure .................................... 51
B. Results and Discussion ................................................................................................ 52
3.4.3. Wash-off Sample Collection ........................................................................................ 53
3.5. Test Methods ...................................................................................................... 62
3.5.1. Heavy Metals ............................................................................................................... 62
A. Method ......................................................................................................................... 63
B. Sampling ...................................................................................................................... 63
C. Quality Control and Quality Assurance ....................................................................... 64
D. Sample Preservation and Storage................................................................................. 66
E. Nitric Acid Digestion ................................................................................................... 67
F. Metals detection by Inductively Coupled Plasma/Mass Spectrometry (ICP/MS) ... 67
3.5.2. Total Petroleum Hydrocarbons .................................................................................... 67
Gasoline Range Organics ...................................................................................................... 68
Diesel Range Organics .......................................................................................................... 68
Sampling ............................................................................................................................... 69
A. GRO in build-up sampling ...................................................................................... 69
B. GRO in wash-off sampling ..................................................................................... 70
C. DRO in build-up sampling ...................................................................................... 71
D. DRO in wash-off sampling ..................................................................................... 71
Quality Control and Quality Assurance................................................................................. 71
Sample Preservation and Storage .......................................................................................... 74
A. GRO in build-up samples ........................................................................................ 74
B. GRO in wash-off samples ....................................................................................... 74
C. DRO in build-up samples ........................................................................................ 74
D. DRO in wash-off samples ....................................................................................... 75
Sample Extraction ................................................................................................................. 75
A. Gasoline Range Hydrocarbons ................................................................................ 75
B. Diesel Range Hydrocarbons .................................................................................... 75
Sample Analyses by Gas Chromatography ........................................................................... 76
3.5.3. Solids ........................................................................................................................... 76
A. Method ......................................................................................................................... 77
B. Determination of Solids by Gravimetric Method......................................................... 77
3.5.4. Organic Carbon ............................................................................................................ 77
A. Method ......................................................................................................................... 77
B. Determination of Organic Carbon ............................................................................... 77
3.5.5. Surface Texture Depth ................................................................................................. 78
3.6. Data Analyses ..................................................................................................... 78
3.6.1. PCA ............................................................................................................................. 78
3.6.2. FC ................................................................................................................................ 79
3.6.3. FA ................................................................................................................................ 80
vi
3.6.4. PROMETHEE ............................................................................................................. 81
3.6.5. GAIA ........................................................................................................................... 82
3.6.6. PLS .............................................................................................................................. 83
3.7. Publication of Results ........................................................................................ 83
3.8. Summary............................................................................................................. 86
Chapter 4 Defining dynamic relationship of heavy metal build-up and
wash-off with urban traffic and climate change .................................. 89
4.1. Introduction ........................................................................................................ 92
4.2. Experimental Section ......................................................................................... 93
4.2.1. Site Selection ............................................................................................................... 93
4.2.2. Build-up and Wash-off Sample Collection .................................................................. 94
4.2.3. Sample Testing ............................................................................................................ 94
4.2.4. Data Analyses .............................................................................................................. 95
4.3. Results and Discussion....................................................................................... 96
4.3.1. Exploratory PCA of Heavy Metals Build-up ............................................................... 96
4.3.2. FC Analysis and PROMETHEE Ranking of Heavy Metals Build-up ......................... 99
4.3.3. Exploratory PCA of Heavy Metals Wash-off ............................................................ 101
4.3.4. FC Analysis and PROMETHEE Ranking of Heavy Metals Wash-off ...................... 103
4.3.5. GAIA Analysis Incorporating Impacts of Traffic and Climate Change .................... 106
4.4. Acknowledgements .......................................................................................... 108
4.5. Supporting Information Available ................................................................. 108
4.6. Brief ................................................................................................................... 108
4.7. References ......................................................................................................... 108
Chapter 5 Characterising the build-up of heavy metals and volatile
organic compounds on urban roads ................................................... 113
5.1. Introduction ...................................................................................................... 116
5.2. Site Selection ..................................................................................................... 117
5.3. Build-up Sample Collection ............................................................................ 118
5.4. Test Results and Data Analyses ...................................................................... 119
5.4.1. Heavy Metals ............................................................................................................. 119
5.4.2. Volatile Organics ....................................................................................................... 124
5.5. Multicriteria Decision Analyses for Heavy Metals and Volatile Organics
Build-up ......................................................................................................................... 127
5.6. Conclusions ....................................................................................................... 129
5.7. References ......................................................................................................... 130
Chapter 6 Characterisation of semi and non volatile organic
compounds build-up on urban roads .................................................. 133
6.1. Introduction ...................................................................................................... 136
6.2. Materials and Methods .................................................................................... 138
6.2.1. Site Selection ............................................................................................................. 138
6.2.2. Key Study Parameters................................................................................................ 140
6.2.3. Build-up Sample Collection ...................................................................................... 141
6.2.4. Sample Preparation .................................................................................................... 142
6.2.5. Sample Testing .......................................................................................................... 142
6.2.6. Data Analyses ............................................................................................................ 147
vii
PCA ..................................................................................................................................... 148
PROMETHEE ..................................................................................................................... 148
GAIA ................................................................................................................................... 149
6.3. Results and Discussion .................................................................................... 149
6.3.1. Trends in the Original Data........................................................................................ 149
6.3.2. Exploratory PCA ....................................................................................................... 150
6.3.3. PROMETHEE ........................................................................................................... 155
6.3.4. GAIA ......................................................................................................................... 158
6.4. Conclusions ...................................................................................................... 159
6.6. References ........................................................................................................ 160
Chapter 7 Characterisation of volatile organic compounds wash-off
from urban roads................................................................................. 165
7.1. Introduction ..................................................................................................... 168
7.2. Materials and Methods ................................................................................... 170
7.2.1. Site selection .............................................................................................................. 170
7.2.2. Wash-off sample collection ....................................................................................... 171
7.2.3. Simulation of rainfall incorporating climate change impacts .................................... 173
7.2.4. Sample testing ............................................................................................................ 174
7.2.5. Data analysis .............................................................................................................. 176
PCA ..................................................................................................................................... 177
PROMETHEE ..................................................................................................................... 178
GAIA ................................................................................................................................... 180
7.3.1. Exploratory PCA ....................................................................................................... 180
7.3.2. PROMETHEE AND GAIA ....................................................................................... 184
7.4. Conclusions ...................................................................................................... 191
7.6. References ........................................................................................................ 192
Chapter 8 Prediction of volatile organic compounds build-up on
urban roads.......................................................................................... 197
8.1. Introduction ..................................................................................................... 200
8.2. Materials and Methods ................................................................................... 202
8.2.1. Site Selection ............................................................................................................. 202
8.2.2. Build-up Sample Collection ....................................................................................... 203
8.2.3. Sample Testing .......................................................................................................... 205
8.2.4. Data Analysis ............................................................................................................. 205
8.3.1. Exploratory PCA ....................................................................................................... 206
8.3.2. Factor Analysis .......................................................................................................... 209
8.3.3. Experimental Design.................................................................................................. 210
8.3.4. PLS Model Validation ............................................................................................... 212
8.5. Supporting Information Available ................................................................. 219
8.6. Brief .................................................................................................................. 219
8.7. References ........................................................................................................ 219
viii
Chapter 9 Characterisation and prediction of semi and non volatile
organic compounds wash-off .............................................................. 225
9.1. Introduction ...................................................................................................... 228
9.2. Materials and Methods .................................................................................... 229
9.2.1. Site Selection ............................................................................................................. 229
9.2.2. Rainfall Simulation Incorporating Climate Change................................................... 230
9.2.3. Wash-off Sample Collection...................................................................................... 231
9.2.4. Sample Testing .......................................................................................................... 233
9.2.5. Data Analysis ............................................................................................................. 234
9.3. Results and Discussion..................................................................................... 234
9.3.1. Exploratory PCA ....................................................................................................... 234
9.3.2. FA .............................................................................................................................. 237
9.3.3. Experimental Design ................................................................................................. 238
9.3.4. PLS Model Validation ............................................................................................... 241
9.4. Acknowledgement ............................................................................................ 248
9.5. Supporting Information .................................................................................. 248
9.6. Brief ................................................................................................................... 249
9.7. References ......................................................................................................... 249
Chapter 10 Conclusions and recommendations................................ 253
10.1. The Object Classification System ................................................................... 253
10.2. Build-up and Wash-off Processes of Pollutants under Dynamic Scenarios 254
Build-up and Wash-off Processes of Heavy Metals ............................................................ 255
Build-up and Wash-off Processes of Volatile Organic Compounds ................................... 256
Build-up and Wash-off of Semi and Non Volatile Organic Compounds ............................ 257
10.3. Prediction Framework for Build-up and Wash-off under Dynamic
Conditions ...................................................................................................................... 258
10.4. Recommendations for Stormwater Quality Mitigation ................................ 258
10.5. Recommendations for Further Research....................................................... 259
Consolidated list of references ............................................................ 261
ix
x
LIST OF TABLES
Table 2.1 Heavy Metals sources from exhaust and non-exhaust

vehicular emissions................................................................................. 18
Table 2.2 Predicted emission scenarios due to climate change............. 34
Table 2.3 Average percentage change in extreme rainfall intensity for
the Gold Coast region: adapted from Abbs et al. (2007) ....................... 37
Table 2.4 Climate change impacts on water quality parameters partially
adapted from Delpla et al. (2009) .......................................................... 39
Table 3.1 Traffic data for the selected study sites .................................. 47
Table 3.2 Traffic data (set 2) from the Gold Coast City Council .......... 47
Table 3.2 Traffic data (set 2) from the Gold Coast City Council (Contd.)
................................................................................................................ 48
Table 3.3 Sample collection data for the wet and dry vacuuming system .
................................................................................................................ 52
Table 3.4 Intensity-Frequency-Duration table for Station ID 40166 in
Gold Coast region (AUSIFD version 2.0) .............................................. 58
Table 3.4 Intensity-Frequency-Duration table for Station ID 40166 in
Gold Coast region (Contd.) .................................................................... 59
Table 3.5 Historical daily and monthly rainfall data for Gold Coast
rainfall stations adapted from the Bureau of Meteorology climate
service (http://www.bom.gov.au/climate/averages/) .............................. 60
Table 3.6 Future simulation events based on the extreme daily rainfall
intensity in the Gold Coast region for 2030 ........................................... 61
Table 3.7 Future simulation events based on the normal daily rainfall
intensity in the Gold Coast region for 2030 ........................................... 62
Table 3.8 Petroleum Hydrocarbon Constituents (adapted from Morrison
and Boyd 1992) ....................................................................................... 67
Table 4.1 Membership values of different objects in heavy metals build-
up after fuzzy clustering........................................................................ 100
Table 4.2 PROMETHEE II net ranking ( φ ) values showing the fuzzy
object in heavy metals build-up could still be classified as a member of
moderate traffic cluster as its value lies within the range of that cluster .
.............................................................................................................. 101
xi
Table 4.3 Membership values of the rainfall events in heavy metals
wash-off after fuzzy clustering ............................................................. 104
Table 4.4 PROMETHEE II net ranking ( φ ) values showing two fuzzy
events, 10 and 21 could still be classified as members of moderate and
extreme clusters, respectively as their φ values fall exclusively within the
ranges of corresponding clusters ......................................................... 105
Table 5.1 Traffic and pavement characteristics data of the selected sites
.............................................................................................................. 118
Table 5.2 Total correlation matrix between heavy metals and other
parameters............................................................................................ 123
Table 5.3 Affinity of individual pollutants towards different chemical
parameters............................................................................................ 126
Table 6.1 Selected road sites with traffic and pavement parameters
(partially adapted from Mahbub et al. 2010a) .................................... 139
Table 6.2 Percent recoveries of spikes applied at 35 mg/L and surrogate
applied at 10 mg/L along with limits of detection for the target
compounds............................................................................................ 145
Table 7.1 Average percentage change in extreme rainfall intensity for
the Gold Coast region: adapted from Abbs et al. (2007) .................... 173
Table 7.2 Future simulation events based on the normal daily rainfall
intensity in the Gold Coast region for 2030: adapted from Mahbub et al.
(2010a) ................................................................................................. 174
Table 8.1 Correlation coefficients matrix after VARIMAX factor rotation
with bold coefficients showing variables strongly associated with
corresponding factors .......................................................................... 209
Table 8.2 PLS Regression parameters for the predictor variables** . 213
Table 9.1 Simulation events based on the daily rainfall intensity at study
sites in the Gold Coast region for 2030: adapted from Mahbub et al.
(14) ....................................................................................................... 231
Table 9.2 New independent variables for each underlying factors
(starting with initials L, H or N) in the data matrices of light SVOC,
heavy SVOC and NVOC....................................................................... 238
Table 9.3 PLS regression parameters for predictor variables** ........ 243
Table A.1.1 IFD data for rain gauge Station 40584 (28.05°S, 153.29°E)
.............................................................................................................. 295
Table A.1.1 IFD data for rain gauge Station 40584 (Contd.) ............. 296
xii
Table A.2.1 Traffic data for the study sites .......................................... 308
Table A.2.2 Simulation rain events for the Gold Coast region at present
and as predicted for year 2030 ............................................................. 309
Table A.2.3 Limits of detection, percent recovery and relative standard
deviation percentage found in the heavy metal analysis with
corresponding molecular weights shown alongside each element ...... 310
Table A.2.4 Possible sources of elements frequently found in exhaust
and non-exhaust emissions of motor vehicle ........................................ 310
Table A.2.5 Chemical compositions (mean±standard deviations) of the
Build-up of Heavy metals in the selected sites ..................................... 311
Table A.2.6 Chemical compositions (mean±standard deviations) of the
wash-off of Heavy metals for the simulated rain events....................... 312
Table A.3.1 Test results for the build-up of SVOCs and NVOCs along
with physico-chemical parameters for the > 300 µm particle size
fraction .................................................................................................. 319
with physico-chemical parameters for the 150- 300 µm particle size
fraction .................................................................................................. 320
with physico-chemical parameters for the 75-150 µm particle size
fraction .................................................................................................. 321
with physico-chemical parameters for the 1-75 µm particle size fraction
.............................................................................................................. 322
with physico-chemical parameters for the <1 µm particle size fraction ...
.............................................................................................................. 323
Table A.3.6 Simple bi-variate correlation matrix between target
variables for the >300 µm particle size fraction from original data
given in supplementary Table A.3.1 ..................................................... 324
variables for the 150-300 µm particle size fraction from original data
variables for the 75-150 µm particle size fraction from original data
xiii
variables for the 1-75 µm particle size fraction from original data given
in supplementary Table A.3.4 .............................................................. 327
variables for the <1 µm particle size fraction from original data given
in supplementary Table A.3.5 .............................................................. 328
Table A.4.1 Traffic and pavement characteristics of eleven study sites ....
.............................................................................................................. 335
Table A.4.2 Chemical composition (mean±standard deviation) of VOCs
at five size fractions; inter-site percentage coefficient of variations
ranging from 29%-44% for eleven sites .............................................. 336
Table A.4.3 Correlation coefficients matrix achieved through VARIMAX
factor rotation including pH and EC as new variables with bold
coefficients showing variables strongly associated with corresponding
factors ................................................................................................... 337
Table A.4.4 PLS calibration set (X1*, X2*: Two independent factors
found in FA method; TOL= Toluene, ETB=Ethylbenzene, MPX= Meta
and Para Xylene, OX= Ortho Xylene; E1-E12 represents the twelve
individual experiments; C1-C3 represents the replicate experiments at
centre) .................................................................................................. 337
Table A.5.1 PLS calibration set for light SVOC: E1-E28 represent the
28 individual experiments; C1-C7 represent the replicate experiments at
centre; L1-L4 represent underlying factors in the light SVOC data
matrix ................................................................................................... 354
Table A.5.2 PLS calibration set for heavy SVOC: E1-E28 represent the
28 individual experiments; C1-C7 represent the replicate experiments at
centre; H1-H4 represent underlying factors in the heavy SVOC data
matrix ................................................................................................... 355
Table A.5.3 PLS calibration set for NVOC: E1-E46 represent the 46
individual experiments; C1-C4 represent the replicate experiments at
centre; N1-N5 represent underlying factors in the NVOC data matrix ....
.............................................................................................................. 356
xiv
LIST OF FIGURES
Figure 2.1 Illustration of pollutant build-up on road surfaces .............. 23

Figure 2.2 Hypothetical representation of build-up and wash-off of
surface pollutant loads for (a) source limited & (b) source unlimited
case (adapted from Vaze and Chiew 2002) ............................................ 25
Figure 3.1 Rainfall Simulator (adapted from Herngren et al. 2005).........
................................................................................................................ 54
Figure 3.2 Logarithmic relationships between percentage change of
intensity and duration in Gold coast region for 2030 ............................ 56
Figure 3.3 Schematic diagram showing detailed representation of
research flow undertaken in this research highlighting the development
of scientific papers .................................................................................. 85
Figure 4.1 PCA biplot of build-up of all heavy metal size fractions;
object identifiers are described in Table 5.1 .......................................... 97
Figure 4.2 PCA biplot for (a) particulate and (b) dissolved fractions for
wash-off of heavy metals; objects are indicated by numbers starting
from 1; numerical object identifiers are described in Table 5.3.......... 102
Figure 4.3 GAIA biplots for (a) build-up :( ) low traffic, ( )moderate
traffic and ( ) high traffic: object identifiers are described in Table
1;and (b) wash-off : ( ) low event, ( ) low to moderate event, ( )
moderate event, ( ) high event and ( ) extreme event: numerical object
identifiers are described in Table 5.3 ................................................... 106
Figure 5.1 Build-up sample collection site .......................................... 118
Figure 5.2 PCA biplot for heavy metals build-up on urban roads at 150-
299 µm fractions (objects are represented with numbered labels with
suffix C=commercial, I=Industrial or R=residential) ......................... 121
Figure 5.3 PCA biplot for heavy metals build-up on urban roads at 1-74
µm fractions (objects are represented with numbered labels with suffix
C=commercial, I=Industrial or R=residential) ................................... 122
Figure 5.4 PCA biplot for heavy metals build-up on urban roads at <1
µm fractions (objects are represented with numbered labels with suffix
C=commercial, I=Industrial or R=residential) ................................... 124
Figure 5.5 PCA biplot for volatile organics build-up on urban roads at
>300 µm fractions (objects are represented with numbered labels with
suffix C=commercial, I=Industrial or R=residential) ......................... 125
xv
75-149 µm fractions (objects are represented with numbered labels with
suffix C=commercial, I=Industrial or R=residential) ........................ 125
<1 µm fractions (objects are represented with numbered labels with
suffix C=commercial, I=Industrial or R=residential) ........................ 126
Figure 5.8 GAIA biplot for predominant chemical parameter scenario
( ) for particulate heavy metals; ( ) pi-decision axis; ( ) >300 µm
fractions; ( )150-299 µm fractions; ( ) 75-149 µm fractions; ( ) 1-74
µm fractions ......................................................................................... 127
Figure 5.9 GAIA biplot for predominant chemical parameter scenario
( ) for the dissolved heavy metals; ( )pi-decision axis; ( ) <1 µm
fraction ................................................................................................. 128
Figure 5.10 GAIA biplot for predominant heavy metals particulate
fraction ( ); ( ) pi-decision axes; ( ) metals’ affinity towards
predominant chemical parameters; ( ) study sites; ( ) TSS’ presence in
particulate fractions ............................................................................. 129
Figure 5.11 GAIA biplot for predominant volatile organics particulate
fraction ( ); ( ) pi-decision axes; ( ) organics’ affinity towards
predominant chemical parameter; ( ) study sites; ( ) TOC’s presence in
particulate fractions ............................................................................. 129
Figure 6.1 PCA biplot of total particulate fractions from <1 µm to >300
µm taken together ................................................................................ 150
Figure 6.2 Individual PCA biplots for (a) >300 µm; (b) 150-300 µm; (c)
75-150 µm; (d) 1-75 µm; and (e) <1 µm size fractions ...................... 152
Figure 6.3 Combined PROMETHEE II net outranking flows of traffic
objects showing commercial sites as predominant sources of SVOCs and
NVOCs build-up ................................................................................... 156
Figure 6.4 GAIA biplot for the build-up of SVOCs and NVOCs
incorporating all size fractions as well as all traffic scenarios .......... 158
Figure 7.1 The relative locations of the four study sites for the VOC
wash-off sample collection ................................................................... 170
Figure 7.2 PCA biplots for (a) >300 µm; (b) 150-300 µm; (c) 75-150
µm; (d) 1-75 µm; and (e) <1 µm size fractions for the wash-off of
toluene (TOL), ethylbenzene (ETB), meta and para xylene (MPX) and
ortho xylene (OX) ................................................................................. 182
Figure 7.3 PROMETHEE partial ranking (a) and complete ranking (b)
for the twenty two different rainfall events in terms of VOC wash-off ......
xvi
.............................................................................................................. 185
Figure 7.4 GAIA biplot of the rain events under the combined scenario
of five size fractions .............................................................................. 186
Figure 7.5 PROMETHEE partial ranking (a) and complete ranking (b)
for the VOC affinity matrix during wash-off from urban roads ........... 187
Figure 7.6 GAIA biplots for five pre-defined rain events clusters (a) and
five different size fractions (b) in terms of VOC’s affinity towards TSS
and TOC................................................................................................ 188
Figure 8.1 PCA biplot of total particulate volatile organic build-up on
urban roads; objects are indicated by labels with the prefix I, C or R for
industrial, commercial and residential sites respectively .................... 207
Figure 8.2 PCA biplot of the calibration set shows the scores of fifteen
experiments and the loadings of nine measured variables .................. 211
Figure 8.3 Prediction results of (a) Toluene; (b) & (c) Ethylbenzene; (d)
Meta & Para Xylene; (e) Orthoxylene at different size fractions showing
most predictions were within 95% confidence limit............................. 215
Figure 8.4 Relative Prediction Error (RPE) at each size fractions and
total RPE at all study sites.................................................................... 217
Figure 8.5 Coefficients of variation (CV) at each size fractions and total
sample at all study sites ........................................................................ 218
Figure 9.1 PCA biplots of particulate (>300µm-1µm combined) and the
dissolved (<1µm) fractions for light SVOCs, heavy SVOCs and NVOCs
for the 22 rain events shown with numerical identifiers ...................... 236
Figure 9.2 PCA biplots of the experimental designs for (a) light SVOCs,
(b) heavy SVOCs and (c) NVOCs with experiments are shown with
initial ‘E’ and replicate experiments with initial ‘C’ ........................... 240
Figure 9.3 Distributions of the box plot statistics at (a) >300 µm and (b)
<1 µm for observed and predicted target compounds ......................... 245
Figure 9.4 Coefficient of Variations (CV %) of the predicted
concentrations at the rain events not used in the calibration .............. 246
Figure A.2.1 PCA biplots of particulate and potential dissolved fraction
for heavy metals build-up for (a) >300µm, (b)150-300 µm, (c)75-150
µm, (d)1-75 µm and (e) <1 µm; objects are indicated by labels with the
prefix I, C or R starting for industrial, commercial and residential sites,
respectively ........................................................................................... 305
Figure A.2.3 Results from Mastersizer S particle size distribution
analysis showing that fractions <1µm constituted around 2% whilst
fractions up to 300µm constituted almost 82% to 96% volume of the
xvii
total wash-off particles in samples collected from the 22 simulated rain
events .................................................................................................... 307
Figure A.4.1 Eleven study sites are shown on the map of selected
suburbs of Coomera, Upper Coomera and Helensvale; Corresponding
traffic parameters and labels of each site are shown alongside the site
name ..................................................................................................... 338
Figure A.4.2 PCA biplots for (a-d) particulate fractions and (e)
potential dissolved fraction with eleven land use scores shown with
initials C, I and R for commercial, industrial and residential
respectively........................................................................................... 339
Figure A.4.3 PCA biplots for (a-d) particulate fractions, (e) dissolved
fraction and (e) total particulates (1->300 µm) incorporating pH and
EC as new variables............................................................................. 340
Figure A.5.1 Four wash-off road sites located within 5 km radius of the
meteorological station 40166 (adapted from the Google Earth map
services)................................................................................................ 350
Figure A.5.2 Distributions of the box plot statistics at 150-300 µm
particulate fraction for observed and predicted target compounds .... 351
xviii
LIST OF APPENDICES
Appendix A.1 Intensity-Frequency-Duration Table for Station 40584

.............................................................................................................. 293
Appendix A.2 Supplementary Information for Chapter 4 ................... 297
Appendix A.6 Regression Equations of the Prediction framework ............
.............................................................................................................. 357
xix
xx
STATEMENT OF ORIGINAL AUTHORSHIP
“The work contained in this thesis has not been previously submitted to meet
requirements for an award at this or any other higher education institution. To the
best of my knowledge and belief, the thesis contains no material previously
published or written by another person except where due reference is made.”
Signature
Date
xxi
xxii
ACKNOWLEDGEMENTS
I would like to sincerely thank my principal supervisor, Professor Ashantha
Goonetilleke, for his dedicated support and guidance throughout the program of
research culminating in this thesis.
I would also thank my co-supervisors, Associate Professor Godwin A. Ayoko, Dr
Tan Yigitcanlar and Dr Prasanna Egodawatta for their part in support and
supervision for this research.
Without the generous financial assistance provided by the Queensland University of
Technology through an Australian Postgraduate Award, I would not have been able
to undertake this postgraduate research. I also wish to extend my gratitude towards
the laboratory technicians of School of Urban Development and Chemistry
Discipline for providing logistic support for this program of study through to
completion.
Special thanks go to the Gold Coast City Council and Queensland Department of
Transports and Main Roads for providing with data and support at various stages of
this research.
xxiii
xxiv
DEDICATION
I would like to dedicate this thesis to my wife Reshmi, daughter Emina and son
Eshan for their unwavering patience, understanding and support during my doctoral
research.
xxv
xxvi
LIST OF PUBLICATIONS FROM THIS
RESEARCH
Monograph:
(1) Mahbub, P., Ayoko, G., Egodawatta, P., Yigitcanlar, T., Goonetilleke, A.
(2010) Traffic and climate change impacts on water quality: measuring
build-up and wash-off of heavy metals and petroleum hydrocarbons. In
Rethinking Sustainable Development: Urban management, Engineering and
Design. Yigitcanlar, T., (Ed.). Engineering Science Reference, New York,
pp. 147-167.
Journals:
(1) Mahbub, P., Ayoko, G. A., Goonetilleke, A., Egodawatta, P., Kokot, S.
(2010). Impacts of Traffic and Rainfall Characteristics on Heavy Metals
Build-up and Wash-off from Urban Roads. Environmental Science &
Technology, 44 (23), 8904-8910. Impact Factor: 5.5; ERA Rank A*.
(2) Mahbub, P., Goonetilleke, A., Ayoko, G. A., Egodawatta, P., Yigitcanlar, T.
(2011). Analysis of Build-up of Heavy Metals and Volatile Organics on
Urban Roads in Gold Coast, Australia. Water Science & Technology, 63(9),
2077-2085. Impact Factor: 1.1; ERA Rank B.
(3) Mahbub, P., Ayoko, G. A., Goonetilleke, A., Egodawatta, P. (2011).

Analysis of the Build-up of Semi and Non Volatile Organic Compounds on
Urban Roads. Water Research, 45(9), 2835-2844. Impact Factor: 4.8; ERA
Rank A*.
(4) Mahbub, P., Goonetilleke, A., Ayoko, G. A., Egodawatta, P. (2011). Effects
of Climate Change on the Wash-off of Volatile Organic Compounds from
Urban Roads. Science of the Total Environment, DOI:
10.1016/j.scitotenv.2011.06.032. Impact Factor: 3.4; ERA Rank A.
(5) Mahbub, P., Goonetilleke, A., Ayoko, G. A. (2011). Prediction Model of the
Build-up of Volatile Organic Compounds on Urban Roads. Environmental
Science & Technology, 45(10), 4453-4459. Impact Factor: 5.5; ERA Rank
A*.
(6) Mahbub, P., Goonetilleke, A., Ayoko, G. A. (2011). Prediction of the Wash-
off of Traffic Related Semi and Non Volatile Organic Compounds from
Urban Roads under Changed Rainfall Characteristics. Journal of Hazardous
Materials (Under Review). Impact Factor: 4.36; ERA Rank A.
xxvii
xxviii
CHAPTER 1 INTRODUCTION
1.1.Background
The water quality in urban areas is being adversely impacted due to rapid
urbanisation and climate change. Urban traffic is identified as one of the most
important sources of pollutants to urban receiving waters. This situation is further
compounded by the fact that the volume and characteristics of urban traffic in
Australia are expected to undergo extensive changes due to the continuous spread of
urbanisation in the continent’s major cities (BITRE 2008).
Additionally, climate change is expected to result in changed rainfall characteristics
throughout Australia (CSIRO 2003, 2007). The changes in the rainfall
characteristics due to climate change can readily affect pollutants wash-off from
urban roads into receiving water bodies. This can impair the water quality and in
turn exert significant impacts on human and ecosystem health in urban areas.
Investigation of pollutant build-up and wash-off under dynamic urban traffic and
climate change scenarios can provide valuable insights into the changing patterns of
urban water quality. Outcomes from such investigations are critical for regulatory
authorities to undertake adaptive mitigation strategies to safeguard the water quality
in urban areas and to maintain ecosystem functions.
1.2.Research Problem
Extensive research studies have been undertaken to characterise pollutant build-up
and wash-off processes under static environmental conditions (for example
Herngren 2005; Herngren et al. 2006; Deletic & Orr 2005; Yuan et al. 2001;
Goonetilleke et al. 2009). Average daily traffic, congestion and surface textures of
1
the urban roads have been found to be the most important parameters generating
traffic related pollutant build-up on urban roads. Likewise, rainfall characteristics
such as intensity, frequency and distribution are the most influential parameters
causing pollutant wash-off from road surfaces. A range of heavy metals and
petroleum hydrocarbons are regarded as pollutants that are generated by traffic.
However, studies which explicitly address the relationships between pollutant build-
up and wash-off processes with changing urban traffic and rainfall characteristics
due to climate change are limited. As a result, a knowledge gap exists in linking
traffic generated pollutant build-up with the subsequent wash-off processes in urban
roads. The limited availability of water quality data under changing traffic and
rainfall characteristics has made the form of relationships between pollutant build-up
and predicted changes in traffic characteristics as well as pollutant wash-off and
predicted changes in rainfall characteristics difficult to determine. A better
understanding of the pollutant build-up and wash-off processes due to changing
urban traffic and climatic conditions is necessary to develop adaptive strategies for
stormwater quality mitigation under such changed situation.
1.3.Aims and Objectives

This research aimed to investigate the following two aspects of pollutant build-up
and wash-off on urban roads under changing urban traffic and climate conditions:
1. The impact of traffic characteristics on pollutant build-up on urban roads.
2. The impact of rainfall characteristics on the pollutant wash-off from urban
roads.
The objectives of this research study were as follows:
2
1. Defining the dynamic relationship of the build-up and wash-off of traffic
generated pollutants with changing urban traffic and climate change,
respectively.
2. Characterisation of the build-up and wash-off processes of stormwater
pollutants generated from urban traffic.
3. Development of prediction frameworks for traffic generated stormwater
pollutants.
4. Provide guidance to stormwater quality mitigation strategies under dynamic
urban traffic and climate change scenarios.
1.4.Research Hypothesis
This research was based on the following two hypotheses:
1. Traffic generated pollutant build-up on urban roads is influenced by changes
in urban traffic characteristics such as average daily traffic, congestion and
road surface texture depth.
2. The pollutant wash-off from urban roads is influenced by changes in rainfall
characteristics such as intensity, frequency and duration due to climate
change.
1.5.Scope of the Research

The pollutant build-up and wash-off studies undertaken in this research study
focused on urban areas with transport infrastructure developed in the past decade.
Three different land uses, namely, residential, commercial and industrial sites were
selected for the study. Even though the research study was undertaken in the
Southeast Queensland region of Australia, the methodologies, prediction
frameworks as well as the pollutant build-up and wash-off characterisations have
generic applicability to urban areas in other geographical locations.
3
The study used simulated rainfall to generate surface runoff from urban roads. The
simulation of natural rainfall was justified from the work of past researchers in
relation to pollutant wash-off studies. The changes in rainfall characteristics due to
climate change were confined to the Southeast Queensland region of Australia for
year 2030. The selection of year 2030 was in conformity with the contemporary
climate change research in Australia that have predicted future changes of rainfall
characteristics for this year.
In relation to pollutant build-up and wash-off, heavy metals, volatile, semi-volatile
and non volatile organic compounds were investigated in this study. Research
studies have identified these pollutants as being mainly generated from traffic and
are quite harmful to both human health and ecosystem. Other common stormwater
pollutants such as, nutrients were not investigated in this study as these pollutants
are not generated by traffic. The build-up and wash-off studies were performed for
both the dissolved and particulate fractions. In build-up, the traffic characteristics
such as, average daily traffic (ADT), volume to capacity ratio (V/C) as well as
pavement characteristic such as surface texture depth (STD) were deemed to be the
influential factors. In this context, only asphalt pavement surfaces were investigated.
Characterisation of the build-up and wash-off of stormwater pollutants from other
types of pavements such as, concrete and gravel are beyond the scope of this
research. In wash-off, the rainfall intensity, frequency and duration as well as the pH
and electrical conductivity of the runoff were considered to be the main influencing
factors.
4
The methodology of this study has been developed in such a way that several journal
paper publications have formed the research outcome in terms of the objectives
described in section 1.3.
1.6.Justification for the Research

To minimise the risks posed by the traffic generated stormwater pollutants to urban
water bodies, adaptive strategies need to be undertaken. However, limited
knowledge on pollutant build-up and wash-off under dynamic scenarios such as,
changed urban traffic and climate change has resulted in significant constraints in
conceptualisation of appropriate adaptive strategies. Research studies have been
undertaken in the past on the process kinetics of several stormwater pollutants.
Unfortunately, these studies have commonly focussed on the characterisation of
pollutant build-up and wash-off from a static point of view. Moreover, the urban
traffic characteristics were not specifically investigated in these studies in relation to
pollutant build-up. Similarly, there is lack of knowledge on the impacts of climate
change induced rainfall characteristics on pollutant wash-off. Even though some
studies have investigated rainfall characteristics whilst characterising pollutant
wash-off, linkages were not established between pollutant build-up and wash-off.
This was mainly due to the fact that parameters influencing the impacts of pollutant
build-up and wash-off such as urban traffic and climate change were not studied
simultaneously in past research. The outcomes from these past studies cannot be
generalised as a result of the lack of linkages between pollutant build-up and wash-
off processes. In the current study, the simultaneous investigations of the build-up
and wash-off processes of traffic generated pollutants under dynamic scenarios have
contributed to helping to overcome these knowledge gaps.
5
Traffic generated pollutants, such as heavy metals and hydrocarbons have the
potential to adversely affect human health and the health of ecosystems. Whilst the
toxic and carcinogenic impacts of these pollutants are well documented in past
research, knowledge on build-up and wash-off processes of traffic generated heavy
metals and hydrocarbons under changed urban traffic and climate change scenarios
is limited. Water quality mitigation strategies, such as water sensitive urban design
(GCCC 2006, 2007) require adaptive pollution control measures to cope with the
changed traffic and climate change scenarios. The outcomes of this study will
contribute to the development of adaptive mitigation strategies for stormwater
quality improvement by providing knowledge of the build-up and wash-off
processes of heavy metals and hydrocarbons under dynamic urban traffic and
climate conditions.
1.7.Innovation and Contribution to the Knowledge

The research study investigated pollutant build-up and wash-off from urban roads
from a dynamic point of view. The changes in urban traffic and climate conditions
were viewed as catalysts influencing stormwater pollutant build-up and wash-off.
Innovative techniques in terms of methodologies were developed for sampling,
laboratory testing and data analyses for traffic generated pollutants. A novel sample
collection system in build-up was proposed in this study. An effective
chromatographic separation technique for semi and non volatile organic compounds
was developed. Additionally, a robust object classification method for classifying
the changing urban traffic scenarios as well as climate change induced rainfall
events has been presented. The prediction methodologies for stormwater pollutants
presented in this study could be used as general prediction frameworks elsewhere
under dynamic urban traffic and climate change scenarios. These techniques along
6
with the outcomes of this study would provide essential knowledge for the
development adaptive mitigation strategies to safeguard the quality and ecosystem
function of urban receiving waters.
1.8.Outline of the Thesis

The thesis was submitted as a combination of six journal papers. The overall
research study was divided into ten chapters. Chapter 1 provides the introduction to
the research topic. Chapter 2 presents a critical review of research literature relating
to water pollution in the context of urban traffic and climate change. In chapter 3,
the detailed methodology adopted in the research study is discussed. Chapter 4 to
Chapter 9 are the scientific papers submitted to international journals for the
dissemination of the study outcomes. Chapter 4 discusses the build-up of heavy
metals and volatile organic compounds on urban roads. Chapter 5 evaluates the
dynamic relationship of heavy metals build-up and wash-off with changing urban
traffic and climate change, respectively. Chapter 6 presents a prediction model for
volatile organic compounds built-up on urban roads. Chapter 7 provides an analysis
of the wash-off of volatile organic compounds under the influence of climate
change. Chapter 8 discusses the characterisation of semi and non- volatile organic
compounds build-up on urban roads. Chapter 9 presents the characterisation and
prediction model for the semi and non- volatile organic compounds wash-off from
urban roads under climate change. Finally, chapter 10 provides the overall
conclusions of this research study in terms of adaptive water quality mitigation
measures under changed urban traffic and climatic conditions together with
recommendations for further research.
7
8
CHAPTER 2
WATER POLLUTION: FROM THE
PERSPECTIVE OF CHANGES IN URBAN
TRAFFIC AND CLIMATE CHANGE
2.1.Introduction
The impacts of changing urban traffic and climate change on urban water quality
have drawn much attention of researchers in recent years. Emissions from an
increased number and different types of motor vehicles influence pollutant build-up
on roads. Likewise, changes in the rainfall characteristics due to climate change
influence pollutant wash-off. This chapter discusses the important concepts
underlining water pollution, pollutant build-up and wash-off in urban areas as well
as climate change, urban traffic changes and their impacts on urban stormwater
quality. Urban traffic and the climate change characteristics have been discussed
from both global and regional perspectives. Additionally, a detailed discussion on
primary water pollutants is included to provide context in relation to their role in
water pollution.
2.2.Urban Water Pollution

Water is fundamental for the long-term sustainability of urban areas. Pollution of
local water bodies in an urban area can pose risks in terms of human and ecosystem
health. The continuous rise in urban population will cause an increase in urban
traffic which in turn can have a significant impact on pollutant build-up on urban
roads. The pollutants that accumulate on urban roads are washed away during
9
surface runoff and eventually transported to local water bodies. However, surface
runoff is not the only factor that impairs urban water quality. According to Shepherd
et al. (2006), groundwater flow as well as baseflow also impacts on urban water
quality. Water pollution may be caused by easily identifiable sources, referred to as
point sources of pollution, or indirectly from multiple sources, referred to as non-
point sources of pollution. According to the United Nations Environmental Glossary
(UNEG 1997), the anthropogenic sources of emissions that are located at spatially
identifiable points such as, sewage treatment plants, power plants cause point source
water pollution. Non-point sources of water pollution are diffused and pollutants
enter the receiving water body from unspecified outlets. Anthropogenic sources such
as, agriculture, urban areas, mining, construction, dams and channels as well as
natural sources such as, forestry and saltwater intrusions are the common non-point
sources of water pollution.
The urban water cycle is comprised of water supply, wastewater disposal and
stormwater drainage (Markopoulos 2008). The management of surface runoff from
urban roads is part of the stormwater drainage component of the urban water cycle.
The increased area of paved surfaces due to expanding urbanisation reduces
infiltration, whilst causing surface runoff to exhibit higher peak flows, shorter times
to peak and accelerated transport of pollutants and sediments (Niemczynowicz
1999). The situation worsens with increased urban traffic volume as well as changed
rainfall patterns due to climate change (Mahbub et al. 2010a).
2.3.Primary Water Pollutants

The water pollutants of concern in urban stormwater runoff include:
• Gross Pollutants
10
• Solids
• Nutrients
• Oxygen-demanding materials
• Toxicants (Heavy metals and Hydrocarbons)
(Adams & Papa 2000)
The following discussion gives an overview on each of these pollutants.
2.3.1. Gross pollutants

Litter or gross pollutants discarded on road surfaces is the most visible matter
identified as water pollutants. This is generally not a major source of water pollution
(Sartor & Boyd 1972). As litter tend to deposit on the road surface, their foremost
impact is related to visual aesthetics. Sartor and Boyd (1972) classified litter as
originating from three major sources: packaging materials, printed materials and
intentionally disposed waste materials. Packaging materials include paper, plastic,
metal and glass. These are discarded either intentionally or otherwise.
2.3.2. Solids
Matter that remains as residue on evaporation and drying at 103° to 105° C is
defined as solids (Sawyer et al. 1994). Solids exist in various forms in nature, e.g.,
dissolved, suspended, volatile, and fixed. The term “sediments” is sometimes used
for suspended solids. The undissolved substance in a solid sample in water on
filtration is referred to as suspended solids (Sawyer et al. 1994). The determination
of the amount of dissolved and undissolved matter is accomplished by undertaking
tests on filtered and unfiltered portions of the samples. Volatile solids are referred to
as organic matter that volatilises at 550°C ignition from a solid sample and the
residual matter remaining as ash is referred to as fixed solids. With regards to urban
stormwater pollution, solids are eroded from pervious surfaces or washed off from
11
impervious surfaces by stormwater. Sewer systems also contribute to the
accumulation of solids at the bed and on the walls of the sewers during dry periods
(Novotny & Olem 1994).
The physical effects of suspended solids on the ambient environment are increased
turbidity, abrasion of fish gills and other sensitive tissues, reduction in visibility, loss
of riparian vegetation leading to reduced shade and refuge, and destruction of
spawning areas (Adams & Papa 2000). However, the chemical effects of suspended
solids on the receiving water are much more adverse in nature. High suspended
solids load increases the portability of various other pollutants by acting as a mobile
substrate through processes such as adsorption and absorption (Sartor & Boyd 1972;
Hoffman et al. 1982; Shinya et al. 2000; Settle et al. 2007). The adsorption
phenomenon concerns the adherence of a chemical substance from a liquid or gas
phase to a solid interface (e.g., onto the surface of a particle). Absorption is the
phenomenon where a chemical substance passes an interface and penetrates into a
different phase (Hvitved-Jacobsen et al. 2010). Organic matter as well as humic
substances can also result in binding of metals through a process known as
complexation which refers to a reaction between metal ions or atoms and naturally
occurring substances/ligands present in the organic matter (Charlesworth & Lee
1999; Ellis & Revitt 1982; Hering & Morel 1988).
2.3.3. Nutrients
Nutrients are defined as substances assimilated by living organisms that promote
growth (EPA 2010). In the context of urban water quality, major elements (e.g.,
nitrogen and phosphorus) and trace elements (e.g., sulphur, potassium, calcium, and
magnesium) are considered as nutrients. Amongst these, nitrogen and phosphorus
12
are the key parameters for the assessment of eutrophication, which is the process of
acquiring high concentrations of nutrients by the receiving water. The processes
taking place in a water body would be virtually impossible to reverse where there is
a significant input of nutrients. A closed cycle originates where the nutrients are
converted to plant matter and released back into the water environment on their
decomposition. Common measures of nutrients are total nitrogen, nitrates, ammonia,
total Kjeldahl nitrogen (TKN), total phosphorus, total organic carbon, and indirectly,
algal mass and chlorophyll a (Wanielista & Yousef 1993).
Adams and Papa (2000) attributed the sources of nutrients to leaching from
vegetation, agricultural fertilisers in runoff, runoff flowing through pastures, parking
lots and lawns and wastewater discharges. Nutrients can stimulate aquatic algal
blooms and excessive macrophytic (aquatic plants) growth, causing depletion of
dissolved oxygen on their death and decay (Wanielista & Yousef 1993). Most
aquatic organisms struggle to survive with depleted levels of dissolved oxygen in
such water bodies suffering from a condition referred to as ‘hypoxia’ (EPA 2010).
Visual impacts of nutrients include colour, turbidity, floating matter and slimes.
Nitrogen in the form of ammonia and nitrates and phosphorus occurring as
orthophosphates are readily available for plant growth. However, Hvitved-Jacobsen
et al. (2010) have reported that excessive amount of the molecular form of ammonia
(NH3), when washed-off into the water body, exerts acute toxic effects by
obstructing the diffusion of ammonia through fish gills. De Jong et al. (2009) have
suggested that drought resulting from climate change can reduce the nitrogen uptake
by plants resulting in increased amount of nitrogen remaining in the soil. Most of
13
this residual nitrogen remains as nitrates which is water soluble. This high nitrate
level can cause algal growth and eutrophication. High nitrate levels in drinking
water can cause methaemoglobinemia (blue baby syndrome) and stomach cancer
(Chambers et al. 2001).
The inorganic compounds of phosphorus, usually referred to as orthophosphates and
polyphosphates, are the principal sources of phosphorus pollution in urban water
(Sawyer et al. 1994). These are water soluble phosphorus compounds and are used
in public water supply as a means of controlling corrosion. Goudier et al. (2009)
have shown that the addition of orthophosphate at a treatment rate of 1 mg PO43-/L
can control the fixed bacterial multiplication (biofilm) on the pipe walls. However,
increase in heavy-duty household synthetic detergents can cause a large amount of
excess polyphosphates in the water supply in urban areas. Phosphorus pollution
manifests in the form of algal or cyanobacterial bloom in surface water and in
extreme cases in fish deaths, and fish and shellfish containing algae toxins fatal to
humans (Heinonen-Tanski & Wijk-Sijbesma 2005; Hwang & Lu 2000).
2.3.4. Oxygen Demanding Materials

Oxygen demanding materials in urban water bodies are derived from plants,
animals, and soil organic matter. Oxygen usage in water systems takes place through
microbiological processes which are particularly important in relation to the
following phenomena (Hvitved-Jacobsen et al. 2010):
• Biodegradation of organic matter
• Dissolved Oxygen (DO) mass balances
• DO depletion
14
In order to characterise the biodegradation of organic matter several parameters,
such as, BOD (biological oxygen demand), COD (Chemical oxygen demand), TOC
(total organic carbon), and DOC (dissolved organic carbon) are used (Schaarup-
Jensen & Hvitved-Jacobsen 1991). Whilst BOD and COD are measures of oxygen
consumption during decomposition of both organic and inorganic substances in
water bodies, total organic carbon (TOC) is a direct expression of the total organic
content present in the water body and is used as a measure of organic carbon content
irrespective of the different oxidation state of organic matter (Xun et al. 2010). The
dissolved organic carbon (DOC) is the filtered fraction of TOC and is regarded as
the most reliable measure of many simple and complex organic molecules making
up the dissolved organic load in a natural water body (Thurman 1985; Kay et al.
2009).
The decomposable organic compounds are discharged into watercourses from
various natural and anthropogenic sources. Thurman (1985) described the natural
sources of organic carbon as precipitation, canopy drip, groundwater, interstitial
water of soil and sediment, snowmelt, as well as phytoplankton, zooplankton, and
bacteria in lake and river water. Industrial activities can also alter the steady state of
the global carbon cycle. For example, incomplete combustion of fossil fuels and
biomass materials are regarded as the source of excessive amount of black carbon
transported through riverine systems or aerosols into the marine environment from
land (Dickens et al. 2004).
An appropriate concentration of dissolved oxygen (DO) is necessary to maintain
aquatic life. Excessive oxidisable matter in water can create a substantial oxygen
15
demand on the water column, causing potential DO level depletion due to
biodegradation. Organic matter also acts as a substrate (i.e., surface or medium) for
invertebrates, bacteria, and fungi and can cause excessive growth of such
microorganisms in the water body (Hvitved-Jacobsen et al. 2010). Garcia-Ochoa et
al. (2010) described that both the oxygen consumption and oxygen transfer rate into
water bodies by microorganisms can affect the dissolved oxygen (DO) mass
balance. The alteration of DO mass balance and DO depletion can affect the
transport and accumulation of dissolved and colloid organic fractions as well as the
particulate organic fractions that might accumulate in sediments in the water bodies.
Substantial loads of oxygen demanding substances often lead to adverse conditions
such as fish kills, foul odours, unsightly discolouration, and slime growth (Sartor &
Boyd 1972).
2.3.5. Toxicants
Toxicants are substances or a mix of substances that exert harmful or injurious
impacts on living organisms (Hvitved-Jacobsen et al. 2010). Heavy metals and
hydrocarbons are the two main groups of toxicants that are of particular interest to
researchers in the context of urban stormwater runoff. Road runoff is a major source
of metals and hydrocarbons to receiving waters, and data from the UK suggest that
runoff contributes to about 10–12% of the total receiving water pollutant budget
(Ellis & Mitchell, 2006).
A. Heavy Metals
Past Studies have confirmed that stormwater runoff from urban areas contain
significant loads of heavy metals (Makepeace et al. 1995; Marsalek et al. 1997;
Davis et al. 2001). Commonly present heavy metals in urban runoff that are of
particular interest are cadmium (Cd), copper (Cu), lead (Pb), nickel (Ni), chromium
16
(Cr) and zinc (Zn) due to their potentially acute or cumulative (chronic) toxic effects
on flora and fauna. The availability and the toxicity of a heavy metal is to a great
extent related to its solubility, speciation, adsorption to particulates and its potential
for formation of complexes with both inorganic and organic substances (Hogland et
al. 1984; Harrison & Wilson 1985; Sartor & Boyd 1972; Shinya et al. 2000;
Hvitved-Jacobsen et al. 2010). Speciation of a substance in an aquatic environment
refers to the phenomenon where the substance depending on external conditions
such as, acidity, alkalinity, concentration and redox potential, might appear in
different chemical forms or species with different characteristics. Metal species can
exist as free ions as well as molecular forms in urban runoff.
The main anthropogenic activities that generate heavy metals in urban runoff are
vehicular traffic, combustion of fossil fuel and lubricants and industrial activities.
Sansalone and Buchberger (1997a) noted that vehicle tyres produce more heavy
metals than any other vehicle component. Cadmium, chromium, copper, iron, lead,
nickel, antimony and zinc are produced from vehicle tyres whereas vehicle brakes,
frames and fuel also play a role in the presence of these toxic products. Table 2.1
illustrates the possible sources of heavy metals from vehicular traffic.
17
Table 2.1 Heavy Metals sources from exhaust and non-exhaust vehicular emissions
Elements Possible Sources
Cu, Sb Bushing, thurstbearing, brake (Sternbeck
et al. 2002; Johansson et al. 2009;
Harrison 1979; Fujiwara et al. 2010; Von
Uexküll 2005)
Zn, Cu Lubricants, engine oil (Lim et al. 2006;
Cadle et al. 1999)
Zn, Cd Tyre (Harrison 1979; Adachi &
Tainosho 2004; Kummer et al. 2009)
Cr Alloy wheel plate, crankshaft, metal
plating, yellow paint of pavement
(Harrison 1979; Adachi & Tainosho
2004)
Pb, Ni Exhaust emission (Johansson et al. 2009)
Cu, Sb, Ba Rush hour stop-start (Grieshop et al.
2006)
The toxic effects of heavy metals in urban water bodies are well documented
(Karlsson et al. 2010; Zheng et al. 2010; Davis & Birch 2010). When stormwater
contaminated with non-biodegradable heavy metals is discharged directly into
natural water bodies, the metals can accumulate in the environment, causing both
short-term (e.g. acute toxicity) and long-term (e.g. carcinogenic damages) adverse
effects on living organisms. For example, neurotoxicity of both the peripheral and
central nervous system of the human body is believed to be the result of chronic
exposure to arsenic compounds (Goyer & Clarsksom 2001). Cadmium is known to
18
enhance lipid peroxidation by increasing the production of free radicals in the lungs,
which leads to tissue damage and cellular death (Méndez-Armenta & Ríos 2007),
and chronic lead toxicity affects gastrointestinal, neuromuscular, renal and
haematological systems (ATSDR 2005). Furthermore, some non-biodegradable
metals such as chromium is considered to be toxic to bacteria, plants and animals
(Richard & Bourg 1991).
B. Hydrocarbons
In the context on urban water pollution, hydrocarbons are referred to as organic
micropollutants that are typically discharged into the environment in trace amounts
(Hvitved-Jacobsen et al. 2010). Numerous hydrocarbons have been identified in
urban runoff. A literature review revealed that at least 656 hydrocarbons could be
present in stormwater runoff (Eriksson et al. 2005). It is generally neither feasible
nor relevant to sample and analyse all potential hydrocarbons. In the case of urban
runoff, the focus on selected substances may therefore be on the basis of their
sources.
Both, anthropogenic and natural sources are responsible for the generation of
hydrocarbons in water bodies (Hunter et al. 1979; Hoffman et al. 1982, 1984;
Herngren 2005; Hojae et al. 2009). Natural sources generally include the anaerobic
degradation of organic materials and forest fires. Urban stormwater runoff have been
found to transport significant amounts of chlorinated hydrocarbons used as
pesticides, polycylic aromatic hydrocarbons (PAHs) and total petroleum
hydrocarbons (TPHs) into surface water bodies (Sartor & Boyd 1974; Hoffman et al.
1982, 1984, 1985). Brown et al. (1985) found that the vehicular crankcase oil is a
significant anthropogenic source of TPH contamination of urban stormwater runoff.
19
Anthropogenic sources also arise from a wide variety of industrial activities,
especially chemical and petrochemical industries (Jo et al. 2008). It has been
reported that almost 74% of oil spill accidents occurred in petroleum refineries, oil
terminals, or storage facilities resulting in increased amounts of petroleum
hydrocarbons entering the surrounding water bodies (Chang & Lin 2006). Although
there is evidence that the number of large scale oil spills has decreased in recent
decades (Burgherr 2007), reports of smaller spills and potential incidents are
occurring on a daily basis. In some areas of the world (e.g. China) spill risk is
increasing due to increased vehicular traffic activities (Woolgar 2008). This is also
the case for European coasts bordering the Baltic, Barents and North Seas and the
English Channel due to the increased transport of heavy and residual fuel oils from
the former Soviet Union.
When TPHs are released directly into water through spills or leaks, certain TPH
fractions float and form thin surface films. Other heavier fractions accumulate in the
bottom sediments due to their hydrophobic and sorption capacity to particles, which
may affect bottom-feeding aquatic organisms. Some organisms found in water
(primarily bacteria and fungi) may break down some of the TPH fractions through
metabolism leading to chemical and biological transformations (Hojae et al. 2009).
Metabolism of such organic compounds typically proceeds in steps and often stable
intermediate compounds which are different in solubility and toxicity from the
original compounds are produced. Bioaccumulation of 16 PAHs through particulate
matter is reported in oysters for human consumption (Cortazar et al. 2009).
20
Persistency of hydrocarbons refers to the low degradability and corresponding
prolonged occurrence in the environment. Kucklick et al. (1997) noted that, PAHs
do not degrade in the environment and represent the largest class of suspected
carcinogens. On the other hand, chlorinated pesticides such as DDDs, DDTs,
Dieldrin are subject to a number of natural degrading actions while they remain on
street surfaces (Sartor & Boyd 1972). Many of these chlorinated pesticides have
long been banned due to concerns about their detrimental effects on ecosystem and
human health. However, residues still persist in the environment as a result of their
long half life, which is about 20 to 30 years (Sericano et al. 1990; Hung et al. 2007).
TPH compounds with lighter molecular weights, such as benzene, toluene and
xylene which are present in gasoline, can affect the human central nervous system.
Prolonged exposure to these TPHs can also cause death. Breathing toluene at
concentrations greater than 100 ppm for more than several hours can cause fatigue,
headache, nausea, and drowsiness (ATSDR 2010). Prolonged exposure to one of the
TPH compounds (n-hexane) can affect the central nervous system causing a nerve
disorder called ‘peripheral neuropathy’ characterised by numbness in the feet and
legs and, in severe cases, paralysis (Hutcheson et al. 1996).
2.4.Pollutant Processes
2.4.1. Build-up of Pollutants
Build-up of pollutants refers to the accumulation of pollutants on pervious and
impervious surfaces through a complex spectrum of dry weather processes such as,
deposition, wind erosion and street cleaning that occur between storm events (Huber
1986). Duncan (1995) described the pollutant build-up process as a dynamic
equilibrium between pollutant deposition and removal between pollutant sources and
sinks. Pollutant build-up in urban areas vary with anthropogenic activities such as
21
concentration of population, commerce and industry, land use and average daily
traffic (Sartor & Boyd 1972; Novotny & Goodrich-Mahoney 1978; Whipple et al.
1974).
Natural sources are also attributed to the build-up of pollutants. Egodawatta et al.
(2009) reported that particles originate mostly from atmospheric sources during their
build-up on roof surfaces. Urban traffic and agricultural activities have been
identified as the principal contributors to pollutant build-up on pervious surfaces as
well (Chen et al. 2010; Oliva & Espinoza 2007). The primary factors influencing
build-up of pollutants can be summarised as follows:
• Climatic conditions
• Land use
• Vehicular traffic
• Pavement texture
• Days since last rainfall or antecedent dry period
• Days since streets were last cleaned
• Method of street cleaning
Changes in climatic conditions affect various weather processes such as rainfall
intensity, frequency and duration, atmospheric temperature and snowfall (IPCC
2000), which in turn, have specific impacts on pollutant build-up. The dry weather
period between two consecutive rain events has been described as a principal factor
relating to the pollutant build-up on urban roads (Sartor et al. 1974). However, the
role of the dry weather period in the accumulation of pollutants has been questioned
by some researchers (Gupta & Saul 1996; Deletic 1998; Deletic & Orr 2005;
22
Novotny et al. 1985). They argued that particle resuspension and redistribution on
road surfaces can cause the antecedent dry period to be only very weakly correlated
with the build-up loading. Egodawatta (2007) found that the rate of increase in
pollutant build-up on road surfaces is very low after seven dry days from initial
build-up and the total build-up varies quite significantly for different land uses.
Kreider et al. (2010) found that the characterisation of roadway particles, tyre wear
particles and tread particles are totally different during their build-up and suggested
that the interactions between tyres and the driving surface may be one of the
principal reasons for such characterisation. The traffic volume, turbulence induced
by the vehicle speed as well as spills from the vehicles also influences pollutants
build-up on urban roads. Sartor and Boyd (1972) illustrated the idealised view of
pollutant build-up on street surfaces as shown in Figure 2.1.
Figure 2.1 Illustration of pollutant build-up on road surfaces
Pollutant accumulation on road surfaces have been considered as a function of all of
the influencing parameters discussed above. It is not possible to derive the slope of
the rising curves in Figure 2.1 due to the uncertainty of the primary build-up
23
parameters (Sartor & Boyd 1972). The mathematical modelling of the process has
typically been treated as a linear, exponential, power, log-normal or stochastic
function (Baffaut & Delleur 1990; Charbeneau & Barrett 1998; Grottker 1987;
Haiping & Yamada 1998; Kuo et al. 1993; Vaze & Chiew 2002). However, limited
data sets and large data scatter has made the form of relationships hard to determine
(Duncan 1995; Whipple et al. 1974). An important point in this respect has been
raised by Bertrand-Krajewski (2006) who noted that the build-up rate as well as the
decay (removal) rate of pollutants are correlated and cannot be calibrated
independently. Therefore, the build-up and wash-off phenomenon must be
considered simultaneously when incorporated into any model.
The discussions on pollutant build-up on road surfaces also lead to another question
as to whether the process starts from zero or not. Irish et al. (1998) supported the
hypothesis that pollutant accumulation starts from zero after a rain event. In this
respect, Vaze and Chiew (2002) illustrated the concepts of “source limited” and
“source unlimited or transport limited”. When the pollutants accumulated are
completely removed due to a rainfall event, the wash-off behaviour is termed as
source limited (Adams & Papa 2000). On the other hand, where the pollutants
accumulated on impervious surfaces is not completely removed or pervious surfaces
that continue to supply pollutants because of erosion may be referred to as source
unlimited or transport limited. The total surface pollutant load is assumed to remain
the same in the source unlimited phenomena. These concepts are useful for
determining the accumulated pollutant load from the governing wash-off behaviour.
The hypothetical representation of source limited and source unlimited build-up and
wash-off of pollutants is illustrated in Figure 2.2.
24
Figure 2.2 Hypothetical representation of build-up and wash-off of surface pollutant loads for
(a) source limited & (b) source unlimited case (adapted from Vaze and Chiew 2002)
Vaze and Chiew (2002) have pointed out that the major limitation in pollutant
accumulation studies was that most have inferred the accumulation from the
measurements of pollutant wash-off. However, this is not the case if wash-off was
transport limited and there was pollutants remaining on the surface after a rainfall
event. Driver and Troutman (1989) found that the wash-off behaviour of pollutants
in urban areas is location specific varying with local climatic characteristics, land
use as well as physical characteristics of the urban catchment. Therefore, specific
build-up and wash-off studies can enhance the understanding of the pollutants build-
up process and relate this to the subsequent wash-off.
2.4.2. Wash-off of Pollutants

Pollutant wash-off is the process of erosion or dissolution of constituents from a
catchment surface during a runoff event (Adams & Papa 2000). According to Bujon
et al. (1992), two simultaneous phenomena control pollutant wash-off. Firstly, as
rain falls on the ground, it initially wets the surface. The resulting impact of the
raindrops and the horizontal overland sheet-like flow provide the necessary
turbulence for dissolving the available soluble fraction of the pollutants. Secondly,
25
the rainfall impact causes detachment of pollutants and their transportation by
surface runoff. The impact of rain drops dislodges the particulate fraction and the
turbulence of the overland sheet-like flow keeps them in suspended form. As the
rainfall intensity increases, surface runoff is initiated, which carries the particulate
fraction and the dissolved pollutants into receiving water bodies. The particulate
pollutants can either be suspended or roll along the ground surface depending on the
flow velocity (Overton & Meadows 1976). Factors affecting the wash-off of
pollutants are essentially rainfall parameters such as, intensity, frequency, duration,
overland flow characteristics that introduce hydrodynamic interactions between
turbulent water flow and settlelable solids, chemical properties such as pH and
electrical conductivity of runoff as well as the impervious surface texture.
The effects of antecedent dry period on pollutant wash-off have been described as
insignificant (Deletic & Maksimovic 1998). It has been suggested by Deletic and
Maksimovic (1998) that enough sediment is always available on the road surface for
wash-off by a rainfall event for its entire duration. Only a few events with large
precipitation volumes and high rainfall intensities have the capacity to wash-off
sediments that have been deposited on the surface, entirely. The transport limiting
cases in wash-off as mentioned by Vaze and Chiew (2002) was in conformity with
these findings. Egodawatta et al. (2007) also supported these studies by introducing
a wash-off capacity factor for different rainfall intensities in describing a
mathematical pollutant wash-off equation. However, the wash-off capacity of a
rainfall event may vary with different pollutants. Soonthornnonda et al. (2008) have
proposed transport coefficients for seven heavy metals along with several physico-
26
chemical and microbiological pollutants in wash-off and ranked the pollutants
according to their availability for wash-off.
Similar to the relationship between antecedent dry period and pollutant wash-off, the
relationship between wash-off and the phenomenon of first flush has also drawn
much attention. The first flush phenomenon refers to the discharge of a large
fraction of the pollutant load in the earlier part of a runoff event. Sansalone and
Christina (2004) defined first flush as the first 20% of the flow volume that contains
80% of the total pollutant load (total dissolved solids and suspended solids). They
found that in several case studies, a relatively large runoff volume must be captured
to effect meaningful reductions in mass and concentration of pollutants, despite a
disproportionately high mass delivery early in the event. A study by Deletic (1998)
on two European catchments found that the presence of first flush is very weak and
rare for suspended solids. In this study, the percentage of total pollutant load in the
first 20% of the runoff was regarded as the first flush. Hall and Ellis (1985)
suggested that the first flush phenomenon is over emphasised and only 60-80% of
storms exhibit an early flushing behaviour. Adams and Papa (2000) suggested that it
may not be prudent to assume the existence of a first flush in all cases as this
phenomenon is more distinct in small catchments with high impervious areas.
The wash-off phenomenon has been formulated primarily as an exponential function
with a first order decay (removal) rate (Sartor & Boyd 1972). The sediment transport
during wash-off has also been formulated as combinations of power and exponential
functions incorporating rainfall energy, critical shear stress, overland flow rate and
spatial density of particles on the surface (Shivalingaiah & James 1984; Tomanovic
27
& Maksimovic 1996; Shaw et al. 2006). These models are generally regarded as
deterministic stormwater quality models (Huber 1986). Recently, another approach
to describe pollutant wash-off has been proposed which is referred to as the
probabilistic approach (Chen & Adams 2007). This approach has incorporated
probability distribution functions of rainfall volume, duration and inter-event time
from historical rainfall records and transformed these using deterministic pollutant
wash-off load models to generate probability distribution functions for event mean
concentrations of pollutants in runoff. However, a single mathematical model or a
simplistic modelling approach (either deterministic or probabilistic) cannot
encompass all of the factors that influence the wash-off processes as the generation
and transport of pollutants in urban systems during a storm event is multifaceted and
it involves many media in both spatial and temporal scales (Ahyerre et al. 1998).
Therefore more research is required to acquire an in-depth understanding of the
processes involved.
2.5.Urban Traffic and its Impact on Water Quality

2.5.1. Road Traffic: Australia Wide Perspective
The Bureau of Infrastructure, Transport, and Regional Economics (BITRE) of
Australia has provided a comprehensive summary of transport activities in Australia
(BITRE 2008). According to the publication, total travel in the urban areas of
Australia has grown remarkably - almost nine-fold over the last 50 years. Almost all
of that growth came from cars and light commercial vehicles used for private
purposes. The total vehicle kilometres travelled (kilometres travelled by a vehicle in
a year) in 2005, by passenger cars was ninety four times higher than that of buses. It
also reported that there were 13.9 million motor vehicles in Australia for the year
2005 which travelled a total of 206 billion kilometres.
28
This trend of dominance in numbers by passenger cars is still continuing as sales of
passenger cars was three times higher than other vehicles in the financial year
2006/07. The stock of light commercial vehicles also increased by four times in
2006 compared to 1971. The total road traffic for cars, light commercial vehicles
and buses in 2004 increased by 2, 1.68 and 1.75 times respectively compared to
1977 in all Australian Metropolitan areas. In the Gold Coast region in South East
Queensland, the percentage increases in the vehicle kilometre travelled (VKT) are
expected to be 39% for 2011 and 74% in 2021 compared to 2000, if current
conditions were to remain unchanged (Brown et al. 2004).
From the statistics relating to road traffic, it is evident that in Australian major cities,
traffic will continue to increase for the foreseeable future. The road traffic is a
constituent of a much broader Australian transport framework which has four main
components (BITRE 2008):
• The relationship between the transport industry and the rest of the Australian
economy;
• Freight and passenger transport activity;
• Transport activity by mode of transport; and
• Impacts of transport - transport safety, transport energy and the environment.
Among these four components, the environmental impact and more specifically the
water quality impact of road transport is the least explored (BITRE 2008; Brown et
al. 2004).
29
2.5.2. Impact of Road Traffic on Water Quality
The major pollutants in water bodies that are generated by transport activities are
polycyclic aromatic hydrocarbons (PAHs), total petroleum hydrocarbons (TPHs),
BTEXs (benzene, toluene, ethylbenzene, xylene) and heavy metals such as
cadmium, chromium, iron, nickel, vanadium, lead, aluminium, zinc and copper
(Peterson & Batley 1992; Sansalone & Buchberger 1997). Automotive parts such as,
tyres and brakes contain a variety of heavy metals listed above. Seen over the entire
lifetime of a passenger car tyre that is in use for 50,000 km, 1 kg of tyre tread rubber
is worn on average. Brake linings are also worn by 12-18 mg/km and have to be
changed after 80,000 km which means that roughly 1 kg of brake lining material is
worn over its lifetime (Öko-Test 2002). Lee and Dong (2010) have reported that
traffic volume as well as vehicle speed are strongly correlated to the PAH
concentrations in urban road dust.
The atmospheric BTEX concentrations are also reported to increase due to
congestion in urban roads (Buczynska et al. 2009). When deposited on the road
surface and subsequently washed-off, these toxic and carcinogenic pollutants can
impair the water quality of the receiving water body. In this respect, the US EPA
(1996) estimates that up to half of suspended solids and a sixth of hydrocarbons
reaching water bodies actually originate from roads. The impacts of urban traffic are
also indirectly related to the costs associated with maintaining the water quality
standards in urban areas. In a study relating to the cleanup costs incurred in relation
to water quality impacted due to motor vehicle operation, Nixon and Saphores
(2007) estimated that 2.9 billion to 15.6 billion US dollar will need to be spent
annually in the United States over a period of next 20 years.
30
When a pollutant is discharged from a motor vehicle, it may be initially emitted to
the atmosphere or deposited on impervious or pervious surfaces. The pollutants in
the atmosphere may fall and accumulate on plants and buildings via interception
processes or on impervious or pervious surfaces through the mechanism of dry
deposition. During a rainfall event, the pollutants that have accumulated in the
atmosphere during the antecedent dry period may be washed out of the atmosphere
via wet deposition. Those that are accumulated on surfaces throughout the
catchment including the roadways are removed via runoff. Pollutants accumulated
on pervious surfaces may be removed via erosion. The accumulated pollutants from
the above processes may find their way into the receiving water bodies, while
resuspension processes also take place by the movement of pollutants from the
catchment surfaces back to the atmosphere. Several other physical processes have
also been described as means of pollutants transport into water bodies such as,
advection, molecular diffusion, dispersion and sedimentation (Walker et al. 2006;
Hvitved-Jacobsen et al. 2010). Pollutants may also undergo chemical and biological
processes during runoff. Given this complexity, simple empirical relationships
cannot fully cover all aspects of the relationship between transport sources and water
quality.
Two separate studies in Australia and New Zealand (Brown et al. 1998; Gardiner &
Armstrong 2007) on the impact of road runoff on water quality have suggested that
the total vehicle kilometres travelled (VKT) is a traffic risk indicator for
understanding the pollution risk to sensitive receiving environments (SREs) such as
water bodies close to a highway. The Australian study proposed a model known as
TRAEMS (Transport Planning Add-on Environmental Modelling System) and the
31
New Zealand study proposed VCLM (Vehicle Contaminant Load Model). Both of
these models assume a simplified pollutant accumulation process to receiving water
bodies from road traffic. The water quality module of TRAEMS model uses:
• The total VKT on roadways within a catchment (or sub-catchment) as a
surrogate measure for substances which may pollute water bodies;
• The assumption that roadway emissions within a particular catchment
will largely be washed off within that catchment;
• The Relative Potential Pollution Load instead of absolute pollution load
as a quantifiable variable amongst different catchments and sub-
catchments.
The VCLM model, on the other hand, estimates the relative pollutant source
strength, expressed as annual mass loads of specific pollutants (particulate matter,
zinc and copper) from road networks. Loads are derived from traffic flow (average
annual daily traffic), service level (speed/congestion), vehicle type and pollutant
emission rates from brake, tyre and road surface wear, oil leakage and exhaust
emissions.
Several other researchers elsewhere have also investigated the characterisation of
pollutant loading from highway runoff (for example Chui et al. 1982; Kayhanian et
al. 2003; Laxen & Harrison 1977; Wu et al. 1998). Wu et al. (1998) used event mean
concentration (EMC), site median EMCs and long term average pollutant loading
rates as the three main parameters for the characterisation of pollutants from
highway runoff. Atmospheric deposition may contribute a significant amount of
pollutant loads that are later transported in highway runoff (Wu et al. 1998).
32
Harrison and Wilson (1985) found that rainfall contributed up to 78% of the major
ionic constituents (Na, K, Mg, Ca, Ca, and SO4) and 48% of suspended solids in
highway runoff. The surrounding land use of a highway corridor may also affect the
amount and type of deposition on roadways during both wet and dry periods. Davis
et al. (2001) reported that pollutant loads typically follow the pattern: Zn (20-5000
µg/L) > Cu ≈ Pb (5-200 µg/L) > Cd (<12 µg/L) and their empirical study revealed
that brake wear is the largest contributor to copper loading (47%) in urban runoff
while tyre wear contribute 25% of zinc loading.
Gupta et al. (1981) reported that highways in or near urban areas carried higher
levels of pollutant loadings originating from dustfall than those in rural areas. In
another study, Kayhanian et al. (2003) found that there was no simple linear
correlation between highway runoff event mean concentrations and AADT (average
annual daily traffic). However, with the inclusion of other factors such as antecedent
dry period, seasonal cumulative rainfall, total event rainfall and maximum rainfall
intensity, drainage area, and land use, Kayhanian et al. (2003) found AADT has a
significant influence on highway runoff pollutants concentration. Chui et al. (1982)
suggested that vehicles during a storm (VDS) are a better independent variable for
estimating total runoff loads for certain pollutants.
2.6.Climate Change and its Impact on Water Quality

2.6.1. Climate Change: Global and Australian Perspective
In 2007, the Intergovernmental Panel on Climate Change (IPCC), a panel of leading
international climate scientists released their fourth assessment report (IPCC 2007),
concluding that:
• Warming of the earth’s climate system is unequivocal;
33
• Humans are very likely to be causing most of the warming that has been
experienced since 1950;
• It is very likely that changes in the global climate system will continue well
into the future, and that these will be larger than those seen in the recent past.
The global climate change is due to the build-up of excessive greenhouse gases
(mainly carbon dioxide, but also methane, chlorofluorocarbons and nitrous oxide) in
the atmosphere that absorbs the outgoing solar radiation (Soh et al. 2008).
According to IPCC (2007) report, most of the observed increase in global average
temperatures since the mid 20th century is due to the increase in greenhouse gas
emissions from human activities.
The IPCC special report on emission scenarios (SRES) has predicted six main
scenarios that will affect the future climate of the world (IPCC 2000). Table 2.2
gives a brief outline of these future scenarios:
Table 2.2 Predicted emission scenarios due to climate change

Scenarios Description
A1F1 A world with rapid economic and global population growth and the energy system is totally
fossil intensive.
A1T A world with rapid economic and global population growth and the energy system is non-fossil
intensive.
A1B A world with rapid economic and global population growth and the energy system is balanced
across all sources.
A2 A world with heterogeneous economic and continuous population growth. Economic
development is regionally oriented.
B1 A world with global population growth as A1 but the economic change is based on clean and
resource efficient technologies.
B2 A world with local solutions to economic, social and environmental sustainability but the
continuous population growth is lower than A2.
For predicting the effect of climate change on rainfall in the Australian continent,
CSIRO has adopted mainly the A1B scenario (CSIRO 2007). The current research
study adopted the CSIRO predictions as basis for rainfall characteristics for South
East Queensland based on the A1B scenario.
34
Studies by CSIRO (2003, 2007) have analysed the impact of this global climate
change on the Australian continent. According to these studies, Australia’s climate is
strongly influenced by the surrounding oceans. Key climatic features in Australia
include tropical cyclones and monsoons in Northern Australia; migratory mid-
latitude storm systems in the south; and the El-Nino – Southern Oscillation (ENSO)
phenomenon, which causes floods, prolonged droughts, and bushfire outbreaks,
especially in eastern Australia. The ENSO phenomenon is a global coupled ocean-
atmospheric phenomenon (Clarke 2008). The CSIRO (2003) report suggests that
ENSO has a strong influence on the climate variability in many parts of Australia,
and this will continue. However, global warming due to increased greenhouse gas
emissions in the atmosphere will enhance the drying of the Australian continent
associated with ENSO events. Hence, the resulting changes to climate across
Australia will lead to variability of the climate characteristics, including the rainfall
characteristics, in Australia.
The envisaged climate change challenges across Australia, including Southeast
Queensland are (CSIRO 2003, 2007):
• Increased average temperature;
• Higher minimum and maximum temperatures and more frequent extreme
heat waves;
• reduced annual average rainfall;
• More frequent storms with heavy rainfall and hail;
• Increased average sea level and more frequent incidence of extreme high sea
levels;
• Degradation of water quality.
35
The CSIRO (2007) report has used a probabilistic approach to infer the Australian
local climate change from widely used general circulation models. For example, the
CSIRO projection tool ‘OZCLIM’ (CSIRO 2007) is based on 23 general circulation
models and 6 future emission scenarios for Australia which can generate future
average yearly or seasonal rainfall for any Australian co-ordinate. It assumes that the
Probability Density Functions (PDFs) of a local climate variable is proportional to
the global variable, which quantifies the likelihood of each possible response
(Whetton et al. 2005).
Currently, the OZCLIM tool (CSIRO 2007) can generate total annual rainfall for any
location on the Australian continent for 2030 and 2070. Research is still being
carried out by CSIRO to generate downscaled regional rainfall patterns. Abbs et al.
(2007) have used a dynamic downscaling technique incorporating the CSIRO CC-
MK3 and CSIRO RAMS model to generate 2030 and 2070 average fractional
change for extreme rainfall intensities at 2, 24 and 72 hour durations in the Gold
Coast area. In the most recent study by CSIRO, Abbs and Rafter (2008) used the
same techniques to generate 2030 and 2070 average fractional change for extreme
rainfall intensities at 2, 24 and 72 hour durations in three catchments at the
Westernport region of Melbourne, Victoria. In all these studies only the mean
percentage change compared to the current observed intensities at 2, 24 and 72 hour
durations have been predicted.
Both Abbs and Rafter (2008) and Abbs et al. (2007) have reported that work is
currently being undertaken by CSIRO to use extreme value statistics to convert these
36
results into more meaningful average recurrence intervals (ARI) used by the
hydrological community. The ongoing research will enable the use of rainfall
extreme value statistics in such a way that future climate change projections can be
applied to a defined ARI rather than the mean values projected so far. These studies
also highlighted that the CSIRO CC-MK3 model is able to predict the likelihood of
occurrence of the summertime extreme rainfall events, but it does not produce the
correct climatology for the wintertime extreme rainfall for the Westernport region of
Victoria. Therefore, the regionalisation and the seasonality effects of these models
need to be further investigated. The published results from the Abbs et al. (2007)
study for the Gold Coast region is summarised in Table 2.3.
Table 2.3 Average percentage change in extreme rainfall intensity for the Gold Coast region:
adapted from Abbs et al. (2007)
Durations Region 2030 2070
% change of Range % change of Range
mean between 10th mean between 10th
intensity and 90th intensity and 90th
percentile percentile
2 All Gold +53 26-89 +48 4-91
Coast
Coastal +50 33-65 +35 6-72
Mountains +56 22-96 +65 27-97
24 All Gold +17 8-29 +16 5-28
Coast
Coastal +15 8-23 +13 0-26
Mountains +19 8-32 +19 9-30
72 All Gold +8 0-17 +14 4-23
Coast
Coastal +6 -2-13 +10 -4-23
Mountains +10 0-20 +17 9-24
It is clear from Table 2.3, that higher changes in the rainfall intensities are projected
for shorter duration events in 2030 and 2070 in the Gold Coast region in Australia.
Even though there is no mention of the shorter duration events in Table 2.3, several
climate change studies (CSIRO 2003, 2007; IPCC 2001, 2007) have predicted that
the likelihood of occurrence of shorter duration events with large change in
precipitation intensities is very high due to climate change across Australia.
37
2.6.2. Impact of Climate Change on Water Quality
The possible effects of climate change on water quality cannot be generalised and
should be assessed on a case-by-case basis (Senhorst & Zwolsman 2005). The
receiving water body temperature and certain ion (chloride, phosphate, zinc and
nickel) concentrations have been found to vary with high flow (due to intense
rainfall) and low flow (due to drought) conditions (Senhorst & Zwolsman 2005).
CSIRO (2003) also reported on the effects of climate change on water quality in
Australia. According to this study, the water quality would be affected by changes in
microfauna and flora, water temperature, carbon dioxide concentration, transport
processes that washes off sediments and chemicals into streams and aquifers, and the
timing and volume of water flow. Intense rainfall events will increase surface runoff,
soil erosion, and sediment loadings. As a result, the combined effects will increase
the risk of flash flooding, sediment load and pollution (Basher et al. 1998).
Recently, a more comprehensive description of the impacts of climate change on
water quality parameters has been provided by Delpla et al. (2009). They have
classified the water quality parameters into three main clusters as, physico-chemical
parameters, micropollutants and biological parameters and reviewed the impacts of
climate change on each of them. The results from their review are partially
reproduced in Table 2.4. Delpla et al. (2009) concluded that amongst different water
quality parameters, dissolved organic matter, micropollutants and pathogens are
susceptible to the rise in concentration as a consequence of temperature increase
(water, air and soil) and heavy rainfalls in temperate countries.
38
Table 2.4 Climate change impacts on water quality parameters partially adapted from Delpla
et al. (2009)
Climate
Change Factors
Water Quality Parameters Water Body
affecting Water
Quality
Draughts Rivers
Temperature Lakes
pH
Increase
Rainfall Rivers
Draughts and Rivers and lakes
Basic Parameters
Temperature
DO Increase
Rainfall River
Draughts
Rivers
Temperature Temperature
Lakes
Physico-Chemical Increase
Temperature and Streams and
DOC
Rainfall Increase Lakes (Peatlands)
River
Droughts
Surface and
Temperature Groundwater
Increase Lakes
Nutrients
Temperature and River basins, Lake
rainfall increase and Groundwater
Heavy Rainfall Streams and

Lakes
River
Droughts
Temperature
High Alpine
Increase
Lakes
Inorganic Metals
Heavy Rainfall
Streams
Temperature and
rainfall Increase
Streams
Micropollutants Temperature and Surface and
rainfall Increase Groundwater
Pesticides Drying and re-

wetting Cycles
Organic
Rainfall Increase Streams
Temperature Groundwater
Increase
Pharmaceuticals
Rainfall Streams
Temperature and
Pathogens Surface Waters
rainfall Increase
Temperature and
Cyanobacteria Lakes
rainfall Increase
Biologicals
Temperature
Cyanotoxins Lakes
Increase
Fish, green algae, Temperature
Freshwaters
diatoms Increase
39
A report by the European Environment Agency (EEA 2003) suggested some
possible effects of climate change on water quality including increased contaminant
discharge (during floods), rising temperature and oxygen depletion (during
droughts) and changes in aquatic ecology. In Australia, the predicted range of
precipitation is annually averaged around -10% to + 5% in northern areas and -10%
to little change in southern areas of Australia taking 1990 as the base year
precipitation (CSIRO 2007). Chiew and McMahon (2002) modelled the surface
runoff from this predicted change in precipitation. According to their study, the
annual runoff for South Eastern Australia could decrease by up to 20%. South
Australia would experience up to 25% decrease in annual runoff and for Western
Australia a change of -25% to +10% is predicted. CSIRO (2007) also predicted that
the mean annual precipitation in South East Queensland will decrease by 1% - 2.5%
while the precipitation intensity (mm of rain per rain day) will increase by 0 - 0.35
mm per day in 2030 for the A1B scenario. These quantitative predictions on
precipitation, intensity and surface runoff will have a qualitative impact on the water
resources of Australia. Therefore, the relationship between pollutant wash-off and
the rainfall characteristics would provide the necessary information on how to
predict pollutant loads and concentrations under changed rainfall conditions due to
climate change.
From the above discussions on climate change and its impact on water quality, it is
clear that rainfall is an important factor in relation to climate change. Hence, the
physical characteristics of rainfall such as intensity (causing high flow and low flow)
and antecedent dry period (causing drought) are of particular interest in relation to
pollutant wash-off. Similarly, the chemical characteristics of runoff such as
40
temperature, certain ion concentrations are also factors for investigation in relation
to the amount of pollutants present in wash-off under changed climatic conditions.
2.7.Conclusions
Water quality in urban areas is impaired due to point source and non-point source
water pollution. Various anthropogenic and natural sources such as, treatment plants,
mining, construction as well as forestry can cause the point and non-point source
water pollution. Amongst these, urban traffic is one of the principal non-point
sources of water pollution and poses a significant threat to urban water quality under
increased urbanisation. Climate change is also responsible for changes in the rainfall
characteristics which can impair urban water quality through the pollutant wash-off
from urban roads.
Primary water pollutants can be classified into five major classes, gross pollutants,
solids, nutrients, oxygen demanding materials and toxicants. However, there are
hundreds of pollutants found in stormwater that might fall into these categories and
it is not feasible to detect and quantify each of them. Therefore, the pragmatic
approach is to focus on individual pollutants and investigate their specific behaviour
in relation to the corresponding research objectives.
The review on the build-up and wash-off of pollutants on the road surface focused
on different concepts presented by past researchers. The primary factors affecting
pollutant build-up on road surfaces are summarised as climate, land use, traffic,
pavement texture, antecedent dry period as well as street cleaning. The concepts of
source limited and source unlimited have also been incorporated into the build-up
process depending on whether the pollutant build-up starts from zero or not. The
41
build-up of pollutants has been characterised as a function of all of the influencing
parameters and several mathematical models such as, linear, exponential, power as
well as stochastic models have been employed.
In pollutant wash-off, the primary influencing factors are identified as rainfall
intensity, frequency and duration as well as overland flow characteristics, pH,
electrical conductivity and surface texture. As opposed to the build-up, the effects of
antecedent dry period on pollutant wash-off were found to be insignificant. Another
important phenomenon known as first flush has drawn much attention from
researchers. Whilst researchers have argued on the extent of first flush in runoff, it
has been commonly accepted that such phenomenon is more distinct in small
catchments with high impervious areas. The pollutant wash-off processes have been
formulated either as deterministic or probabilistic.
The review of urban traffic aspects focused on the increase in traffic volume in the
Australian continent and its environmental impacts especially in relation to water
quality. The percentage increase in vehicle kilometres travelled in South East
Queensland is expected to increase by 39% for 2011 compared to 2000. Some
models that have attempted to relate urban traffic with selected water quality
parameters have also been presented. The water quality impacts of such an increase
in traffic volume are envisaged through substantial heavy metals loadings from
vehicle components, emissions of petroleum hydrocarbons from vehicles as well as
high atmospheric BTEX concentrations in highways subject to intensive traffic.
Researchers have found that the average annual daily traffic (AADT), combined
with other important pollutant build-up and wash-off parameters such as, antecedent
42
dry period, seasonal cumulative rainfall, intensity, land use as well as drainage area,
can significantly affect highway runoff pollutant concentrations.
The effect of climate change on rainfall characteristics in Australia has been
reviewed in this research. The global climate change scenarios that have potential to
affect the regional rainfall characteristics especially in South-East Queensland have
been discussed. The Commonwealth Scientific and Industrial Research Organisation
(CSIRO), which is the leading scientific organisation in Australia, has adopted the
A1B emission scenario to predict the effect of climate change on rainfall in the
Australian continent. This review also covered discussions on several research
reports on the future predictions of rainfall intensities in various Australian regions
including South-East Queensland. The effects of climate change on water quality
have been reviewed in this research and different water quality parameters such as,
dissolved organic matter, micropollutants and pathogens are found to be susceptible
to rise in concentration as a result of temperature increase as well as changes in the
rainfall characteristics. Additionally, a 20 % decrease of the annual runoff for South
Eastern Australia, 25% decrease in annual runoff for South Australia and a change
of -25% to +10% for Western Australia are predicted as a result of climate change.
The review of research literature has covered relevant aspects of the study objectives
and forms the foundation for the research study. The field data and sample
collection, laboratory analyses, data interpretation, and publication of results were
undertaken in the next stages of this research. The research methodology
encompasses discussions on these subsequent steps.
43
44
CHAPTER 3 RESEARCH DESIGN AND
METHODS
3.1.Background
In order to study the impacts of urban traffic and climate change on stormwater
quality, a robust research methodology that incorporates the build-up and wash-off
of traffic related pollutants on urban roads was needed. The overall research study
was formulated in terms of three distinct phases, namely, literature review,
characterising and predicting selected pollutants under dynamic conditions and
research outcome. Whilst the previous chapter was dedicated to the literature review
phase, the research methodology was concerned with logical steps that enabled the
characterisation and prediction of selected stormwater pollutants under changed
urban traffic and climate change. This chapter outlines the detailed steps that were
undertaken to formulate the methodology of this research study. The outcome of the
research was influenced by the careful selection of methodologies as well as the
research strategy that was followed throughout the study.
3.2.Research Methodology
The research methodology consisted of five steps. These were as follows:
• Site Selection
• Sample Collection
• Sample Testing
• Data Analysis
• Publication of Results
45
The research study commenced with selecting suitable study sites for build-up and
wash-off sampling. An efficient methodology was developed in order to collect the
build-up samples more efficiently than a conventional vacuuming system. To
understand the specific impact of changed rainfall conditions on wash-off, the
research methodology adopted a wash-off sample collection method that
incorporated future rainfall scenarios. Additionally, detailed discussions on the test
methods for various water quality parameters as well as the data analyses techniques
are discussed. Finally, a brief discussion on the publication of results that helped to
disseminate the tangible outcome of this research is presented.
3.3.Site Selection
The site selection criteria for this research study was based on a suburb based
approach. Two suburbs in the Gold Coast region in Southeast Queensland, Australia
were chosen. The two selected suburbs also represented the transport infrastructures
that were developed within the Gold Coast City Council area in the past decade. The
suburbs were Helensvale and Coomera. The sites selected for sample collection for
build-up and wash-off incorporated three types of land uses, namely, residential,
commercial and industrial. As the objective of this study was to relate pollutant
build-up on road surfaces with urban traffic, the selection of different land uses
provided a cross-section of traffic activities on road surfaces within the Gold Coast
region. Eleven road sites in the two suburbs were selected. Table 3.1 shows the
selected sites and their respective traffic parameters as provided by the Gold Coast
City Council.
46
Table 3.1 Traffic data for the selected study sites
Volume Volume Volume
Predicted to Predicted to Predicted to
Site Name Daily Traffic Capacity Daily Traffic Capacity Daily Traffic Capacity
2011 Ratio 2016 Ratio 2021 Ratio
V/C V/C V/C
Abraham Road 8149 0.57 11414 0.86 10314 0.83
Reserve Road 8144 0.61 8733 0.62 10713 0.26
Peanba Park road 6420 0.76 8652 1.03 4423 0.6
Billinghurst Cres 628 0.14 1411 0.33 812 0.25
Beattie Road 3822 0.39 4117 0.55 2935 0.36
Shipper Drive 2501 0.39 4074 0.49 3765 0.57
Hope Island Road 26506 0.64 27619 0.71 29697 0.76
Lindfield Road 14091 1.21 14698 1.25 15175 1.28
Town Centre Drive 9860 0.31 15318 0.45 19023 0.43
Dalley Park Drive 2888 0.26 3080 0.27 3063 0.27
Discovery Drive 6856 0.46 6454 0.47 6334 0.45
Table 3.2 below shows the Average Annual Daily Traffic (AADT) and average
traffic speed for the selected sites which were also provided by the Gold Coast City
Council.
Table 3.2 Traffic data (set 2) from the Gold Coast City Council
AADT -
Specific Data Average
Survey Date
Land Use Suburb Site name Set from Vehicle
& Data
Survey Speed
Totals
Commercial Coomera Abraham Road May 05 North 3,523 N 81km
8,755 South 5,232 S 80km
Jun 05 North 3,817 N 60km
7,955 South 4,138 S 63km
Aug 06 East 4,145 E 94km
9,727 West 5,582 W 74 km
Residential Coomera Reserve Road Nov 02 East 5,294 E 68km
10,643 West 3,694 W 61 km
Jun 04 East 3,369 E 42km
7,063 West3,694 W 46km
Aug 06 East 4,878 E 68km
9,674 West 4,796 W 68km
Residential Coomera Peanba Park Rd NO SURVEY NO SURVEY NO SURVEY
Residential Coomera Billinghurst Cres Aug 07 East 268 E 56km

499 West 231 W 54km
Industrial Coomera Beattie Rd Feb 01 East 2,084 E 71km
4,168 West 2,084 W 71km
Feb 03 East 2,758 E 77km
5,460 West 2,702 W 79km
Industrial Coomera Shipper Dr NO SURVEY NO SURVEY NO SURVEY
Commercial Helensvale Hope Island Rd DEPT OF DEPT OF DEPT OF

MAIN ROADS MAIN ROADS MAIN
ROADS
47
Table 3.2 Traffic data (set 2) from the Gold Coast City Council (Contd.)
AADT -
Specific Data Average
Survey Date
Land Use Suburb Site name Set from Vehicle
& Data
Survey Speed
Totals
Commercial Helensvale Lindfield Rd NO SURVEY NO SURVEY NO SURVEY
Commercial Helensvale Town Centre Dr NO SURVEY NO SURVEY NO SURVEY
Residential Helensvale River Links Blvd NO SURVEY NO SURVEY NO SURVEY
Residential Helensvale Dalley Park Dr NO SURVEY NO SURVEY NO SURVEY
Residential Helensvale Discovery Dr May 08 North 7,283 N 60km

15,152 South 7,869 S 62km
Aug 08 West 6,039 W 69km
11,784 East 5,745 E 66km
Industrial Helensvale Shelter Rd NO SURVEY NO SURVEY NO SURVEY
The build-up sample collection on selected road sites was undertaken with a view to
relating the pollutant load and concentration to the traffic data provided by the Gold
Coast City Council. This study incorporated average daily traffic (ADT) instead of
average annual daily traffic (AADT), as the former is predicted by a sophisticated
transport model called ZENITH (GCCC 2006) which is currently being used by the
Gold Coast City Council. AADT does not have any direct correlation with pollutant
build-up on road surfaces (Kayhanian et al. 2003 & 2007). Gardiner and Armstrong
(2007) also found that traffic levels measured as AADT were a poor proxy for
runoff quality.
The site selection criteria for this study did not incorporate the vehicle speed as an
independent variable to the pollutant build-up on the road surface. This was due to
the fact that the mean vehicle speed has been found to be constant at the maximum
capacity of a traffic lane or a roadway (Ogden & Taylor 1999). As the service flow
rate (SF) decreases, the vehicle speed tends to increase and attain its maximum
designated value at the lowest flow rate. Except for the peak hour capacity, the
variation in speed from its maximum designated value over a stretch of roadway is
48
insignificant. The volume to capacity ratio (V/C) of a roadway describes the traffic
activities on the stretch of road during the peak hour (Ogden & Taylor1999). This
parameter was found to vary quite significantly for the different sites that were
selected for the study. Past researchers have used a vehicle kilometre travelled
(VKT) in catchment based approaches as a measure of total traffic activities within
the catchment (Tomerini & Brown 1998; Gardiner & Armstrong 2007). However, it
was hypothesised in this study that vehicle congestion due to increased traffic
volumes in urban areas would directly affect the pollutant build-up on urban roads.
Hence, this study adopted a research methodology selecting average daily traffic
(ADT) and volume to capacity ratio (V/C) as the two principal traffic characteristics
that affected the pollutant build-up.
Surface roughness of the road can affect the vehicle dynamics in terms of potential
body damage and variation in speed (Shahin 2005). The US federal highway
administration recommended specific surface texture depths to be applied to
pavements so that current and predicted traffic needs could be accommodated in a
safe, durable, and cost effective manner (FHWA 2005). Texture depths in the
pavement surface can affect pollutant build-up and wash-off from these surfaces.
Legret and Colandini (1999) found that a porous asphalt pavement affects runoff
water quality by reducing the mean pollutant loads in the runoff. In their study,
Legret and Colandini (1999) reported higher accumulation of heavy metals in porous
asphalt pavements in a comparison with a nearby catchment drained by a
conventional stormwater drainage system.
49
Even though the selected study sites were regarded as impervious asphalt
pavements, the texture depths of the pavement can also affect the pollutant build-up.
The road texture was found to affect the interactions between the vehicle tyres and
the driving surface (Kreider et al. 2010). Hence, the surface texture depths of the
pavement at the selected road sites were also investigated in the study.
3.4.Sample Collection
3.4.1. Build-up Sample Collection
Deletic and Orr (2005) used a “Wet Method” for build-up sample collection. They
used a commercially available sprayer to deliver sufficient pressure to dislodge fine
sediment particles without destroying the road surface. However, in their method
they have not mentioned how to achieve the maximum recommended water pressure
for different road surfaces. Vaze and Chiew (2002) adopted a different approach by
separating the build-up samples into “free load” and “fixed load”. They used a
conventional vacuum cleaner to collect whatever was available on the road surface
as “free load”. This was followed by a scrubbing technique using a fibre brush to
dislodge the “fixed load”. The scrubbing with brushes has the potential to disturb the
surface coating of the asphalt.
An improved procedure for sample collection referred to as “wet and dry vacuum
system” was proposed in this investigation. An experiment to standardise the
maximum recommended pressure using a commercially available sprayer was also
undertaken. The improved build-up sample collection procedure is described below.
3.4.2. The Wet and Dry Vacuum System
A. Equipment
• Vacuum Cleaner (Delonghi Aqualand Water Filter System).
50
• 12 volt electric sprayer (Swift 60 L Compact Sprayer with pressure control
module and pressure gauge attached).
B. Definitions
• Dry samples: Samples collected by the vacuuming system before spraying
water on the road surface.
• Wet samples: Samples collected by the vacuuming system after spraying
water.
A. Standardisation of the Build-up Sample Collection Procedure
1. Three 1 m x 1 m plots on a bituminous surface adjacent to each other were
selected.
2. Using a water hose, each of these plots were cleaned of pollutants.
3. The plots were allowed to dry for an hour. It was assumed that during the 1
hour dry period, the pollutant build-up on the three plots were similar.
4. In the first plot (Plot 1), 100.00 gm fine sand that passed 100% through 420
µm sieve and retained 100% on 0.7 µm Whatman® GF F glass fiber filter
was applied. Then the vacuum system was used to collect the dry sample.
Deionised water was sprayed from the sprayer at 3 bar pressure for 3
minutes. Then the wet sample was collected by the vacuum cleaner.
5. On the second plot (Plot 2), 100.00 gm fine sand was applied again. Then the
vacuuming system was used to collect the dry samples. Deionised water was
sprayed from the sprayer at 2 bar pressure for 3 minutes. Then the wet
sample was collected by the vacuum cleaner.
51
6. In the third plot (Plot 3), the dry sample was collected by the vacuuming
system. Deionised water was sprayed from the sprayer at 1 bar pressure from
very close to the road surface for 3 minutes. Then the wet sample was
collected by the vacuum cleaner.
7. The difference in weight between the dry samples in step 4 and step 6 as well
as step 5 and step 6 were determined. At the same time, the difference in
weight between the wet samples from step 4 and step 6 as well as step 5 and
step 6 were determined. The total weights of the dry and wet samples were
compared to the 100.00 gm spread at each plot provided the efficiency of the
vacuum system.
8. To obtain the optimum spraying pressure, the percentage collection
efficiency of the wet samples at plot 1 and plot 2 were compared.
B. Results and Discussion
Table 3.3 gives the results of this experiment on an asphalt road surface:
Table 3.3 Sample collection data for the wet and dry vacuuming system
Parameters Plot 1 Plot 2 Plot 3
Plot Drying Time after Washing (hr) 1 1 1
Applied Sand (gm) 100.00 100.00 0.00
Dry Sample (gm) 79.14 66.14 2.2
Wet Sample (gm) 14.16 26.11 0.65
Sprayer Pressure (bar) 3 2 1
Efficiency of total collection (%) 90.45 89.4 -
Efficiency of wet sample collection (%) 58.6 70.6 -
Based on the total collection efficiency, it was evident that the vacuum equipment
was working in the range of 90% efficiency (from 89.4% to 90.45%). It was also
evident that the wet sample collection efficiency increased from 58.6% to 70.6%
when the spraying pressure reduced from 3 to 2 bar. As the total collection
efficiency was slightly lower at 2 bar pressure, it was decided not to apply a pressure
52
less than 2 bar. The field data collection used exactly 2 bar pressure with the sprayer
for 3 minutes from a standing position by keeping the nozzle horizontal.
It was also evident that a combination of vacuuming and spraying was essential to
achieve an acceptable efficiency in the field sample collection. A controlled
laboratory environment could give a higher efficiency using only the vacuum
system. However, a laboratory environment does not represent the actual field sites.
This experiment was done on an actual asphalt pavement surface subject to
atmospheric wear and tear as well as daily traffic. Hence, the results obtained closely
resembled the actual sampling environment of the study sites. In all subsequent
discussions, the build-up samples refer to the samples collected by this method.
Additionally, this research adopted a seven day antecedent dry period before
collecting any build-up or wash-off samples in conformity with the findings of
Egodawatta (2007) who noted that the pollutant build-up on road surfaces asymptote
to an almost constant value after a seven day dry period.
3.4.3. Wash-off Sample Collection

The research study used a rainfall simulator (Herngren et al. 2005) to replicate the
required rainfall on road surfaces and a commercially available vacuum cleaner to
collect the wash-off samples. The rainfall simulator was based on the design of
simulators used in agricultural research as described by Floyd (1981), Silburn et al.
(1996) and Loch et al. (2001). It consisted of an A-frame structure made of
aluminium tubing of 40-mm diameter, as shown in Figure 3.1. Three Veejet 80100
nozzles, spaced 1 m apart, were mounted on a stainless steel boom at a height of 2.4
m. This was the prescribed height for creating terminal velocities similar to natural
rainfall for all rain drop sizes (Duncan 1972; Herngren 2005).
53
Figure 3.1 Rainfall Simulator (adapted from Herngren et al. 2005)
The plot area over which the simulated rainfall was created consisted of a frame
connected to a collecting trough that holds 100 L of runoff water. The runoff water
in the collecting trough was vacuumed continuously into 25 L plastic containers and
was sub-sampled into 1 L glass bottles. The water pressure through the Veejet 80100
nozzles was kept at 41 kPa which was found to be the most appropriate pressure to
create drop size distribution near natural rainfall (Bubenzer 1979).
54
To conform to the approach in water sensitive urban design practice, this research
focused on the simulation of current and future rainfall events with ARI varying
from 1 to 100 year. Amongst the eleven study sites described earlier in Table 3.1 for
collection of build-up samples, four sites within 5 km distance from a rain gauging
station were chosen as the wash-off sample collection sites. These were namely,
Billinghurst Crescent, Discovery Drive, Shipper Drive and Lindfield Road. The
rainfall intensity-frequency-duration (IFD) data for this station (station ID 40166)
was generated according to the guidelines stipulated by the Australian Rainfall and
Runoff Volume 1 and 2 (IEAUST 2001). It was assumed that the IFD data for
Station 40166 would uniformly apply to all four wash-off sites due to their close
proximity to the gauging station.
The Third Assessment Report of the Intergovernmental Panel on Climate Change
(IPCC 2001) stated, “Precipitation extreme are projected to increase more than the
mean and the intensity of precipitation events are projected to increase. The
frequency of extreme precipitation events was projected to increase almost
everywhere.” Recently, the IPCC released its “Summary for Policy Makers” (IPCC
2007) that states, “It is very likely that hot extremes, heat waves, and heavy
precipitation events will continue to become more frequent.” These climate change
projections clearly indicate the fact that while intensities of extreme events will
increase more than the mean, the frequencies of these events will also increase due
to climate change. In other words the ARIs of the extreme events will decrease due
to climate change. From the above discussion, the following four possibilities were
summarised as possible scenarios of future climate change effects on rainfall in the
Gold Coast region:
55
1. Shorter duration, with higher intensity with ARI fixed;
2. Lower ARI, with higher intensity with duration fixed;
3. Lower ARI, shorter duration with intensity fixed; and
4. Lower ARI, with higher intensity while duration gets shorter.
The possibility of occurrence of scenario 2 is unlikely because at a fixed duration,
the intensity always gets lower at shorter ARIs in the IFD table. Therefore, the
remaining three scenarios have been simulated to replicate the future climate change
effects on rainfall. In this context, this study chose the year 2030 for future rainfall
simulation as this was one of the years for which the percentage change in rainfall
intensities in the Gold Coast region were predicted by Abbs et al. (2007).
From Table 2.2 in Chapter 2, this study established a logarithmic relationship as
shown in Figure 3.2, between the percentage change of intensity and duration for
2030 in Gold Coast region, for use in this research.
60
y = -12.886Ln(x) + 60.998
50
% change of intensity ofy
R2 = 0.9871
40
30
20
10
0
0 20 40 60 80
Duration, hr
Figure 3.2 Logarithmic relationships between percentage change of intensity and duration in
Gold coast region for 2030
56
It is clear from Figure 3.2 that the percentage change in intensity could be as high as
50% for 5 minute duration of an extreme rainfall event. In fact it was shown in the
study by Abbs et al. (2007), that greater changes in rainfall intensities are projected
for shorter duration events in 2030 and 2070 across the Australian continent
including Gold Coast in South East Queensland. The current intensities at 2, 24 and
72 hours can be easily calculated from the IFD curves for the Gold Coast region for
any ARI. Even though there is no mention of events shorter than 2 hour duration in
Chapter 2, Table 2.2, it can be inferred from the climate change studies (Abbs et al.
2007; CSIRO 2007; IPCC 2007) that the shorter duration events will definitely
undergo the largest amount of percentage change in their corresponding mean
intensities. Figure 3.2 provides a logarithmic equation to calculate this percentage
change in the intensity.
For rainfall simulation, both the normal rainfall events and the extreme rainfall
events were selected from the Intensity-Frequency-Duration (IFD) data (Table 3.4)
generated by AUSIFD version 2.0 software (AUSIFD 2005) for the rain gauging
Station 40166.
57
Table 3.4 Intensity-Frequency-Duration table for Station ID 40166 in Gold Coast region
(AUSIFD version 2.0)
ARI 1 ARI 2 ARI 5 ARI 10 ARI 20 ARI 50 ARI 100
Duration
(min) Intensity Intensity Intensity Intensity Intensity Intensity Intensity
(mm/hr) (mm/hr) (mm/hr) (mm/hr) (mm/hr) (mm/hr) (mm/hr)
5 124 155 186 202 226 256 279
5.5 120 150 180 196 219 248 270
6 116 145 174 190 212 240 262
6.5 113 141 169 184 206 233 254
7 110 137 164 179 200 227 247
7.5 107 134 160 174 195 221 241
8 104 131 156 170 190 216 235
8.5 102 127 153 166 186 211 229
9 99 125 149 162 182 206 224
9.5 97 122 146 159 178 201 219
10 95 119 143 156 174 197 215
11 91 115 137 150 167 190 206
12 88 111 132 144 161 183 199
13 85 107 128 139 155 176 192
14 82 103 124 135 150 171 186
15 80 100 120 130 146 165 180
16 77 97 116 127 141 160 175
17 75 94 113 123 138 156 170
18 73 92 110 120 134 152 165
19 71 90 107 117 130 148 161
20 70 87 105 114 127 144 157
21 68 85 102 111 124 141 153
22 67 83 100 109 121 138 150
23 65 82 98 106 119 135 147
24 64 80 96 104 116 132 144
25 62 78 94 102 114 129 141
26 61 77 92 100 112 127 138
27 60 75 90 98 110 124 135
28 59 74 88 96 108 122 133
29 58 73 87 95 106 120 130
30 57 71 85 93 104 118 128
32 55 69 82 90 100 114 124
34 53 67 80 87 97 110 120
36 52 65 77 84 94 107 116
38 50 63 75 82 92 104 113
40 48.8 61 73 80 89 101 110
45 45.8 57 69 75 84 95 103
50 43.2 54 65 71 79 89 97
58
Table 3.4 Intensity-Frequency-Duration table for Station ID 40166 in Gold Coast region
(Contd.)
ARI 1 ARI 2 ARI 5 ARI 10 ARI 20 ARI 50 ARI 100
Duration
55 40.9 51 61 67 75 85 92
60 39 48.9 58 64 71 81 88
75 33.7 42.4 51 56 63 71 78
90 29.9 37.7 45.7 50 56 64 70
105 27 34 41.5 45.6 51 59 64
120 24.6 31.2 38.2 42 47.5 54 60
135 22.8 28.8 35.4 39.1 44.2 51 56
150 21.2 26.9 33.2 36.7 41.5 47.8 53
165 19.9 25.2 31.2 34.6 39.2 45.2 49.8
180 18.7 23.8 29.6 32.8 37.2 43 47.4
195 17.7 22.6 28.1 31.2 35.5 41 45.2
210 16.9 21.5 26.8 29.8 34 39.3 43.4
225 16.1 20.5 25.7 28.6 32.6 37.8 41.7
240 15.4 19.7 24.6 27.5 31.3 36.4 40.2
270 14.2 18.2 22.9 25.6 29.2 33.9 37.6
300 13.2 16.9 21.4 24 27.4 31.9 35.4
360 11.7 15 19.1 21.4 24.6 28.7 31.9
420 10.5 13.5 17.3 19.5 22.4 26.2 29.2
480 9.62 12.4 15.9 18 20.7 24.3 27.1
540 8.88 11.5 14.8 16.7 19.3 22.7 25.3
600 8.27 10.7 13.8 15.7 18.1 21.3 23.8
660 7.76 10 13 14.8 17.1 20.2 22.6
720 7.31 9.47 12.3 14 16.2 19.2 21.5
840 6.76 8.76 11.4 13 15.1 17.8 19.9
960 6.31 8.18 10.7 12.2 14.1 16.7 18.7
1080 5.94 7.7 10.1 11.5 13.3 15.8 17.7
1200 5.63 7.3 9.54 10.9 12.6 15 16.8
1320 5.36 6.95 9.09 10.4 12 14.3 16
1440 5.12 6.64 8.69 9.92 11.5 13.7 15.3
1800 4.55 5.91 7.75 8.85 10.3 12.2 13.7
2160 4.12 5.36 7.04 8.05 9.37 11.1 12.5
2520 3.79 4.92 6.48 7.41 8.63 10.3 11.5
2880 3.51 4.57 6.02 6.89 8.03 9.55 10.7
3240 3.28 4.27 5.63 6.45 7.52 8.95 10.1
3600 3.08 4.01 5.29 6.07 7.08 8.44 9.48
3960 2.91 3.79 5 5.74 6.7 7.98 8.98
4320 2.75 3.59 4.75 5.45 6.36 7.58 8.53
For normal rainfall, suitable event durations for the simulator were selected from
Table 3.4. For extreme rainfall, Table 3.5 was used to obtain the historical highest
59
daily rainfall intensities for five rainfall stations around the Gold Coast region. Each
of these five stations including Station 40166 are geographically located close to
each other as shown in Table 3.5. It was hypothesised in this study that an extreme
event in one of these stations would occur at each of the selected wash-off sites.
Table 3.5 also shows both the historical highest daily as well as the highest monthly
intensities at these stations and the monthly intensities were higher than the daily
intensities in most stations. However, due to the uncertainties involved in
determining the mean number of days of significant rain in the month, only the daily
intensities were compared to identify the extreme events.
Table 3.5 Historical daily and monthly rainfall data for Gold Coast rainfall stations adapted
from the Bureau of Meteorology climate service (http://www.bom.gov.au/climate/averages/)
Mean
no. of
Elevation days of
Highest Highest Highest Highest
from Data rain≥25
Station Geographic Monthly Daily Daily Monthly
Mean Length mm
ID Location Rainfall Rainfall Intensity Intensity
Sea level (yr) during
(mm) (mm) (mm/hr) (mm/hr)
(m) highest
Monthly
Rain
28.05ºS 1975- 767.1 307
40584 110 1.2 12.79 26.6
153.29ºE 2008 May 1996 11Feb1976
27.94ºS 1994- 471.2 350.8
40764 3 1.9 14.61 10.3
153.43ºE 2008 Feb 2003 30Jun 2005
27.98ºS 1881- 880.1 426.2
40190 17 1.9 17.76 19.3
153.41ºE 2008 Feb 1893 6Feb 1931
27.97ºS 1888- 1790 563.2
40197 515 2 23.50 37.3
153.20ºE 2008 Jan 1974 27Jan 1974
27.90ºS 1894- 921.6 367
40166 6 2 15.29 19.2
153.31ºE 2008 Jan 1974 27Jan 1974
From Table 3.5, it is evident that in Station 40197, there was one single daily rainfall
event of 23.5 mm/hr intensity which exceeded all other intensities in that region over
the past 120 years. Therefore, this event was regarded as an extreme event for all
four selected sites with very high ARI (≥100 year).
60
In Table 3.4, a 100 year ARI event with an intensity of 23.8 mm/hr has a duration of
600 min. The relationship shown in Figure 3.2 between percentage changes in
intensity with duration for year 2030 was used to infer the percentage change of this
extreme intensity and Table 3.4 was used to obtain the corresponding duration for
the event. As a conservative approach conforming to the water sensitive urban
design practices, the ARI for these extreme future events were restricted to 1 year.
Hence, the 23.8 mm/hr with 600 minute duration event was simulated for year 2030
as an event of 1 year ARI with an intensity of 31.25 mm/hr and duration of 85 min.
This simulation satisfied the above mentioned future climate Scenario 4. Table 3.6
shows the simulation plan for the future rainfall events based on future climate
Scenario 1, 3, and 4 for the Gold Coast region for year 2030.
Table 3.6 Future simulation events based on the extreme daily rainfall intensity in the Gold
Coast region for 2030
Current event from Table 3.4 Future event for 2030
Scenario Intensity Intensity
ARI (yr) Duration (mn) ARI (yr) Duration (mn)
(mm/hr) (mm/hr)
Shorter 1 120 24.6 1 65 37.39
Duration,
with Higher
Intensity
while ARI
fixed
Shorter ARI, 10 300 24 1 120 24.6
shorter
Duration
while
Intensity
fixed
Shorter ARI, 100 600 23.8 1 85 31.25
with Higher
Intensity
while
Duration
getting
shorter
The study also used the IFD data for Station 40584 (Table A1; Appendix A) to
generate the normal rainfall prediction for year 2030 satisfying all the three possible
rainfall scenarios. This was done in order to incorporate homogeneity in the rainfall
prediction for the Gold Coast region. The normal rain events were selected in such a
way that each scenario included at least a 100 year ARI current event with long
61
duration (≥45minutes). For want of more sophisticated information, it was assumed
that the relationship between percentage changes in extreme intensity with duration
in Figure 3.2 applies uniformly to all the different ARI events in each scenario.
Table 3.7 shows the planned simulation for normal rainfall for year 2030.
Table 3.7 Future simulation events based on the normal daily rainfall intensity in the Gold
Coast region for 2030
Current event from IFD Table of Future event for Gold Coast region for
Station ID 40584 2030
Scenario
Duration Intensity Duration Intensity
ARI (yr) ARI (yr)
(mn) (mm/hr) (mn) (mm/hr)
Shorter 1 60 39.3 1 25 63
Duration, 2 90 39.3 2 42.5 61.2
with Higher 5 133 39.3 5 69 59.2
Intensity 10 160 39.3 10 85 58.3
while ARI
fixed 100 105 75 100 49 115
Shorter ARI,
shorter
Duration
100 45 125 1 5 125
while
Intensity
fixed
Shorter ARI, 10 52.5 77 5 16 125
with Higher 20 67.5 77 10 21 122
Intensity 50 86.7 77 2 10.5 120
while
Duration
getting 100 101.25 77 1 5.75 119
shorter
Table 3.6 and 3.7 gives the selected rainfall events incorporating the future climate
change scenarios that were simulated at the selected study sites.
3.5.Test Methods
3.5.1. Heavy Metals
The heavy metals selected for this study were cadmium (Cd), chromium (Cr), nickel
(Ni), lead (Pb), zinc (Zn), antimony (Sb), copper (Cu), manganese (Mn), aluminium
(Al) and iron (Fe). This selection was based on the conclusions from the literature
review on heavy metal pollution due to vehicular traffic (Drapper et al. 2000;
Sansalone and Buchberger 1997a; Deletic and Orr 2005; Sartor and Boyd 1972;
Herngren et al. 2006; Fujiwara et al. 2010). In order to differentiate the sources,
common earth elements such as iron, aluminium and manganese were also chosen.
62
A. Method
The methods used for heavy metal sample collection, digestion and determination
were covered in USEPA 200.8 (EPA 1994). The detailed description of the
sampling, preparation and determination of heavy metals is given below.
B. Sampling
The total recoverable heavy metal analytes were tested for both the build-up and
wash-off samples. The proposed wet and dry vacuuming method was used to collect
the sample in de-ionised water in 25 L plastic containers. Samples were transferred
to the laboratory as soon as possible after field collection. 500 mL homogeneous
sub-samples were prepared in deionised water using a churn splitter. The particle
size distributions of the suspended solids in the sub-samples were determined using
a Malvern Mastersizer S particle size analyser capable of analysing particles
between 0.05 to 900 µm diameter. Based on the particle size distribution, the total
particulate analytes were fractioned into four size ranges, namely, >300 µm, 150-
300 µm, 75-150 µm, 1-75 µm using wet sieving. The filtrate passing through a 1 µm
Whatman® GF B glass fiber filter was considered as the potential total dissolved
fraction. The choice of these size fractions was in conformity with the studies by
Herngren (2005) who used such partitioning of samples for characterising polycyclic
aromatic hydrocarbons (PAH) in build-up and wash-off from roads. In each case,
sub-samples were stored in 500 mL amber glass bottles with PTFE seals, preserved
at 4°C in the laboratory.
For wash-off sample collection, the runoff water was vacuumed continuously into
25 L plastic containers using the vacuum cleaner and 500 mL event mean
63
concentration (EMC) samples were collected using a churn splitter. The EMC
represented a flow weighted average concentration of the pollutants computed as the
total pollutant mass divided by the total runoff volume for a simulated rainfall event.
As pollutant concentrations can vary by orders of magnitude during a runoff event,
the flow weighted average or event mean concentration samples (EMC) were found
to be appropriate for evaluating the impacts of stormwater runoff on receiving
waters (Sansalone & Buchberger 1997b). The required volumes for a particular
duration constituting an EMC sample was calculated from the percentages of the
total runoff for that duration and mixed together to obtain a 500 mL EMC sample for
an event.
C. Quality Control and Quality Assurance
For quality control and assurance, calibration standards, internal standards, blanks
and certified reference materials were used using the following standards:
Internal Standards
• Scandium ICP-MS Standard 100 µg/mL in 2% HNO3 prepared by
Accustandard®
• Bismuth ICP-MS Standard 100 µg/mL in 2% HNO3 prepared by

Accustandard®
• Indium ICP-MS Standard 100 µg/mL in 2% HNO3 prepared by

Accustandard®
• Terbium ICP-MS Standard 100 µg/mL in 2% HNO3 prepared by

Accustandard®
• Yttrium ICP-MS Standard 100 µg/mL in 2% HNO3 prepared by

Accustandard®
External Standards
64
• ICP Quality Control Standard #3 100 ppm in 5% HNO3 prepared by
Accustandard®
Certified Reference Material

• Multielement standard solution V for ICP prepared by TRACECERT
The calibration standard supplied by Accustandard® contained each of the target
heavy metal analyte at a concentration of 100 mg/L. Six different calibration
standards at concentrations of 20, 10, 5, 1, 0.1 and 0.01 mg/L were prepared. The
internal standard containing indium (In), bismuth (Bi), terbium (Tb), scandium (Sc)
and yttrium (Y) were prepared at a concentration of 1 mg/L.
The certified reference material (TraceCERT, Sigma-Aldrich®) contained 10 mg/L
of each target analytes except iron (100 mg/L). The volumes of samples, standards
and blanks were kept at 50 mL after digestion while the concentration of internal
standards was kept at 0.02 mg/L for analysis. The laboratory fortified blanks were
prepared by adding the certified reference materials to the deionised water to obtain
a concentration of 0.1 mg/L for each target analyte except iron (1 mg/L) and were
treated exactly as samples for analysis. The analytical technique used for the trace
analysis of the metal analytes was inductively coupled plasma/mass spectrometry
(ICP/MS). The percentage recoveries of the spiked blanks with known concentration
of analytes were estimated using the following equation:
R = ( LFB − LRB ) / C × 100 ---------------------------------------- (3.1)
where R= percent recovery, LFB= laboratory fortified blank, LRB= blank and C=
stated concentrations of analytes in the LFB.
65
To determine the repeatability of the process, seven replicates of a randomly chosen
sample from each batch were analysed. The relative standard deviations (RSD) of
the above samples were measured using the equation:
RSD = ( S rsd / X rsd ) × 100 -------------------------------------------(3.2)
where S rsd = Standard deviation of the replicate samples and X rsd = mean of the
replicate samples.
The reporting limits of the method were established by estimating the limits of
detection (LOD) using seven separate blank samples. The LOD was the lowest
concentration of an analyte measured by a method that could be reliably
distinguished from zero. The LOD was calculated using the equation:
LOD = X LOD + 3S LOD -----------------------------------------------(3.3)
where X LOD = mean of the seven blanks and S LOD = standard deviation of the seven
blanks.
D. Sample Preservation and Storage

The total particulate heavy metal analytes were preserved in 1 L polyethylene bottle
with 3 mL (1+1) reagent grade nitric acid. The samples were held for 16 hours and
then verified for pH< 2 just prior to withdrawing an aliquot for nitric acid digestion.
For the potential dissolved heavy metal analytes, the samples were passed through a
1 µm Whatman® GF B glass fiber filter as soon as possible after the time of
collection. The filtrates were then preserved in a 1 L polyethylene bottle with 3 mL
(1+1) reagent grade nitric acid. The samples were held for 16 hours and then
verified for pH< 2 just prior to withdrawing an aliquot for nitric acid digestion.
66
E. Nitric Acid Digestion
The nitric acid digestion was performed according to Method 3030E (APHA 2005).
Concentrated ultra pure nitric acid was used as reagent.
F. Metals detection by Inductively Coupled Plasma/Mass Spectrometry
(ICP/MS)
The trace metal determination was performed using USEPA method 200.8 (EPA
1994). Samples were introduced into an argon-based, high-temperature radio
frequency plasma and the ions passing through a mass spectrometer were counted by
an electron multiplier detector, and the resulting information was processed by a
computer.
3.5.2. Total Petroleum Hydrocarbons

Total petroleum hydrocarbons (TPH) are a large family of hydrocarbons that
originate from fossil fuels containing high percentages of carbon which include coal,
petroleum and natural gas (Morrison and Boyd 1992). As these sources of petroleum
hydrocarbons contain several hundred chemical compounds all known as TPH, it is
not practical to measure each one separately. Table 3.8 shows the classification of
petroleum hydrocarbons based on distillation temperature and carbon number.
Table 3.8 Petroleum Hydrocarbon Constituents (adapted from Morrison and Boyd 1992)
Fraction Distillation Temperature, ºC Carbon Number
Gas Below 20 ºC C1-C4
Petroleum Ether 20-60 ºC C5-C6
Ligroin (light Naphtha) 60-100 ºC C6-C7
Natural Gasoline 40-205 ºC C5-C10, and cycloalkanes
Kerosine 175- 325 ºC C12-C18, and aromatics
Gas Oil Above 275 ºC C12 and higher
Probably long chains attached
Lubricating Oil Non-volatile liquids
to cyclic structures
Asphalt or Petroleum Coke Non-volatile solids Polycyclic structures
For simplicity in measuring TPH, a feasible approach is to group them according to
volatility. Two composite methods based on several USEPA methods (EPA 2008)
67
have differentiated TPH into the gasoline range and diesel range organics. The
gasoline range organics (GROs) were identified as volatile TPHs eluting between
hexane and decane and the diesel range organics (DROs) were identified as semi and
non-volatile hydrocarbons eluting between decane and octacosane (API 1994). The
two methods for GROs and DROs cover the determination of hydrocarbons for a
carbon number up to C28. Draper et al. (1996) proposed a modification to the DRO
method to include the determination of motor oil for a carbon number up to C38.
Gasoline Range Organics

The test methods adopted for the determination of gasoline range organics were
USEPA methods 5035, 5030B, 8015, 8021, and 8260 (EPA 2008). The target
analytes in GRO were as follows:
• Ethylbenzene
• Toluene
• Ortho-Xylene
• Meta-Xylene
• Para-Xylene
Diesel Range Organics

The methods adopted for the determination of diesel range organics were USEPA
method 3510C, 8015, 8021, and 8260 (EPA 2008). The target analytes in DRO
followed by their respective carbon number were as follows:
• Octane-C8
• Decane-C10
• Dodecane-C12
• Tetradecane-C14
• Hexadecane-C16
• Octadecane-C18
• Eicosane-C20
68
• Docosane-C22
• Tetracosane-C24
• Hexacosane-C26
• Octacosane-C28
• Triacontane-C30
• Dotriacontane-C32
• Tetratriacontane-C34
• Hexatriacontane-C36
• Octatriacontane-C38
• Tetracontane-C40
Sampling
The sampling methodology followed the collection procedure for different sample
matrices such as, solids or aqueous as described in the EPA (2008). According to the
sample collection techniques described for organic analytes by EPA (2008), the
research study identified that build-up samples for the gasoline range and diesel
range petroleum hydrocarbons remained as solids or sediments matrix and the wash-
off samples for the gasoline range and diesel range petroleum hydrocarbons
remained as aqueous matrix. Hence, the proposed vacuuming and spraying
technique for the sampling of these different matrices needed further adaptations to
incorporate the EPA approved collection methods. The following discussion is on
the adaptation to the EPA collection methods for GROs and DROs in different
phases:
A. GRO in build-up sampling

• Collected the build-up samples in plastic containers using the proposed
vacuuming and spraying method.
69
• Samples were transferred immediately to the laboratory and fractioned into
four particulate fractions, namely, >300 µm, 150-300 µm, 75-150 µm, 1-75
µm using wet sieving. The filtrate passing through a 1 µm Whatman® GF B
glass fiber filter was considered as the potential total dissolved fraction.
• Immediately 38 mL aliquots of samples were collected in 40 mL amber glass
vials which were large enough to contain at least 5 gm of solid material and
at least 10 mL of water as recommended by the EPA (2008). Each vial was
then sealed with a screw-cap containing a PTFE-faced silicone septum.
These represented the particulate and the potential dissolved GRO analytes
in the build-up.
B. GRO in wash-off sampling

• Collected the wash-off samples by only vacuuming the surface runoff into
plastic containers.
• Samples were transferred immediately to the laboratory and 500mL event
mean concentration (EMC) sub-samples were prepared using a churn splitter
as described above. The 500 mL EMC samples were then fractioned into
four particulate fractions, namely, >300 µm, 150-300 µm, 75-150 µm, 1-75
µm using wet sieving. The filtrate passing through a 1 µm Whatman® GF B
glass fiber filter was considered as the potential total dissolved fraction.
• Immediately 38 mL aliquots of samples were collected in 40 mL amber glass
vials with PTFE lined septum caps. These represented the particulate GRO
analytes in the wash-off. Also 38 mL aliquots of the filtrates were collected
in other 40 mL amber glass vials with PTFE lined septum caps. These
represented the dissolved GRO analytes in the wash-off.
70
C. DRO in build-up sampling
• Collected the build-up samples in plastic containers using the vacuuming and
spraying method outlined.
• Samples were transferred to the laboratory and fractioned into five size
fractions as described above. Aliquots of the collected sample were filtered
through 1 µm Whatman® GF B glass fiber filter.
• Samples were then transferred to 500 mL clear glass containers with PTFE
lined lids. Also, 500 mL of filtrates were collected in the same glass bottles.
These represented both the particulate DRO analytes and the potential
dissolved DRO analytes in the build-up.
D. DRO in wash-off sampling

• Collected the wash-off samples by only vacuuming the surface runoff into
plastic containers.
• Samples were transferred to the laboratory, 500 mL event mean
concentration (EMC) samples were prepared using a churn splitter and
fractioned into five size fractions mentioned above. A prepared EMC sample
of 500 mL was filtered through 1 µm Whatman® GF B glass fiber filter.
• Samples were then transferred to 500 mL clear glass containers with PTFE
lined lids. Also, 500 mL of filtrates were collected in the same glass bottles.
These represented both the particulate and the dissolved DRO analytes in the
wash-off.
Quality Control and Quality Assurance

For quality control, calibration standards, internal standards, spikes and blanks were
prepared using the following standards:
71
Internal Standards
• EPA 8260 Internal Standards Mix- GRO internal standard prepared by
Chemservice®
• EPA 8270 Semivolatile Internal Standards Mix-DRO internal standard

prepared by Sigma-Aldrich®
Certified Reference Material used as Quality Control Standards

• Revised PVOC/GRO Mix- GRO certified reference materials prepared by
Sigma-Aldrich®
• TPH Mix-1-DRO certified reference materials prepared by Sigma-Aldrich®
Surrogate Standards
• EPA 8260 Surrogate Standard Mix- GRO surrogate standard prepared by
Sigma-Aldrich®
• N-Triacontane-d62 5.0 mg/mL in THF- DRO surrogate standard prepared by

Accustandard®
Calibration Standards
• GRH-003S (10 components) - GRO calibration standards prepared by
Chemservice®
• FTRPH calibration standard (17 component) - DRO calibration standards

prepared by Accustandard®
For the GRO, tests were undertaken according to the USEPA Method 5035 and
8260B (EPA 2008) using purge and trap extraction followed by Gas
Chromatography/Mass Spectrometry (GC/MS). Ten different calibration standards
(Chemservice® THM501 – 1RPM) at 1, 2, 5, 10, 20, 50, 100, 150, 200 and 250 µg/L
concentrations were prepared for each target analyte. Volatile internal standards
(Chemservice® IS-8260ARPM) consisting of flourobenzene, chlorobenzene-d5 and
1, 4- dichlorobenzene-d4 were added to each sample and standards at 50 µg/L
72
concentration. Field blanks were used during each field trip and all results were
blank corrected.
Three quality control standards at 10, 50 and 100 µg/L concentrations were prepared
independently of the calibration standards and were included in each batch for
comparison with the calibration standards. One sample from each batch was spiked
with another quality control standard at a concentration of 20 µg/L. The percentage
recoveries of the spikes were estimated using the following equation:
R = (C1 − C 2) / C1 × 100 -------------------------------------------------------------------- (3.4)
where R= percent recovery, C1= initial spike concentration before extraction, C2=
final spike concentration.
For DRO, tests were undertaken according to the USEPA Method 3510C, 8015,
8021 and 8260B (EPA 2008) using separatory funnel liquid-liquid extraction,
Kuderna-Danish concentration, nitrogen blowdown and Gas Chromatography/Mass
Spectrometry (GC/MS). Nine different calibration standards (Sigma-Aldrich®) were
prepared at 0.1, 0.5, 0.7, 1, 1.4, 7, 10, 28, 50 mg/L concentrations for each target
analyte. The DRO internal standard consisting of acenaphthene-d10, chrysene-d12,
naphthalene-d8, perylene-d12, phenanthrene-d10, 1, 4-dichlorobenzene-d4 was
added to each sample and standards at 5 mg/L concentration. Field blanks were used
during each field trip and all results were blank corrected.
Three quality control standards at 1, 10 and 50 mg/L concentrations were prepared
with another quality control standard at a concentration of 35 mg/L. Surrogate
standards consisting of 10 mg/L of n-triacontane-d62 were added to seven randomly
73
chosen samples. The spike or surrogate recoveries were calculated using equation 4
given above.
Sample Preservation and Storage

A. GRO in build-up samples
Samples were preserved with 500 µL of 50% HCl in 40 mL glass vials. The pH
measurement was performed in the left over sample in the plastic container. Samples
were flagged for pH adjustment if the pH was greater than 2. Samples were
preserved under 4°C immediately after collection and the extraction and sample
determination were performed within 14 days from the date of collection. All the
internal standards and matrix spikes were added to the vials inside the laboratory by
puncturing the septum with small-gauge needles just before sample introduction to
the Gas Chromatograph (GC) detector for determination.
B. GRO in wash-off samples

Samples were preserved with 500 µL of 50% HCl in 40 mL glass vials and cooled to
4°C immediately after collection. Acid was added to the vials prior to adding the
sample. The pH was measured in the left over sample in the plastic container.
Samples were flagged for pH adjustment if the pH was greater than 2. All the
extraction and sample determination were performed within 14 days from the date of
collection. The internal standards, matrix spikes and surrogates were added to the
vials inside the laboratory by puncturing the septum with small-gauge needles just
before the sample introduction to GC for determination.
C. DRO in build-up samples

Samples were preserved with 5 mL of 50% HCl in 500 mL wide mouth glass
containers. Acid was added to the container prior to adding the sample. All samples
were kept under 4°C immediately after collection. The pH was measured in the left
74
over sample in the plastic container. Samples were flagged for pH adjustment if the
pH was greater than 2. Samples were extracted within 14 days from the date of
collection and extracts were analysed within 40 days following extraction.
D. DRO in wash-off samples

Samples were preserved with 5 mL of 50% HCl in 500 mL amber glass containers.
Acid was added to the container prior to adding the sample. All samples were kept
under 4°C immediately after collection. The pH measurement was performed in the
left over sample in the plastic container. Sample results were flagged for samples
with pH greater than 2. Samples were extracted within 14 days from the date of
collection and extracts were analysed within 40 days following extraction.
Sample Extraction
A. Gasoline Range Hydrocarbons
To extract the sample, USEPA methods 5035 and 5030B (EPA 2008) were used.
These are purge and trap methods for solids and aqueous samples. The purge and
trap system consists of three units referred to as, sample purger, the trap and sample
desorber. This system adds water, surrogates and internal standards to the vials
containing the sample, purges the volatile organic hydrocarbons using an inert gas
stream, and traps the released volatile organic compounds for subsequent desorption
into the gas chromatograph.
B. Diesel Range Hydrocarbons

For the extraction of DROs, USEPA method 3510C (EPA 2008) was used. This
technique uses the separatory funnel liquid-liquid extraction. 250 mL Hexane was
used as the exchange solvent for this extraction. The pH of the aqueous phase was
adjusted during initial extraction for <2 using 1:1 (v/v) sulphuric acid and during the
secondary extraction for >11 using 10 N sodium hydroxide. Further concentrations
75
were carried out by using the Kuderna-Danish apparatus followed by nitrogen
blowdown technique (EPA 2008). The extractions and concentrations were carried
out until a final extracted volume of 1 mL was achieved for GC analyses.
Sample Analyses by Gas Chromatography

The Gas Chromatographic (GC) technique according to methods 8015 and 8260
(EPA 2008) was used with a capillary column specially built for TPH analyses. The
column was temperature programmed to separate the analytes, which were then
detected by a mass spectrometer or a flame ionisation detector interfaced to the GC.
A HP5MS Agilent® column of 30 m length, 0.32 mm internal diameter and 0.25 µm
film thickness was used. A splitless sample injection of 2 µL at an inlet temperature
of 280°C, inlet pressure of 5.16 psi (35.58 kN/m2) and a flowrate of 2.4 mL/min was
°
used. The initial oven temperature was set at 40 C, holding for 12 minutes, then
increased at 10°C per minute until temperature reached 300°C and held for 20
minutes during sample introduction to GC column. The identification of target
analytes was performed by comparing their mass spectra with the electron impact
spectra of authentic standards.
3.5.3. Solids
Total solids in all build-up and wash-off samples collected from the study sites were
measured. Total solids refer to the material residue left in the vessel after
evaporation of a sample and its subsequent drying in an oven at a defined
temperature (APHA 2005). Total solids included, total suspended solids, the portion
of total solids that was retained in a filter, and the total dissolved solids, the portion
that passed through the filter.
76
A. Method
Method 2540C was used for determining total dissolved solids and the method
2540D for total suspended solids (APHA 2005). All of these were gravimetric
methods that determined the total concentration of solids in different phases.
B. Determination of Solids by Gravimetric Method
Samples were analysed in batches of 15. Each batch had at least three blanks for
quality control purposes. The oven temperature for drying of suspended solids was
103-105°C and for dissolved solids was 180°C. The difference in weight between
the dried residue along with the petri dish and the weight of the petri dish indicated
the concentration of solids per sample volume.
3.5.4. Organic Carbon

Organic carbon is composed of a variety of organic compounds in various oxidation
states. Total organic carbon (TOC) is a direct expression of the total organic content
irrespective of the different oxidation state of organic carbon compounds. In this
study, the fraction of TOC that passed through a 1 µm Whatman® GF B glass fiber
filter was considered as the potential dissolved organic carbon (DOC) fraction. TOC
and DOC were measured in the all the build-up and wash-off samples.
A. Method
APHA method 5310B was adopted to determine TOC and DOC concentrations in
the samples (APHA 2005).
B. Determination of Organic Carbon
Total organic carbon (TOC) content of the different particle sizes and the dissolved
organic carbon (DOC) content of the filtrate were measured using a Shimadzu TOC-
77
5000A analyser according to the manufacturer’s instructions. The concentration of
TOC was determined from the difference in concentration of the total carbon and the
concentration of inorganic carbon.
3.5.5. Surface Texture Depth

The surface texture depths (STD) of the selected study sites were measured
according to method T250 (Main Roads 2008). 75 m lengths of straight road lengths
at each site were selected to carry out the measurement of STD.
3.6.Data Analyses
Chemometric multivariate data analyses approaches were used to interpret the
results. These techniques are useful for processing large volumes of data in order to
explore and understand relationships between variables (Kokot et al. 1998).
Amongst the different chemometrics tools, principal component analysis (PCA),
fuzzy clustering (FC), two phase factor analysis (FA), preference ranking
organisation method for enrichment evaluation (PROMETHEE), graphical analysis
for interactive aid (GAIA) and partial least square regression (PLS) were employed.
Software including, ‘SIRIUS’ (Sirius 2008), ‘DecisionLab’ (Decision 2000) and
‘SPSS’ (SPSS 2009) were used for undertaking these data analyses techniques. Brief
descriptions of the data analysis techniques are given below.
3.6.1. PCA
PCA is a pattern recognition technique employed to understand the correlations
among different variables and clusters among objects. The PCA technique is used to
transform the original variables to a new orthogonal set of Principal Components
(PCs) such that the first PC contains most of the data variance and the second PC
contains the second largest variance and so on. Though PCA produces the same
amount of PCs as the original variables, the first few contain most of the variance.
78
Therefore, the first few PCs are often selected for interpretation. This reduces the
number of variables without losing useful information contained in the original data
set. The number of PCs to be used for interpretation is typically selected using the
Scree Plot method described by Jackson (1993).
The application of PCA to a data matrix generates a loading for each variable and a
score for each object on the principal components. Consequently, the data can be
presented diagrammatically by plotting the loading of each variable in the form of a
vector and the score of each object in the form of a data point. This type of plot is
referred to as a ‘Biplot’. The angle between variable vectors is the indicator of the
degree of correlation. Clustered data points in a biplot indicate objects with similar
characteristics. Detailed descriptions of PCA can be found elsewhere (Adams 1995;
Massart et al 1997). In this study, ‘SIRIUS’ software (Sirius 2008) was used to
perform the exploratory PCA procedures.
3.6.2. FC
Fuzzy clustering is an object classification method that assigns a degree of class
membership for a given object over several classes (Bezdek 1982, Otto 1988). The
classification is performed with a user specified membership function which, in the
case of ‘SIRIUS’ software (Sirius 2008) is similar to that described by Bezdek
P
(1982). An example of a membership function is m( x) = 1 − c x − a , where a and
c are constants and p is called cluster exponent with a suggested value between 1 to
3 (Sirius 2008). Values closer to 1 result in hard clustering where the objects are
placed into their most preferred classes while values closer to 3 result in soft
clustering where the objects are allowed to spread over as many classes as possible.
The sum of the membership values of each object is 1. The main advantage of the
79
fuzzy clustering is that it facilitates the distinction between the objects that clearly
belong to one cluster and those that are members of several clusters. A class
membership threshold is defined as 1/ n (n = number of clusters).
3.6.3. FA
Factor analysis (FA) is another type of data extraction method that has been used in
prediction models for spatially different data observations (Christensen and
Amemiya 2003) including hydrocarbon classifications (Zhu et al. 1999). As opposed
to PCA, the factors in the FA method are extracted as independent variables that
account only for the shared variance, namely, the covariance of the variables in the
data matrix.
The two phases of the FA method are factor extraction and factor rotation. Amongst
several extraction processes described in Meyers et al. (2006), the Principal
Component Extraction (PCE) is the most efficient extraction process that could be
performed by microprocessors. Factor rotation, on the other hand, is used to achieve
simple structures where the measured variables could be associated with each factor
in terms of their strong correlations. Amongst several factor rotation methods, the
orthogonal factor rotation with maximum variance (varimax) has been used in
several environmental studies (e.g., Yidana et al. 2008). This rotation method allows
a factor to correlate quite strongly with some variables (correlations closer to 1), but
more weakly with the other variables (correlations closer to 0) in the data matrix.
The total variance explained by the factors before and after rotation remains the
same. The study adopted the FA method including the PCE followed by varimax
with a view to strengthen a validation strategy for the PCA components through the
80
factor extraction process. This study used ‘SPSS’ software (SPSS 2009) to perform
FA procedures.
3.6.4. PROMETHEE
PROMETHEE is designed to rank a number of objects in terms of the data criteria
(Brans et al. 1986; Keller et al. 1991). The ranking for each variable or criterion is
performed by a user specified preference function. The study used ‘DecisionLab’
software (Decision 2000) to perform PROMETHEE analysis. The steps involved in
the application of PROMETHEE are as follows:
1. For each variable all objects or actions in the data matrix are compared pairwise,
in all possible combinations by subtraction, and thus a difference, d , matrix is
generated;
2. A preference function P ( a , b ) is chosen for each variable. It describes to what
extent outcome a is preferred to outcome b . In the ‘DecisionLab’ software, one
of six such functions along with corresponding threshold values may be chosen
by the user. It is also necessary to specify whether top-down (maximized) or
bottom-up (minimized) ranking of objects for each variable is preferred.
Additionally, each variable can be weighted in importance, but in general, most
modelling initially uses the default weighting of 1.
3. The products of each preference function P ( a , b ) and the weights for the
corresponding variables were summed up to generate a preference index table.
These indices correspond to the pairwise comparison of the objects or actions.
4. To compare each action one-to-one with the others systematically, preference
flows are computed. The ‘DecisionLab’ software supports three types of
preference flows, namely, positive outranking flow ( φ + ), negative outranking
flow ( φ − ) and the net outranking flow ( φ ). The positive outranking flow ( φ + ) is
81
associated with the degree of preference with which one action is preferred on
average over the other actions and the negative outranking flow ( φ − ) is
associated with the degree of preference with which the other actions are
preferred on average to that action. The higher the φ + and lower the φ − , the
more preferred is the action. This procedure results in a partial pre-order, called
PROMETHEE I ranking.
5. There are certain circumstances as described by Keller et al. (1991) when two
actions a and b may not be comparable. The net outranking flow ( φ ), which is
the algebraic difference between the positive and negative outranking flows, is
then calculated. This procedure, known as PROMETHEE II, was used in this
study to eliminate any incomparability between actions or objects.
3.6.5. GAIA
GAIA is a sensitivity analysis technique for multicriteria decision methods such as
PROMETHEE (Keller et al. 1991; Mareschal & Brans 1988). GAIA provides a
graphical view of the actions and variables for net outranking flow ( φ ) in the form
of a PCA biplot by decomposing the φ values from PROMETHEE II into
unicriterion flows for each variable. The advantages of GAIA over a PCA biplot is
that it also produces a decision axis that takes into account the weights associated
with the variables. This helps the decision-maker with an enriched understanding of
the problem in terms of the detection of clusters of actions, conflicts in variables,
inability to compare between actions and so on (Mareschal and Brans 1988). The
study used ‘DecisionLab’ software (Decision 2000) to perform PROMETHEE and
GAIA analysis.
82
3.6.6. PLS
Partial least squares regression (PLS) is a multivariate regression method that is used
to predict response variables (R) from predictors or independent variables (C). This
method simultaneously estimates the underlying factors in both the response and the
predictor data matrix. The matrices are decomposed as follows:
R = TP + E ----------------------------------------------------------------------------------- (3.5)
C = UQ + F ----------------------------------------------------------------------------------- (3.6)
where the elements of T and U are the scores of R and C respectively, the elements
of P and Q are the loadings. E and F are the errors associated with the estimation of
underlying factors of R and C in equations 3.5 and 3.6 respectively.
Detailed descriptions of PLS estimation of responses from predictors can be found
in Beebe and Kowalski (1987). Several environmental studies (for example Sotelo
2008) and analytical chemistry studies (for example Ni et al. 2001) have applied
PLS along with other prediction methods such as PCR (principal component
regression) and established that PLS is a better tool than PCR when there are
independently varying major response components. In this study, PLS was
performed using the ‘SIRIUS’ software (Sirius 2008).
3.7.Publication of Results
The publication of scientific papers was undertaken as part of the characterisation
and prediction of hazardous stormwater pollutants under dynamic conditions. The
research methodology presented in this chapter provided the foundation for a series
of journal articles that were submitted to several international journals. The build-up
and wash-off characteristics of the heavy metals, volatile, semi-volatile and non-
volatile organic compounds and their corresponding relationships with changed
urban traffic and climatic conditions are discussed in these publications. Figure 3.3
83
provides a schematic diagram of the research design and highlighting the scientific
publications. The manuscripts of these publications are presented in the subsequent
Chapters 4 to 9 and their relevant positions within the research flow are highlighted
in Figure 3.3. The last phase of this research study was the tangible outcome in
terms of specifications, guidance as well as adaptive measures for stormwater
quality mitigation under changed urban traffic and climatic conditions. These were
presented as the conclusions at the end of this thesis.
84
Generalised study on Identifying impacts of
pollutant build-up traffic and climate change
Identifying hazardous pollutants
and wash-off on water quality
on urban roads, their sources and
the justification for investigating
the dynamic relationships with
build-up and wash-off
Literature Review
and Desktop Study
Characterising and Site selection Sample collection Sample testing

Predicting
Hazardous
Pollutants under
Dynamic
Conditions
Data analysis
Defining the Characteris- Characterisa- Characterisa- Prediction of Characterisa-

dynamic ing the build- tion of SVOC tion of VOC VOC build-up tion and
relationship of HM up of HM and NVOC wash-off from on urban roads prediction of
build-up and wash- and VOC on build-up on urban roads (Paper 5; SVOC and
off with urban urban roads urban roads (Paper 4; Chapter 8) NVOC wash-
Traffic and climate (Paper 2; (Paper 3; Chapter 7) off from urban
change respectively Chapter 5) Chapter 6) roads (Paper 6;
(Paper 1; Chapter Chapter 9)
4)
Characterisation of the Development of prediction

build-up and wash-off methodologies and
of HM, VOC, SVOC adaptive stormwater
and NVOC under the quality mitigation
dynamic urban traffic guidelines during the
and climate change build-up and wash-off of
scenarios HM, VOC, SVOC and
NVOC under dynamic
urban traffic and climate
change scenarios
Outcomes
Figure 3.3 Schematic diagram showing detailed representation of research flow undertaken in
this research highlighting the development of scientific papers
85
3.8.Summary
The methodology for the research study included sampling, preservation, testing and
data analyses techniques for the analysis of selected heavy metals, total petroleum
hydrocarbons, organic carbons in total and dissolved form, and solids in suspended
and dissolved forms in relation to their build-up and wash-off from roads under
changed urban traffic and climate change. Methodologies to simulate natural rainfall
on the selected study sites incorporating both extreme and normal rainfall events
with ARI 1 to 100 years was developed. Similarly, different future scenarios of
wash-off for 2030 in the South-East Queensland region were reviewed and
incorporated into the research methodology. The study selected the year 2030 for
future rainfall simulation as this was one of the years for which the percentage
changes of rainfall intensities in the Gold Coast region were predicted by climate
change researchers.
The site selection methodology was based on the availability of urban traffic data
from the Gold Coast City Council and Queensland Department of Transport. A
suburb based selection of sites in residential, commercial and industrial land uses
was undertaken with a view to incorporating a mix of traffic and pavement
characteristics into the research study. The selected sites represented transport
infrastructures developed in the past decade in the Gold Coast City Council region.
The required urban traffic data necessary to relate the build-up of water pollutants to
traffic parameters were identified as average daily traffic (ADT) and volume to
capacity ratio (V/C). Additionally, the surface texture depth of the pavement (STD)
was also considered to influence pollutant build-up on road surfaces.
86
As part of the research methodology, an efficient build-up sample collection system
known as ‘The Wet and Dry Vacuuming System’ was developed. It was established
that 90% sample collection efficiency could be achieved when vacuuming was
combined with a spraying technique. An efficient wash-off sample collection
methodology was also developed to incorporate the effects of climate change on
rainfall characteristics in South-East Queensland. It incorporated three different
future climate change scenarios with different combinations of rainfall intensity,
frequency and duration for year 2030.
A number of test methods were reviewed in order to identify the most appropriate
chemical and physical tests required for analyses of the pollutants. Detailed
methodology focusing on sample collection, particulate and dissolved sample
preparation, preservation, storage and analysis were identified based on these
reviews. Reviews also assisted in the simplified classifications of petroleum
hydrocarbons such as gasoline and diesel range organics and the subsequent
analyses. The temperature programme was established for the chromatographic
separation of the target hydrocarbons in this research study.
The data analyses methodologies were established to meet the project objectives. A
schematic framework is provided highlighting the key steps in the research study
along with key outcomes which have been disseminated in the form of research
publications.
87
88
CHAPTER 4 DEFINING DYNAMIC
RELATIONSHIP OF HEAVY METAL BUILD-UP
AND WASH-OFF WITH URBAN TRAFFIC AND
CLIMATE CHANGE
Manuscript Title
Impacts of traffic and rainfall characteristics on heavy metals build-up and wash-off
from urban roads
Parvez Mahbub1*, Godwin A. Ayoko2, Ashantha Goonetilleke1, Prasanna

Egodawatta1, Serge Kokot2
1
School of Urban Development, Queensland University of Technology, GPO Box
2434, Brisbane 4001, Queensland, Australia
2
School of Physical and Chemical Sciences, Queensland University of Technology,
GPO Box 2434, Brisbane 4001, Queensland, Australia
s.mahbub@qut.edu.au; p.egodawatta@qut.edu.au; g.ayoko@qut.edu.au;
a.goonetilleke@qut.edu.au; s.kokot@qut.edu.au
*Corresponding Author: Parvez Mahbub;Tel: 61 7 3138 1540;Fax: 61 7 3138
1170;email: s.mahbub@qut.edu.au
Published (2010): Environmental Science & Technology, 44(23): 8904-8910.

Impact Factor 5.5; ERA ranking A*.
Abstract
An investigation into the effects of changes in urban traffic characteristics due to
rapid urbanisation and the predicted changes in rainfall characteristics due to climate
change on the build-up and wash-off of heavy metals was carried out in Gold Coast,
Australia. The study sites encompassed three different urban land uses. Nine heavy
metals commonly associated with traffic emissions were selected. The results were
interpreted using multivariate data analysis and decision making tools, such as
principal component analysis (PCA), fuzzy clustering (FC), PROMETHEE and
GAIA. Initial analyses established high, low and moderate traffic scenarios as well
89
as low, low to moderate, moderate, high and extreme rainfall scenarios for build-up
and wash-off investigations. GAIA analyses established that moderate to high traffic
scenarios could affect the build-up while moderate to high rainfall scenarios could
affect the wash-off of heavy metals under changed conditions. However, in wash-
off, metal concentrations in 1-75µm fraction were found to be independent of the
changes to rainfall characteristics. In build-up, high traffic activities in commercial
and industrial areas influenced the accumulation of heavy metal concentrations in
particulate size range from 75 - >300 µm, whereas metal concentrations in finer size
range of <1-75 µm were not affected. As practical implications, solids <1 µm and
organic matter from 1 - >300 µm can be targeted for removal of Ni, Cu, Pb, Cd, Cr
and Zn from build-up whilst organic matter from <1 - >300 µm can be targeted for
removal of Cd, Cr, Pb and Ni from wash-off. Cu and Zn need to be removed as free
ions from most fractions in wash-off.
Keywords: Climate change, Heavy metals, Build-up, Wash-off, Water pollution
Statement of Contributions of Joint Authorship

Parvez Mahbub (Principal Author)
Writing and compilation of the manuscript; establishing methodology, data analysis;
preparation of figures, tables and supplementary information.
Godwin A. Ayoko (Co-author)

Assisted in manuscript compilation and editing
Ashantha Goonetilleke (Co-author)
Prasanna Egodawatta (Co-author)

Assisted in manuscript editing
Serge Kokot (Co-author)

This chapter is an exact copy of the published journal paper.
90
Linkage of the Paper to the Research Methodology and
Development
The data analyses of this journal paper were formulated on the basis of the
objectives of the overall research study. The first objective as mentioned in section
1.3, Chapter 1, was to define the dynamic relations between the build-up of
pollutants with changing urban traffic as well as the wash-off of pollutants with
changing rainfall characteristics induced by climate change. In this context, an
object classification system capable of identifying the dynamic scenarios of
changing urban traffic and the changes in rainfall characteristics was developed in
this journal paper. The method adopted the fuzzy clustering (FC) technique as well
as the preference ranking organisation method for enrichment evaluation
(PROMETHEE). This classification enabled to study the build-up and wash-off of
traffic generated heavy metals, volatile organic compounds, semi volatile organic
compounds as well as non volatile organic compounds under the changing urban
traffic and rainfall scenarios in the current and the subsequent chapters. Figure 3.3 in
section 3.7, Chapter 3 provides a schematic flow diagram of this research study
where the publication of this journal paper was highlighted as an integral process
that contributes to the overall research outcome.
91
4.1. Introduction
Rapid urbanisation and climate change are two global phenomena that are attracting
increased debate amongst the scientific community throughout the world in terms of
their impacts on the environment. Rapid urbanisation and the consequent changes to
urban traffic characteristics such as increased volume and congestion will in turn
affect pollutant build-up on road surfaces (1, 2). Additionally, predicted changes in
rainfall characteristics (3) due to climate change can readily affect pollutant wash-
off from urban road surfaces. An in-depth knowledge of the build-up and wash-off
processes of pollutants under such phenomena is the key to developing adaptive
measures to mitigate the impacts of these changes on the water environment and to
create a sustainable urban environment. For example, Delpla et al. (4) suggested that
monitoring and analysis of the occurrence and fate of micro-pollutants including
heavy metals must be considered in adaptive measures in the treatment of drinking
water with regards to the climate change impacts.
In this regard, the environmental impacts and more specifically the water quality
impacts of road transport in combination with climate change have received limited
attention in research literature (2, 5). Pollutant accumulation and its subsequent
wash-off in the environment are complex processes. Pollutant build-up in urban
areas is influenced by anthropogenic activities related to population density,
commerce and industry, land use, average daily traffic (ADT) and traffic volume to
capacity ratio (V/C) (2, 5, 6). Pollutant wash-off is mainly influenced by the rainfall
characteristics and the subsequent surface runoff from both pervious and impervious
surfaces (7).
92
Heavy metals are among the most important pollutants in stormwater runoff that are
generated by transport activities (8, 9). These pollutants have significant human and
ecosystem health impacts (10-12). Several research studies have been undertaken in
the past to characterise the build-up and wash-off of heavy metals under static
environmental conditions (13, 14). However, in order to develop improved
management practices for urban water quality, an in-depth understanding of the
impacts of dynamic scenarios under changing urban traffic and rainfall conditions on
the build-up and wash-off processes of heavy metals is critical. The research study
discussed in this paper investigated the impacts of such changes in urban traffic and
rainfall characteristics on the build-up and wash-off of heavy metals. A fuzzy
clustering classification system is presented to characterise these dynamic scenarios.
The outcome of this study is expected to contribute to the development of adaptive
mitigation measures for the improvement of urban water quality.
4.2. Experimental Section

4.2.1. Site Selection
The research study was undertaken in the Gold Coast region of Southeast
Queensland, Australia. The urban areas in this region have undergone rapid
development. A suburb based approach was followed by selecting 11 road-sites in
two suburbs, namely Helensvale and Coomera, which reflected the transport
infrastructure that were developed during the last decade in the region. The sites
selected for sample collection encompassed typical urban land uses, namely,
residential, commercial and industrial, in order to obtain a cross-sectional view of
traffic activities on road surfaces in the region. The residential sites comprised of
single detached houses as well as multi-storied apartment blocks, the commercial
sites were located near shopping mall car parks and the industrial sites were located
in light industrial areas.
93
4.2.2. Build-up and Wash-off Sample Collection
A build-up sample collection method referred to as ‘Wet and Dry Vacuum System’
(15) was adopted for this study. A domestic vacuum cleaner with a water filtration
system was used to collect the road dust from a 2 x 1.5 m plot area in the middle of
the traffic lane in 25L plastic containers. Immediately afterwards de-ionised water
was sprayed at 2 bar pressure on the collection plots and vacuuming was undertaken
again to collect any remaining dust into the plastic containers.
Based on the climate change studies (3, 16), the predicted rainfall characteristics for
year 2030 were identified for the Gold Coast region. This study used a specially
designed rainfall simulator (13) to replicate the design rainfall events common to the
study region. It consisted of an A-frame structure with three Veejet 80100 nozzles,
mounted on a stainless steel boom at a height of 2.4 m. In this paper, a total of 22
wash-off events were simulated. The runoff water was vacuumed continuously into
25L plastic containers using a domestic vacuum cleaner. For both build-up and
wash-off, the total particulate analytes were fractioned into four size ranges, namely,
300 µm, 150-299 µm, 75-149 µm, 1-74 µm using wet sieving. The filtrate that
passed through a 1 µm membrane filter was considered as the potential dissolved
fraction. Detailed sample collection techniques are described in the supporting
information.
4.2.3. Sample Testing

The heavy metals selected for analysis of build-up and wash-off samples were
cadmium (Cd), chromium (Cr), nickel (Ni), lead (Pb), zinc (Zn), copper (Cu),
antimony (Sb), manganese (Mn), aluminium (Al) and iron (Fe). These are mainly
vehicle generated heavy metals (13, 17, 18). Samples were preserved in 1 L
polyethylene bottles with 3 mL (1+1) reagent grade nitric acid in 4˚C for at least 16
94
hours and then verified for pH< 2 just prior to withdrawing 50 mL aliquot for nitric
acid digestion. The trace metal determination was performed by Inductively Coupled
Plasma/Mass Spectrometry (ICP/MS).The test methods used are described in
USEPA 200.8 (19).
As a measure of quality control, the percentage recovery, relative standard deviation
and limit of detection for the target heavy metals were established. Details of these
results are discussed in the Supporting Information.
The total and dissolved organic carbon (TOC and DOC), total and dissolved
suspended solid (TSS and TDS) concentration, pH and electrical conductivity (EC)
were also determined in all samples. The surface texture depth (STD) of pavements
was measured according to the recommendations of the US Federal Highway
Administration (20).
4.2.4. Data Analyses

In the build-up data matrices, the 11 sites were designated with object identifiers
CH, IS, RB, RP, RD, CL, CT, RA, RDS, IBT and RR where the initials C, I and R
were used to represent commercial, industrial and residential sites respectively. The
22 wash-off events were designated as rainfall objects assigned with numerical
object identifiers starting from 1. The attributes of the objects in build-up and wash-
off are described in the supporting information. After initial observation of the
probability distribution of the build-up and wash-off objects, normalisation of each
object was undertaken as a pre-treatment measure.
95
Considering the number of variables involved and the large amount of data
generated, a range of multivariate analytical methods including principal component
analysis (PCA), fuzzy clustering (FC) and the multicriteria decision making methods
(MCDM) PROMETHEE and GAIA were employed for in-depth analyses of the
data. A brief description of these data analyses techniques is outlined in the
Supporting Information. Detailed discussion of these techniques can be found in the
referenced literature (21-24).
4.3. Results and Discussion

4.3.1. Exploratory PCA of Heavy Metals Build-up
PCA was performed on the pre-treated build-up data matrices. The resulting PCA
biplot given in Figure 4.1 shows the patterns of variation for all particle size
fractions taken together with the heavy metal elements. This initial PCA
investigation revealed that Mn, Fe and Al form a group (A) of variables that are
strongly correlated to each other while Ni, Cr, Zn, Pb, Cu, Sb and Cd form a
separate group (B). These two groups are relatively orthogonal to each other which
suggest that the two groups are uncorrelated and originating from different sources.
Sb is found to have the strongest correlation with Cu amongst other group (B)
variables. Vehicle brakes, bushings and thrust-bearings are described as the possible
sources of both Cu and Sb with a very high Cu ratio. For example, 90% Cu in brake
wear dust has been reported by Johansson et al. (17) and a Cu:Sb ratio of 4.6±2.3:1
by Sternbeck et al. (18).
96
3
2 IBT
ADT
CL
RP group B
1 Ni Cr STD
RD Pb-Cd
CT RA IS V/C
Sb
Cu
0 CH Zn
RR TOC
PC 2 (21.1%)
-1 RB
Mn-Fe
Al
TSS
group A
-2
-3
-4
RDS
-5
-4 1 6
PC 1 (71.8%)
Figure 4.1 PCA biplot of build-up of all heavy metal size fractions; object identifiers are
described in Table 5.1
Figure 4.1 also includes all the traffic related attributes of the objects. It is evident
that STD and V/C are strongly correlated to group (B). Hence, it is postulated that
the elements in group (B) are more likely to be generated from traffic activities. The
elements in group (A) could be originating from the surrounding area as Mn, Fe and
Al are among the most common elements in soil. Two objects with positive scores
on PC1 (IBT and RDS) are strongly correlated with the groups (A) and (B) variables
while objects IBT, RP and CL with positive scores on PC2 are strongly correlated to
most of the group (B) variables and ADT, V/C and STD. These objects are a mix of
residential, commercial and industrial land uses with varying traffic related
attributes.
In order to better understand the relationship between the object scores and variable
loadings, individual PCA biplots of each fraction were analysed similarly as
97
described above. Considering the individual biplots, it was observed that for particle
fractions from 75 µm to >300 µm, the traffic related heavy metals are primarily
generated in the commercial and industrial sites (objects CL, CT and IBT). High
traffic densities in these areas are attributed to this fact. For the finer fractions such
as 1 to 75 µm and the potential dissolved fraction of <1 µm, the traffic related heavy
metals are present in all three types of land uses (objects CL, RDS, CT, IS, IBT,
RR).
In a study by Patra et al. (25), the coarser particles were found to resuspend faster
than finer particles due to vehicle induced turbulence on urban roads. They also
found that fine particles remained on the ground for a longer time with very slow
decay whilst the grinding of coarser materials under the wheels of traffic replenished
the fine material reservoir on the road surface which further slowed the fine particle
decay pattern. Manoli et al. (26) also observed strong contribution of resuspended
road dust to the coarse particles on urban roads. Thus, in the current study, the
predominance of coarser particles from 75 µm to >300 µm in commercial and
industrial areas would mean their relatively rapid re-suspension and redistribution by
traffic activities in these areas. However, finer particles of 1-75 µm and the potential
dissolved fraction of <1 µm in all three land uses indicate that their presence is not
affected by traffic related activities.
Another important observation was the significant presence of TOC in the
particulate fractions >300 µm, 150-300 µm, 75-150 µm and 1-75 µm with group (B)
variables containing Ni, Cr, Zn, Pb, Cu and Cd whilst TSS was weakly present in
these particulate fractions. Charlesworth and Lees (27) studied the association of
98
heavy metals in group (B) with organic matter at source, transport and deposition
phases. They found that irrespective of particle size between 63 µm and 2 mm,
organic matter acted as the most dominant binding material with heavy metals at
source. In the current research study, the particulate fraction was investigated further
down to 1 µm. Thus, the presence of TOC with the group (B) heavy metals pointed
to the fact that organic matter was acting as the binding agent for traffic related
heavy metals at fractions higher than 1 µm. In the case of the potential dissolved
fraction, the traffic related group (B) heavy metals were positively correlated with
total dissolved solids (TDS) whilst dissolved organic carbon (DOC) was negatively
correlated on PC1. This suggested that solids act as the binding agent for traffic
related heavy metals for the potential dissolved fraction <1 µm. Therefore, as a
practical implication of the build-up study, solids can be targeted up to <1 µm and
organic matter can be targeted from 1 µm to >300 µm for the removal of traffic
generated heavy metals on urban roads.
4.3.2. FC Analysis and PROMETHEE Ranking of Heavy Metals Build-up

In order to classify the study objects consisting of traffic parameters such as daily
traffic and congestion and hence better understand the correlation between the
objects and the traffic generated variables discussed above, fuzzy clustering (FC)
and PROMETHEE II net rankings were undertaken. The FC model consisted of
three clusters (in keeping with the three land uses) and the cluster exponent p was set
to 1.9 (moderately soft clustering). Thus, the class membership threshold value was
set at 0.33 for an object to qualify as a member of a class. Table 4.1 shows the
membership values obtained by fuzzy clustering of each object and the clusters are
defined accordingly.
99
Table 4.1 Membership values of different objects in heavy metals build-up after fuzzy
clustering
Object names(labels or identifiers)* Cluster 1 Cluster2 Cluster 3 Cluster names
Hope Island Road(CH) 0.651 0.343 0.006 fuzzy
Shipper Drive(IS) 0.723 0.272 0.005 High traffic
Moderate
Billinghurst Crescent(RB) 0.012 0.988 0.001
traffic
Peanba Park Road(RP) 0.013 0.021 0.966 Low traffic
Dalley Park Drive(RD) 0.045 0.102 0.853 Low traffic
Moderate
Lindfield Road(CL) 0.052 0.937 0.011
traffic
Town Centre Drive(CT) 0.956 0.041 0.003 High traffic
Abraham Road(RA) 0.998 0.002 0.000 High traffic
Discovery Drive(RDS) 0.998 0.002 0.000 High traffic
Moderate
Beattie Road(IBT) 0.008 0.992 0.000
traffic
Moderate
Reserve Road(RR) 0.000 1.000 0.000
traffic
*
The initials C, I and R were used in each object identifier to represent commercial,
industrial and residential sites respectively
Cluster 1 is associated with high traffic volume ranging from 9000 to 24000 ADT
with relatively high congestion. Cluster 2 is associated with moderate ADT values
ranging from 2300 to 5900 with moderate congestion whilst cluster 3 is mainly
associated with low traffic volume ranging from 500 to 3500 ADT with low
congestion. In Table 4.1 there is one object (CH) left unclassified (fuzzy) as this
object did not belong to a particular cluster according to the set membership
threshold. The net ranking information produced by PROMETHEE II was used to
generate the φ ranking values of all objects. The φ value determines how one object
outranks the others in terms of its preference. The PROMETHEE II net ranking flow
values of both the classified and fuzzy objects are given in Table 4.2.
100
Table 4.2 PROMETHEE II net ranking ( φ ) values showing the fuzzy object in heavy metals
build-up could still be classified as a member of moderate traffic cluster as its value lies within
the range of that cluster
Net ranking values of Net ranking value of
Object names (clusters) Fuzzy Object
classified objects fuzzy object
Town Centre Drive (high) 0.94
Shipper Drive (high) 0.67
Discovery Drive (high) 0.58
Abraham Road (high) 0.56
Billinghurst Crescent
-0.03
(moderate)
Hope Island Road -0.25
Beattie Road (moderate) -0.06
Reserve Road (moderate) -0.22
Lindfield Road
-0.25
(moderate)
Dalley Park Drive (low) -0.67
Peanba Park Road (low) -0.86
It was found that the PROMETHEE II net ranking values could be used to classify
the fuzzy object in Table 4.2. In this case the ‘Hope Island’ object could be
classified into cluster 2 even though it had a slightly higher ADT value than this
cluster. As the ‘Hope Island’ object had a relatively low congestion and similar
texture depth as the other objects in cluster 2, the PROMETHEE II net ranking
allowed this object to fall into this cluster. This type of classification simplified the
analysis by decomposing the objects into high, low and moderate traffic objects.
4.3.3. Exploratory PCA of Heavy Metals Wash-off

The simulated rainfall events causing the wash-off of heavy metals were
characterised by three attributes, namely, duration, intensity and average recurrence
interval (ARI). PCA was performed on both particulate and dissolved wash-off data
matrices as shown in Figures 4.2a and 4.2b.
101
4 3
Particulate 5 Dissolved
17
3
6
2 14 1618 22
Pb-Cd
Cu EC 15 20
1
21 2
pH 19 3
2 1 DOC
21 12 Zn
32
group A 11
PC 2 (27.5%)
PC 2 (24.1%)
1 pH 17 13 TDS
Mn-Fe 0
Al 1
Pb-Cd
13 11 Al 10 8
4 Mn-Fe 7 EC
0 TOC 20
18 -1 Ni
22 Cr
10 Cu 9
Cr Ni
-1
14
9 6Zn
-2
group B 7 5 4
16
-2
15 -3
19 TSS 8 12
-3
-4
-7 -5 -3 -1 1 3
-6 -1 4
PC 1 (39.9%)
PC 1 (32.8%)
(a) (b)
Figure 4.2 PCA biplot for (a) particulate and (b) dissolved fractions for wash-off of heavy
metals; objects are indicated by numbers starting from 1; numerical object identifiers are
described in Table 5.3
The PCA biplot of the total particulate wash-off given in Figure 4.2a revealed that
pH and TOC have very strong correlation with all of the group (B) heavy metals
except Cu and Zn. TSS was moderately correlated to the group (B) heavy metals
except for Cu which has a very strong negative correlation with TSS. Charlesworth
and Lees (27) found that organic matter and carbonates dominate the binding of
particulate associated Cd, Cu, Ni, Zn and Pb during their transport from source to
deposit. However, during the transport phase Cu and Zn were found to be bound
more abundantly with carbonates than organic matter. Hence, the strong correlation
between TOC and the group (B) metals except Cu and Zn reflected the fact that the
organic matter could be acting as the binding agent for the particulate associated Cd,
Cr, Pb and Ni during their wash-off.
In a sorption test using road dust leachates from roads with heavy traffic, Murakami
et al. (28) found high concentrations of Cu and Zn in the soakbeds, suggesting traffic
102
activities were contributing to the accumulation of these pollutants and that Zn
might be released into the groundwater as free ions. The electrical conductivity (EC)
of Cu and Zn is higher than that of Pb, Cd, Cr and Ni. In the case of the current
study, the strong correlation of Cu and Zn with EC and their weak correlation with
TOC in the particulate fraction meant that Cu and Zn could remain as free ions and
were less likely to be associated with organic matter than the other group (B) heavy
metals in particulate wash-off.
The PCA biplot for the dissolved fraction in Figure 4.2b shows that TDS has a
relatively small loading vector and hence does not play a significant role in the
dissolved fraction. Furthermore, in the dissolved fraction, Zn was found to have
positive correlation with EC. Figure 4.2b suggests that organic matter is acting as the
predominant binding agent for Ni, Cr, Cu, Pb and Cd whilst Zn still remained as free
ions in the dissolved fraction. Therefore, as a practical implication of these findings,
organic matter can be targeted from 1 µm to >300 µm size fraction for the removal
of Cd, Cr, Pb and Ni, while Cu and Zn will need to be removed as free ions. In the
dissolved fraction of <1 µm, organic matter can be targeted for the removal of Cd,
Cr, Pb, Cu and Ni, while Zn still needs to be removed as free ions.
4.3.4. FC Analysis and PROMETHEE Ranking of Heavy Metals Wash-off

From Figure 4.2, the impacts of rainfall attributes such as the intensity, frequency
and duration on the wash-off of heavy metals were not easily discernible. Hence,
fuzzy clustering (FC) coupled with PROMETHEE II net ranking was undertaken in
order to further clarify the heavy metals wash-off. The cluster exponent was set to
1.9 and the number of clusters to four. This approach was chosen in order to
incorporate low, moderate, high and extreme rainfall events. The events with
103
intensity <40 mm/hr with relatively low average recurrence interval (ARI) were
classified as low events; those having intensity between 50 to 100 mm/hr but with
relatively higher ARIs of up to 50 years were classified as moderate events; events
with intensities >100 mm/hr with very high frequency were classified as high events
whilst events with similar intensities to moderate and high with extremely rare
occurrence (ARI ≥ 100 years) were classified as extreme events. The membership
threshold was set to 0.25. Table 4.3 shows the membership values obtained by fuzzy
clustering of each object and the clusters were defined accordingly.
Table 4.3 Membership values of the rainfall events in heavy metals wash-off after fuzzy
clustering
Rainfall Duration, Intensity, Cluster
ARI Cluster 1 Cluster 2 Cluster 3 Cluster 4
Events* minutes mm/hr names
1 60 39.3 1 0.035 0.840 0.082 0.043 low
2 65 37.39 1 0.018 0.929 0.032 0.020 low
3 90 39.3 2 0.049 0.836 0.065 0.050 low
4 120 24.6 1 0.258 0.386 0.201 0.155 fuzzy
5 133 39.3 5 0.047 0.812 0.054 0.086 low
6 160 39.3 10 0.030 0.886 0.031 0.053 low
7 5 125 1 0.048 0.053 0.874 0.026 high
8 5.75 119 1 0.069 0.042 0.859 0.030 high
9 10.5 120 2 0.055 0.066 0.838 0.041 high
10 16 125 5 0.105 0.100 0.531 0.264 fuzzy
11 21 122 10 0.256 0.125 0.122 0.497 fuzzy
12 45 125 100 0.105 0.090 0.050 0.755 extreme
13 49 115 100 0.050 0.014 0.010 0.926 extreme
14 52.5 77 10 0.494 0.032 0.034 0.440 fuzzy
15 67.5 77 20 0.915 0.015 0.032 0.039 moderate
16 86.8 77 50 0.955 0.007 0.009 0.029 moderate
17 101.25 77 100 0.011 0.005 0.003 0.981 extreme
18 105 75 100 0.409 0.051 0.036 0.504 fuzzy
19 25 63 1 0.831 0.032 0.073 0.064 moderate
20 42.5 61.2 2 0.362 0.053 0.036 0.550 fuzzy
21 69 59.2 5 0.250 0.092 0.049 0.609 fuzzy
22 85 58.3 10 0.279 0.048 0.038 0.636 fuzzy
*
Numerical object identifiers used elsewhere are same as the rainfall events numbers
As some of the rain events remained unclassified (fuzzy) after clustering, the
PROMETHEE II routine was used to obtain their φ net ranking flow. The range of
this value was used to determine the extent to which one cluster outranks another
104
cluster. Hence, a fuzzy rain event from Table 4.3 could still be classified if its φ
value falls within the range of a cluster. Table 4.4 presents the φ ranking values for
both the classified and fuzzy events.
Table 4.4 PROMETHEE II net ranking ( φ ) values showing two fuzzy events, 10 and 21 could
still be classified as members of moderate and extreme clusters, respectively as their φ values
fall exclusively within the ranges of corresponding clusters
Net ranking values of Net ranking values of
Rain events (clusters) Fuzzy Rain Events
classified rain events fuzzy rain events
2(low) 0.18 4 0.18
1(low) 0.09 20 0.14
3(low) 0.06 11 0.02
6(low) 0.02 10 0.20
5(low) -0.22 14 0.00
19(moderate) 0.20 22 -0.07
15(moderate) 0.03 18 -0.14
16(moderate) -0.17 21 -0.43
13(extreme) -0.24 - -
17(extreme) -0.44 - -
12(extreme) -0.50 - -
7(high) 0.55 - -
8(high) 0.50 - -
9(high) 0.12 - -
It is evident from Table 4.4 that the low and moderate clusters do not decisively
outrank each other as the ranges of their φ values overlap. Hence, some fuzzy
events (events 4, 11, 14, 18, 20 and 22) were inclusive to both low and moderate
clusters after PROMETHEE II ranking. These events were further classified into a
new sub-cluster called “low to moderate”. Consequently, event 10 could now be
classified as moderate while event 21 as extreme as their φ values fall exclusively
within the ranges of the corresponding clusters. Due to changes in rainfall
characteristics, this approach decomposed the heavy metal wash-off scenarios into
105
five easily identifiable clusters, namely, low, moderate, low to moderate, high and
extreme. Further heavy metal wash-off analyses were based on these clusters.
4.3.5. GAIA Analysis Incorporating Impacts of Traffic and Climate Change

As the final step, the study incorporated the impact of traffic and climate change into
the heavy metal build-up and wash-off analyses by considering together all the
clusters obtained. The five size fractions (>300µm, 150-300µm, 75-150µm, 1-75µm
and <1µm) were considered as separate scenarios and were combined together for
both build-up and wash-off. Under the existing conditions of heavy metals build-up
and wash-off, the dominant build-up and wash-off clusters (which include the
objects or actions) were investigated. The magnitude and inclination of the π
decision vector indicated the dominance of a particular action or a group of actions
under the combined scenarios. GAIA multicriteria routine was used to generate the
biplots of the build-up and wash-off scenarios shown in Figure 4.3a and 4.3b
respectively.
(a) (b)
Figure 4.3 GAIA biplots for (a) build-up :( ) low traffic, ( )moderate traffic and ( ) high
traffic: object identifiers are described in Table 1;and (b) wash-off : ( ) low event, ( ) low to
moderate event, ( ) moderate event, ( ) high event and ( ) extreme event: numerical object
identifiers are described in Table 5.3
106
In Figure 4.3a the decision vector π was located between moderate and high traffic
(between the objects CL and IS). The particle size >150µm was strongly correlated
with moderate to high traffic (CL and IS) whilst smaller particle sizes from 1 to
150µm were strongly correlated with moderate to low traffic (RB and RP). The
rapid re-suspension of the larger particles (>150µm) in high traffic areas was
attributed to this fact. The potential dissolved fraction (<1µm) and the finer fraction
of 1-75µm were also strongly correlated with both moderate and high traffic (IS and
IBT for <1µm; CL for 1-75µm). This pointed to the fact that the attributes of the
moderate to high traffic sites will dominate heavy metals build-up in a mixture of
urban traffic clusters for both particulate and dissolved fractions.
The GAIA biplot for heavy metals wash-off (Figure 4.3b) shows that the decision
axis is situated between moderate (events 10 and 19) and high (events 7 and 8) rain
events. The high rain events and moderate rain events have significant correlations
with both, particulate and dissolved fractions. The low to moderate rain events
(events 1 and 20) have limited impacts on these fractions as evident from their low
correlations with both particulate and dissolved fractions. The extreme and low rain
events did not have any noteworthy correlations with the wash-off of any heavy
metal species. These findings suggest that the attributes of the moderate to high
rainfall events will dominate heavy metal wash-off arising from predicted changes to
rainfall characteristics due to climate change. Another important point to note in the
GAIA biplot for wash-off is the magnitude of the loading vector for the 1-75µm
fraction which is greater than all other fractions with little or no correlations with
most of the rain events. Hence, it is postulated that the wash-off of heavy metals in
this fraction is independent of any rain event and will not be affected by the
107
predicted changes to rainfall characteristics due to climate change. Therefore, any
adaptive measure incorporating the impacts of climate change on heavy metals
wash-off must include the 1-75 µm fraction irrespective of the classified rain events
described in this paper.
4.4. Acknowledgements
The research study was undertaken as a part of an Australian Research Council
funded Linkage project (LP0882637). The first author gratefully acknowledges the
postgraduate scholarship awarded by the Queensland University of technology to
conduct his doctoral research. The help and support from Gold Coast City Council
and Queensland Department of Transport and Main Roads is also gratefully
acknowledged.
4.5. Supporting Information Available

Detailed description of sample collection, quality control and data analyses
techniques are presented in the Appendix A.2 along with additional tables and
figures for interested readers. This information is available free of charge via the
internet at http://pubs.acs.org.
4.6. Brief
The build-up and wash-off of heavy metals on urban roads have been investigated
from a dynamic point of view incorporating the changes in urban traffic and rainfall
characteristics.
4.7. References
(1) Zhao, H.; Li, X.; Wang, X.; Tian, D. Grain size distribution of road-deposited
sediment and its contribution to heavy metal pollution in urban runoff in Beijing,
China. J. Hazard. Mater. 2010, doi: 10.1016/j.jhazmat.2010.07.012.
108
(2) Bureau of Infrastructure, Transport, and Regional Economics: Australian
transport statistics yearbook 2007. Canberra, ACT.
http://www.btre.gov.au/statistics/statsindex.aspx
(3) CSIRO Division of Marine and Atmospheric Research: The impact of climate
change on extreme rainfall and coastal sea levels over south-east Queensland. Part 2:
A high-resolution modelling study of the effect of climate change on the intensity of
extreme rainfall events. 2007, Aspendale, Victoria.
http://www.csiro.au/resources/Publications.html
(4) Delpla, I.; Jung, A.-V.; Baures, E.; Clement, M.; Thomas, O. Impacts of
climate change on surface water quality in relation to drinking water production.
Environ. Int. 2009, 35, 1225-1233.
(5) Brown, L.; Affum, J.; Chan, A. Transport pollution futures for Gold Coast city
2000, 2011, 2021, based on the Griffith University Transport Pollution Modelling
System (TRAEMS). In Urban Policy Program Research Monographs; Dodson, J.
Ed.; Urban Policy Program, Griffith University, Brisbane, 2004.
(6) U.S. Environmental Protection Agency: Water pollution aspects of street
surface contaminants. Report no. EPA-R2-72/081, Washington, D. C. 1972.
(7) Bujon , G.; Herremans, L.; Phan, L. Flupol: A forecasting model for flow and
pollutant discharge from sewerage systems during rainfall events. Water Sci.
Technol. 1992, 25, 207-215.
109
(8) Harrison, R. M.; Wilson, S. J. The chemical composition of high-way drainage
waters. Sci. Total Environ.1985, 43, 63-77.
(9) Grottker, M. Runoff quality from a street with medium traffic loading. Sci.
Total Environ.1987, 59, 457-466.
(10) Karlsson, K.; Viklander, M.; Scholes, L.; Revitt, M. Heavy metal
concentrations and toxicity in water and sediment from stormwater ponds and
sedimentation tanks. J. Hazard. Mater. 2010, 178, 612-618.
(11) Zheng, N.; Liu, J.; Wang, Q.; Liang, Z. Health risk assessment of heavy metal
exposure to street dust in the zinc smelting district, northeast of China. Sci. Total
Environ.2010, 408, 726-733.
(12) Birch, G. F.; McCready, S. Catchment condition as a major control on the
quality of receiving basin sediments (Sydney harbour, Australia). Sci. Total
Environ.2009, 407, 2820-2835.
(13) Herngren, L. Build-up and wash-off process kinetics of PAHs and heavy
metals on paved surfaces using simulated rainfall. PhD Dissertation, Queensland
University of Technology, Brisbane, 2005.
(14) Yuan, Y.; Hall, K.; Oldham, C. A preliminary model for predicting heavy
metal contaminant loading from an urban catchment. Sci. Total Environ. 2001, 266,
299-307.
110
(15) Mahbub, P.; Ayoko, G.; Egodawatta, P.; Yigitcanlar, T.; Goonetilleke, A.
Traffic and climate change impacts on water quality: measuring build-up and wash-
off of heavy metals and petroleum hydrocarbons. In Rethinking Sustainable
Development: Urban management, Engineering and Design; Yigitcanlar, T. Ed.;
Engineering Science Reference, New York, 2010.
(16) CSIRO: Climate change in Australia. Technical Report 2007,
http://www.csiro.au/resources/Publications.html.
(17) Johansson, C.; Norman, M.; Burman, L. Road traffic emission factors for
heavy metals. Atmos. Environ.2009, 43, 4681-4688.
(18) Sternbeck, J.; Sjödin, Å.; Andréasson, K. Metal emissions from road traffic
and the influence of resuspension – results from two tunnel studies. Atmos.
Environ.2002, 36, 4735-4744.
(19) US Environmental Protection Agency: Determination of trace elements in
waters and wastes by Inductively Coupled Plasma – Mass Spectrometry. Method
200.8, 1994.
(20) Federal Highway Administration: Technical Advisory Report T5040.36. US
Department of Transportation, 2005,
http://www.fhwa.dot.gov/pavement/t504036.cfm.
(21) Massart, D. L.; Vandeginste, B. G. M.; Buydens, L. M. C.; De Jong, S.; Lewi,
P. J.; Smeyers-Verbeke, J. Handbook of Chemometrics and Qualimetrics Part A;
Elsevier: Amsterdam, 1997.
111
(22) Bezdek, J. C. Pattern Recognition with Fuzzy Objective Function Algorithms;
Plenum Press: New York, 1982.
(23) Otto, M. Fuzzy theory explained. Chemom. Intell. Lab. Syst.1988, 4, 101-120.
(24) Keller, H. R.; Massart, D. L.; Brans, J. P. Multicriteria decision making: a case
study. Chemom. Intell. Lab. Syst.1991, 11, 175-189.
(25) Patra, A.; Colvile, R.; Arnold, S.; Bowen, E.; Shallcross, D.; Martin, D.; Price,
C.; Tate, J.; Apsimon, H.; Robbins, A. On street observations of particulate matter
movement and dispersion due to traffic on an urban road. Atmos. Environ.2008, 42,
3911-3926.
(26) Manoli, E.; Voutsa, D.; Samara, C. Chemical characterization and source
identification/ apportionment of fine and coarse air particles in Thessalonoki,
Greece. Atmos. Environ.2002, 36, 949-961.
(27) Charlesworth, S. M.; Lees, J. A. Particulate associated heavy metals in the
urban environment: their transport from source to deposit, Coventry, UK.
Chemosphere. 1999, 39, 833-848.
(28) Murakami, M.; Fujita, M.; Furumai, H.; Kasuga, I.; Kurisu, F. Sorption
behavior of heavy metal species by soakaway sediment receiving urban road runoff
from residential and heavily trafficked areas. J. Hazard. Mater. 2009, 164, 707-712.
112
CHAPTER 5 CHARACTERISING THE BUILD-
UP OF HEAVY METALS AND VOLATILE
ORGANIC COMPOUNDS ON URBAN ROADS
Manuscript Title
Analysis of Build-up of Heavy Metals and Volatile Organics on Urban Roads in
Gold Coast, Australia
Parvez Mahbub*, Ashantha Goonetilleke*, Godwin A. Ayoko**, Prasanna

Egodawatta*, Tan Yigitcanlar*
*School of Urban Development, Faculty of Built Environment and Engineering

(E-mail : s.mahbub@qut.edu.au; a.goonetilleke@qut.edu.au;
p.egodawatta@qut.edu.au; tan.yigitcanlar@qut.edu.au)
**School of Physical and Chemical Sciences, Faculty of Science and Technology
(Email :g.ayoko@qut.edu.au)
Queensland University of Technology, Brisbane, Australia
Published (2011): Water Science & Technology, 63(9), 2077-2085. Impact

Factor 1.09; ERA ranking B.
Abstract
Urban water quality can be significantly impaired by the build-up of pollutants such
as heavy metals and volatile organics on urban road surfaces due to vehicular traffic.
Any control strategy for the mitigation of traffic related build-up of heavy metals
and volatile organic pollutants should be based on the knowledge of their build-up
processes. In the study discussed in this paper, the outcomes of a detailed
experimental investigation into build-up processes of heavy metals and volatile
organics are presented. It was found that traffic parameters such as average daily
traffic (ADT), volume over capacity ratio (V/C) and surface texture depth (STD) had
similar strong correlations with the build-up of heavy metals and volatile organics.
Multicriteria decision analyses revealed that that the 1 to 74 µm particulate fraction
of total suspended solids (TSS) could be regarded as a surrogate indicator for
113
particulate heavy metals in build-up and this same fraction of total organic carbon
(TOC) could be regarded as a surrogate indicator for particulate volatile organics
build-up. In terms of pollutants affinity, total suspended solids (TSS) was found to
be the predominant parameter for particulate heavy metals build-up and total
dissolved solids (TDS) was found to be the predominant parameter for the potential
dissolved fraction in heavy metals build-up. It was also found that land use did not
play a significant role in the build-up of traffic generated heavy metals and volatile
organics.
Keywords Heavy metals, pollutant build-up, traffic pollutants, urban water

quality, volatile organics

preparation of figures and tables.



Tan Yigitcanlar (Co-author)

This chapter is an exact copy of the accepted manuscript of the journal paper.

Development
objectives of the overall research study. Amongst several objectives, characterising
the build-up of traffic generated pollutants and accordingly, providing guidance to
stormwater quality mitigation strategies were mentioned in section 1.3, Chapter 1.
114
The methodologies to achieve such objectives were formulated in this journal paper
using principal component analysis (PCA), preference ranking organisation method
for enrichment evaluation (PROMETHEE) and geometrical analysis for interactive
aid (GAIA). The fulfillment of the overall objectives of this research study were
influenced by the findings of this journal paper. The journal paper has identified the
traffic and pavement parameters that influence the build-up of heavy metals and
volatile organic compounds on roads. In addition to that, surrogate parameters and
predominant particulate fractions during build-up of these traffic generated
pollutants were also identified. Figure 3.3 in section 3.7, Chapter 3 provides a
schematic flow diagram of this research study where the publication of this journal
paper was highlighted as an integral process that contributes to the overall research
outcome.
115
5.1. Introduction
Rapid urbanisation is a global phenomenon that is happening as a result of increased
demand in urban activities throughout the world. One of the significant impacts of
urbanisation is the increase in vehicle usage on urban roads. The scenario of changes
in urban traffic due to increased urbanisation can readily affect the pollutant build-
up on road surfaces. In this regard, the environmental impacts and more specifically
the water quality impacts of road transport and their mitigation strategies have
received limited attention (BITRE 2008; Brown et al. 2004; Tomerini & Brown
1998). Some proposed models such as Transport Planning Add-on Environmental
Modelling System (Brown et al. 1998) and Vehicle Contaminant Load Model
(Gardiner & Armstrong 2007) have assumed a simplified pollutant accumulation
process to describe pollutant inputs into receiving water bodies from road traffic.
Wu et al. (1998) used long term average pollutant loading rates to characterise
highway pollutant loading. Charlesworth and Lees (1999) studied particulate
associated heavy metal pollutants and identified the dominant heavy metal species in
the source-transport-deposition cascade. Pollutant accumulation in the urban
environment is a complex process and an in-depth understanding of the process of
build-up of significant pollutants on the road surface will strengthen the knowledge
base leading to improved stormwater quality mitigation strategies.
The major pollutants to water bodies that are generated by transport activities
include polycyclic aromatic hydrocarbons (PAHs), total petroleum hydrocarbons
(TPHs), volatile organics and heavy metals (Hoffman et al. 1982, 1984; Sansalone &
Buchberger 1997). These have significant human health impacts. This paper
investigates the build-up processes of heavy metals and volatile organics generated
116
by urban traffic on road surfaces. The outcome of this study will contribute to the
development of robust mitigation measures of improve urban water quality in terms
of vehicle generated heavy metals and volatile organics in build-up on urban road
surfaces.
5.2. Site Selection

The research study was undertaken in the Gold Coast region of south-east
Queensland, Australia. The study adopted a suburb based approach by selecting ten
road sites which represented a combination of residential, commercial and industrial
land uses in two suburbs. These two suburbs reflected the transport infrastructures
that were developed over the last decade in the Gold Coast region. The selected
suburbs were Helensvale and Coomera. The selection of different land uses provided
a cross-section of traffic activities on road surfaces in the Gold Coast region. The
average daily traffic (ADT) and congestion are two important traffic parameters in
terms of characterising urban roads. Congestion on urban roads mainly occurs
during peak hours. The volume to capacity ratio (V/C) describes the traffic activities
on the stretch of a road during peak hours (Ogden & Taylor1999). Hence, ADT and
V/C were selected as traffic parameters in this study. A number of pavement
parameters such as surface texture depth (STD) and lane width were also used to
characterise the sample collection sites. A sophisticated transport model called
‘ZENITH’ (GCCC 2006), which is currently being used by the Gold Coast City
Council for their pavement infrastructure planning and design activities, was used to
predict the current average daily traffic (ADT) and volume over capacity ratios
(V/C) for the selected sites. The surface texture depth (STD) and the lane widths of
the roads were measured at the study sites. Figure 5.1 shows the ground level photo
of one of the build-up sites. This was a typical residential site having DG14 grade
117
asphalt with 5.1 % aggregate binder. The surface texture depth of this site is 0.75
mm. Table 5.1 shows the selected sites and their corresponding traffic and pavement
data.
Figure 5.1 Build-up sample collection site
Table 5.1 Traffic and pavement characteristics data of the selected sites
Volume
Surface
to Lane
Site Name (Labels) Average Daily Texture
Land Use Capacity Width
Traffic(ADT)* Depth
Ratio (m)
(mm)
(V/C)*
Abraham Road (1C) Commercial 13028 1.11 0.6467 3.5
Reserve Road (2R) Residential 6339 0.45 0.7505 3.5
Peanba Park road (3R) Residential 581 0.15 0.6844 2.8
Beattie Road (4I) Industrial 2670 0.24 0.7074 3.5
Shipper Drive (5I) Industrial 7530 0.55 0.6788 3.5
Hope Island Road (6C) Commercial 7534 0.57 0.7254 3.4
Lindfield Road (7C) Commercial 2312 0.33 0.9417 3.3
Town Centre Drive (8C) Commercial 24506 0.62 0.6416 3.5
Dalley Park Drive (9R) Residential 3534 0.42 0.8342 2.9
Discovery Drive (10R) Residential 9116 0.25 0.6957 2.9
*GCCC (2006)
5.3. Build-up Sample Collection

A sample collection method referred to as ‘Wet and Dry Vacuum System’ (Mahbub
et al. 2009) was used in this study. Deionised water was sprayed using a high
118
pressure sprayer on a 2×1.5 m plot followed by a thorough vacuuming using a
domestic vacuum cleaner. In terms of collecting samples from an actual road surface
subject to atmospheric wear and tear as well as daily traffic, this method achieved
the same level of efficiency as described in earlier studies performed on synthetic
surfaces (Deletic and Orr 2005; Egodawatta 2007). The build-up samples were
collected in 25 L plastic containers containing deionised water. Homogeneous 500
mL subsamples were transferred to high density 1 L polyethylene bottles using a
churn splitter. The particulate analytes were fractioned into four sizes namely 300
µm, 150-299 µm, 75-149 µm, 1-74 µm using wet sieving. The filtrate that passed
through a 1 µm membrane filter was considered to contain the potential dissolved
analytes. Samples were collected from the road surfaces after a minimum antecedent
dry period of 7 days (Egodawatta 2007).
5.4. Test Results and Data Analyses

The methods used for sample collection, digestion and determination are covered in
USEPA 200.8 (EPA 1994). The methods followed for the determination of volatile
range organics were USEPA 5035, 5030B, 8015, 8021, and 8260 (EPA 2008). The
road surface texture depth was measured according to the recommendations of the
US Federal Highway Administration (FHWA 2005).
5.4.1. Heavy Metals

The heavy metals selected for this investigation were cadmium (Cd), chromium
(Cr), nickel (Ni), lead (Pb), zinc (Zn), copper (Cu), manganese (Mn), aluminium
(Al) and iron (Fe). Iron and lead respectively had further two and three species in
terms of different isotopes. The selection of heavy metals for analysis was based on
a detailed literature review on heavy metal pollution generated by road traffic (for
example Drapper et al. 2000; Deletic and Orr 2005; Herngren et al. 2006). The
119
particulate and potentially dissolved heavy metals were tested in the build-up
samples. For quality control, calibration standards, internal standards, blanks and
certified reference materials were used. The trace metal detection was performed
using inductively coupled plasma/mass spectrometry (ICP/MS). The percentage
recovery of the target heavy metals ranged within 85% to 115%. The relative
standard deviations of the repetitive samples were found within 1.5% to 15% for
different heavy metals.
Principal component analysis (PCA) is regarded as an effective tool for pattern
recognition (Massart et al. 1997) and hence used in this study to identify inherent
patterns in the data. All twelve heavy metals along with other parameters including
total and dissolved organic carbon (TOC and DOC), particle size distribution (PSD),
pH, electrical conductivity (EC), average daily traffic (ADT), volume to capacity
ratio (V/C), total and dissolved suspended solid (TSS and TDS) and surface texture
depth (STD) were considered as variables. The ten study sites were considered as
objects. The pollutant concentrations were expressed in mg/m2 of the road surface.
The PCA biplots in Figures 5.2 and 5.3 show the patterns observed in two of the
four particulate size fractions investigated. To differentiate the isotopes of iron and
lead, corresponding molecular weights are shown alongside the biplots.
120
Figure 5.2 PCA biplot for heavy metals build-up on urban roads at 150-299 µm fractions
(objects are represented with numbered labels with suffix C=commercial, I=Industrial or
R=residential)
In Figure 5.2, the 150-299 µm particulate fraction shows two groups of variables
that are negatively correlated with ADT. In this fraction, TSS has a strong positive
correlation with the iron species, manganese, TOC and aluminium in one group,
whilst copper, zinc, nickel and cadmium has strong positive correlation with STD,
pH, EC and V/C. This indicates that the iron species, manganese and aluminium
could be from sources other than traffic, whilst copper, zinc, nickel and cadmium
would be generated from traffic. It is also noticeable that these two groups are nearly
perpendicular to each other which indicate that they were independent of each other.
The lead species had moderate positive correlation with TOC. Nonetheless, the lead
species were significant as they had large loadings in the biplot. In Figure 5.2, only
three objects (3R, 4R and 5C) had positive scores along with positive loadings of all
variables except ADT on PC1 whilst the rest of the objects had negative scores
along with negative loading of ADT on PC1. None of the objects except 3R and 4R
had noteworthy scores on PC2. Similar findings regarding other fractions (e.g., Fig.
5.3) underlined the fact that traffic related heavy metals build-up was not directly
121
influenced by the land use; rather the traffic and pavement characteristics were more
directly correlated with the heavy metals build-up.
Figure 5.3 PCA biplot for heavy metals build-up on urban roads at 1-74 µm fractions (objects
are represented with numbered labels with suffix C=commercial, I=Industrial or R=residential)
Initially, the fact that the decrease in ADT was related to the increase of heavy
metals build-up on roads appeared to be contradictory. However, after a close
examination of V/C and STD, it was found that these traffic parameters were also
negatively correlated with ADT. This can be explained by the fact that as the
capacity of a lane is fixed, the increase in V/C indicates congestion on the road with
low movement of traffic. Hence, it is postulated that the decrease in ADT as noted
indicates lane congestion which in turn caused the increase in trace element build-up
on the road surface. Also, low traffic movement during congestion could affect the
texture of the road surface in a different way than high traffic movement during little
or no congestion. Table 5.2 gives the correlation matrix between target heavy metals
and traffic, pavement and other significant chemical parameters.
122
Table 5.2 Total correlation matrix between heavy metals and other parameters
Heavy ADT V/C STD pH EC PSD TSS TOC
Metals
Al -0.82 0.47 0.39 0.37 0.50 0.36 0.97 0.98
Cr 0.13 -0.04 -0.10 -0.11 -0.04 -0.15 -0.10 -0.08
Mn -0.63 0.25 0.12 0.10 0.25 0.08 0.99 0.98
Fe/56 -0.82 0.47 0.39 0.37 0.50 0.36 0.97 0.98
Ni -0.79 0.82 0.94 0.96 0.94 0.99 0.15 0.20
Cu -0.53 0.76 0.71 0.64 0.53 0.55 0.20 0.31
Zn -0.82 0.93 0.95 0.91 0.86 0.85 0.32 0.43
Cd -0.83 0.91 0.99 0.99 0.96 0.99 0.19 0.27
Pb/206 -0.60 0.67 0.55 0.47 0.44 0.38 0.52 0.61
In Figure 5.3, the 1-74 µm fraction also revealed two distinct groups of variables
that were again pointing towards different sources. In this fraction, TSS has a strong
correlation with copper, zinc, aluminium, manganese, iron and lead species. In the
other group, cadmium and chromium has a very strong positive correlation with
TOC, V/C, pH, EC, PSD and STD. Nickel has an insignificant loading score in this
fraction. There was no impact of land use on traffic related heavy metal build-up
detected in this fraction as well.
The PCA of the potential dissolved fraction of heavy metals in the build-up is shown
in Figure 5.4. Unlike the particulate fractions discussed so far, the potential
dissolved fraction did not show any particular groups of variables as evident in
Figure 5.4. However, two important similarities with the particulate fractions were
still found. These were as follows:
• ADT has negative correlation with all the variables including V/C and STD.
• No impact of land use could be found in the potential dissolved fraction.
In this fraction, zinc, copper, cadmium, chromium and lead species were very
strongly correlated with V/C, STD and TDS.
123
Figure 5.4 PCA biplot for heavy metals build-up on urban roads at <1 µm fractions (objects are
represented with numbered labels with suffix C=commercial, I=Industrial or R=residential)
5.4.2. Volatile Organics

The volatile organics investigated were toluene (TLE), ethylbenzene (ETB), meta
and para xylene (MPX) and ortho xylene (OX). A purge and trap system along with
gas chromatograph mass spectrometry were used for sample extraction and
determination. The lower reporting limit for each analyte was 0.001 mg/L. The
particulate and potential dissolved fractions were prepared same as for the heavy
metals. For quality control, calibration standards, internal standards and surrogates
were used as recommended. Analyses of all particulate and the dissolved fractions of
volatile organics revealed that the target volatiles form a group of variables that has
very strong positive correlations with TOC for all particulate fractions. The high
percentage of carbon in the molecular structures of the volatile organics is attributed
to be the reason. This also indicates that the target volatiles which are hydrophobic
in nature generally inclined towards the organic carbons primarily in particulate
form. Figures 5.5 and 5.6 show two PCA biplots for the particulate fractions of
volatile organics.
124
Unlike for some heavy metals, it was noted that TSS has no impact on the volatile
organic build-up in any particulate fraction.
Figure 5.5 PCA biplot for volatile organics build-up on urban roads at >300 µm fractions
R=residential)
Figure 5.6 PCA biplot for volatile organics build-up on urban roads at 75-149 µm fractions
R=residential)
The result for the potential dissolved fraction is shown in Figure 5.7. The potential
dissolved fraction, showed opposite results to the particulate fractions. In Figure 5.7,
the volatile organics have moderately positive correlations with TDS and weak
positive correlations with DOC. No impact of land use on traffic related volatile
organic build-up was found for any fraction. A summary of the outcomes of the
PCA is given in Table 5.3 below.
125
Figure 5.7 PCA biplot for volatile organics build-up on urban roads at <1 µm fractions (objects
are represented with numbered labels with suffix C=commercial, I=Industrial or R=residential)
Table 5.3 Affinity of individual pollutants towards different chemical parameters

Affinity in Particulate
Principal Affinity in potential
fractions (1µm to
Pollutant Individual Pollutant dissolved fraction
300µm and higher) in
Category (<1µm) in build-up
build-up
Lead (Pb) - TDS
Nickel (Ni) TOC -
Cadmium (Cd) TOC DOC
Chromium (Cr) TOC DOC
Heavy Metal Zinc (Zn) - TDS
Copper (Cu) - TDS
Iron (Fe) TSS, TOC -
Aluminium (Al) TSS, TOC -
Manganese (Mn) TSS, TOC -
Toluene (TLE) TOC TDS
Volatile
Ethylbenzene (ETB) TOC TDS
Organic
Meta and Para Xylene (MPX) TOC TDS
Carbon
Ortho Xylene (OX) TOC TDS
Table 5.3 is a generalised view of a pollutant’s affinity during build-up. A closer
analysis of Table 5.3 points to two specific issues that merited further investigation
in order to better understand the build-up processes of heavy metals and volatile
organics on urban roads. These issues are:
• For heavy metals, which chemical parameter out of TOC, DOC, TSS and
TDS is predominant in terms of pollutant’s affinity towards them, both in
particulate and dissolved form;
• For both heavy metals and volatile organics, which particle fraction is
predominant in terms of the pollutant’s affinity towards predominant
126
chemical parameters in particulate form.
5.5. Multicriteria Decision Analyses for Heavy Metals and Volatile

Organics Build-up
Multicriteria decision analyses (Keller et al. 1991) incorporating geometrical
analysis for interactive aid (GAIA) was used to investigate the two issues
highlighted in Table 5.3. In order to determine the affinity of heavy metals and
volatile organics towards different chemical parameters such as TSS, TDS, TOC and
DOC, the concentrations were expressed as loadings (mg of target analytes/mg of
predominant chemical parameter). According to Figure 5.8, for heavy metals the
combined particulate fraction from 1 to 300 µm and higher showed that TSS was the
predominant parameter rather than TOC and in Figure 5.9, the combined potential
dissolved fraction revealed that TDS as the predominant parameter rather than DOC
in terms of their build-up loadings. The inclination of the pi-decision axis towards a
chemical parameter determined its predominance.
1-74 µm
>300 µm
75-149 µm
150-299 µm
Figure 5.8 GAIA biplot for predominant chemical parameter scenario ( ) for particulate heavy
metals; ( ) pi-decision axis; ( ) >300 µm fractions; ( )150-299 µm fractions; ( ) 75-149 µm
fractions; ( ) 1-74 µm fractions
127
Figure 5.9 GAIA biplot for predominant chemical parameter scenario ( ) for the dissolved
heavy metals; ( )pi-decision axis; ( ) <1 µm fraction
The predominant particulate fraction in build-up was also analysed for both heavy
metals and volatile organics. Figure 5.10 shows that the particle size fraction 1 to 74
µm was predominant for heavy metals build-up on the road surfaces. As TSS was
primarily represented by the 1 to 74 µm particulate fraction in Figure 5.10, TSS can
be regarded as a surrogate indicator for particulate heavy metals in build-up. In the
case of volatile organics, the particle size fraction 1 to 74 µm was also the
predominant fraction for the volatile organics in build-up as shown in Figure 5.11.
As this study found that TOC was mainly present with this fraction, TOC can be
regarded as a surrogate indicator for volatile organics in build-up on urban road
surfaces.
Herngren et al. (2006) found that 0.45 to 75 µm range contained the highest heavy
metal concentration in road deposited sediments whilst Deletic and Orr (2005) found
fractions less than 63 µm had maximum concentration. In this regard, the finding of
this study is significant as it has identified the predominant particulate fraction for
128
heavy metals build-up and characterised the affinity of heavy metals in terms of
predominant chemical parameters both in particulate and the potential dissolved
fraction.
Figure 5.10 GAIA biplot for predominant heavy metals particulate fraction ( ); ( ) pi-decision
axes; ( ) metals’ affinity towards predominant chemical parameters; ( ) study sites; ( ) TSS’
presence in particulate fractions
Figure 5.11 GAIA biplot for predominant volatile organics particulate fraction ( ); ( ) pi-
decision axes; ( ) organics’ affinity towards predominant chemical parameter; ( ) study sites;
( ) TOC’s presence in particulate fractions
5.6. Conclusions
This study has undertaken an in-depth investigation into the inherent processes in
the build-up of heavy metals and volatile organics on urban roads due to vehicular
traffic. It was found that the decrease in average daily traffic (ADT) was associated
129
with the increase in volume over capacity ratio (V/C) and surface texture depth
(STD) for both heavy metals and volatile organics in build-up. Hence, the
congestion of vehicles in a traffic lane was found to be primarily responsible for the
pollutants build-up rather than the vehicle counts during a specified time period. It
was also found that land use did not affect the build-up of traffic related heavy
metals and volatile organics on urban road surfaces.
Multicriteria decision analyses revealed that total suspended solids (TSS) in the 1
to74 µm fraction could be regarded as a surrogate indicator for particulate build-up
of heavy metals whilst total organic carbon (TOC) in the 1 to 74 µm fraction could
be regarded as a surrogate indicator for particulate build-up of volatile organics.
Total suspended solids (TSS) was found to be the predominant parameter in
particulate heavy metals build-up whilst total dissolved solids (TDS) was found to
be the predominant chemical parameter in dissolved heavy metals build-up in terms
of pollutants affinity towards them.
5.7. References
BITRE (2008). Australian Transport Statistics Yearbook 2007. Canberra, ACT,
Bureau of Infrastructure, Transport, and Regional Economics: pp. 167.
Brown, A. L., Affum, J. K, & Tomerini, D. (1998). TRAEMS:The Transport

Planning Add-on Environmental Modelling System. Proceedings of the 19th ARRB
Transport Research Conference.
Brown, L., Affum, J., & Chan, A. (2004). Transport pollution Futures for Gold
Coast City 2000, 2011, 2021, based on the Griffith University Transport Polution
Modelling System (TRAEMS). Urban Policy Program, Griffith University,
Brisbane, QLD. pp: 75.
Charlesworth, S. M., & Lees, J. A. (1999). Particulate Associated Heavy Metals in

the Urban Environment: Their Transport from Source to Deposit, Coventry, UK.
Chemosphere 39(5): 833-848
130
Deletic, A., & Orr, D., W. (2005). Pollution Buildup on Road Surfaces. Journal of
Environmental Engineering 131(1): 49-59.
Drapper, D., Tomlinson, R., & Williams, P. (2000). Pollutant Concentration in Road
Runoff: SouthEast Queensland Case Study. Journal of Environmental Engineering
126(4): 313-320
Egodawatta, P. K. (2007). Translation of small-plot scale pollutant build-up and

wash-off measurements to urban catchment scale. Queensland University of
Technology, Brisbane. PhD Thesis.
EPA(1994). Determination of Trace Elements in Waters and Wastes by Inductively

Coupled Plasma – Mass Spectrometry.US Environmental Protection Agency,
Method 200.8.
EPA (2008). Test Methods for Evaluating Solid Waste, Physical/Chemical Methods:
SW-846. 3rd Edition.
http://www.epa.gov/epawaste/hazard/testmethods/sw846/online/index.htm. United
States Environmental protection Agency (accessed 20 August 2009).
FHWA (2005). Federal Highway Administration Technical Advisory Report

T5040.36. US Department of Transportation.
GCCC (2006). Gold Coast Priority Infrastructure Plan- Derivation of Trunk Road
Infrastructure Network Charges:
FinalReport,http://www.goldcoast.qld.gov.au/gcplanningscheme_0509/attachments/
planning_scheme_documents/part8_infrastructure/reference_documentation.pdf
,Veitch Lister Consulting pp:8 (accessed 21 September 2009).
Gardiner, L. R., & Armstrong, W. (2007). Identifying Sensitive Receiving

Environments at Risk from Road Runoff. Land Transport New Zealand: Research
report No 315. pp. 68.
Herngren, L., Goonetilleke, A., & Ayoko, G. A. (2006). Analysis of Heavy Metals
in Road-Deposited Sediments. Analytica Chimica Acta 571: 270-278.
Hoffman, E. J., Latimer, J. S., Mills, G. L., & Quinn, J. G. (1982). Petroleum
Hydrocarbons in urban runoff from a commercial land use area. Journal Water
Pollution Control Federation 54(11): 1517-25.
Hoffman, E. J., Mills, G. L., Latimer, J. S., & Quinn, J. G. (1984). Urban Runoff as
a Source of Polycyclic Aromatic Hydrocarbons to Coastal Waters. Environmental
Science and Technology 18: 580-587.
Keller, H. R., Massart, D. L., Brans, J. P. (1991). Multicriteria Decision Making: A

Case Study. Chemometrics and Intelligent laboratory Systems 11(2): 175-189.
Mahbub, P., Ayoko, G., Egodawatta, P., Yigitcanlar, T., & Goonetilleke, A. (2009).
Development: Planning, Designing, Engineering and Managing Urban Infrastructure
131
and Development. Yigitcanlar, T., (Ed.). Hersey, PA: Information Science
Reference, accepted for publication Sep 2009.
Massart, D. L., Vandeginste, B. G. M., Buydens, L. M. C., De Jong, S., Lewi, P. J.

Smeyers-Verbeke, J. (1997). Handbook of Chemometrics and Qualimetrics Part A,
Elsevier: 771-804.
Ogden, K. W., Taylor, S. Y. (1999). Traffic Engineering and Management. Institute

of Transport Studies, Monash University, Australia. pp. 592- 594.
Sansalone, J. J., & Buchberger, S. G. (1997). Characterization of Solid and Metal

Element Distributions in Urban Highway Stormwater. Water Science and
Technology 36(8-9): 155-160.
Tomerini, D. M., Brown, A. L. (1998). Predicting the Impacts of Road Transport on

Urban Water Quality. The International Conference on Integrated Modelling of the
Urban Environment, Sydney, Australia, July 1998.
Wu, J. S., Allan, C. J., Saunders, W. L., & Evett, J. B. (1998). Characterization and
Pollutant Loading Estimation for Highway Runoff. Journal of Environmental
Engineering 124(7): 584-592.
132
CHAPTER 6 CHARACTERISATION OF SEMI
AND NON VOLATILE ORGANIC COMPOUNDS
BUILD-UP ON URBAN ROADS
Manuscript Title
Analysis of the Build-up of Semi and Non Volatile Organic Compounds on
Urban Roads

Egodawatta1
1
2
Chemistry Discipline, Queensland University of Technology, GPO Box 2434,
Brisbane 4001, Queensland, Australia
s.mahbub@qut.edu.au; g.ayoko@qut.edu.au; a.goonetilleke@qut.edu.au;
p.egodawatta@qut.edu.au
*
Corresponding Author: Parvez Mahbub;Tel: 61 7 3138 9945;Fax: 61 7 3138 1170;
email: s.mahbub@qut.edu.au
Published (2011): Water Research, 45(9), 2835-2844. Impact Factor 4.8; ERA
ranking A*.
Abstract
Vehicular traffic in urban areas may adversely affect urban water quality through the
build-up of traffic generated semi and non volatile organic compounds (SVOCs and
NVOCs) on road surfaces. The characterisation of the build-up processes is the key
to developing mitigation measures for the removal of such pollutants from urban
stormwater. An in-depth analysis of the build-up of SVOCs and NVOCs was
undertaken in the Gold Coast region in Australia. Principal Component Analysis
(PCA) and Multicriteria Decision tools such as PROMETHEE and GAIA were
employed to understand the SVOC and NVOC build-up under combined traffic
scenarios of low, moderate, and high traffic in different land uses. It was found that
congestion in the commercial areas and use of lubricants and motor oils in the
133
industrial areas were the main sources of SVOCs and NVOCs on urban roads,
respectively. The contribution from residential areas to the build-up of such
pollutants was hardly noticeable. It was also revealed through this investigation that
the target SVOCs and NVOCs were mainly attached to particulate fractions of 75 to
300 µm whilst the redistribution of coarse fractions due to vehicle activity mainly
occurred in the >300 µm size range. Lastly, under combined traffic scenario,
moderate traffic with average daily traffic ranging from 2300 to 5900 and average
congestion of 0.47 was found to dominate SVOC and NVOC build-up on roads.
Keywords Semi volatile organic compounds, non volatile organic compounds,

traffic pollutants, pollutant build-up, multicriteria decision tools

preparation of figures, tables and supplementary documents.



This chapter is an exact copy of the accepted manuscript of the journal paper.

Development
the build-up of traffic generated pollutants and accordingly, providing guidance to
134
The methodologies to achieve such objectives were formulated in this journal paper
using principal component analysis (PCA), preference ranking organisation method
for enrichment evaluation (PROMETHEE) and geometrical analysis for interactive
aid (GAIA) with a view to characterise the build-up of semi and non volatile organic
compounds (SVOCs and NVOCs) under the dynamic changes of urban traffic as
decribed in Chapter 4. This journal paper has contributed to the overall outcome of
this research project by establishing the predominant traffic scenarios that affect the
build-up of such pollutants. The multicriteria decision tools used in this study also
confirmed the sources of such pollutants as well as predominant fractions that need
to be removed from the build-up as a mitigation measure. Figure 3.3 in section 3.7,
Chapter 3 provides a schematic flow diagram of this research study where the
publication of this journal paper was highlighted as an integral process that
contributes to the overall research outcome.
135
6.1. Introduction
Urban traffic activities are one of the predominant sources of stormwater pollutants
that accumulate on urban roads and are eventually transported to receiving water
bodies. In the context of traffic generated pollutants on urban roads, semi volatile
organic compounds (SVOCs) are mainly associated with diesel, fuel oil 1-6 and
kerosene, whilst the non volatile organic compounds (NVOCs) are mainly
associated with motor oils and lubricants (Draper et al. 1996). In a broader sense,
these pollutants are part of a larger family of hydrocarbons which are assessed as
total petroleum hydrocarbon (Morrison & Boyd 1992).
According to the criteria stipulated by the American Petroleum Institute (API),
products such as diesel fuels, fuel oils 1-6 and heavier engine oils and lubricants are
classified as diesel range organics (DROs) (API 1994). These are the most widely
used and distributed traffic related products. Homologous series of n-alkanes from
decane to tetracontane are amongst the most common constituents of these products
(Draper et al. 1996). In this context, particulate n-alkane concentrations on roads can
also result from tyre abrasion and brake lining dust (Rogge et al. 1993). Brown et al.
(1985) reported significant concentrations of vehicle generated SVOCs and NVOCs
in urban runoff which may alter the quality of the receiving water, thus harming the
endemic biological community. Whilst, both petrol and diesel engine vehicles emit
gaseous and particulate hydrocarbons as a result of incomplete combustion (Neeft et
al. 1996), Andreou and Rapsomanikis (2009) noted that past studies mainly
characterised only one organic group (e.g., polycyclic aromatic hydrocarbons). As
the characteristics of urban traffic in terms of traffic volume and congestion is
rapidly changing with increased urbanisation throughout the world, an in-depth
136
understanding of the impacts of traffic generated semi and non volatile organic
compounds on the urban water environment is needed in order to develop
appropriate mitigation measures.
The characterisation of the build-up of semi and non volatile organic compounds on
urban roads due to changing traffic characteristics under rapid urbanisation is the
key to the formulation of appropriate mitigation measures. In this context, the
current state of knowledge on the build-up processes of semi and non volatile
organic compounds on urban roads is limited. Brandenberger et al. (2005) in their
investigation of the emissions of diesel fuels and lubricating oils under different
driving conditions found that poor combustion, reduced conversion efficiency of the
oxidation catalyst, and increased mean load of the vehicle driving cycle were the
primary reasons for increased particulate emissions of lubricating oils and diesel
fuels. However, while their results represented the effects of different driving cycles
of the motor vehicles on the ambient concentrations of particulate pollutants, it is
important to note that not all of the vehicular emissions are necessarily deposited on
impervious surfaces. Ning et al. (2005) reported that the initial pollutant
concentration at the exhaust pipe, exit velocity, exit angle, and crosswind intensity
affect the pollutant dispersion pattern significantly even at the idle condition.
Traffic parameters such as, average daily traffic (ADT) and congestion on the road
(volume to capacity ratio, V/C) along with pavement characteristics such as surface
texture depth (STD) are reported to significantly influence pollutant build-up on
urban roads (Mahbub et al 2010a; Brown et al. 2004; Pitt et al. 1995). The dynamic
variability of the traffic characteristics mentioned above poses a significant threat to
137
urban water bodies through the accumulation of semi and non volatile organic
compounds in the urban environment. In this study, the build-up processes of semi
and non volatile organic pollutants have been characterised with respect to physico-
chemical (e.g., particle size distribution), traffic and land use parameters, and
pavement characteristics. The outcome of this study is expected to provide guidance
for mitigating the impacts of semi and non volatile organic pollutants transported by
urban stormwater runoff to receiving waters.
6.2. Materials and Methods

The site selection criteria were formulated using a suburb based approach. Two
suburbs namely, Helensvale and Coomera in the Gold Coast region in Southeast
Queensland, Australia were selected. The two selected suburbs also represent the
transport infrastructure developed within the Gold Coast City region in the past
decade. Eleven road sites (Table 6.1) located in three different land uses, namely,
residential, commercial and industrial were selected for build-up sample collection.
The selection of different land uses ensured a cross-section of traffic activities on
road surfaces within the Gold Coast region.
138
Table 6.1 Selected road sites with traffic and pavement parameters (partially adapted from
Mahbub et al. 2010a)
Average Top Coat
Daily Surface Age of the Material
Volume to Road
Site Name Geo- Traffic Texture
Land Use Capacity Section, % of
Identifier Coordinates (ADT), Depth
Ratio (V/C) (yrs) Aggregate
vehicles/day (STD), mm
Binder
Abraham Road DG14a
27.865°S
Commercial 13028 1.11 0.6467 3
153.307°E
CA 5.1
Reserve Road DG14a

27.870°S
Residential 6339 0.45 0.7505 3
153.301°E
RR 5.1
Peanba Park Road DG10b

27.851°S
Residential 581 0.15 0.6844 4
153.281°E
RP 5.3
Billinghurst Cres DG10b

27.856°S
Residential 5936 0.74 0.7015 10
153.298°E
RB 5.3
Beattie Road DG14a

27.868°S
Industrial 2670 0.24 0.7074 2
153.324°E
IBT 5.1
Shipper Drive DG14a

27.861°S
Industrial 7530 0.55 0.6788 6
155.332°E
IS 5.1
Hope Island Road
DG14a
27.882°S
Commercial 7534 0.57 0.7254 3
153.328°E
5.1
CH
Lindfield Road DG10b

27.922°S
Commercial 2312 0.33 0.9417 10
153.334°E
CL 5.3
Town Centre Drive DG14a

27.929°S
Commercial 24506 0.62 0.6416 4
153.337°E
CT 5.1
Dalley Park Drive DG10b

27.887°S
Residential 3534 0.42 0.8342 10
153.346°E
RD 5.3
Discovery Drive DG14a

27.899°S
Residential 9116 0.25 0.6957 2
153.327°E
RDS 5.1
a
Dense Grade Bitumen Asphalt with 5.1% aggregate binder
b
139
6.2.2. Key Study Parameters
In the study, the key traffic parameter used was the Daily Traffic (ADT) instead of
Average Annual Daily Traffic (AADT), as the former is predicted by a sophisticated
transport model called ZENITH (GCCC 2006) which is currently being used by the
Gold Coast City Council. Gardiner and Armstrong (2007) have found that traffic
levels measured as AADT are a poor proxy for stormwater runoff quality.
Kayhanian et al. (2003) also reported that AADT itself does not have any direct
correlation with pollutant build-up on road surfaces.
The Volume to Capacity ratio (V/C) of a roadway describes the traffic
characteristics on the stretch of road during the peak hour (Ogden & Taylor1999).
This parameter was found to vary quite significantly for the different sites that were
selected for the study. Studies have shown that vehicle congestion due to increased
traffic volumes in the urban areas had a direct influence on pollutant emission levels
on roads (Smit et al.2008). As such, Average Daily Traffic (ADT) and Volume to
Capacity Ratio (V/C) were incorporated as the two principal traffic parameters that
would influence the build-up of semi and non volatile organic compounds on urban
roads.
The US Federal Highway Administration recommend specific pavement surface
texture depths so that current and predicted traffic needs could be accommodated in
a safe, durable, and cost effective manner (FHWA 2005). The texture depth can
influence pollutant build-up and wash-off from pavement surfaces (Pitt et al. 1995;
Legret & Colandini 1999). The road texture also affects the interactions between the
vehicle tyres and the driving surface (Kreider et al. 2010). Hence, the surface texture
depth of the pavement surfaces at the selected road sites were also incorporated into
140
the study. Table 6.1 lists the selected sites with the identifiers adopted and the
corresponding traffic and pavement characteristics.
6.2.3. Build-up Sample Collection

The pollutant build-up process was characterised as having four main functional
forms such as, linear, power, exponential, and Michaelis-Menton (Huber 1986).
Amongst these, the non-linear asymptotic form proposed by Sartor et al. (1974) has
been most often cited and also used in several stormwater quality models such as,
DR3-QUAL, FHWA, SWMM (Huber 1986). In this context, Egodawatta (2007)
noted that pollutant build-up on road surfaces asymptote to an almost constant value
after a seven day antecedent dry period. Hence, in this study, seven dry days were
allowed at each site prior to any sample collection. Samples were collected over a
two month period in April and May 2009. The weather was dry and the temperature
during the sampling ranged between 22°C to 25°C. Three different time periods
including 8 to 9 am in the morning, 12 to 1 pm at noon as well as 3 to 4 pm in the
afternoon were chosen as sample collection time from the eleven sites to incorporate
both rush hour and normal traffic.
A pilot study, reported in Mahbub et al. (2010b), was undertaken and 90% sample
collection efficiency was achieved through a domestic vacuum cleaner with a water
filtration system. This collection efficiency of the vacuum cleaner was for sand dust
that passed 100% through 420 µm sieve and retained 100% on 0.7 µm Whatman®
GF F glass fiber filter. The test was performed on the middle of the lanes of actual
road surface subject to daily traffic. Three build-up plots of 2 × 1.5 m2 area were
initially cleaned with deionised water and allowed to dry up for 1 hour. It was
assumed that the build-up of pollutants during 1 hr was uniform for the three plots.
141
Two of the plots were applied with 100 gm sand dust and the third plot was kept
without applying any sand dust. The ‘wet and dry vacuum’ system (Mahbub et al.
2010b), which incorporates vacuuming of the build-up plot in dry and subsequently
in wet condition was then applied at different combinations of pressure and time.
The wet condition was created by a sprayer. The difference in the collected sand
dust from the first two plots compared with the third plot at various combinations
indicated that optimum pressure of 2 bar for 3 minutes was required to achieve to
90% collection efficiency. The total build-up sample was collected in 8 L deionised
water.
6.2.4. Sample Preparation

The collected samples were transported to the laboratory and 500 mL sub-samples
were prepared using a churn splitter. The total particulate analytes were fractioned
into four size ranges, namely, >300 µm, 150-300 µm, 75-150 µm, 1-75 µm using
wet sieving. The filtrate passing through a 1 µm Whatman® GF B glass fiber filter
was considered as the potential total dissolved fraction. In each case, 500mL
homogeneous sub-samples were prepared by mixing with deionised water, stored in
500 mL amber glass bottles with PTFE seals, preserved with 5 mL of 50% HCl at
4°C in the laboratory and analysed within 40 days of collection.

The target SVOCs for the study were octane (OCT), decane (DEC), dodecane
(DOD), tetradecane (TED), hexadecane (HXD), octadecane (OCD), Eicosane (EIC),
docosane (DOC), tetracosane (TTC), hexacosane (HXC), and octacosane (OCC)
having boiling points ranging from 125° C to 432° C. The target NVOCs were
triacontane (TCT), dotriacontane (DTT), tetratriacontane (TRT), hexatriacontane
(HXT), octatriacontane (OTT), and tetracontane (TTT) with boiling points ranging
142
from 449° C to 525° C (Kudchadker & Zwolinski 1966). The test methods adopted
for the determination of SVOCs were USEPA 3510C, 8015, 8021, and 8260 (EPA
2008). Draper et al. (1996) proposed modifications to the EPA methods to include
the determination of motor oil with a carbon number up to C38 which was used as a
guide to establish the Gas Chromatographic (GC) temperature programme in this
study for the determination of both SVOC and NVOC simultaneously.
Calibration standards, internal standards, surrogate spikes and blanks were used in
order to maintain quality control and quality assurance of the testing. Nine different
calibration standards (17 component FTRPH calibration standards from
Accustandard®) were prepared at 0.1, 0.5, 0.7, 1, 1.4, 7, 10, 28, 50 mg/L
concentrations for each target analyte. The DRO internal standard (Sigma-Aldrich®)
consisting of acenaphthene-d10, chrysene-d12, naphthalene-d8, perylene-d12,
phenanthrene-d10, 1, 4-dichlorobenzene-d4 were added to each sample and
standards at 5 mg/L concentration. Field blanks were used for each field sample
collection episode and all results were blank corrected.
Three quality control standards (TPH Mix-1-DRO certified reference materials from
Sigma-Aldrich®) at 1, 10 and 50 mg/L concentrations were prepared independently
of the calibration standards and were included in each batch for comparison with the
calibration standards. The sample batch was reanalysed if deviation of >10% from
the certified value was observed for at least half of the target analytes in the quality
control standards. One sample from each batch was spiked with another quality
control standard at a concentration of 35 mg/L. Surrogate standards
(Accustandard®) consisting of 10 mg/L of n-triacontane-d62 were added to seven
143
randomly chosen samples. Seven field blanks were used to establish the limits of
detection (LOD) for each analyte. Values less than LODs were replaced by half of
the LOD values and values above the highest concentration limit of the calibration
standard were discarded as outliers. Seven replicate sub-samples were prepared from
randomly chosen samples from each of the eleven sites. The intra-site relative
standard deviation was found within the range of 8%-19% for each replicate. The
inter-site relative standard deviation was found within the range of 15%-21% for
each analyte. This was within the range of the relative standard deviation suggested
by Horwitz (1982) for ppm level concentrations. Table 6.2 shows the recoveries of
the surrogates and the spikes. The test results for each of the five size fractions are
provided as supplementary data available online with this study.
144
Table 6.2 Percent recoveries of spikes applied at 35 mg/L and surrogate applied at 10 mg/L along with limits of detection for the target compounds
Limits of % recovery of spikes % recovery of surrogate in randomly chosen samples
detection
Analytes Batch Batch Batch
(LOD), Sample1 Sample2 Sample3 Sample4 Sample5 Sample6 Sample7
1 2 3
mg/L
n-triacontane-
Surrogate - - - - 76 79 104 88 81 96 113
d62
Octane 0.54 128 110 101
Decane 0.32 80 82 83
Dodecane 0.44 131 124 97
Tetradecane 0.38 71 76 88
Hexadecane 0.85 75 88 84
Spiked
Octadecane 0.71 72 79 91
SVOCs
Eicosane 0.34 90 86 104
Docosane 0.54 124 129 -
Tetracosane 0.05 110 91 92
Hexacosane 1.07 79 84 86
Octacosane 1.16 78 83 97
Triacontane 1.02 84 70 71
Dotriacontane 1.32 69 79 75
Spiked Tetratriacontane 1.23 85 78 74
NVOCs Hexatriacontane 1.06 98 84 65
Octatriacontane 0.60 - 82 71
Tetracontane 0.85 80 81 86
145
USEPA method 3510C (EPA 2008) was used to extract SVOCs and NVOCs using
the separatory funnel liquid-liquid extraction technique with 250 mL hexane as the
exchange solvent. The samples were cleaned using standard column cleanup
protocol with 5 cm silica gel and 5 cm pyrex® glass wool topped with 5 cm
anhydrous Na2SO4 (EPA2008). Further concentration was carried out using the
Kuderna-Danish apparatus followed by the nitrogen blowdown technique (EPA
2008). The extractions and concentrations were carried out until a final extracted
volume of 1 mL was achieved for Gas Chromatographic (GC) analyses.
A specially built HP5MS Agilent® capillary column of 30 m length, 0.32 mm
internal diameter and 0.25 µm film thickness was used in the GC analyses. The
column was temperature programmed to separate the analytes, which were then
detected by a mass spectrometer interfaced to the GC. A splitless sample injection of
2 µL at an inlet temperature of 280°C, inlet pressure of 35.58 kN/m2 (5.16 psi) and a
flowrate of 2.4 mL/min was used. The initial oven temperature was set at 40°C, held
at that temperature for 12 min., followed by an increase of 10°C per min. until the
oven temperature reached 300°C and finally the temperature was held at 300°C for
20 min. Hence, the total GC runtime was 58 min. per sample. The identification of
target analytes was performed by comparing their mass spectra with the electron
impact spectra of authentic standards.
Other physico-chemical variables such as particle size distribution (PSD) of the sub-
samples were determined using a Malvern Mastersizer S particle size analyser
capable of analysing particle size between 0.05 and 900 µm diameter (Malvern
1994). Total suspended solids (TSS) and total organic carbon (TOC) were analysed
146
by methods 2540D and 5310B (APHA 2005). The PSD of the sub-samples were
compared with each other and used as a guide for homogeneity maintained in the
sub-sampling process. The surface texture depths (STD) of the pavement surface at
the selected road sites were determined according to method T250 (Main Roads
2009). Additionally, the pH and electrical conductivity (EC) of each sample were
measured using standard pH and EC probes in the laboratory according to methods
4500-H+ B and 2510B respectively (APHA 2005).
6.2.6. Data Analyses

The data matrices consisted of 11 objects and 25 variables for each of the five
particle size fractions noted above. The 11 road sites were considered as the 11
objects with identifiers listed in Table 6.1 with the prefixes C, I, or R for
commercial, industrial, and residential land uses, respectively. Variables such as,
ADT, V/C, STD, pH, EC, PSD, TSS, and TOC were considered as attributes of the
objects responsible for the build-up of the target SVOCs and NVOCs, and hence
considered as independent variables. After initial investigation of the probability
distribution of the objects and variables in the data matrices, standardisation of the
variables was performed as a pre-treatment measure so that each variable could be
treated with equal importance in the data analysis.
Chemometric multivariate data analyses techniques such as, principal component
analysis (PCA), preference ranking organisation method for enrichment evaluation
(PROMETHEE) and geometric analysis for interactive aid (GAIA) were employed.
Component extraction processes such as PCA and multicriteria decision making
processes such as PROMETHEE and GAIA have been used recently to characterise
147
the incorporation of pollutants in stormwater runoff from urban roads (Jartun et al.
2008; Ayoko et al. 2007). A brief description of these techniques is discussed below.
PCA
The principal component analysis (PCA) is a data pattern recognition technique that
extracts information from a data matrix by the projection of objects and variables to
the principal components (PCs). The PCs are considered as the latent variables
which are linear combinations of the original variables of the dataset. The PCA
technique transforms the original variables to a new orthogonal set of PCs in such a
way that they contain the data variance in a decreasing order, i.e., the first PC
contains most of the data variance and the second PC contains the second largest
variance and so on. Consequently, the data can be presented diagrammatically by
plotting the loading of each variable in the form of a vector and the score of each
object in the form of a data point. This type of plot is referred to as a ‘Biplot’. More
insight into the PCA technique can be found in Massart et al. (1997). In this study,
SIRIUS2008 software (Sirius 2008) was used to perform the PCA procedures.
PROMETHEE
PROMETHEE is an object ranking technique based on data criteria that uses some
user defined preference functions to prioritise objects (Keller et al. 1991). The
PROMETHEE method calculates the positive and negative outranking flows,
φ + and φ − , respectively based on the preference functions in order to rank the
objects. The φ + value indicates how each object outranks all the others, whilst the
φ − value indicates how each object is outranked by all the others. This procedure is
known as PROMETHEE I ranking. However, in some instances, two objects cannot
be compared as they perform equally on different criteria. In these cases, the net
148
outranking flow, φ which is the algebraic difference between φ + and φ − , is
calculated in order to facilitate the comparison. This procedure is known as
PROMETHEE II ranking.
GAIA
GAIA is essentially a PCA biplot which facilitates a sensitivity analysis for
multicriteria decision methods such as PROMETHEE (Keller et al. 1991). GAIA
provides a graphical view of the objects and variables for net outranking flow φ in
the form of a PCA biplot by decomposing the values from PROMETHEE II into
that it produces a decision axis that takes into account the weights associated with
the variables. These weights can be interactively adjusted for maximum achievable
‘ φ ’ net ranking values obtained by PROMETHEE II. This helps the decision-maker
with an enriched understanding of the problem in terms of the detection of clusters
of objects, conflicts in variables, inability to compare objects and so on. More details
on the PROMETHEE and GAIA methods are discussed in Keller et al. (1991) and
Ayoko et al. (2004).The DecisionLab 2000 software (Decision 2000) was used to
perform PROMETHEE and GAIA analysis.

6.3.1. Trends in the Original Data
The bulk volume of the original data (presented in Appendix A.3 as Tables A.3.1-
A.3.5) makes it hard to discern any meaningful trends. Simple bi-variate correlations
between the target variables at each of the five size fractions in supplementary
Tables A.3.6-A.3.10 showed that the correlation of PSD, pH, EC, solids and organic
carbon with the target SVOCs and NVOCs were very low, within a range of ±0.2 in
the dissolved fraction of <1 µm. With the exception of EC, the correlations of PSD,
149
pH, TSS and TOC with the target compounds started to increase from 1µm to 300
µm size fractions. This suggested that the target compounds were mainly associated
in non-ionic form with the particulate fraction 1-300 µm. More intrusive data
analyses techniques such as, PCA, PROMETHEE and GAIA were employed to
further investigate the trends noted in the original data matrices.
6.3.2. Exploratory PCA

Initially PCA was performed on the pre-treated data matrices starting with the total
particulate fractions from 1 µm - >300 µm as well as the potential dissolved fraction
of <1 µm taken together as shown in Figure 6.1. All the physico-chemical, traffic,
pavement, and land use variables were included along with the target semi volatile
and non volatile compounds.
RD
3 DOD
TSS
DTT
2
TOC
CH RB
1 EC
RR OCC
CT
CL CA TRT
TCT
PC 2 (12.4%)
RDS
0 OCD
STD
TTC TTT
RP TED DOC
HXD
pH
EIC
-1 OTT
HXC
ADT
-2 IBT
HXT
-3
V/C
-4 IS
DEC OCT
-5
-5 0 5 10
PC 1 (70.8%)
Figure 6.1 PCA biplot of total particulate fractions from <1 µm to >300 µm taken together
The traffic parameters V/C and STD were found to be more strongly correlated with
the target SVOCs and NVOCs than ADT in Figure 6.1. This suggested that
150
congestion on the road as well as the road texture conditions affected the build-up of
SVOCs and NVOCs directly whilst ADT may have influenced the redistribution of
particles on the road surface. Whilst the bulk of the free-flowing traffic was in the
commercial and most of the residential areas, low traffic volumes were noted in the
industrial areas. This explains the strong association of ADT with commercial and
residential sites on PC1 in Figure 6.1. The age and the grade of the top coat on the
road as described in Table 6.1 was also found to be important as the STD in Figure
6.1 positively correlates with most of the target variables.
In Figure 6.1, only four objects (two residential and two industrial) were found to be
associated with the target pollutants. This suggested that there is little or varying
influence exerted by the land use parameters on the build-up of SVOCs and NVOCs.
However, without detailed studies on the individual particle size fractions, these
findings could not be validated. Figure 6.2 shows biplots of the build-up of five
individual size fractions from >300 µm to <1 µm.
151
RB >300 µ m
5
5
150-300 µm
TRT
4 4
OTT
IBT
3
3
TCT TTT
Comm ercia l DOD TED
sites correla ted HXT CH
DEC
with SVOCs DOD 2 DTT OCT
2 HXD
DEC
PC 2 (23.3%)
CH IS
PC 2 (20.6%)
1 Industrial sites
RDS correlated with
1 Industria l sites
NVOCs
correla ted with
NVOCs
OCT 0 CA
RDS
0 RB RP HXD
OCC RR
CA RR
TTT TRT -1 RD
TED OTT
CT OCD
OCD RD CT
CL DTT
RP IBT TTC
-1
EIC
TTC DOC HXT -2 EIC
Commercial
TCT
DOC sites correlated
-2 with SVOCs
IS -3
OCC CL HXC
HXC
-3 -4
-5 0 5 -6 -4 -2 0 2 4
PC 1 (35.6%) PC 1 (39.1%)
(a) (b)
CL
4
RD OCC IS
OCD RP 1-75 µm
HXC
2
3
HXD
TRT
DOD
HXD TCT
1
TED 2
RR RP TTT
DOC RD RR
HXC
DTT
CH OTT
TTC 1 OTT
DOD
EIC
0 CA CT HXT TTT
RB TED
PC 2 (19.0%)
PC 2 (20.4%)
TTC CL IS
RDS 0 CA
TCT HXT
CT CH DTT
-1 -1 IBT
OCC
IBT
TRT
DOC
OCD
-2
-2 OCT
DEC
EIC
-3 Industrial
DEC sites
Industrial sites Commercial OCT correla ted
-3
correlated to sites to NVOCs
75-150 µm NVOCs
-4
correlated to
RDS
RB SVOCs
-4 -5
-3 -1 1 3 5 -4 -2 0 2 4 6
PC 1 (39.4%) PC 1 (33.8%)
(c) (d)
4
<1 µm
RD
TRT
DOD
3 HXT
DTT
TCT CH
2
RB
HXC OTT
TED
1 TTT RR
OCD
CT
RP
PC 2 (17.5%)
CL
-1
HXD
IBT
DEC
IS
-2 OCC
DOC OCT
EIC
-3 TTC
-4
CA
-5
-5 0
PC 1 (42.6%)
(e)
Figure 6.2 Individual PCA biplots for (a) >300 µm; (b) 150-300 µm; (c) 75-150 µm; (d) 1-75 µm;
and (e) <1 µm size fractions
152
In Figure 6.2(a), the higher molecular weight NVOCs (422 gmol-1 - 562 gmol-1)
with boiling points ranging from 449°C to 525°C are strongly associated with the
industrial sites whilst the comparatively lighter molecular weight SVOCs (114 gmol-
1
- 394 gmol-1) with boiling points ranging from 125°C to 432°C are mainly
associated with the commercial sites for the >300 µm particulate fraction on PC1.
There are some associations of residential sites (RDS, RP, RD, and RR) with octane
(OCT) and tetradecane (TED) in Figure 6.2(a). However, the association of
residential sites with the build-up of either SVOCs or NOVCs was found to be
generally negligible on both PCs for the >300 µm fraction.
In Figures 6.2(b), 6.2(c) and 6.2(d), similar findings suggested that the semi volatile
components of petrol and diesel fuels are predominantly associated with the
commercial areas whilst the non volatile heavier compounds were mainly associated
with the industrial areas. The commercial areas in this study were close to carparks,
shopping centres as well as service stations and the industrial areas mainly
comprised of marine and light metal industries.
According to Table 6.1, the average congestion (0.66±0.33) in the commercial areas
was much higher than the average congestions (0.40±0.22) in both the industrial and
residential areas. The average volume of traffic in the commercial areas is almost
twice the volumes for the residential and industrial areas. This suggested that slow
moving traffic in the commercial areas were contributing significantly towards the
build-up of SVOCs through exhaust and non-exhaust emissions, whereas, the strong
correlations between NVOCs and the industrial sites observed in particulate
fractions from >300 µm to 1 µm in Figure 6.2 suggested that these NVOCs in the
153
industrial areas may not necessarily originate from traffic alone. Usage of different
types of motor oils and lubricants by machinery in the industrial areas may also
contribute to the build-up of NVOCs in these areas. In either case, the contribution
of residential areas to the build-up of such pollutants on urban roads is hardly
noticeable. It is important to note that traffic generated SVOCs are prominently
associated with particulate matter from 1 µm to >300 µm in the commercial areas in
Figures 6.2(a) through to 6.2(d).
In Figure 6.2(e), for the potential dissolved fraction of <1 µm, the three different
land uses are not directly associated with the build-up of SVOCs and NVOCs as no
clear separation of land use with target variables were identified in either of the PCs.
There are some associations of residential objects (e.g., RB, RD, and RP) with the
build-up of a few SVOCs and NVOCs in Figure 6.2(e). However, as the average
volume of daily traffic in the residential study areas was quite similar (around 5100
vehicles per day) to the industrial areas, it is understandable that the traffic in the
residential areas did not directly influence the build-up of SVOCs and NVOCs.
Patra et al. (2008) noted that coarser particles resuspend and redistribute faster than
the finer particles due to vehicle induced turbulence and the reservoir of finer
particles get replenished by grinding of the coarser particles under the vehicle
wheels. The role of organic matter as a binding agent between solids and other
pollutants has been discussed by Charlesworth and Lees (1999). They reported that
organic matter act as a predominant binder for particle sizes ranging from 63 µm to
2 mm during build-up. Hence, the ‘land use independent’ loadings of SVOCs and
NVOCs in the potential dissolved fraction of <1 µm in Figure 6.2(e) suggest that
154
traffic may have caused the resuspension and redistribution of coarser particles
generated elsewhere and replenished the fine particles of <1 µm size which has
adsorbed the target organics which is independent of the land use. However, the
extent of adsorption of target pollutants by the finer fraction of <1 µm may be very
limited as the loading vectors of the target pollutants in the dissolved fraction of <1
µm are quite similar in magnitude to the particulate fractions in Figures 6.2(a)
through to 6.2(d). This suggests that the variances of target pollutant concentrations
in the dissolved fraction are quite similar to the particulate fractions. This is
attributed to the fact that organic compounds have very limited solubility in most of
the solvents which may cause the target pollutants to remain free without adsorbing
to the fine particles and hence the potential dissolved fraction manifested similar
variances as the particulate fractions.
The PCA analysis provides a fundamental characterisation of the build-up of traffic
related SVOCs and NVOCs for different land uses. In order to characterise such
build-up in terms of particle size fractions as well as the predominant urban traffic
scenarios that influences build-up, PROMETHEE ranking and GAIA analysis were
employed.
6.3.3. PROMETHEE
The preference ranking organisation method for enrichment evaluation
(PROMETHEE) was applied to the same data matrices that were used for the PCA.
In the context of ranking the study sites as urban traffic objects with variable traffic
parameters, Mahbub et al. (2010a) proposed high, moderate, and low urban traffic
scenarios based on a moderately soft fuzzy clustering technique that allows traffic
attributes of different scenarios to intersect with each other. The high traffic scenario
155
comprised of traffic volumes ranging from 9000 to 24000 ADT with relatively high
congestion; moderate traffic scenario comprised of ADT values ranging from 2300
to 5900 with moderate congestion whilst low traffic scenario was associated with
low traffic volume ranging from 500 to 3500 ADT with low congestion.
This study adopted the same classification of urban traffic scenarios to interpret the
PROMETHEE ranking. According to this urban traffic classification system, high
traffic scenario comprised of objects IS, CT, CA, and RDS; moderate traffic
scenario comprised of CH, CL, IBT, RB, and RR whilst low scenario comprised of
RD and RP. Figure 6.3 shows the ‘ φ ’ net outranking flows of the 11 traffic objects.
The three different land use types and the five different size fractions were
incorporated in the ranking. The Gaussian preferential function (Brans et al. 1986)
with the threshold value set equal to the standard deviation of each criterion was
used in the PROMETHEE model. This function was chosen according to the
suggestion of Brans et al. (1986) who showed that the Gaussian function provided
the least discontinuities and guaranteed the most stable results out of the six different
preference functions in PROMETHEE.
Figure 6.3 Combined PROMETHEE II net outranking flows of traffic objects showing
commercial sites as predominant sources of SVOCs and NVOCs build-up
156
In Figure 6.3, all the objects with positive outranking flows are from commercial
and residential sites. These along with the negatively ranked industrial sites suggest
that traffic related build-up of SVOCs and NVOCs mainly occur in commercial and
residential sites. However, the objects with negative outranking flows in Figure 6.3
comprised of all three land uses (e.g., CA, RDS, IS, and CT). According to the
above noted classification of traffic scenarios, most of the negatively ranked objects
fall into the high traffic scenario. To the contrary, the top three objects (CL, CH, and
RR) are from the moderate traffic cluster. Therefore, it is evident from the
PROMETHEE ranking that the moderate traffic scenario with ADT values ranging
from 2300 to 5900 with average congestion of 0.47 would dominate the SVOCs and
NVOCs build-up.
The low traffic scenario (objects: RD and RP with very low positive outranking flow
values) may have some impacts on such build-up through the resuspension and
redistribution of coarse particles as both of them fall into residential land use.
However, the high traffic scenario (objects: CA, CT, IS, and RDS) did not affect the
build-up. Whilst the high traffic scenario had the highest average traffic volume and
congestion, the role of texture depths may also play an important role in the SVOCs
and NVOCs build-up. The average texture depths of the high traffic objects was 0.67
mm which was comparatively lower than the moderate and low traffic objects (0.77
mm). This difference could have led to weaker correlations between high traffic
objects and the different particle size fractions investigated in the study. In order to
facilitate the sensitivity of the findings derived through the PROMTHEE ranking,
the geometric application for interactive aid (GAIA) was performed on the same
data matrices used for PCA and PROMETHEE.
157
6.3.4. GAIA
The GAIA method provided a PCA biplot with a decision axis (pi) for all traffic
scenarios and size fractions. The quality of the decision axis was tested for its
stability by interactively changing the weights of the different variables in the data
matrix for the maximum achievable ‘ φ ’ net ranking values and the optimised GAIA
biplot is shown in Figure 6.4.
Figure 6.4 GAIA biplot for the build-up of SVOCs and NVOCs incorporating all size fractions
as well as all traffic scenarios
The GAIA biplot in Figure 6.4 isolates most of the moderate traffic objects from the
high traffic objects. Additionally, the decision axis (pi) is strongly correlated with
the higher particulate fractions of 75 to 300 µm fractions as well as the moderate
traffic objects on both axes. This suggests that the target organic compounds are
predominantly present in the 75 to 300 µm particulate fractions. The low traffic
objects RD and RP as well as moderate objects IBT are also strongly correlated with
158
the particulate fraction >300 µm, suggesting that the redistribution of particulate
matter occurred in this fraction. The potentially dissolved fraction <1 µm is not
correlated with any of the traffic objects even though the magnitude of its loading
vector is significant. This suggested that the presence of the fine fraction <1 µm did
not contribute to the build-up of SVOCs or NVOCs and only the resuspension and
the replenishment of the finer materials as described earlier are active in this
fraction.
6.4. Conclusions
The build-up of traffic generated semi and non volatile organic compounds under
combined traffic scenarios of low, moderate, and high has been characterised in this
study. The key findings can be summarised as follows:
• The build-up of lighter semi volatile compounds is mainly associated with the
commercial areas whilst non volatile lubricants and motor oil compounds are
associated with the industrial areas. The residential areas do not significantly
contribute to the build-up of such pollutants on urban roads. Congestion in the
commercial areas appears to be the main source of build-up of SVOCs whilst
industrial usage of lubricants and heavier oils may also contribute to the build-up
of NVOCs in industrial areas.
• Moderate traffic scenario with ADT ranging from 2300 to 5900 and average
congestion of 0.47 would predominate SVOCs and NVOCs build-up on urban
roads under combined traffic scenarios. As a practical outcome of this finding, a
moderate traffic scenario in any type of land use can be targeted as a significant
source of such pollutants.
159
• Amongst the different size fractions, the particulate fraction 75 - 300 µm is the
most predominant in associating with the SVOCs and NVOCs build-up.
Particulate fraction >300 µm primarily influences the redistribution of coarser
particle due to vehicular activities. The potential dissolved fraction <1 µm is not
associated with the build-up of SVOCs and NVOCs in any of the land uses
investigated in the study. Therefore, mitigation measures for removal of SVOCs
and NVOCs from build-up should target the 75 - 300 µm particulate fractions.
This research study was undertaken as a part of an Australian Research Council
funded Linkage project (LP0882637). The first author gratefully acknowledges the
postgraduate scholarship awarded by Queensland University of Technology to
conduct his doctoral research. The help and support from Gold Coast City Council,
Queensland Department of Transport and Main Roads as well as Queensland Police
is also gratefully acknowledged.
6.6. References
Andreou,G., Rapsomanikis, S. (2009). Origins of n-alkanes, carbonyl compounds
and molecular biomarkers in atmospheric fine and coarse particles of Athens,
Greece. Science of the Total Environment. 407 (21) 5750-5760.
APHA (2005). Standard methods for the examination of water and wastewater, 21st
Edition, American Public Health Association, Washington DC.
API (1994). Interlaboratory study of three methods for analysing petroleum

hydrocarbons in soils. API publication no. 4599, American Petroleum Institute,
Washington DC.
Ayoko, G. A., Morawska, L., Kokot, S., Gilbert, D. (2004). Application of

multicriteria decision making methods to air quality in the microenvironments of
residential houses in Brisbane, Australia. Environmental Science and Technology,
38(9), 2609-2616.
Ayoko, G. A., Singh, K., Balerea, S. and Kokot, S. (2007). Exploratory multivariate
modeling and prediction of the physico-chemical properties of surface water and
groundwater. Journal of Hydrology, 336(1-2), 115-124.
160
Brandenberger, S., Mohr, M., Grob, K., Neukom, H. P. (2005). Contribution of
unburned lubricating oil and diesel fuel to particulate emission from passenger cars.
Atmospheric Environment, 39 (37), 6985-6994.
Brans, J., Vincke, P. and Mareschal, B. (1986). How to select and how to rank
projects: the PROMETHEE method. European Journal of Operational Research
24(2), 228-238.
Brown, L., Affum, J., Chan, A. (2004). Transport pollution futures for Gold Coast
city 2000, 2011, 2021, based on the Griffith University Transport Pollution
Modelling System (TRAEMS). In Urban Policy Program Research Monographs;
Dodson, J. Ed.; Urban Policy Program, Griffith University, Brisbane.
Brown, R. C., Pierce, R. H., Rice, S. A. (1985). Hydrocarbon contamination in

sediments from urban stormwater runoff. Marine Pollution Bulletin, 16 (6), 236-240.
Charlesworth, S. M., Lees, J. A. (1999). Particulate associated heavy metals in the

urban environment: their transport from source to deposit, Coventry, UK.
Chemosphere, 39(5), 833-848.
Decision (2000). Visual Decision Inc., Getting Started Guide, DecisionLab 2000,
Version 1.01.0388, Copyright© 1998 – 2000, Visual Decision Inc., Montreal,
Quebec, Canada.
Draper, W.M., Dhaliwal, J. S., Perera, S. K., Bauman, F. J. (1996). Determination of

diesel fuel and motor oil in water and wastes by a modified diesel-range organics
total petroleum hydrocarbon method. Journal of AOAC International, 79(2), 508-
519.

EPA (2008). Test methods for evaluating solid waste, physical/chemical methods:
SW-846. 3rd Edition. United States Environmental protection Agency. Methods:
8000, 3510C,8015, 8021, 8260.
FHWA (2005). Federal Highway Administration Technical Advisory Report

Gardiner, L. R., Armstrong, W. (2007). Identifying sensitive receiving environments

at risk from road runoff. Land Transport New Zealand: Research report No 315. pp.
68.
GCCC. (2006). Gold Coast Priority Infrastructure Plan- Derivation of Trunk Road
Infrastructure Network Charges: FinalReport . Veitch Lister Consulting, pp:8.
http://www.goldcoast.qld.gov.au/gcplanningscheme_0509/attachments/planning_sch
eme_documents/part8_infrastructure/reference_documentation.pdf . Accessed 2
September 2010.
Horwitz, W. (1982). Evaluation of analytical methods used for regulation of foods

and drugs. Analytical Chemistry, 54(1), 67A-76A.
161
Huber, W. C. (1986). Deterministic modeling of urban runoff quality. In Urban
runoff pollution. Torno, H. C., Desbordes, M., Marsalek, J. (Eds.), NATO ASI
series, Springer-Verlag, Berlin, New York; pp. 167-242.
Jartun, M., Ottesen, R. T., Steinnes, E. and Volden, T. (2008). Runoff of particle
bound pollutants from urban impervious surfaces studied by analysis of sediments
from stormwater traps. Science of the Total Environment 396(2-3),147-163.
Kayhanian, M., Singh, A., Suverkropp, C., & Borroum, S. (2003). Impact of annual
average daily traffic on highway runoff pollutant concentrations. Journal of
Environmental Engineering, 129(11), 975-990.
Keller, H. R., Massart, D. L. and Brans, J. P. (1991). Multicriteria decision making:

A case study. Chemometrics and Intelligent laboratory Systems, 11(2), 175-189.
Kreider, M. L., Panko, J. M., McAtee, B.,L., Sweet, L. I., Finley, B. L. (2010).
Physical and chemical characterization of tire-related particles: Comparison of
particles generated using different methodologies. Science of the Total Environment,
408 (3), 652-659.
Kudchadker, A. P., Zwolinski, B. J. (1966). Vapour pressures and boiling points of

normal alkanes, C21 to C100. Journal of Chemical and Engineering Data, 11(2),
253-255.
Legret, M., Colandini, V. (1999). Effects of a porous pavement with reservoir

structure on runoff water: Water quality and fate of heavy metals. Water Science
Technology, 39(2), 111-117.
Mahbub, P., Ayoko, G. A., Goonetilleke, A., Egodawatta, P., Kokot, S. (2010a).
from urban roads. Environmental Science and Technology, 44(23), 8904-8910.
Mahbub, P., Ayoko, G., Egodawatta, P., Yigitcanlar, T., Goonetilleke, A. (2010b).
Development: Urban management, Engineering and Design. Yigitcanlar, T., (Ed.).
Engineering Science Reference, New York, pp. 147-167.
Main Roads (2009). Austraods Test Manual, Method AGPT/T250,

http://www.austroads.com.au/pavement/testmethods.html, Accessed 05 January
2009.
Malvern (1994). Mastersizer S Long Bed Version 2.19 © Malvern Instruments Ltd.
1992-1994.
Massart, D. L., Vandeginste, B. G. M., Buydens, L. M. C., De Jong, S., Lewi, P. J.

Smeyers-Verbeke, J., 1997. Handbook of Chemometrics and Qualimetrics Part A,
Elsevier: 771-804.
Morrison, R.T., Boyd, R. N. (1992). Organic Chemistry 6th edition. Prentice-Hall

Inc. pp.94-97.
162
Neeft, J. P. A., Makkee, M., Moulijn, J. A. (1996). Diesel particulate emission
control. Fuel Processing Technology, 47 (1), 1-69.
Ning, Z., Cheung, C. S., Lu, Y., Liu, M. A., Hung, W. T. (2005). Experimental and
numerical study of the dispersion of motor vehicle pollutants under idle condition.
Atmospheric Environment, 39(40), 7880-7893.
Ogden, K. W., Taylor, S. Y. (1999). Traffic Engineering and Management. Institute

of Transport Studies, Monash University, Australia. pp. 592- 594.
Patra, A., Colvile, R., Arnold, S., Bowen, E., Shallcross, D., Martin, D., Price, C.,
Tate, J., Apsimon, H., Robbins, A. (2008). On street observations of particulate
matter movement and dispersion due to traffic on an urban road. Atmospheric
Environment, 42(17), 3911-3926.
Pitt, R., Field, R., Lalor, M., Brown, M. (1995). Urban stormwater toxic pollutants:
assessment, sources, and treatability. Water Environment Research, 67 (3), 260-275.
Rogge, W.F., Hildemann, L.M., Mazurek, M.A., Cass, G.R., Simoneit, B.R.T.
(1993). Sources of fine organic aerosol. 3. Road dust, tire debris, and organometallic
brake lining dust: roads as sources and sinks. Environmental Science and
Technology, 27 (9), 1892-1904.
Sartor, J. D., Boyd, G. B., Agardy, F. J. (1974). Water pollution aspects of street
surface contaminants. Journal Water Pollution Control Federation, 46(3), 458-467.
Sirius (2008). SIRIUS version 7.1© copyright 1987-2008, Pattern Recognition

Systems AS, http://www.prs.no, Help Topics.
Smit, R., Brown, A. L., Chan, Y. C. (2008). Do air pollution emissions and fuel
consumption models for roadways include the effects of congestion in the roadway
traffic flow? Environmental Modelling & Software, 23 (10-11), 1262-1270.
163
164
CHAPTER 7 CHARACTERISATION OF
VOLATILE ORGANIC COMPOUNDS WASH-OFF
FROM URBAN ROADS
Manuscript Title
Effects of Climate Change on the Wash-off of Volatile Organic Compounds from
Urban Roads
Parvez Mahbub1*, Ashantha Goonetilleke1, Godwin A. Ayoko2, Prasanna

Egodawatta1
1
2
s.mahbub@qut.edu.au; a.goonetilleke@qut.edu.au; g.ayoko@qut.edu.au;
*
Accepted for Publication (2011): Science of the Total Environment, DOI:

10.1016/j.scitotenv.2011.06.032. Impact Factor 3.4; ERA ranking A.
Abstract
The predicted changes in rainfall characteristics due to climate change could
adversely affect stormwater quality in highly urbanised coastal areas throughout the
world. This in turn will exert a significant influence on the discharge of pollutants to
estuarine and marine waters. Hence, an in-depth analysis of the effects of such
changes on the wash-off of volatile organic compounds (VOCs) from urban roads in
the Gold Coast region in Australia was undertaken. The rainfall characteristics were
simulated using a rainfall simulator. Principal Component Analysis (PCA) and
Multicriteria Decision tools such as PROMETHEE and GAIA were employed to
understand the VOC wash-off under climate change. It was found that low, low to
moderate and high rain events due to climate change will affect the wash-off of
165
toluene, ethylbenzene, meta-xylene, para-xylene and ortho-xylene from urban roads
in Gold Coast. Total organic carbon (TOC) was identified as predominant carrier of
toluene, meta-xylene and para-xylene in <1µm to 150µm fractions and for
ethylbenzene in 150µm to >300µm fractions under such dominant rain events due to
climate change. However, ortho-xylene did not show such affinity towards either
TOC or TSS (total suspended solids) under the simulated climatic conditions.
KeywordsVolatile organic compounds, rainfall characteristics, climate change,

total organic carbon, wash-off

preparation of figures and tables.



This chapter is an exact copy of the submitted manuscript of the journal paper.

Development
the wash-off of traffic generated pollutants and accordingly, providing guidance to
The methodologies to accomplish such objectives were formulated in this journal
paper using principal component analysis (PCA), preference ranking organisation
method for enrichment evaluation (PROMETHEE) and geometrical analysis for
166
interactive aid (GAIA) with a view to characterise the wash-off volatile organic
compounds (VOCs) under the dynamic changes of rainfall characteristics induced
by climate change as described in Chapter 4. This journal paper has contributed to
the overall outcome of this research project by establishing the predominant rain
events that affect the wash-off of such pollutants. The multicriteria decision tools
used in this study has identified the surrogate parameters as well as dominant
particle size fractions to which the VOCs remain attached during their wash-off.
These findings strengthened the knowledge in adopting stormwater quality
mitigation strategies in relation to the removal of VOCs during wash-off. Figure 3.3
in section 3.7, Chapter 3 provides a schematic flow diagram of this research study
where the publication of this journal paper was highlighted as an integral process
that contributes to the overall research outcome.
167
7.1. Introduction
Rainfall characteristics such as intensity, duration and frequency are predicted to
change throughout the world due to climate change. In Australia, longer periods of
dry weather with fewer, but more intense storms are forecast (CSIRO 2007). This is
compounded by the fact that population growth rates in coastal regions of Australia
are consistently high (ABS 2010). In recent years, many coastal local government
authorities in Australia have experienced growth rates in the range of 50% to 60%
higher than the national average.
This rapid population growth coupled with changes in rainfall characteristics will
significantly impact on coastal communities. In this respect, the effects of population
growth and land use on ecosystem health, particularly on coastal water quality, have
drawn much attention both locally (Burnley and Murphy 2004; Gurran et al 2006) as
well as globally (Kimura 1988; Decembrini et al. 1995; Benson et al. 2008).
However, current knowledge on the dynamic effects of climate change on the water
quality in highly urbanised coastal areas is very limited.
Abbs et al. (2007) predicted high risk of floods spanning large areas of developed
flood plains in the Gold Coast region due to extreme rainfall events resulting from
climate change. A report by the European Environment Agency, EEA (2003)
suggested some possible effects of climate change on water quality including
increased pollutant discharges (during floods), rising temperature and oxygen
depletion (during droughts) and changes in aquatic ecology. This in turn will exert a
significant influence on the discharge of pollutants to receiving estuarine and marine
waters. CSIRO (2007) also noted some possible effects of climate change on water
168
quality in Australia in terms of changes to micro fauna and flora as well as the
processes that transport pollutants into streams and aquifers. Most of these studies
are generalised without particularly focussing on the coastal regions.
Vehicular traffic is one of the most significant sources of toxic and carcinogenic
pollutants deposited on urban roads. These pollutants are washed-off from road
surfaces during rainfall-runoff events and eventually transported to water bodies.
The major pollutants that are generated by vehicular traffic include polycyclic
aromatic hydrocarbons (PAHs), total petroleum hydrocarbons (TPHs), volatile
organics and heavy metals (Hoffman et al. 1982, 1984; Sansalone & Buchberger
1997a). Amongst these different pollutants, benzene, toluene, ethylbenzene and
xylene commonly referred to as BTEX are a special family of pollutants that are
volatile organic compounds or VOCs with boiling points ranging from 50-260 ºC
(Ayoko 2004) and primarily generated in the urban environment from vehicle
exhaust, brakes, engine oils, evaporative emissions, indoor air pollution activities
and equipments. These compounds are listed as carcinogenic or possibly
carcinogenic to humans (IARC 2009). Herngren et al. (2005a) studied the wash-off
relationships of traffic generated pollutants with solids and organic carbon.
However, the wash-off relationships between VOCs and the physico-chemical
parameters of runoff have not featured prominently in the literature.
The Gold Coast region of Australia is a popular tourist destination and subject to
high population growth. From the period of 2001 to 2007 the region recorded a 3.7%
annual population growth, which is higher than the current Australian population
growth and a concomitant increase in vehicle usage (ABS 2010). An in-depth
169
understanding of traffic generated pollutant wash-off processes in Gold Coast will
help in the development of wash-off models influenced by climate change. This
paper discusses a research study undertaken to investigate the wash-off processes of
toluene, ethylbenzene, meta-xylene, para-xylene and ortho-xylene under predicted
changes in rainfall characteristics due to climate change in the Gold Coast region.
The wash-off relationships of these VOCs with total suspended solids (TSS) and
total organic carbon (TOC) was also investigated to establish surrogate parameters
during VOC wash-off.

7.2.1. Site selection
Four road sites in three suburbs in the Gold Coast were selected as the wash-off
study sites. Figure 7.1 shows the study sites and their relative locations with respect
to their distance from the Meteorological Gauging Station 40166. This station has
recorded daily rainfall data since 1894 and is located at 27.90° S and 153.31° E at an
elevation of 6 m. The shoreline is at a straight line distance of 12.76 km from the
furthest site, namely Billinghurst Crescent.
Figure 7.1 The relative locations of the four study sites for the VOC wash-off sample collection
170
In Figure 7.1, the sites named as Billinghurst Crescent and Discovery Drive are
situated in residential areas, Shipper Drive is situated in an industrial area, whilst
Lindfield Road is situated in a commercial area. The selected sites also represent the
transport infrastructure developed in the Gold Coast region in the past decade.
Egodawatta (2007) found that the pollutant build-up on road surfaces asymptote to
an almost constant value after a minimum antecendent dry period of seven days.
Hence, seven dry days were allowed at each site before collecting the wash-off
samples for this study. A total of twenty two rain events were simulated at the four
sites. The distributions of these events per site are discussed in the subsequent
sections.
7.2.2. Wash-off sample collection

The research study used a rainfall simulator (Herngren et al. 2005b) to replicate the
design rainfall events on the road surfaces and a commercially available vacuum
cleaner was used to collect the wash-off samples. The rainfall simulator was based
on the design of simulators used in agricultural research as described by Floyd
(1981) and Loch et al. (2001). It consisted of an A-frame structure made of
aluminium tubing of 40-mm diameter. Three Veejet 80100 nozzles, spaced 1 m
apart are mounted on a stainless steel boom at a height of 2.4 m. This is the
prescribed height for creating terminal velocities similar to natural rainfall for all
drop sizes (Herngren 2005b). Further details on the design of the simulator can be
found in Herngren et al. (2005b) and Loch et al. (2001).
The plot area for rainfall simulation was separated by a frame connected to a
collecting trough. The runoff water in the collecting trough was vacuumed
171
continuously into 25 L plastic containers. The water pressure through the nozzles
was maintained at 41 kPa which has been found to be the most appropriate pressure
to create drop size distribution near natural rainfall for the given height of the
nozzles above the road surface (Bubenzer 1979).
The runoff samples were transported to the laboratory for sub-sampling immediately
after collection. As pollutant concentrations can vary by orders of magnitude during
a runoff event, the flow weighted average or event mean concentration samples
(EMC) were found to be appropriate for evaluating the impacts of stormwater runoff
on receiving waters (Sansalone & Buchberger 1997b). In this study, 500 mL EMC
samples were prepared in the laboratory using a churn splitter. The required volumes
at a particular duration constituting an EMC sample was calculated from the
percentages of the total runoff at that duration and mixed together to get the 500 mL
EMC sample for an event.
The particle size distributions of the suspended solids in the subsamples were
determined using a Malvern Mastersizer S Particle Size Analyser capable of
analysing particles between 0.05 to 900 µm diameter. Based on the particle size
distribution, the total particulate analytes were fractioned into four size ranges,
namely, >300 µm, 150-300 µm, 75-150 µm, 1-75 µm using wet sieving. The filtrate
passing through a 1 µm membrane filter was considered as the total dissolved
fraction. In each case, 500 mL homogeneous sub-samples were prepared using
deionised water, collected in 500 mL amber glass bottles with a PTFE seal,
preserved at 4°C in the laboratory and analysed within 14 days of collection.
172
7.2.3. Simulation of rainfall incorporating climate change impacts
The rainfall simulation was based on the studies of Abbs et al. (2007) who used a
dynamic downscaling technique incorporating the CSIRO CC-MK3 and CSIRO
RAMS climate models to generate 2030 and 2070 average fractional change for
extreme rainfall intensities at 2, 24 and 72 hour durations for the Gold Coast area.
The published results from that study for the Gold Coast region are summarised in
Table 7.1 below.
Table 7.1 Average percentage change in extreme rainfall intensity for the Gold Coast region:
adapted from Abbs et al. (2007)
2030 2070
Range between Range between
Duration Region % change of % change of
10th and 90th 10th and 90th
mean intensity mean intensity
percentile percentile
All Gold Coast 53 26-89 48 4-91
2 Coastal 50 33-65 35 6-72
Mountains 56 22-96 65 27-97
All Gold Coast 17 8-29 16 5-28
24 Coastal 15 8-23 13 0-26
Mountains 19 8-32 19 9-30
All Gold Coast 8 0-17 14 4-23
72 Coastal 6 -2-13 10 -4-23
Mountains 10 0-20 17 9-24
It is clear from Table 7.1 that higher changes in the rainfall intensities are projected
for shorter duration events in the 2030 and 2070 predictions for the Gold Coast
region. Several climate change studies (CSIRO 2007; IPCC 2007) have predicted
that the probability of occurrence of shorter duration (<2 hr) events with large
change in precipitation intensities is very high.
173
Mahbub et al. (2010a) used the outcome from the Abbs et al. (2007) study to
propose three scenarios of changes in rainfall characteristics for the Gold Coast
region. The current study used all of these scenarios to simulate the 2030 rainfall in
Gold Coast. Table 7.2 shows the rainfall simulation plan according to the study by
Mahbub et al. (2010a). For simplicity and due to Gold Coast City Council
restrictions on lane closure times, the simulation events were distributed in the four
study sites as per their intensities which ranged from 24.6-39.3, 58.3-63, 75-77 and
119-125 mm/hr.
Table 7.2 Future simulation events based on the normal daily rainfall intensity in the Gold
Coast region for 2030: adapted from Mahbub et al. (2010a)
Current simulation events for Gold Coast Future simulation events for Gold Coast
region region for 2030
Scenario Simulation Simulation

ARI Duration Intensity ARI Duration Intensity
Event Event
(year) (min) (mm/hr) (year) (min) (mm/hr)
Number Number
Shorter 1 60 39.3 1 1 25 63 19
Duration, 2 90 39.3 3 2 42.5 61.2 20
with 5 133 39.3 5 5 69 59.2 21
higher 10 160 39.3 6 10 85 58.3 22
intensity 100 105 75 18 100 49 115 13
while ARI
constant - - - 1 65 37.39 2
Shorter 100 45 125 12 1 5 125 7
ARI,
shorter
duration
- - - 1 120 24.6 4
while
intensity
constant
Shorter 10 52.5 77 14 5 16 125 10
ARI, with 20 67.5 77 15 10 21 122 11
Higher 50 86.7 77 16 2 10.5 120 9
Intensity
while
Duration 100 101.25 77 17 1 5.75 119 8
getting
shorter
7.2.4. Sample testing

The pH and electrical conductivity (EC) of the wash-off samples were determined
initially. Subsequently, the samples were subjected to testing for VOCs namely,
toluene, ethylbenzene, ortho-xylene, meta-xylene and para-xylene. These tests were
undertaken according to the USEPA Method 5035 and 8260B (EPA 2008) using
174
purge and trap extraction followed by Gas Chromatography/Mass Spectrometry
(GC/MS). Ten different calibration standards (Chemservice® THM501 – 1RPM) at
1, 2, 5, 10, 20, 50, 100, 150, 200 and 250 µg/L concentrations were prepared for
each target analyte. Volatile internal standards (Chemservice® IS-8260ARPM)
consisting of flourobenzene, chlorobenzene-d5 and 1, 4- dichlorobenzene-d4 were
added to each sample and standards at 50 µg/L concentration. Field blanks were
used during each field trip and all results were blank corrected.
Three quality control standards at 10, 50 and 100 µg/L concentrations were prepared
with another quality control standard at a concentration of 20 µg/L. The percentage
R = (C1 − C 2) / C1 × 100 -------------------------------------------------------------------- (7.1)
final spike concentration.
The percentage recoveries were found to be within 90%-95%. The limit of detection
was established as 0.01 µg/L for the test method.
The total and dissolved organic carbon (TOC and DOC) concentrations and total and
dissolved suspended solids (TSS and TDS) concentrations were also determined
according to methods 2540C and 2540D in APHA (2005). The particle size
percentages (PSP) were calculated using the Particle Size Analyser software
(Malvern 1994).
175
7.2.5. Data analysis
The data matrices consisted of twenty two objects and twelve variables for each of
the five size fractions. The twenty two objects were numerically defined starting
from 1. After initial observation of the probability distribution of the objects,
normalisation of all objects was undertaken so that each object had same relative or
absolute size. As the variables were measured in different units, standardisation of
each variable was also undertaken as a pre-treatment measure so that each variable
could be treated with equal importance at the beginning of the data analysis.
The data analysis was designed to investigate the impact on wash-off of volatile
organics from urban roads due to changes in rainfall characteristics. Hence, the
twenty two rainfall simulation events described above were taken as objects. The
attributes of rainfall events such as intensity, duration and average recurrence
interval (ARI) along with the target VOCs, total suspended solids (TSS), total
organic carbon (TOC), particle size percentages (PSP), electrical conductivity (EC)
and pH were considered as variables in the data analysis. Multivariate chemometrics
methods such as principal component analysis (PCA), preference ranking
organisation method for enrichment evaluation (PROMETHEE) and geometrical
analysis for interactive aid (GAIA) were employed for the data analysis.
Factor /component extraction processes such as PCA (Jartun et al. 2008) and
multicriteria decision making processes such as PROMETHEE and GAIA
(Herngren et al. 2005a; Ayoko et al. 2007) have been used widely to characterise the
incorporation of pollutants in stormwater runoff from urban roads, to correlate
suspended solids with heavy metals in runoff and to model the pollution impacts on
physico-chemical properties of surface water and groundwater. In this study, the
176
combined use of these three methods applied to investigate stormwater quality from
road runoff were expected to provide the generic patterns of behaviour of
stormwater pollutants underpinned by specific decisions based on different criteria.
Detailed discussion of these techniques can be found in the referenced literature
(e.g., Keller et al. 1991; Mareschal & Brans 1988).
PCA
PCA is a pattern recognition technique employed to understand the correlations
contains the second largest variance and so on. Though PCA produces the same
amount of PCs as the original variables, the first few contain most of the variance.
Therefore, the first few PCs are often selected for interpretation. This reduces the
number of variables without losing useful information contained in the original data
set. The number of PCs to be used for interpretation is typically selected using the
Scree Plot method described by Jackson (1993).
The application of PCA to a data matrix generates a loading for each variable and a
score for each object on the principal components. Consequently, the data can be
presented diagrammatically by plotting the loading of each variable in the form of a
vector and the score of each object in the form of a data point. This type of plot is
referred to as a ‘Biplot’. The angle between variable vectors is the indicator of the
degree of correlation. Clustered data points in a biplot indicate objects with similar
characteristics. Detailed descriptions of PCA can be found elsewhere (Jackson
177
1993). In this study, SIRIUS2008 software (Sirius 2008) was used to perform the
exploratory PCA procedures.
PROMETHEE
PROMETHEE is designed to rank a number of objects in terms of the data criteria
(Brans et al. 1986; Keller et al. 1991). The ranking for each variable or criterion is
performed by a user specified preference function. The study used DecisionLab
2000 software (Decision 2000) to perform PROMETHEE analysis. The steps
involved in the application of PROMETHEE are as follows:
6. For each variable all objects or actions in the data matrix are compared pairwise,
in all possible combinations by subtraction, and thus a difference, d , matrix is
generated;
7. A preference function P ( a , b ) was chosen for each variable. It describes how
much outcome a is preferred to outcome b . In the DecisionLab2000 software,
one of six such functions along with corresponding threshold values may be
chosen by the user. It was also necessary to specify whether top-down
(maximized) or bottom-up (minimized) ranking of objects for each variable was
preferred. In addition, each variable can be weighted in importance, but in
general, most modelling initially uses the default weighting of 1. In this work,
the affinity of the target VOCs with the TSS and TOC was studied and hence,
variables were maximised. The Gaussian preference function ( P ) described
below was applied:
P ( a , b ) = 0 for d ≤ 0 --------------------------------------------------------(7.2)
2
/ 2σ 2
P (a, b) = 1 − e− d for d ≥ 0 --------------------------------------------(7.3)
178
where d is the difference for each pairwise comparison and σ is the threshold,
which was set at the value of the standard deviation of each criterion (Brans et
al. 1986). This was the smallest deviation in terms of the target VOCs affinity
towards TSS or TOC that would be considered as decisive by the
DecisionLab2000 software. The choice of the Gaussian function was based on
the suggestion by Brans et al. (1986) as they showed that the function provides
the least discontinuities and guaranteed the most stable results out of the six
different preference functions. All the variables were weighted equally at the
beginning of the procedure.
8. The products of each preference function P ( a , b ) and the weights for the
corresponding variables were summed up to generate a preference index table.
These indices corresponded to the pairwise comparison of the objects or actions.
9. To compare each action one-to-one with the others systematically, preference
flows were computed. The DecisionLab2000 software supported three types of
preference flows namely, positive outranking flow ( φ + ), negative outranking
flow ( φ − ) and the net outranking flow ( φ ). The positive outranking flow ( φ + ) is
associated with the degree of preference with which one action is preferred on
average over the other actions and the negative outranking flow ( φ − ) is
associated with the degree of preference with which the other actions are
preferred on average to that action. The higher φ + and the lower φ − , the more
preferred is the action. This procedure resulted in a partial pre-order, called
PROMETHEE I ranking.
10. There were certain circumstances as described by Keller et al. (1991) when two
actions a and b may not be comparable. The net outranking flow ( φ ), which is
179
the algebraic difference between the positive and negative outranking flows, is
then calculated. This procedure, known as PROMETHEE II, was used in this
study to eliminate any incomparability between actions or objects.
GAIA
GAIA facilitates a sensitivity analysis technique for multicriteria decision methods
such as PROMETHEE (Keller et al. 1991; Mareschal & Brans 1988). GAIA
provides a graphical view of the actions and variables for net outranking flow ( φ ) in
the form of a PCA biplot by decomposing the φ values from PROMETHEE II into
that it also produces a decision axis that takes into account the weights associated
with the variables. This helps the decision-maker with an enriched understanding of
the problem in terms of the detection of clusters of actions, conflicts in variables,
inability to compare between actions and so on (Mareschal & Brans 1988). This
study used DecisionLab 2000 software (Decision 2000) to perform PROMETHEE
and GAIA analysis.

In the five wash-off data matrices (four for different particulate size fractions and
one for the dissolved fraction described above), the twenty two rainfall objects were
considered having object attributes such as intensity, frequency and duration that
were responsible for generating the wash-off of volatile organics, TSS and TOC
from urban roads. Electrical conductivity (EC), pH and the particle size percentages
(PSP) were also considered as parameters that influence the wash-off of volatile
organics. The concentration ranges were 0.03 to 0.13 ppb for toluene, 0.01 to 0.03
ppb for ethylbenzene, 0.02 to 0.06 ppb for meta and para-xylene and 0.01 to 0.03
180
ppb for ortho-xylene. The pH ranged between 6.99 to 7.2 and EC ranged between
22.8 to 63.4 microsiemens/cm. In order to get a better understanding of the generic
data patterns, PCA was performed on each of the pre-treated wash-off data matrices.
Figure 7.2 shows the PCA biplots of the five wash-off size fractions.
181
5
4 150-300µm
8
8 4
>300µm
3 10 7
9 9
3
TSS
Intensity
PC 2 (23.8%)
2 11 EC Intensity
7 10
PSP
2 TSS
EC
TOC
4
PC 2 (20.6%)
1 14
12 3 1 1
TOC 19
13 pH
1
0 13
12 11
16 5 2 19 pH 0 4 2
15 3
6 22 20 PSP
OX 17 16
-1 ARI
15 -1 18 20 OX
ARI 6 21 ETB
ETB
22 TOL MPX
1718 21
MPX
-2 -2 5
Duration
14 TOL Duration group A
group A
-3 -3
-5 0 5 -5 0 5
PC 1 (43.7%) PC 1 (45.2%)
(a) (b)
5
4 8 1-75µm
4 8
75-150µm
7
3 79
9
3
10 10 Intensity
2 19 TOC
11 Intensity 11
TSS 2
PC 2 (27.9%)
TSS
TOC PSP
PC 2 (25.5%)
1 EC EC 19
12 1
14 15 20
13 15
16 pH
0 14 pH
ARI ETB 0
2 13
OX
TOL MPX 12 16 1 TOL
1 3 2 OX
17 18 3 group A MPX
-1 PSP 17ARI 18 22
20 -1 21 ETB
22
6 5 4 21 group A
5 4
-2 Duration
-2
6
Duration
-3 -3
-5 0 5 -5 0 5
PC 1 (47.6%) PC 1 (43.4%)
(c) (d)
4
<1µm
9
3 78
Intensity DOC
14
10
PSP
2 11 19
1
15
EC 20
PC 2 (27.4%)
pH
TDS
0 12
13 16 TOL
1 21 MPX
ETB
ARI 2 OX
-1
17
18 22 group A
3
-2 4
5
6
Duration
-3
-4
-5 0 5
PC 1 (41.0%)
(e)
Figure 7.2 PCA biplots for (a) >300 µm; (b) 150-300 µm; (c) 75-150 µm; (d) 1-75 µm; and (e) <1
µm size fractions for the wash-off of toluene (TOL), ethylbenzene (ETB), meta and para xylene
(MPX) and ortho xylene (OX)
182
In Figure 7.2, it was identified that the target volatile organics formed a group
(group A) in wash-off from urban roads for all the five size fractions. In Figure
7.2(a), 7.2(b), 7.2(c) and 7.2(d), the correlations of the group (A) variables with
TOC are stronger than with TSS. However, in the dissolved form in Figure 7.2(e),
the correlation of the group (A) variables with TDS is stronger than with DOC. This
suggests that organic carbon acts as the predominant medium for the target volatile
organics wash-off in the particulate fraction from >300 µm to 1 µm, but in the
dissolved fraction of <1 µm, TDS is the predominant medium of volatile organics.
In all of the fractions in Figure 7.2, pH has much stronger correlations with the
group (A) organics than EC. This suggests that VOCs are washed-off as non-ionic
compounds for all fractions. Amongst the three rainfall attributes, average
recurrence interval (ARI) has the strongest correlation with the group (A) organics
than duration and intensity. However, ARI always has negative correlations with the
group (A) organics for all fractions whilst PSP has positive correlations with them in
the four particulate fractions. This suggested that for the four particulate size
fractions, low ARI events were able to significantly wash-off the corresponding
particle fractions of the group (A) organics from urban roads. The influence of
duration and intensity on the particulate wash-off of group (A) organics was not
significant in Figures 7.2(a) through to 7.2(d) as they had low correlation with the
group (A) organics. In Figure 7.2(e), the PSP as well as the intensity and duration
are nearly orthogonal to the group (A) organics suggesting that in the wash-off of
the dissolved fraction of VOCs, these parameters are not significant.
183
The scores for the events 19, 20 and 21 were significantly positive along with
positive loadings from the group (A) organics on PC1 for all five size fractions. This
suggested that the occurrence of VOC wash-off during these events were
predominant than in the case of the other events. The investigation of these events in
Table 7.2 revealed that intensity ranging from 59 to 63 mm/hr, duration ranging
from 25 to 69 minutes and ARI from 1 to 5 years were the attributes of these
predominant events in the wash-off of VOCs. However, the PCA did not incorporate
all the five size fractions together. Hence, this PCA outcome is suggestive of
important rain events during VOC wash-off. However, a detailed investigation of the
predominant rain event clusters under changed climatic conditions is beyond the
scope of PCA. Mahbub et al. (2010b) suggested several rain event clusters under
changed rainfall characteristics due to climate change. This study has incorporated
those clusters and used multicriteria decision making analysis to investigate their
effect on VOC wash-off.
7.3.2. PROMETHEE AND GAIA

The PROMETHEE rankings were performed in order to establish the predominant
rain events as well as the predominant surrogate carrier (either TSS or TOC) in
terms of VOC wash-off. Subsequently GAIA was used to perform sensitivity
analysis of the PROMETHEE outcome. The twenty two rain events were classified
into five clusters namely, low, low to moderate, moderate, high and extreme events.
This classification was based on a study by Mahbub et al. (2010b). The events with
intensity <40 mm/hr with relatively low ARI were classified as low events; those
having intensity between 50 to 100 mm/hr but with relatively higher ARIs of up to
50 years were classified as moderate events; events having intensities >100 mm/hr
with very high frequency were classified as high events whilst events with similar
184
intensities to moderate and high with extremely rare occurrence (ARI ≥ 100 years)
were classified as extreme events. Events which manifested the attributes of both
low and moderate events were classified as low to moderate events.
Consequently, events 1, 2, 3, 4, 5, 6 were classified as low events; events 19, 20, 21
were classified as low to moderate events; events 14, 15, 16, 22 were classified as
moderate events; 7, 8, 9, 10, 11 were classified as high events; 12, 13, 17, 18 were
classified as extreme events. The study also defined the affinity of the target VOCs
towards TSS and TOC as ‘µg of VOC/ mg of TSS or TOC’. Accordingly, two
PROMETHEE data matrices were constructed.The first one consisted of the
classified rain events described above as actions and all the variables described in
the PCA as criteria while taking all five size fractions as different scenarios. Figure
7.3 presents the PROMETHEE outranking flows for VOC wash-off.
(a)
(b)
Figure 7.3 PROMETHEE partial ranking (a) and complete ranking (b) for the twenty two
different rainfall events in terms of VOC wash-off
185
In Figure 7.3(a), events 1, 2, 3, 8 and 19 were not comparable in the partial
outranking flows of PROMETHEE I and as such are shown at the beginning. Event
7 was outranked by events 8 and 19 whilst event 4 was outranked by events 1 and 2.
In Figure 7.3(b), the complete outranking flow of PROMETHEE II is presented.
This confirms that the top ten rain events in terms of VOC wash-off are 8, 19, 1, 2,
3, 4, 9, 7, 10 and 20. Interestingly, these events are composed of low, low to
moderate and high rain events. Hence, the PROMETHEE rankings show that the
attributes of low and high rain events would affect VOC wash-off from urban roads
more predominantly than moderate and extreme events. The quality of these
decisions was investigated using the GAIA biplot as shown in Figure 7.4.
Figure 7.4 GAIA biplot of the rain events under the combined scenario of five size fractions
In Figure 7.4, the decision axis (pi) is located within the vicinity of the low, low to
moderate and high rain objects. This suggests that low, low to moderate and high
rain events are the predominant events in the combined rainfall scenario due to
186
climate change. This decision was tested for its stability by changing the weights of
the criteria interactively for the maximum achievable net outranking flows of the
actions as shown in Figure 7.3(b). It is also evident from the loadings of the five size
fractions in Figure 7.4 that these fractions were mainly present in the low, low to
moderate and high rain events.
The second PROMETHEE data matrix was constructed by taking the affinity of
VOCs towards TSS or TOC (expressed as µg of VOC/ mg of TSS or TOC) as the
actions, whilst the previously described five rain events cluster as criteria for all five
size fractions. The PROMETHEE outranking flows are shown in Figure 7.5.
(a)
(b)
Figure 7.5 PROMETHEE partial ranking (a) and complete ranking (b) for the VOC affinity
matrix during wash-off from urban roads
187
In Figure 7.5(a), the partial outranking flows of the VOC affinity towards TSS or
TOC revealed that the affinity of toluene and meta and para-xylene towards TOC
were stronger than towards TSS. The affinity of ethylbenzene towards TOC and
meta and para-xylene towards TSS were not comparable. However, the complete
outranking flows in Figure 7.5(b) revealed that toluene, meta and para-xylene as
well as ethylbenzene are more strongly associated with TOC than TSS. The affinity
of ortho-xylene towards either TSS or TOC were very weak.
GAIA biplots for the five rain events clusters as well as the five different size
fractions were analysed for the affinity of VOCs towards TSS and TOC. Figure 7.6
presents the GAIA outcomes.
(a) (b)
Figure 7.6 GAIA biplots for five pre-defined rain events clusters (a) and five different size
fractions (b) in terms of VOC’s affinity towards TSS and TOC
In Figure 7.6(a), five rain clusters (low, high, low to moderate, moderate and
extreme) are presented as the criteria whilst the VOC affinity is presented as the
actions with the size fractions combined. The decision axis is very strongly inclined
188
towards the affinity of toluene towards TOC indicating that toluene might form the
strongest bond with organic carbon during wash-off. Moreover, the affinity of
toluene, meta and para-xylene as well as ethylbenzene towards TOC are mainly
present in low, low to moderate, or high events in Figure 7.6(a). This finding is quite
significant in the sense that initially in Figure 7.4, these rain events were found to be
the predominant clusters for the transport of the five different size fractions during
VOC wash-off. Here in Figure 7.6(a), these are also the predominant rain event
clusters confirming the affinity of VOCs towards TOC except for ortho-xylene. The
wash-off of ortho-xylene was found to be independent of any of the rain event
clusters.
In Figure 7.6(b), the affinity of VOCs were analysed for five different size fractions.
The correlations of the affinity of toluene, meta and para-xylene towards TOC with
the loading vectors for <1 µm, 1-75 µm and 75-150 µm were stronger than those for
150-300 µm and >300 µm on the horizontal axis. This suggested the fact that the
two finer fractions 1-75 µm and 75-150 µm as well as the dissolved fraction <1 µm
represent the affinity of toluene, meta and para-xylene towards TOC during wash-
off. The affinity of ethylbenzene towards TOC was mainly present in 150-300 µm
and >300 µm size fractions in Figure 7.6(b). Ortho-xylene’s affinity towards either
TSS or TOC could not be observed in any of the size fractions. In both cases of
Figure 7.6(a) and 7.6(b), the decisions were tested for their stability by changing the
weights of the criteria interactively for the maximum achievable net outranking
flows of the actions as shown in Figure 7.5(b).
189
This study has characterised the wash-off of the toluene, ethylbenzene, meta and
para-xylene and ortho-xylene under simulated rainfall characteristics on urban roads
due to climate change, in the Gold Coast region. The following findings were drawn
from this investigation:
• Low, low to moderate and high rain events due to climate change would be
the predominant events in terms of toluene, ethylbenzene, meta and para-
xylene and ortho-xylene wash-off from urban roads in the Gold Coast
region. Highly urbanised coastal areas with similar effects due to climate
change on rainfall characteristics are expected to demonstrate similar wash-
off phenomena.
• Toluene, ethylbenzene and meta and para-xylene show a stronger affinity
towards TOC than TSS during their wash-off under the changed climatic
conditions for the particulate fraction 1µm to > 300 µm. Hence, TOC could
be regarded as the predominant carrier of VOCs in wash-off under changed
climatic conditions in these particulate fractions. The removal of these
pollutants from stormwater runoff could be achieved by specifically
targeting the removal of TOC in the corresponding particulate fractions.
• Ortho-xylene did not show any affinity towards either TSS or TOC during
wash-off. Hence, no surrogate carrier could be established for the wash-off
of ortho-xylene. The removal of ortho-xylene from stormwater runoff may
require independent measures from that of toluene, ethylbenzene and meta
and para-xylene for both particulate and dissolved fractions. Due to very low
concentration of ortho-xylene at ppb level in the stormwater runoff, this
study proposes only periodic monitoring scheme at the inlet to storage
reservoirs at this stage. However, further studies on mitigation measures
190
need to be undertaken if the ortho-xylene concentration in the stormwater
runoff exceeds the safe concentration limit.
• Under the combined size fraction scenario, the affinity of toluene and meta
and para-xylene towards TOC were more predominant in the two finer
fractions of 75-150 µm and 1-75 µm as well as the dissolved fraction of <1
µm and the affinity of ethylbenzene towards TOC was mainly present in
150-300 µm and >300 µm. Hence, the effectiveness of the removal of these
pollutants could be enhanced by targeting only the removal of the
corresponding fractions. This approach enables a priority based removal of
toluene, ethylbenzene and meta and para-xylene from stormwater runoff
depending on the existing concentration levels.
• Under combined size fraction scenario, VOCs affinity towards TOC was
predominantly observed in low, low to moderate and high rain events
clusters. The importance of this finding could be realised by targeting the
removal of TOC in <1µm to 300µm fraction from the runoff from such rain
events as described in this study. This type of adaptation will enable
authorities involved in stormwater quality mitigation strategies to cope with
the changed rainfall characteristics due to climate change in highly
urbanised coastal regions.
7.4. Conclusions
The low, low to moderate and high rain events under changed climatic conditions
will affect the wash-off of toluene, ethylbenzene, meta-xylene, para-xylene and
ortho-xylene. TOC is found to be predominant carrier of toluene, meta- xylene and
para- xylene in the two finer fractions of 75-150 µm and 1-75 µm as well as the
dissolved fraction of <1 µm and for ethylbenzene in 150-300 µm and >300 µm
191
under such rain events due to climate change. Ortho-xylene did not show such
affinity towards either TSS or TOC during wash-off under such changes.
This study was undertaken as a part of an Australian Research Council funded
linkage project (LP0882637). The first author gratefully acknowledges the
postgraduate scholarship awarded by the Queensland University of Technology to
conduct his doctoral research. Contributions from Vince Alberts at Queensland
Health for VOC analysis and the support from Gold Coast City Council and
Queensland Department of Transport and Main Roads are also gratefully
acknowledged.
7.6. References
ABS (2010). Australian Bereau of Statistics. Catalogue No. 1318.3 - Qld Stats, Feb
2010.
http://www.abs.gov.au/AUSSTATS/abs@.nsf/Lookup/1318.3Main%20Features3Fe
b%202010. Access date 10 May 2010.
Abbs, D., McInnes, K. and Rafter, T. (2007). The Impact of Climate Change on
Extreme Rainfall and Coastal Sea Levels over South-East Queensland, Part 2: A
High-Resolution Modelling Study of the Effect of Climate Change on the Intensity
of extreme Rainfall Events. CSIRO Division of Marine and Atmospheric Research,
Aspendale, Victoria, Australia: pp. 35.
APHA (2005). Standard Methods for the Examination of Water and Wastewater,
21st Edition, American Public Health Association. Methods: 2:55-58, 3:7-8, 5:19-
21.
Ayoko, G. A. (2004). Volatile Organic Compounds in Indoor Environments, in:

Pluschke, P. (Ed.), Indoor Air Pollution: The Handbook of Environmental
Chemistry. Springer, pp. 1-35.
Ayoko, G. A., Singh, K., Balerea, S. and Kokot, S. (2007). Exploratory multivariate
modeling and prediction of the physico-chemical properties of surface water and
groundwater. Journal of Hydrology 336(1-2), 115-124.
Benson, N. U., Essien, J. P., Ebong, G. A. and Williams, A. B. (2008). Petroleum

hydrocarbons and limiting nutrients in Macura reptantia, Procambarus clarkii and
benthic sediment from Qua Iboe Estuary, Nigeria. The Environmentalist 28(3), 275-
282.
192
Brans, J., Vincke, P. and Mareschal, B. (1986). How to select and how to rank
projects: the PROMETHEE method. European Journal of Operational Research
24(2), 228-238.
Bubenzer, G.D. (1979). Inventory of rainfall simulators. Proceedings of the Rainfall

Simulator Workshop. Tucson: U.S. Department of Agriculture, pp. 120-130.
Burnley, I. and Murphy, P. (2003). Seachange: Movement from metropolitan to

arcadian Australia. UNSW Press, Sydney.
CSIRO (2007). Climate Change in Australia. Technical Report 2007: pp.148.
Decembrini, F., Azzaro, F. and Crisafi, E. (1995). Distribution of Chemical

Polluting Factors in South Italian Seas along Calabria Coastal Waters (Low
Tyrrhenian Sea, High Ionian Sea and Straits of Messina). Water Science and
Technology 32(9-10), 231-237.
Decision (2000). Visual Decision Inc., Getting Started Guide, DecisionLab 2000,
Version 1.01.0388, Copyright© 1998 – 2000, Visual Decision Inc., Montreal,
Quebec, Canada.
EEA (2003). Mapping the Impacts of Recent Natural Disasters and Technological
Accidents in Europe. E. E. Agency. Copenhagen, Denmark.

EPA (2008). Test Methods for Evaluating Solid Waste, Physical/Chemical Methods:
SW-846. 3rd Edition. United States Environmental protection Agency. Methods:
5035, 8260B.
Floyd, C.N. (1981). Mobile rainfall simulator for small plot field experiments.
Journal of Agricultural Engineering Research 26(4), 307-314.
Gurran, N., C. Squires, and E. Blakley (2006). Meeting the sea change challenge:
Best Practice Models of Local and Regional Planning for Sea Change Communities
in Coastal Australia. Report No. 2 for the National Seachange Taskforce. The
University of Sydney Planning Research Centre, Sydney.
Herngren, L., Goonetilleke, A. and Ayoko, G. A. (2005a). Understanding heavy

metal and suspended solids relationships in urban stormwater using simulated
rainfall. Journal of Environmental Management 76(2), 149-158.
Herngren, L., Goonetilleke, A., Sukpum, R. and De Silva, D. Y. (2005b). Rainfall

simulation as a tool for urban water quality research. Environmental Engineering
Science 22(3), 378-383.
Hoffman, E. J., Latimer, J. S., Mills, G. L. and Quinn, J. G. (1982). Petroleum

hydrocarbons in urban runoff from a commercial land use area. Journal Water
Pollution Control Federation, 54(11), 1517-25.
193
Hoffman, E. J., Mills, G. L., Latimer, J. S. and Quinn, J. G. (1984). Urban runoff as
a source of polycyclic aromatic hydrocarbons to coastal waters. Environmental
Science and Technology 18(8), 580-587.
IARC (2009). IARC Monographs on the Evaluation of Carcinogenic Risks to

Humans. http://monographs.iarc.fr/ENG/Classification/crthgr01.php. Access date 20
March 2010.
IPCC (2007). Climate Change 2007: The Physical Science Basis. S. Solomon, Qin,
D., Manning, M., Chen, Z., Marquis, M., Averyt, K.B., Tignor, M., Miller, H.L
(Eds.) Cambridge, United Kingdom and New York, USA, Cambridge University
Press: pp. 996.
Jackson, D. A. (1993). Stopping Rules in Principal Component Analysis: A

Comparison of Heuristical and Statistical Approach. Ecology 74(8), 2204-2214.
Jartun, M., Ottesen, R. T., Steinnes, E. and Volden, T. (2008). Runoff of particle
bound pollutants from urban impervious surfaces studied by analysis of sediments
from stormwater traps. Science of the Total Environment 396(2-3),147-163.
Kimura, I. (1988). Aquatic pollutions problems in Japan. Aquatic Toxicology 11(3-

4), 287- 301.
Keller, H. R., Massart, D. L. and Brans, J. P. (1991). Multicriteria decision making:

A case study. Chemometrics and Intelligent laboratory Systems, 11(2), 175-189.
Loch, R.J., Robotham, B.G., Zeller, L., Masterman, N., Orange, D.N., Bridge, B.,
Sheridan, G. and Bourke, J.J. (2001). A multi-purpose rainfall simulator for field
infiltration and erosion studies. Australian Journal of Soil Research 39(3), 599-610.
Mahbub, P., Ayoko, G., Egodawatta, P., Yigitcanlar, T., and Goonetilleke, A.
(2010a). Traffic and climate change impacts on water quality: measuring build-up
and wash-off of heavy metals and petroleum hydrocarbons, in: Yigitcanlar, T., (Ed.),
Rethinking Sustainable Development: Urban management, Engineering and Design.
Engineering Science Reference, New York, pp. 147-167.
Mahbub, P., Ayoko, G. A., Goonetilleke, A., Egodawatta, P. and Kokot, S. (2010b).
from urban roads. Environmental Science and Technology (accepted for publication
October 2010).
Malvern (1994). Mastersizer S Long Bed Version 2.19 © Malvern Instruments Ltd.
1992-1994.
Mareschal, B. and Brans, J. P. (1988). Geometrical representations for MCDA.

European Journal of Operational Research 34(1), 69-77.
Sansalone, J. J. and Buchberger, S. G. (1997a). Characterization of solid and metal

element distributions in urban highway stormwater. Water Science and Technology
36(8-9), 155-160.
194
Sansalone, J. J. and Buchberger, S. G. (1997b). Partitioning and first flush of metals
in urban roadway storm water. Journal of Environmental Engineering 123(2), 134-
143.
SIRIUS (2008). SIRIUS version 7.1© copyright 1987-2008, Pattern Recognition

Systems AS, http://www.prs.no, Help Topics.
195
196
CHAPTER 8 PREDICTION OF VOLATILE
ORGANIC COMPOUNDS BUILD-UP ON URBAN
ROADS
Manuscript Title
Prediction Model of the Build-up of Volatile Organic Compounds on Urban Roads
Parvez Mahbub1*, Ashantha Goonetilleke1, Godwin A. Ayoko2

1
2
s.mahbub@qut.edu.au; a.goonetilleke@qut.edu.au; g.ayoko@qut.edu.au
*
Corresponding Author: Parvez Mahbub; Tel: 61 7 3138 9945; Fax: 61 7 3138 1170;
Published (2011): Environmental Science & Technology, 45(10), 4453-4459.

Impact Factor 5.5; ERA ranking A*.
Due to copyright restrictions, this article is

not available here. Please consult the
hardcopy thesis available from QUT Library
or view the published version online at:
http://dx.doi.org/10.1021/es200307x
197
CHAPTER 9 CHARACTERISATION AND
PREDICTION OF SEMI AND NON VOLATILE
ORGANIC COMPOUNDS WASH-OFF
Manuscript Title
Prediction of the Wash-off of Traffic Related Semi and Non Volatile Organic
Compounds from Urban Roads under Changed Rainfall Characteristics
Parvez Mahbub1*, Ashantha Goonetilleke1, Godwin A. Ayoko2

1
2
s.mahbub@qut.edu.au; a.goonetilleke@qut.edu.au; g.ayoko@qut.edu.au
*
Under Review (2011): Journal of Hazardous Materials. Impact Factor 4.36;

ERA ranking A.
Abstract
Traffic generated semi and non volatile organic compounds (SVOCs and NVOCs)
pose a serious threat to human and ecosystem health when washed-off into receiving
water bodies by stormwater. Climate change influences rainfall characteristics which
makes the estimation of these pollutants in stormwater quite complex. The research
study discussed in the paper present a prediction framework of such pollutants under
the dynamic influence of climate change on rainfall characteristics. The framework
incorporates orthogonal rotatable central composite experimental design to set up
calibration matrices and partial least square regression to identify significant factors
in predicting the target SVOCs and NVOCs in four particulate and one dissolved
fraction. The methodology overcomes the limitation of stringent laboratory
preparation of calibration matrices by extracting uncorrelated underlying factors in
225
the data matrices through systematic application of principal component analysis
and factor analysis. For particulate fractions of >300-1 µm, similar distributions of
predicted and observed concentrations of the target compounds from minimum to
75th percentile were achieved using the datasets which were not used in the
calibration. The inter-event coefficient of variations for particulate fractions of
>300-1 µm were 5% to 25%. The limited solubility of the target compounds in
stormwater restricted the predictive capacity of the proposed method for the
dissolved fraction of <1 µm.
Keywords: Semi volatile organic compounds, non volatile organic compounds,

pollutant wash-off, climate change

preparation of figures, tables and supporting information.


This chapter is an exact copy of the submitted manuscript of the journal paper.

Development
objectives of the overall research study. Amongst several objectives, the
characterisation of the wash-off of traffic generated stormwater pollutants on urban
roads, development of prediction frameworks and accordingly, providing guidance
to the stormwater quality mitigation strategies were mentioned in section 1.3,
Chapter 1. The methodologies to accomplish such objectives were formulated in this
journal paper using principal component analysis (PCA), orthogonal experimental
226
design, factor analysis (FA) and partial least square regression (PLS) model with a
view to predict the concentrations of seventeen semi and non volatile organic
compounds (VOCs) on urban roads during wash-off. The knowledge gained in
Chapter 4 in terms the dynamic changes in the rainfall characteristics due to climate
change as well as the PCA techniques employed in this chapter were used to
formulate the calibration matrices for PLS prediction model. Additionally, this
chapter confirmed the fact that using orthogonal experimental design and
VARIMAX rotation through FA can overcome the difficulties in stringent
laboratory conditions to prepare optimised calibration matrices. The acceptable
prediction results in both Chapter 8 and 9 for a total of twenty one organic
compounds in their build-up and wash-off have established the fact that this
prediction framework can be generalised to predict the concentrations of other traffic
generated pollutants on urban roads. Figure 3.3 in section 3.7, Chapter 3 provides a
schematic flow diagram of this research study where the publication of this journal
paper was highlighted as an integral process that contributes to the overall research
outcome.
227
9.1. Introduction
Traffic related semi and non volatile organic compounds (SVOCs and NVOCs) are
primarily associated with diesel fuels, fuel oils, heavier engine oils and lubricants
(1). Homologous series of n-alkanes from decane to tetracontane are amongst the
most common constituents of these products (2). These are widely used in motor
vehicles and have the potential to pollute the urban water environment through
deposition and wash-off from urban roads. In this context, the rainfall characteristics
such as, intensity, duration and frequency or average recurrence interval (ARI) are
predicted to undergo significant changes resulting from climate change. The
commonwealth scientific and industrial research organisation (CSIRO) has
forecasted longer periods of dry weather with fewer, but more intense storms in
Australia due to climate change (3). These climate change driven changes in the
rainfall characteristics will affect the wash-off processes of various stormwater
pollutants including the SVOCs and NVOCs.
The detrimental effects of SVOCs and NVOCs on human health have been widely
reported in research literature. Mutagenic evidence in mammalian cells caused by
diesel engine exhaust particles has been cited by Bao et al. (4). Morgan et al. (5)
attributed the long term exposure to diesel engine exhaust particles to respiratory
allergy, cardiopulmonary mortality and risk of lung cancer. Petroleum related
activities have been attributed to significant wetland loss in the Mississippi Delta
(6). To-date, studies have been commonly undertaken to characterise the impacts of
volatile compounds such as BTEXs (benzene, toluene, ethylbenzene and xylene)
generated from traffic on urban roads (7) and ambient atmosphere (8, 9). Vehicle
generated organic pollutants has also been characterised in terms of concentrations
228
and modelled for the ambient atmosphere by researchers (10). However, the current
state of knowledge on traffic generated semi and non volatile organic compounds
(SVOCs and NVOCs) available on roads for wash-off is very limited. Additionally,
it is important to note that pollutants present in the urban atmosphere do not
necessarily get deposited on the urban roads due to various climatic factors. This
situation becomes complex when the changed rainfall characteristics due to climate
change affects the wash-off processes of such pollutants from roads. Therefore,
accurate estimations of the concentrations of available SVOCs and NVOCs on roads
in wash-off under climate change is required in order to undertake mitigation
measures for the management of such pollutants in the stormwater runoff. Instead of
focussing on the ambient atmosphere, this research study presents a framework for
predicting the concentrations of traffic generated SVOCs and NVOCs in wash-off
under climate change influenced rainfall characteristics. This approach is expected
to provide rationalisation for the uncertainties involved in the wash-off processes of
these pollutants and consequently help to contributing to developing appropriate
mitigation measures for pollution management.

Four road sites within a 5 km radius from a meteorological gauging station were
selected as the wash-off study sites. The station was located at 27.90° S and 153.31°
E at an elevation of 6 m above mean sea level with daily rainfall data recorded since
1894. It was hypothesized in this study that the predicted changes in the rainfall
characteristics at the study sites due to climate change are similar to that at the rain
gauging station due to their close proximity. The selected road sites were situated in
three relatively new suburbs in the Gold Coast region, Australia with the transport
infrastructure developed in the last decade. The sites were in different land uses such
229
as residential, commercial and industrial in order to incorporate a mix of vehicular
traffic characteristics. The locations of the sites are shown in the Appendix A.5.
9.2.2. Rainfall Simulation Incorporating Climate Change

The research study used a rainfall simulator (11) to replicate the design rainfall
events resulting from climate change. The rainfall simulation was based on the
studies of Abbs et al. (12) who predicted the average fractional change for extreme
rainfall intensities at 2, 24 and 72 hour durations for the Gold Coast area for 2030
and 2070 using dynamic downscaling techniques incorporating the CSIRO CC-MK3
and CSIRO RAMS climate models. Several climate change studies (3, 13) have
predicted that the probability of occurrence of shorter duration (<2 hr) events with a
large change in precipitation intensities is very high.
Mahbub et al. (14) used the outcome from the Abbs et al. (12) study and proposed
the following three scenarios to describe the climate change influenced rainfall
characteristics in the Gold Coast region:
• Shorter duration, with higher intensity with ARI constant;
• Shorter ARI, shorter duration with intensity constant; and
• Shorter ARI, with higher intensity while duration becomes shorter.
The current study incorporated these scenarios to simulate the predicted 2030
rainfall characteristics in the Gold Coast. Table 9.1 gives the rainfall simulation
strategy according to the study by Mahbub et al. (14).
230
Table 9.1 Simulation events based on the daily rainfall intensity at study sites in the Gold Coast
region for 2030: adapted from Mahbub et al. (14)
Scenario Current simulation events for Gold Coast Future simulation events for Gold Coast region
region for 2030
Simulation Duration Intensity ARI Simulation Duration Intensity ARI
Event (min) (mm/hr) (year) Event (min) (mm/hr) (year)
Shorter 1 60 39.3 1 19 25 63 1
Duration, 3 90 39.3 2 20 42.5 61.2 2
with 5 133 39.3 5 21 69 59.2 5
higher 6 160 39.3 10 22 85 58.3 10
intensity 18 105 75 100 13 49 115 100
with ARI - - - 2 65 37.39 1
constant
Shorter 12 45 125 100 7 5 125 1
ARI, - - - 4 120 24.6 1
shorter
duration
with
intensity
constant
Shorter 14 52.5 77 10 10 16 125 5
ARI, with 15 67.5 77 20 11 21 122 10
Higher 16 86.7 77 50 9 10.5 120 2
Intensity 17 101.25 77 100 8 5.75 119 1
while
Duration
becomes
shorter
A total of twenty two rain events were simulated in the four selected road sites. For
simplicity and due to City Council restrictions on road lane closure, the simulation
events were distributed in the four study sites as per their intensities which ranged
from 24.6-39.3, 58.3-63, 75-77 and 119-125 mm/hr.
9.2.3. Wash-off Sample Collection

The wash-off samples resulting from the simulated rainfall events were collected
from a 3 m2 collection plot using a commercially available vacuum cleaner. The
collection plots were located in the middle of the traffic lanes at the study sites,
marked with permanent markers, and thoroughly cleaned with deionised water. Then
the plots were left for seven antecedent dry days to allow for traffic generated
pollutants to build-up. This allowance of seven dry days was in conformity with the
findings of Egodawatta (15) who noted that the pollutant build-up on road surfaces
asymptote to an almost constant value after an antecedent dry period of seven days.
231
The plot area for rainfall simulation was connected to a collection trough (11). The
runoff water in the collection trough was vacuumed continuously into 25 L plastic
containers. The runoff samples were transported to the laboratory for sub-sampling
immediately after collection. As pollutant concentrations can vary by orders of
magnitude during a runoff event, the flow weighted average or event mean
concentration samples (EMC) were found to be appropriate for evaluating the
impacts of stormwater runoff on receiving waters (16). In this study, 500 mL EMC
samples were prepared in the laboratory using a churn splitter. The required volumes
at a particular duration constituting an EMC sample were determined from the
percentages of the total runoff collected in different containers for that duration and
mixed together to obtain the 500 mL EMC sample for an event.
The particle size distribution of the suspended solids in the subsamples were
determined using a Malvern Mastersizer S Particle Size Analyser capable of
analysing particles between 0.05 to 900 µm diameter. The particle size distributions
of the sub-samples were used as a guide for maintaining homogeneity in the sub-
samples throughout the sample splitting process. Based on the particle size
distribution, the total particulate analytes were fractioned into four size ranges,
namely, >300 µm, 150-300 µm, 75-150 µm, 1-75 µm using wet sieving. The filtrate
passing through a 1 µm membrane filter was considered as the total dissolved
fraction. In each case, 500 mL homogeneous sub-samples were prepared using
deionised water, collected in 500 mL amber glass bottles with a PTFE seal,
preserved with 5 mL of 50% HCl at 4°C in the laboratory and analysed within 40
days of collection. A total of 110 wash-off samples were prepared for the 22
232
simulated rain events with each event having 5 samples for the five size fractions
mentioned above.

The target SVOCs for the study were octane (OCT), decane (DEC), dodecane
(DOD), tetradecane (TED), hexadecane (HXD), octadecane (OCD), eicosane (EIC),
docosane (DOC), tetracosane (TTC), hexacosane (HXC), and octacosane (OCC)
having boiling points ranging from 125° C to 432° C. For the convenience of the
predictive framework proposed in the study, the target SVOCs were further
separated into two groups based on their molecular weights, namely ‘Light SVOC’
and ‘Heavy SVOC’. The ‘Light SVOC’ group consisted of four SVOCs from octane
to tetradecane whilst the ‘Heavy SVOC’ group consisted of the remaining seven
SVOCs from hexadecane to octacosane. The target NVOCs were triacontane (TCT),
dotriacontane (DTT), tetratriacontane (TRT), hexatriacontane (HXT),
octatriacontane (OTT), and tetracontane (TTT) with boiling points ranging from
449° C to 525° C (17).
USEPA methods 3510C, 8015, 8021, and 8260 (18) were adopted for the
determination of SVOCs. Draper et al. (2) proposed modifications to the USEPA
methods to determine motor oils with carbon numbers up to C38. This study used
such modifications as a guide to establish the Gas Chromatographic (GC)
temperature program for simultaneous determination of both SVOCs and NVOCs.
Details of SVOC and NVOC test methods are given in the Appendix A.5.
Other physico-chemical variables such as total suspended solid (TSS) and total
organic carbon (TOC) were determined by methods 2540D and 5310B (19).
233
Additionally, the pH and electrical conductivity (EC) of each sample were measured
using standard pH and EC probes in the laboratory according to methods 4500-H+ B
and 2510B respectively (19).
9.2.5. Data Analysis

Data matrices were constructed for light SVOCs, heavy SVOCs and NVOCs at five
size fractions noted above. Each matrix consisted of twenty two objects with
numerical object identifiers (same as the simulation events in Table 9.1) starting
with 1. Rainfall characteristics such as, intensity, frequency, and duration as well as
the physico-chemical characteristics such as TSS, TOC, pH, and EC were
considered to be the independent variables causing the wash-off of the target SVOCs
and NVOCs. After initial observation of the probability distribution of the objects
and variables, standardisation of each variable and normalisation of each object were
undertaken as pre-treatment measures.
The data analysis was designed to investigate the wash-off process of SVOCs and
NVOCs under climate change conditions and then to apply the findings from the
initial investigations to develop a prediction framework for light SVOCs, heavy
SVOCs and NVOCs wash-off. Multivariate chemometrics methods such as principal
component analysis (PCA), factor analysis (FA), experimental design, and partial
least squares regression (PLS) were employed for the data analysis. Discussions of
these techniques are given in the Supporting Information.

Wash-off data matrices for light SVOCs, heavy SVOCs and NVOCs were analysed
for all five size fractions. Figure 9.1 shows the PCA biplots for total particulate
(<300µm-1µm) and dissolved fractions (<1µm). This study adopted the rain events
234
classification under climate change proposed by Mahbub et al. (20). The events with
intensity <40 mm/hr with relatively low ARI were classified as low events; those
having intensity between 50 to 100 mm/hr but with relatively higher ARIs of up to
50 years were classified as moderate events; events having intensities >100 mm/hr
with very high frequency were classified as high events whilst events with similar
intensities to moderate and high with extremely rare occurrence (ARI ≥ 100 years)
were classified as extreme events. Events which manifested the attributes of both
low and moderate events were classified as low to moderate events.
235
3
Light SVOC Particulate 4
Low
1 Light SVOC Dissolved
DOD
2
2 8 ARI
7 3 13 12
5 17
3 9
1 6 4 EC 18
TSS 2
15 14 16 Low to
OCT 10 moderate PSP
TOC
Duratio n In ten sity
11
0 18
19
16 11 1 20 14 9
17 Moderate Intensity
PC 2 (21.7%)
pH DEC
10
PC 2 (25.0%)
ARI
-1 13 21
DOC
0
87
TED 22 OCT
DEC 4 21
12
-2 TED
-1
5 15
Duration
22 6 19
-3 DOD
-2
3
1
Low 2
pH EC
TDS
-4 20
-3
-5 -4
-4 1 6 -4 -2 0 2 4
PC 1 (37.4%) PC 1 (32.5%)
(a) (b)
4
3
Heavy SVOC Particulate Heavy SVOC Dissolved
12
3 19
In ten sity
OCD DOC 2 14 21
13 8
7 PSP 9
20
12 TOC
16
HXC
EIC
2 17
ARI
20
1
ARI 13 21 11 HXC
11 HXD
9 22 Intensity
18
1 17
PC 2 (21.5%)
PC 2 (23.1%)
HXD
15
10 16 Moderate 0
7
22 1
0
EIC
8 OCD
EC DOC
TSS
-1
TSS
OCC
-1 pH
TOC pH
19
18 TTC
OCC
14 5 15
Low to moderate 6
Duration 4 -2
-2 3 Duration
Low 6
5 2
1 4 3
TTC 2
EC
10
-3 -3
-5 0 5 -4 1
PC 1 (32.8%) PC 1 (28.1%)
(c) (d)
3
7 Intensity Moderate 5
NVOC Particulate
16
TSS NVOC Dissolved
2 EC
1
4
13 12
1 17 9 11
TOC
ARI 5 8 4 15 3
1
3 10 Low to moderate
0 Duratio n 18
2
6 14 TCT
2 5 3
OTT
21
PC 2 (17.2%)
-1
PC 2 (21.7%)
6 15 OTT
19 1 DTT
HXT 22
-2 pH 2 4 pH
Duration TRT TDS
TCT
TTT
20 EC
18 22 20 0
TTT
-3
16 19
Low to moderate HXT
21
-1 ARI 10
17 13
-4 12 11
14 8
DOC
-5 -2 7
DTT
In ten sity
TRT 9
-6 -3
-5 0 5 -4 -2 0 2 4
PC 1 (35.3%) PC 1 (28.2%)
(e) (f)
Figure 9.1 PCA biplots of particulate (>300µm-1µm combined) and the dissolved (<1µm)
fractions for light SVOCs, heavy SVOCs and NVOCs for the 22 rain events shown with
numerical identifiers
236
Based on the classification, events 1, 2, 3, 4, 5, 6 were grouped as low events; events
19, 20, 21 were grouped as low to moderate events; events 14, 15, 16, 22 were
grouped as moderate events; 7, 8, 9, 10, 11 were grouped as high events; 12, 13, 17,
18 were grouped as extreme events. In Figure 9.1, two important facts were noted.
Firstly, the average recurrence intervals (ARI) are uncorrelated to both the
intensities and durations of events as the loading vector of ARI is nearly
perpendicular to those of intensities and durations (Figs. 9.1b-9.1f). Therefore, any
prediction framework for SVOCs and NVOCs should not include all three of them
together as measured variables. As the intensities and durations were more strongly
correlated with the target variables than ARI (Figs. 9.1a-9.1f), the analysis excluded
ARI from the measured variable list in the subsequent analysis. The relative
importance of other variables such as, pH, EC, TSS, TOC in the prediction of the
target compounds during wash-off was susbstantiated based on their positive
correlations with these compounds in Figures 9.1a to 9.1f.
The second important fact evident in the biplots of Figure 9.1 was that the low, low
to moderate and moderate rain events formed clusters strongly correlated to the
target compounds during wash-off except in Fig. 9.1d. This suggested that the low,
low to moderate and moderate rain events primarily caused the wash-off of the light
SVOCs, heavy SVOCs and NVOCs. These preliminary findings were useful in
selecting the experiments (i.e., rain events) to construct the calibration matrices in
the experimental design.
9.3.2. FA
Factor analysis in two phases namely, factor extraction and orthognal varimax
rotation was performed to identify the underlying independent factors of the data
237
matrices for light SVOCs, heavy SVOCs and NVOCs. After careful investigation of
the rotated component matrices for light SVOCs, heavy SVOCs and NVOCs which
consisted of the correlations between the measured variables and the factors, four
underlying factors were found sufficient for the light SVOC and heavy SVOC
matrices whilst five factors were deemed necessary for the NVOC matrix. These
independent factors were extracted based on the initial eigenvalue criteria ≥ 1. The
underlying factors were assigned with numerical identifiers each starting from 1
with initials ‘L’, ‘H’ and ‘N’ for light SVOCs, heavy SVOCs and NVOCs
respectively. New variables for each factor corresponding to the twenty two rain
events were then created by the regression method (21) as shown in Table 9.2.
Table 9.2 New independent variables for each underlying factors (starting with initials L, H or
N) in the data matrices of light SVOC, heavy SVOC and NVOC
Underlying Factors
Rain
Light SVOC Heavy SVOC NVOC
Events
L1 L2 L3 L4 H1 H2 H3 H4 N1 N2 N3 N4 N5
1 -1.26 -.061 1.173 .195 -.522 -.840 -.232 1.037 -1.194 1.598 -.342 -.966 .708
2 -1.1 .043 .987 .516 -.668 -.881 -.278 .951 -1.037 1.671 -.148 .088 .423
3 -.871 -.554 .376 .083 -.471 -1.034 .618 .897 -1.141 .448 -.498 1.109 -.976
4 -1.48 -.262 .527 .257 -.545 -1.355 .229 .704 -1.516 .575 .004 .052 .172
- -.412 -.192 .366 -.958 -1.294 -.119 .086 -1.423 -.286 -1.154 -.337 -.293
5
1.245
6 -1.7 -.519 -.308 -.281 -.913 -1.610 -.340 .048 -1.510 -.776 -.508 -.580 -1.358
7 1.466 -.443 2.265 -.071 -.747 1.708 -.008 2.027 1.232 1.651 -.777 .583 -1.126
8 1.234 -.517 2.166 -.605 -.276 1.554 .600 1.981 1.361 1.877 -.233 .151 -1.012
9 1.052 .138 .390 .974 .562 1.165 -.448 .235 1.363 .205 .206 -.724 -.075
10 1.128 -.120 .078 -.052 -.510 1.137 -.757 -.072 1.233 -.088 .083 -.162 -.439
11 1.377 .278 -.910 -.267 -.251 1.052 -1.039 -.608 1.262 -.564 -.457 -.513 .479
12 1.113 .326 -1.384 -.469 .042 .818 -1.213 -1.117 .827 -1.148 -.683 .796 -.209
13 .978 -.004 -.949 3.348 .008 .303 -.668 -.678 .419 -1.029 -.550 1.998 -.894
14 .293 -.677 -.647 -.816 -.628 .696 .891 -1.411 .488 -.712 -.427 -1.516 .710
15 .212 -.974 -.558 -.892 .548 .322 2.140 -.676 .147 -.540 -.049 -.900 .338
16 .144 -.845 -.650 -.875 .741 .035 1.415 -.382 .383 -1.026 -.825 -.495 1.105
17 .062 -.972 -1.120 -.806 -.521 .300 1.013 -1.638 -.141 -1.157 -.166 -.876 -1.051
18 -.255 -1.256 -.784 -.311 -.240 -.170 1.711 -.630 -.342 -.552 3.688 -.236 -1.589
19 -.074 1.602 .773 -.684 3.162 -.019 -.369 .696 .522 .961 1.208 .980 1.171
20 -.163 3.101 -.147 -.862 .124 -.158 -1.362 -.606 .121 .190 1.121 -.452 2.076
21 -.322 1.516 -.892 -.519 2.260 -1.227 -.095 .151 -.590 -.922 .270 2.618 1.630
22 -.625 .612 -.194 1.771 -.197 -.504 -1.686 -.996 -.463 -.376 .237 -.621 .206
The new variables (i.e., factor scores) generated in Table 9.2 were used in the
subsequent PLS regression models to predict the corresponding target variables.
9.3.3. Experimental Design

Three calibration sets for the PLS model were optimised with two level orthogonal
rotatable central composite design for light SVOCs, heavy SVOCs and NVOCs. As
238
the number of factor levels and their values were unknown in the design, the study
incorporated the Sirius software (22) generated coded values for the two levels,
namely, high and low and incorporated 35 experiments (28 individual experiments
and 7 replicate experiments at centre) for the light SVOC and heavy SVOC data
matrices. Similarly, the study incorporated 50 experiments (46 individual
experiments and 4 replicate experiments at centre) for the NVOC data matrix. A
higher number of experiments for NVOC were required due to the large number of
underlying factors in the NVOC data matrix. Experiments were only chosen from
low, low to moderate and moderate rain events as these were found to be the
predominant events causing the wash-off of the target compounds. It was ensured
that each of the five size fractions contributed to the calibration matrices by selecting
at least seven experiments from each fraction. In Figure 9.2, PCA biplots of three
calibration sets are shown.
239
3
3
Light SVOC
Heavy SVOC E11
E5 E17
E17
EC
2 E10
E22
2 E12 E22 E5 TTC
Duration
E9
E4 E19 HXC
pH
E21 E10 E19 OCT
E2 E23
E25 EIC
E8 C1 C7
1 1
E11 E7
C2 E13 E15 E27
C3 C3 C2 OCD
Intensity C5 E2 E21
C4 OCC
DOC
PC 2 (19.0%)
C1 E4
PC 2 (18.3%)
E16 E18 C6 E13

E12 E1
0 E15 0
E6 C4
C6 DEC
DOD C5
E20
Duration
E3 TOC
TED E9
C7
-1
E14
E8 -1 TSS
E1 Intensity
E6 HXD
E23 E28 E14

E7 E3
-2
TSS
E26 -2 E18
E24 E20
TOC E26
E28 E24 EC
E25 E16
pH
E27
-3
-3
-4 -2 0 2 4 -4 1 6
PC 1 (35.3%) PC 1 (26.4%)
(a) (b)
5
NVOC TSS
TOC
E41
4
E44
3 E38 E42 E25

E37 E45
HXT
2 E43
E46
PC 2 (18.1%)
E40
1 E39
E24 C1
Intensity
E22
E23
E8 C4 E7E31 E15
0 C2
E19 C3
E33
E21 E3 E5 DTT
E30 E20 E29 E36 EC
E32 TRT
E27 TTT pH
E11 E10 E9 E28 E13 E35 E6 E34
-1 E1 E17
E2E26
E18
E4
E12
Duration
OTT
-2 E16
E14
TCT
-3
-4 1
PC 1 (26.4%)
(c)
Figure 9.2 PCA biplots of the experimental designs for (a) light SVOCs, (b) heavy SVOCs and
(c) NVOCs with experiments are shown with initial ‘E’ and replicate experiments with initial
‘C’
240
With few exceptions, in Figure 9.2a to 9.2c, most of the central experiments were
found close to origin of the biplots, which meant that these were replicates of the
same or similar experiments and did not need to be included in the design. The
central or replicate experiments were chosen in order to identify any curvature
present on the response surface by comparing their mean values with that of the rest
of the experiments. In Figure 9.2(a), 12 experiments were found to be very strongly
correlated with target compound dodecane (DOD) and octane (OCT), in Figure
9.2(b), 13 experiments were found to be strongly correlated with all target heavy
SVOCs whilst in Figure 9.2(c), 19 experiments were found to be strongly correlated
with all target NVOCs. This suggested that the calibration matrices closely
corresponded towards the wash-off of the target compounds under climate change
influenced rainfall characteristics even though the total variances explained by the
PCs in Figure 9.2 were around 45% - 53%. The calibration sets are provided in the
Appendix A.5.
9.3.4. PLS Model Validation

In the PLS regression, the target compounds OCT, DEC, DOD, TED, HXD, OCD,
EIC, DOC, TTC, HXC, OCC, TCT, DTT, TRT, HXT, OTT and TTT were
considered as dependent or measured variables whilst the factors extracted in the
factor analysis process along with Intensity, Duration, TSS, TOC, pH and EC were
considered as the predictor variables. A cross validation method (23) that left one
experiment out at a time from the calibration set was used to measure the standard
error in cross validation (SECV). The following three criteria were employed to
determine the required number of PLS components for regression:
• SECV≤1;
241
• 10% maximum difference between the percentage variance explained by the
predictor and the measured variables;
• There is no significant change in the percentage variance explained by the
predictor with the inclusion of an additional PLS component.
Table 9.3 gives the outcome of the PLS regression based on the above criteria.
242
Table 9.3 PLS regression parameters for predictor variables**
variance variance Regression Coefficients for predictor variables
Coefficient of
Measured PLS explained by explained by Underlying Factors
Determination, SECV
variables components predictor measured TSS TOC pH EC Intensity Duration L1a L2a L3a L4a
r2
variables, % variables, %
OCT 1 31.87 39.40 0.50 0.94 -0.29 -0.28 -0.18 0.22 I.F.* I.F.* -0.21 I.F.* I.F.* I.F.*
DEC 1 65.51 57.55 0.57 0.71 I.F.* -0.03 -0.04 I.F.* -0.04 0.03 I.F.* I.F.* I.F.* I.F.*
DOD 1 46.28 45.17 0.50 0.86 I.F.* -0.09 -0.27 0.01 -0.27 0.20 -0.06 I.F.* I.F.* I.F.*
TED 1 50.96 46.48 0.50 0.96 I.F.* 0.26 0.29 I.F.* I.F.* -0.24 I.F.* 0.13 0.03 I.F.*
H1b H2b H3b H4b
HXD 1 55.24 51.08 0.60 0.95 0.04 I.F.* 0.50 I.F.* I.F.* I.F.* I.F.* I.F.* 0.29 I.F.*
OCD 1 61.50 57.07 0.67 1.00 I.F.* I.F.* 0.08 -0.19 0.22 -0.19 -0.22 -0.13 I.F.* I.F.*
EIC 1 52.96 45.03 0.60 1.00 -0.20 -0.10 0.15 -0.23 I.F.* I.F.* -0.39 I.F.* I.F.* I.F.*
DOC 1 47.89 42.61 0.63 0.97 I.F.* I.F.* 0.20 -0.19 0.47 I.F.* I.F.* -0.23 I.F.* I.F.*
TTC 1 55.33 54.25 0.67 1.00 I.F.* I.F.* I.F.* I.F.* -0.21 I.F.* -0.40 I.F.* -0.36 I.F.*
HXC 1 46.82 47.64 0.70 0.96 -0.13 -0.10 -0.27 -0.27 I.F.* -0.08 I.F.* I.F.* I.F.* I.F.*
OCC 1 39.09 30.71 0.51 1.00 0.17 I.F.* I.F.* -0.17 I.F.* -0.09 I.F.* I.F.* I.F.* -0.31
N1c N2c N3c N4c N5c
TCT 1 60.93 62.19 0.71 1.00 I.F.* 0.08 -0.23 -0.18 0.13 -0.11 I.F.* -0.13 I.F.* I.F.* -0.18
DTT 1 65.69 69.02 0.81 0.84 0.60 I.F.* -0.04 I.F.* I.F.* I.F.* I.F.* I.F.* I.F.* I.F.* -0.08
TRT 1 66.51 63.67 0.80 1.00 0.31 -0.17 I.F.* I.F.* I.F.* I.F.* 0.37 I.F.* I.F.* I.F.* -0.16
HXT 1 69.56 64.51 0.78 1.00 0.17 I.F.* 0.25 I.F.* I.F.* -0.22 -0.07 I.F.* I.F.* 0.17 I.F.*
OTT 1 73.96 76.76 0.80 0.94 0.27 I.F.* -0.16 I.F.* I.F.* -0.16 0.30 I.F.* I.F.* I.F.* I.F.*
TTT 1 65.79 66.04 0.82 0.97 0.23 I.F.* -0.05 I.F.* I.F.* -0.23 0.19 I.F.* 0.16 0.15 I.F.*
*Insignificant factors in the PLS prediction model for corresponding measured variables
a
Underlying factors in the light SVOC matrix
b
Underlying factors in the heavy SVOC matrix
c
Underlying factors in the NVOC matrix
**Regression equations are given in Appendix A.6
243
The outcomes of the PLS regression model was optimised with a reduced number of
predictor variables in Table 9.3. Therefore, not all of the predictor variables were
required to predict the individual target components in Table 9.3. As a final step in
the model validation, data matrices were constructed from the remaining rain events
that were not used in the construction of the calibration matrices. In this study, the
validation of the PLS model was performed by comparing the distributions of the
box plot statistics of observed and predicted data matrices for the five size fractions.
Figure 9.3 shows the distributions for >300 µm and <1 µm size fractions. The
remaining particulate fractions are given in the Appendix A.5.
244
(a)
(b)
Concentration, ppm
Concentration, ppm
0.5
1.5
2.5
3.5
4.5
5.5
6.5
7.5
8.5
-0.5
-0.5
0
1
2
3
4
0.5
1.5
2.5
3.5
OCT(predicted)
OCT(observed) OCT(predicted)
OCT(observed)
media n
DEC(predicted)
minimum
DEC(predicted)
maximum
DEC(observed)
75th quartile
25th quartile
median
DEC(observed)
minimum
DOD(predicted)
maximum
DOD(predicted)
75th quartile
25th quartile
DOD(observed)
DOD(observed)
TED(predicted)
TED(predicted)
TED(observed)
TED(observed)
HXD(predicted)
predicted target compounds

HXD(predicted)
HXD(observed)
HXD(observed)
OCD(predicted)
OCD(predicted)
OCD(observed)
OCD(observed)
EIC(predicted)
EIC(predicted)
EIC(observed) EIC(observed)
DOC(predicted) DOC(predicted)
DOC(observed) DOC(observed)
TTC(predicted) TTC(predicted)
TTC(observed) TTC(observed)
245
HXC(predicted) HXC(predicted)
HXC(observed) HXC(observed)
OCC(predicted) OCC(predicted)
OCC(observed) OCC(observed)
TCT(predicted) TCT(predicted)
TCT(observed) TCT(observed)
DTT(predicted) DTT(predicted)
<1 µ m
DTT(observed) DTT(observed)
TRT(predicted) TRT(predicted)
TRT(observed) TRT(observed)
HXT(predicted) HXT(predicted)
HXT(observed) HXT(observed)
OTT(predicted) OTT(predicted)
OTT(observed)
>300 µ m
OTT(observed)
TTT(predicted) TTT(predicted)
TTT(observed) TTT(observed)
Figure 9.3 Distributions of the box plot statistics at (a) >300 µm and (b) <1 µm for observed and
In Figure 9.3(a), it is evident that except for DOC, HXC and OTT, the distributions
of the concentrations of the remaining 14 target compounds are quite similar from
minimum to 75th quartile in the observed and predicted data matrices. Other
particulate fractions also showed similar results with very few exceptions. However,
the dissolved fraction of <1 µm did not show any such similarity among the box plot
statistics in Figure 9.3(b). This is attributed to the fact that the solubilities of the
target compounds in water are very low and these compounds are mainly attached to
the particulate solid fractions during their wash-off. In order to derive a more
comprehensive outlook on the validation of the PLS model, the coefficient of
variation (CV %) of the predicted concentrations for the remaining rain events were
analysed and the results are shown in Figure 9.4.
60
50
Inter-Event Coefficient of Variation, CV (%)
40 CV >300 µ m (%)
CV 150-300 µ m (%)
CV 75-150 µ m (%)
30 CV 1-75 µ m (%)
CV <1 µ m (%)
20
10
Figure 9.4 Coefficient of Variations (CV %) of the predicted concentrations at the rain events
not used in the calibration
Horwitz (24) suggested a range of ±20% for the coefficient of variation at the ppm
level concentrations estimation of organic compounds in different laboratories. In
246
Figure 9.4, the inter-event percentage CV are compared between the five size
fractions and it is clearly evident that the CV % were as high as 55% for the
dissolved fraction of <1 µm. However, the CV % were in the range of 5-25% for the
particulate fractions from >300 µm to 1 µm with very few exceptions. This also
confirmed the fact that the solubility of the target compounds were very low in water
and hence, compromises the predictive capacity of the PLS framework for the
dissolved fraction of <1 µm. The PLS framework performed with acceptable
predictions within the range of minimum to 75th quartile of the observed
concentration values at particulate fractions from >300-1 µm for the different
rainfall characteristics influenced by climate change.
As the current study focused on the available traffic generated SVOCs and NVOCs
on urban roads, the prediction of the wash-off of these pollutants was strongly
correlated with their accumulation on road surfaces. As a result, such a prediction
framework could be generalised and used in predicting the accumulation of traffic
generated organic compounds on roads (25). In several chemometric studies where
experiments were conducted under stringent laboratory conditions, the experimental
design of the calibration matrices gave better prediction results. For example,
Sivakumar et al. (26) achieved ≤3% coefficient of variation in an optimisation study
aimed at commercial pharmaceutical preparation with three independent factors
assumed significant priori. In another study, Ni et al. (27) reported up to 36% error
during the prediction of nitrobenzene and nitro-substituted phenols using the single
component PLS method with the number of measured variables taken as significant
factors in the data matrices. Even though these methods produced acceptable results,
there was a possibility of introducing bias into the prediction as it was also found in
247
the current study that not all of the measured variables were significant for
predicting a particular compound.
Another advantage of the current study over the past studies is that this framework
allows the introduction of the underlying uncorrelated factors into the data matrices.
Therefore, it is not necessary to assume significant factors priori. Considering the
fact that stringent laboratory conditions could not be applied into the experimental
design of the calibration matrices as the wash-off sample collection was field based,
the model’s ability to predict most of the light SVOCs, heavy SVOCs and NVOCs
within an acceptable range provide researchers a robust tool to forecast the
concentrations of these pollutants in particulate fractions of wash-off due to climate
change driven rainfall characteristics.
9.4. Acknowledgement
This study was undertaken as a part of an Australian Research Council funded
linkage project (LP0882637). The first author gratefully acknowledges the
postgraduate scholarship awarded by Queensland University of Technology to
conduct his doctoral research. Support from the Gold Coast City Council and
Queensland Department of Transport and Main Roads are also gratefully
acknowledged.
9.5. Supporting Information

Test methods of SVOCs and NVOCs are described in the Appendix A.5 along with
detailed description of data analyses techniques. Additional Figures showing the
selected sites as well as the box plot statistics of observed and predicted
concentrations of target analytes for 150-300 µm, 75-150 µm and 1-75 µm
particulate fractions are given in Figures A.5.1, A.5.2, A.5.3 and A.5.4 respectively.
248
Additional tables showing the calibration matrices of light SVOCs, heavy SVOCs
and NVOCs are given in Tables A.5.1, A.5.2 and A.5.3 respectively. This
information is available free of charge via the internet at http://pubs.acs.org.
9.6. Brief
A robust tool to predict the semi and non volatile organic compounds under the
dynamic influence of climate change on rainfall characteristics has been presented in
the paper.
9.7. References
(1) Interlaboratory study of three methods for analysing petroleum hydrocarbons
in soils. API publication no. 4599, American Petroleum Institute, Washington
DC 1994; www.api.org/Publications. Accessed 10 September 2010.
(2) Draper, W.M.; Dhaliwal, J. S.; Perera, S. K.; Bauman, F. J. Determination of

diesel fuel and motor oil in water and wastes by a modified diesel-range
organics total petroleum hydrocarbon method. J. AOAC Int. 1996, 79, 508-
519.
(3) Climate Change in Australia. CSIRO Technical Report 2007, Bureau of

Meteorology, Australia; www.climatechangeinaustralia.gov.au. Accessed 12
September 2010.
(4) Bao, L.; Chan, S.; Wu, L.; Hei, T. K.; Wu, Y.; Yu, Z.; Xu, A. Mutagenicity of
diesel exhaust particles mediated by cell-particle interaction in mammalian
cells. Toxicology, 2007, 229, 91-100.
(5) Morgan, W. K. C.; Reger, R. B.; Tucker, D. M. Health effects of diesel

emissions. Ann. Occup. Hyg. 1997, 41, 643-658.
(6) Ko, J.-Y.; Day, J. W. A review of ecological impacts of oil and gas
development on coastal ecosystems in the Mississippi Delta. Ocean Coast.
Manage. 2004, 47, 597-623.
(7) Caselli, M.; Gennaro, G. D.; Marzocca, A.; Trizio, L.; Tutino, M. Assessment
of the impact of the vehicular traffic on BTEX concentration in ring roads in
urban areas of Bari (Italy). Chemosphere, 2010, 81, 306-311.
(8) Buczynska, A. J.; Krata, A.; Stranger, M.; Godoi, A. F. L.; Kontozova-
Deutsch, V.; Bencs, L.; Naveau, I.; Roekens, E.; Grieken, R. V. Atmospheric
BTEX-concentrations in an area with intensive street traffic. Atmos. Environ.
2009, 43, 311-318.
249
(9) Wang, P.; Zhao, W. Assessment of ambient volatile organic compounds
(VOCs) near major roads in urban Nanjing, China. Atmos. Res. 2008, 89, 289-
297.
(10) Smith, L.; Mukerjee S.; Gonzales M.; Stallings C.; Neas L.; Norris, G.;
Özkaynak, H. Use of GIS and ancillary variables to predict volatile organic
compound and nitrogen dioxide levels at unmonitored locations. Atmos.
Environ. 2006, 40, 3773-3787.
(11) Herngren, L.; Goonetilleke, A.; Sukpum, R.; De Silva, D. Y. Rainfall

simulation as a tool for urban water quality research. Environ. Eng. Sci. 2005,
22, 378-383.
(12) Abbs, D.; McInnes, K.; Rafter, T. The Impact of Climate Change on Extreme
Rainfall and Coastal Sea Levels over South-East Queensland, Part 2: A High-
Resolution Modelling Study of the Effect of Climate Change on the Intensity of
extreme Rainfall Events. CSIRO Division of Marine and Atmospheric
Research, Aspendale, Victoria, Australia, 2007.
(13) Climate Change 2007: The Physical Science Basis. S. Solomon, Qin, D.,
Manning, M., Chen, Z., Marquis, M., Averyt, K.B., Tignor, M., Miller, H.L.
Eds.; Intergovernmental Panel on Climate Change, Cambridge, United
Kingdom and New York, USA, Cambridge University Press, 2007.
Traffic and climate change impacts on water quality: measuring build-up and
wash-off of heavy metals and petroleum hydrocarbons. In Rethinking
Sustainable Development: Urban management, Engineering and Design;
Yigitcanlar, T., Ed.; Engineering Science Reference, New York 2010; pp. 147-
167.
(15) Egodawatta, P. K. Translation of small-plot scale pollutant build-up and wash-

off measurements to urban catchment scale. Ph.D. Dissertation, Queensland
University of Technology, Brisbane, 2007.
(16) Sansalone, J. J. ; Buchberger, S. G. Partitioning and first flush of metals in

urban roadway storm water. J. Environ. Eng. 1997, 123, 134-143.
(17) Kudchadker, A. P.; Zwolinski, B. J. Vapour pressures and boiling points of

normal alkanes, C21 to C100. J. Chem. Eng. Data, 1966, 11, 253-255.
(18) Test methods for evaluating solid waste, physical/chemical methods: SW-846.
3rd Edition. United States Environmental protection Agency. Methods:
3510C,8015, 8021, 8260. www.epa.gov/wastes/hazard/testmethods/sw846/.
Accessed 20 October 2009.
250
(19) Eaton, A. D.; Clesceri, L. S.; Rice, E. W.; Greenberg, A. E., Eds. Standard
Methods for the Examination of Water and Wastewater; 21st Edition,
American Public Health Association, Washington DC, 2005.
(20) Mahbub, P.; Ayoko, G. A.; Goonetilleke, A.; Egodawatta, P.; Kokot, S.
Impacts of traffic and rainfall characteristics on heavy metals build-up and
wash-off from urban roads. Environ. Sci. Technol. 2010, 44, 8904-8910.
(21) SPSS (2009). PASW Statistics 18, Release 18.0.1, Tutorials,

http://www.spss.com.
(22) SIRIUS (2008). SIRIUS version 7.1© copyright 1987-2008, Pattern

Recognition Systems AS, Help Topics, http://www.prs.no.
(23) Wold, S. Cross-Validatory Estimation of the Number of Components in Factor

and Principal Components Models. Technometrics, 1978, 20, 397.
(24) Horwitz, W. Evaluation of Analytical Methods Used for Regulation of Foods

and Drugs. Anal. Chem. 1982, 54, 67A-76A.
(25) Mahbub, P.; Goonetilleke, A.; Ayoko, G. A.; Egodawatta, P. A Prediction

Model of the Build-up of Volatile Organic Carbons on Urban Roads. Environ.
Sci. Technol. 2010. (Under Review).
(26) Sivakumar, T.; Manavalan, R.; Muralidharan, C.; Valliappan, K. Multi-criteria

decision making approach and experimental design as chemometric tools to
optimize HPLC separation of domperidone and pantoprazole. J. Pharm.
Biomed. Anal. 2007, 43, 1842-1848.
(27) Ni, Y.; Wang, L.; Kokot, S. Simultaneous Determination of nitrobenzene and
nitro-substituted phenols by differential pulse voltammetry and chemometrics.
Anal. Chim. Acta, 2001, 431, 101-113.
(28) Massart, D. L.; Vandeginste, B. G. M.; Buydens, L. M. C.; De Jong, S.; Lewi,
P. J.; Smeyers-Verbeke, J. Handbook of Chemometrics and Qualimetrics Part
A; Elsevier, Amsterdam, 1997.
(29) Meyers, L. S.; Gamst, G.; Guarino, A. J. Applied Multivariate Research:

Design and Interpretation. Sage Publications Inc. California, 2006.
(30) Peña-Méndez, E. M.; Astorga-España, M. S.; García-Montelongo, F. J.

Interpretation of analytical data on n-alkanes and polynuclear aromatic
hydrocarbons in Arbacia lixula from the coasts of Tenerife (Canary Islands,
Spain) by multivariate data analysis. Chemosphere, 1999, 39, 2259-2270.
(31) Yidana, M. S.; Ophori, D.; Yakubo, B. A multivariate statistical analysis of

surface water chemistry data—The Ankobra Basin, Ghana. J. Environ.
Manage. 2008, 86, 80-87.
251
(32) Zhu, X.; Zhang, L.; Che, X.; Wang, L. The classification of hydrocarbons with
factor analysis and the PONA analysis of gasoline. Chemom. Intell. Lab. Syst.
1999, 45, 147-155.
(33) Christensen, E. R.; Arora, S. Source apportionment of PAHs in sediments

using factor analysis by time records: Application to Lake Michigan, USA.
Water Res. 2007, 41, 168-176.
(34) Deming, S, N.; Morgan, S, L. Experimental Design: A Chemometric

Approach. Second Edn. Elsevier Science Publishers, Amsterdam, 1993.
(35) Tarley, C. R. T.; Silveira, G.; dos Santos, W. N. L.; Matos, G. D.; da Silva, E.
G. P.; Bezerra, M. A.; Miró, M.; Ferreira, S. L. C. Chemometric tools in
electroanalytical chemistry: Methods for optimization based on factorial
design and response surface methodology. Michrochem. J. 2009, 92, 58-67.
(36) Beebe, K. R.; Kowalski, B. R. An Introduction to Multivariate Calibration and

Analysis. Anal. Chem. 1987, 59, 1007A-1017A.
(37) Wang, H.; Liu, Q.; Tu, Y. Interpretation of partial least-squares regression
models with VARIMAX rotation. Comput. Stat. Data Anal. 2005, 48, 207-
219.
252
CHAPTER 10 CONCLUSIONS AND
RECOMMENDATIONS
This research study hypothesised that increased urbanisation and climate change are
the two most important phenomena that influence urban traffic and rainfall
characteristics, respectively. This, in turn, affects pollutant build-up and wash-off
from urban roads. Hence, pollutant build-up and wash-off processes on urban roads
were investigated from a dynamic point of view under changing urban traffic and
climate change. The research aims and objectives were achieved by classifying
traffic and rainfall scenarios and applying this classification to study the build-up
and wash-off of traffic generated pollutants under dynamic conditions and finally,
by formulating prediction frameworks for such pollutants based on the knowledge
derived. In this chapter, the key findings from the research study are discussed
starting with an object classification system for changed urban traffic and the rainfall
characteristics. This was followed by the findings on build-up and wash-off
processes of traffic generated pollutants under dynamic conditions. The robust
frameworks for the prediction of build-up and wash-off of pollutants under dynamic
conditions are presented with recommendations for further research.
10.1. The Object Classification System

This research established an effective object classification system based on fuzzy
clustering. The study successfully identified high, low and moderate traffic scenarios
as well as low, low to moderate, moderate, heavy and extreme rainfall scenarios
based on current observed traffic characteristics as well as current and future rainfall
characteristics. In this context, different combinations of climate change induced
253
rainfall characteristics such as, intensity, frequency and durations were incorporated
into the object classification system for year 2030 based on the contemporary
climate change research. The methodology discussed in Chapter 5 demonstrated the
systematic analysis of the impacts of urban traffic and rainfall characteristics on the
build-up and wash-off of traffic generated pollutants on roads.
In relation to build-up, this study has classified the high traffic scenario as
comprising of traffic volumes ranging from 9000 to 24000 average daily traffic
(ADT) with relatively high congestion; moderate traffic scenario comprise of ADT
values ranging from 2300 to 5900 with moderate congestion whilst low traffic
scenario relates to low traffic volume ranging from 500 to 3500 ADT with low
congestion. This classification was based on a moderately soft fuzzy clustering
technique.
For wash-off, the study classified the rainfall events with intensity <40 mm/hr with
relatively low ARI as low events; those having intensity between 50 to 100 mm/hr
but with relatively higher ARIs of up to 50 years were classified as moderate events;
events having intensities >100 mm/hr with very high frequency were classified as
high events whilst events with similar intensities to moderate and high with
extremely rare occurrence (ARI ≥ 100 years) were classified as extreme events.
Events which manifested the attributes of both low and moderate events were
classified as low to moderate events.
10.2. Build-up and Wash-off Processes of Pollutants under Dynamic

Scenarios
The study established that urban traffic characteristics such as average daily traffic
(ADT), volume to capacity ratio (V/C), and surface texture depth (STD) influence
254
heavy metals and total petroleum hydrocarbons build-up whilst rainfall
characteristics under climate change such as intensity, frequency, and duration
influence the wash-off of these pollutants from urban roads. Cadmium (Cd),
chromium (Cr), copper (Cu), antimony (Sb), nickel (Ni), zinc (Zn), lead (Pb),
manganese (Mn), aluminium (Al) and iron (Fe) were the target heavy metals. The
volatile, semi volatile and non volatile hydrocarbon compounds which constitute the
petroleum products used in motor vehicles were the target total petroleum
hydrocarbon compounds. The volatile compounds included toluene, ethylbenzene,
meta-xylene, para-xylene and ortho-xylene; the semi volatile compounds comprised
of octane, decane, dodecane, tetradecane, hexadecane, octadecane, eicosane,
docosane, tetracosane, hexacosane and octacosane whilst the non volatile
compounds consisted of triacontane, dotriacontane, tetratriacontane,
hexatriacontane, octatriacontane, and tetracontane.
Build-up and Wash-off Processes of Heavy Metals

The study established that the moderate to heavy traffic scenarios affect the build-up
of Cd, Cr, Cu, Ni, Zn and Pb whilst Mn, Al and Fe remain uncorrelated with traffic
sources. It was found that congestion expressed as volume to capacity ratio (V/C)
and the surface texture depth (STD) of the road surfaces were more strongly
correlated to traffic generated heavy metals build-up than the average daily traffic
(ADT). In build-up, high traffic activities in commercial and industrial areas
influence the accumulation of heavy metal concentrations in the particulate size
range from 75 - >300 µm, whereas metal concentrations in finer size range of <1-75
µm were not affected. The study also established through multicriteria decision
analyses that the 1-74 µm particulate fractions of total suspended solids (TSS) could
be regarded as a surrogate indicator for particulate heavy metals in build-up. In
255
terms of pollutants affinity, total suspended solids (TSS) was found to be the
predominant parameter for particulate heavy metals in build-up and total dissolved
solids (TDS) was found to be the predominant parameter for dissolved heavy metals
in build-up.
The research study also found that moderate to high rainfall scenarios influenced the
wash-off of traffic generated heavy metals from roads. However, it was established
that metal concentrations in the 1-75µm fraction were independent of the changes to
rainfall characteristics due to climate change. The study also found that organic
matter from <1 - >300 µm size could be targeted for the removal of Cd, Cr, Pb and
Ni from wash-off whilst Cu and Zn need to be removed as free ions from most
fractions in wash-off.
Build-up and Wash-off Processes of Volatile Organic Compounds

The research study established that the build-up of toluene, ethylbenzene, meta and
para-xylene, and ortho-xylene was mainly present in the particulate fractions of 1-
150 µm and total organic carbon (TOC) could be regarded as a surrogate indicator
for particulate volatile organic compounds in build-up. Additionally, commercial,
industrial and residential land use did not play a significant role in the build-up of
volatile organic compounds (VOCs). Furthermore, the highly congested traffic lane
of a road with high V/C was more likely to cause volatile organic build-up than the
ADT on the road. An important fact in the build-up of volatile organics was that the
rapid re-suspension and re-distribution of 75 - >300 µm particulates due to traffic
activities created an inverse relationship between VOC concentrations in build-up
and free flowing traffic.
256
The research study established that low, low to moderate and high rain events due to
climate change were the predominant events that influenced the wash-off of toluene,
ethylbenzene, meta-xylene, para-xylene and ortho-xylene from urban roads. TOC
was identified as the predominant carrier of toluene, meta-xylene and para-xylene by
the <1-150 µm particulate fraction and ethylbenzene by the 150 µm - >300 µm
particulate fraction under such predominant rain events due to climate change.
However, ortho-xylene was not found to show such affinity towards either TOC or
TSS.
Build-up and Wash-off of Semi and Non Volatile Organic

Compounds
The research study established that the build-up of semi volatile organic compounds
(SVOCs) and non volatile organic compounds (NVOCs) were primarily caused by
moderate traffic. This translates to average daily traffic ranging from 2300 to 5900
and average congestion of 0.47. It was also found that traffic congestion in the
commercial areas and the use of lubricants and motor oils in the industrial areas
were the main sources of SVOCs and NVOCs on urban roads, respectively. The
study outcomes also revealed that the target SVOCs and NVOCs were mainly
attached to particulate fractions in the range of 75-300 µm whilst the re-distribution
of coarse fractions due to vehicle activity mainly occurred in the particulate fraction
>300 µm.
Low, low to moderate and moderate rain events were the predominant events that
caused the wash-off of the target SVOCs and NVOCs in the particulate and
dissolved fractions taking into consideration the influence of climate change on
rainfall characteristics. It was also established that amongst the different rainfall
characteristics affected by climate change, intensity and duration are more strongly
257
correlated with the wash-off of target SVOCs and NVOCs than the rainfall
frequency.
10.3. Prediction Framework for Build-up and Wash-off under

Dynamic Conditions
The research study has proposed prediction frameworks for the build-up of VOCs as
well as wash-off of light SVOCs, heavy SVOCs and NVOCs. It was further
demonstrated that the use of factor analysis and experimental design can overcome
the limitations inherent in the preparation of calibration matrices under stringent
laboratory conditions. In the prediction of light SVOCs, heavy SVOCs and NVOCs,
the study outcomes confirmed that the bias in prediction could be minimised by
considering only the underlying significant factors of the build-up and wash-off data
matrices in the PLS prediction model. In chapter 9, it has been demonstrated that
these unknown underlying factors do not need to be assumed priori into the data
matrices.
10.4. Recommendations for Stormwater Quality Mitigation

Stormwater quality mitigation strategies are commonly implemented for a specific
design period which does not include future changes in urban traffic and climate
change. This in turn can lead to accelerated obsolescence of these systems which can
be costly to implement. The outcomes of the current study provide guidance on the
development of measures for adaptation under changed urban traffic and climate
change conditions. The following recommendations have been made in this context:
• The object classification system discussed in Section 10.2 should be
incorporated into the design process of a stormwater quality management
strategy. This will ensure the identification of possible changes in urban
258
traffic and changes to rainfall characteristics due to climate change at the
planning phase of any mitigation strategy.
• At the implementation phase, appropriate measures should be identified to
target solids (TSS) of 1-75 µm particle size to remove the traffic generated
Cd, Cr, Cu, Ni, Zn and Pb from road surfaces. Additionally, organic matter
from <1 - >300 µm size fraction should be targeted for the removal of Cd,
Cr, Pb and Ni from wash-off whilst Cu and Zn need to be removed as free
ions in wash-off.
• Appropriate technologies for the removal of volatile organic compounds
(VOC) from build-up should target total organic carbon (TOC) of 1-150 µm
fraction to remove toluene, ethylbenzene, ortho, meta and para-xylene. In
wash-off, TOC <1 - >300 µm size fraction should be targeted for the
removal of these VOCs.
• Appropriate technologies for the removal of semi and non volatile organic
compounds (SVOC and NVOC) from build-up should target particulate
fractions of 75-300 µm size. In wash-off, mitigation strategies should target
1-300 µm particulate fractions for the removal of SVOC and NVOC from
wash-off.
10.5. Recommendations for Further Research

The future research directions based on this study can be three fold, namely
theoretical, experimental and computer modelling. The theoretical research can
extend the object classification system incorporating more complex dynamic
systems such as the impacts of combined urban rail and traffic network due to
259
increased urbanisation on water quality. Additionally, this may also extend to the
inclusion of the impacts of temperature, bushfire, flash-flooding and greenhouse gas
emissions due to climate change on water quality. The influence of source
identification and apportionment of traffic generated pollutants on the object
classification system can be investigated and the prediction frameworks can be
further improved based on such enhanced classifications.
The experimental research can be directed into setting up pilot studies for the
removal of heavy metals and petroleum hydrocarbons from urban roads based on the
build-up and wash-off processes discussed in this investigation. The pilot studies can
also be extended to semi-urban and rural roads. Different types of pavements such
as, concrete pavements, gravel pavements, walkways and bridges can be included in
the build-up and wash-off studies. In this context, synthetic pavements of different
textures can be fabricated in the laboratory and placed on the actual roads to be
collected at a later time. Seasonal effects on the build-up of traffic generated
pollutants can be investigated by extending the sample collection to different
seasons. These experimental set-ups can also target speciation of different heavy
metals on the pavements during build-up and wash-off which can help in the
prioritisation of the removal of heavy metals under different dynamic situations.
Finally, the predictive frameworks proposed in this study can be performed through
computer modelling and decision support systems. Large datasets will need to be
generated for this purpose and experimental pilot studies can provide such datasets.
260
CONSOLIDATED LIST OF REFERENCES
(1) Abbs, D., McInnes, K., Rafter, T. (2007). The impact of climate change on
extreme rainfall and coastal sea levels over South-East Queensland, Part 2: A
high-resolution modelling study of the effect of climate change on the
intensity of extreme rainfall events. CSIRO Division of Marine and
Atmospheric Research, Aspendale, Victoria, Australia: pp. 35.
(2) Abbs, D., Rafter, T. (2008). The effect of climate change on extreme rainfall
events in the Westernport region. CSIRO Division of Marine and
Atmospheric Research, Aspendale, Victoria, Australia: pp. 32.
(3) ABS (2010). Australian Bereau of Statistics. Catalogue No. 1318.3 - Qld
Stats,Feb2010.http://www.abs.gov.au/AUSSTATS/abs@.nsf/Lookup/1318.3
Main%20Features3Feb%202010. Access date 10 May 2010.
(4) Adachi, K., Tainosho, Y. (2004). Characterization of heavy metal particles
embedded in tire dust. Environment International, 30, 1009-1017.
(5) Adams, B. J., Papa, F. (2000). Runoff quality data analysis: Urban
stormwater management planning with analytical probabilistic models, John
Wiley & Sons, Inc. pp. 81-111.
(6) Adams, M. J. 1995. Chemometrics in analytical spectroscopy. Springer
Verlag, New York.
(7) Ahyerre, M., Chebbo, G., Tassin, B., Gaume, E. (1998). Storm water quality
modelling, an ambitious objective? Water Science and Technology, 37, 205-
213.
(8) Andreou,G., Rapsomanikis, S. (2009). Origins of n-alkanes, carbonyl
compounds and molecular biomarkers in atmospheric fine and coarse
261
particles of Athens, Greece. Science of the Total Environment, 407 (21)
5750-5760.
(9) APHA (2005). Standard methods for the examination of water and
wastewater, 21st Edition, Eaton, A. D., Clesceri, L. S., Rice, E. W.,
Greenberg, A. E., (Eds.), American Public Health Association, Washington
DC.
(10) API (1994). Interlaboratory study of three methods for analysing petroleum
hydrocarbons in soils. API publication no. 4599, American Petroleum
Institute, Washington DC 1994; www.api.org/Publications. Accessed 10
September 2010.
(11) ATSDR (2005). Agency for toxic substance and disease registry,
toxicological profile for Lead. U.S. Department of Health and Humans
Services, Public Health Service, Atlanta, GA.
(12) ATSDR (2010). Agency for toxic substances and disease registry. Total
Petroleum Hydrocarbons (TPH): Public Health Statement.
http://www.atsdr.cdc.gov/
(13) AUSIFD (2005). Design rainfall analysis made easy with AUS-IFD Version
2.0, © 2005 Dr G. A. Jenkins.
http://www.ens.gu.edu.au/eve/Research/AusIfd/AusIfdVer2.htm
(14) Ayoko, G. A. (2004). Volatile Organic Compounds in Indoor Environments,
in: Pluschke, P. (Ed.), Indoor Air Pollution: The Handbook of Environmental
Chemistry. Springer, pp. 1-35.
(15) Ayoko, G. A., Morawska, L., Kokot, S., Gilbert, D. (2004). Application of
multicriteria decision making methods to air quality in the
262
microenvironments of residential houses in Brisbane, Australia.
Environmental Science and Technology, 38, 2609-2616.
(16) Ayoko, G. A., Singh, K., Balerea, S. and Kokot, S. (2007). Exploratory
multivariate modeling and prediction of the physico-chemical properties of
surface water and groundwater. Journal of Hydrology, 336, 115-124.
(17) Baffaut, C., Delleur, J. W. (1990). Calibration of SWMM runoff quality
model with expert system. Journal of Water Resources Planning and
Management, 116, 247-261.
(18) Bao, L., Chan, S., Wu, L., Hei, T. K., Wu, Y., Yu, Z., Xu, A. (2007).
Mutagenicity of diesel exhaust particles mediated by cell-particle interaction
in mammalian cells. Toxicology, 229, 91-100.
(19) Barbe, D. E., Cruise, J. F., Mo, X. (1996). Modeling the build-up and wash-
off of pollutants on urban watersheds. Water Resources Bulletin, 32, 511-
519.
(20) Basher, R. E., Pittock, A.B., Bates, B., Done, T., Gifford, R. M., Howden, S.
M., Sutherst, R., Warrick, R., Whetton, P., Whitehead, D., Williams, J.E.,
Woodward, A. (1998). The regional impacts of climate change: An
assessment of vulnerability. R. T. Watson, Zinyowera, M. C., Moss, R. H.,
Dokken, D. J. (Eds). Cambridge, United Kingdom and New York, USA,
Cambridge University Press: pp. 878.
(21) Beebe, K. R., Kowalski, B. R. (1987). An introduction to multivariate
calibration and analysis. Analytical Chemistry, 59, 1007A-1017A.
(22) Bencs, L., Naveau, I., Roekens, E., Grieken, R. V. (2009). Atmospheric
BTEX- concentrations in an area with intensive street traffic. Atmospheric
Environment, 43, 311-318.
263
(23) Benson, N. U., Essien, J. P., Ebong, G. A., Williams, A. B. (2008).
Petroleum hydrocarbons and limiting nutrients in Macura reptantia,
Procambarus clarkii and benthic sediment from Qua Iboe Estuary, Nigeria.
The Environmentalist, 28, 275-282.
(24) Bertrand-Krajewski, J.-L. (2006). Influence on field data sets on calibration
and verification on stormwater pollutant models. Proceedings from the 7 th
International Conference on Urban Drainage Modeling (UDM), Melbourne,
Australia, April 2-7.
(25) Bezdek, J. C. (1982). Pattern recognition with fuzzy objective function
algorithms. Plenum Press, New York.
(26) Birch, G. F., McCready, S. (2009). Catchment condition as a major control
on the quality of receiving basin sediments (Sydney harbour, Australia).
Science of the Total Environment, 407, 2820-2835.
(27) BITRE (2008). Australian transport statistics yearbook 2007. Canberra,
ACT, Bureau of Infrastructure, Transport, and Regional Economics, pp. 167.
(28) Bradford, W. L. (1977). Urban stormwater pollutant loadings: A statistical
summary through 1972. Journal Water Pollution Control Federation, 49,
613-622.
(29) Brandenberger, S., Mohr, M., Grob, K., Neukom, H. P. (2005). Contribution
of unburned lubricating oil and diesel fuel to particulate emission from
passenger cars. Atmospheric Environment, 39, 6985-6994.
(30) Brans, J., Vincke, P. and Mareschal, B. (1986). How to select and how to
rank projects: the PROMETHEE method. European Journal of Operational
Research, 24, 228-238.
264
(31) Brown, A. L., Affum, J. K, Tomerini, D. (1998). TRAEMS:The Transport
Planning Add-on Environmental Modelling System. Proceedings of the 19th
ARRB Transport Research Conference.
(32) Brown, L., Affum, J., Chan, A. (2004). Transport pollution futures for Gold
Coast city 2000, 2011, 2021, based on the Griffith University Transport
Polution Modelling System (TRAEMS) Brisbane, QLD, Urban Policy
Program, Griffith University: pp. 75.
(33) Brown, R. C., Pierce, R. H., Rice, S. A. (1985). Hydrocarbon contamination
in sediments from urban stormwater runoff. Marine Pollution Bulletin, 16,
236-240.
(34) Bubenzer, G.D. (1979). Inventory of rainfall simulators. Proceedings of the
Rainfall Simulator Workshop. Tucson: U.S. Department of Agriculture, pp.
120-130.
(35) Buczynska, A. J., Krata, A., Stranger, M., Godoi, A. F. L., Kontozova-
Deutsch, V., Bencs, L., Naveau, I., Roekens, E., Grieken, R. V. (2009).
Atmospheric BTEX-concentrations in an area with intensive street traffic.
Atmospheric Environment, 43, 311-318.
(36) Bujon , G., Herremans, L., Phan, L. (1992). Flupol: A forecasting model for
flow and pollutant discharge from sewerage systems during rainfall events.
Water Science and Technology, 25, 207-215.
(37) Burgherr, P. (2007). In-depth analysis of accidental oil spills from tankers in
the context of global spill trends from all sources. Journal of Hazardous
Materials, 140, 245-256.
(38) Burnley, I., Murphy, P. (2003). Seachange: Movement from metropolitan to
arcadian Australia. UNSW Press, Sydney.
265
(39) Cadle, S.H., Mulawa, P.A., Hunsanger, E.C. (1999). Composition of light
duty motor vehicle exhaust particulate matter in the Denver, Colorado area.
Environmental Science and Technology, 33, 2328-2339.
(40) Caselli, M., Gennaro, G. D., Marzocca, A., Trizio, L., Tutino, M. (2010).
Assessment of the impact of the vehicular traffic on BTEX concentration in
ring roads in urban areas of Bari (Italy). Chemosphere, 81, 306-311.
(41) Chambers, P.A., Guy, M., Roberts, E.S., Charlton, M.N., Kent, R., Gagnon,
C., Grove, G., Foster, N. (2001). Nutrients and their impact on the canadian
environment. Agriculture and Agri-Food Canada, Environment Canada,
Fisheries and Oceans Canada, Health Canada and Natural Resources Canada,
Ottawa, ON.
(42) Chang, J. I., Lin, C. (2006). A study of storage tank accidents. Journal of
Loss Prevention in the Process Industries, 19, 51-59.
(43) Charbeneau, R. J., Barrett, M. E. (1998). Evaluation of methods for
estimating stormwater pollutant loads. Water Environment Research, 70,
1295-1302.
(44) Charlesworth, S. M., Lees, J. A. (1999). Particulate associated heavy metals
in the urban environment: their transport from source to deposit, Coventry,
UK. Chemosphere, 39, 833-848.
(45) Chen, J., Adams, B. J. (2007). A derived probability distribution approach to
stormwater quality modeling. Advances in Water Resources, 30, 80-100.
(46) Chen, X., Xia, X., Zhao, Y., Zhang, P. (2010). Heavy metal concentrations in
roadside soils and correlation with urban traffic in Beijing, China. Journal of
Hazardous Materials, 181, 640-646.
266
(47) Chiew, F. H. S., McMahon, T. A. (2002). Modelling the impacts of climate
change on Australian streamflow. Hydrological Processes, 16, 1235-1245.
(48) Christensen, E. R., Arora, S. (2007). Source apportionment of PAHs in
sediments using factor analysis by time records: Application to Lake
Michigan, USA. Water Research, 41, 168-176.
(49) Christensen, W. F., Amemiya, Y. (2003). Modeling and prediction for
multivariate spatial factor analysis. Journal of Statistical Planning and
Inference, 115, 543-564.
(50) Chrysikopoulos,C. V., Hildemann,L. M., Roberts, P. V. (1992). Modeling
the emission and dispersion of volatile organics from surface aeration
wastewater treatment facilities. Water Research, 26, 1045-1052.
(51) Chui, T. W., Mar, B. W., Horner, R. R. (1982). Pollutant loading model for
highway runoff. Journal of Environmental Engineering, 108, 1193-1210.
(52) Clarke, A. J. (2008). An introduction to the dynamics of El-Nino and the
southern oscillation. Academic Press, pp. 308.
(53) Cortazar, E., Bartolomé, L., Arrasate, S., Usobiaga, A., Raposo, J. C.,
Zuloaga, O., Etxebarria, N. (2008). Distribution and bioaccumulation of
PAHs in the UNESCO protected natural reserve of Urdaibai, Bay of Biscay.
Chemosphere, 72, 1467-1474.
(54) CSIRO (2003). Climate change: An Australian guide to the science and
potential impacts. B. Pittock. Canberra, ACT, The Australian Greenhouse
Office: pp. 238.
(55) CSIRO (2007). Climate change in Australia. Technical Report 2007 Bureau
of Meteorology, Australia. www.climatechangeinaustralia.gov.au. Accessed
12 September 2010..
267
(56) Davis, A. P., Shokouhian, M., Ni, S. (2001). Loading estimates of lead,
copper, cadmium, and zinc in urban runoff from specific sources.
Chemosphere, 44, 997-1009.
(57) Davis, B., Birch, G. (2010). Comparison of heavy metal loads in stormwater
runoff from major and minor urban roads using pollutant yield rating curves.
Environmental Pollution, 158, 2541-2545.
(58) De jong, R., Drury, C. F., Yang, J. Y., Campbell, C.A. (2009). Risk of water
contamination by nitrogen in Canada as estimated by the IROWC-N model.
Journal of Environmental Management, 90, 3169-3181.
(59) Decembrini, F., Azzaro, F., Crisafi, E. (1995). Distribution of Chemical
Polluting Factors in South Italian Seas along Calabria Coastal Waters (Low
Tyrrhenian Sea, High Ionian Sea and Straits of Messina). Water Science and
Technology, 32, 231-237.
(60) Decision (2000). Visual Decision Inc., Getting Started Guide, DecisionLab
2000, Version 1.01.0388, Copyright© 1998 – 2000, Visual Decision Inc.,
Montreal, Quebec, Canada.
(61) Deletic, A. (1998). The first flush load of urban surface runoff. Water
Research, 32, 2462-2470.
(62) Deletic, A. B., Maksimovic, C. T. (1998). Evaluation of water quality factors
in storm runoff from paved areas. Journal of Environmental Engineering,
124, 869-879.
(63) Deletic, A., Orr, D., W. (2005). Pollution buildup on road surfaces. Journal
of Environmental Engineering, 131, 49-59.
268
(64) Delpla, I., Jung, A. -V., Baures, E., Clement, M., Thomas, O. (2009).
Impacts of climate change on surface water quality in relation to drinking
water production. Environment International, 35, 1225-1233.
(65) Deming, S, N., Morgan, S. L. (1993). Experimental Design: A Chemometric
Approach. 2nd Edition, Elsevier Science Publishers, Amsterdam, pp. 262-
264.
(66) Dickens, A. F., Gélinas, Y., Masiello, C. A., Wakeham, S., Hedges, J. H.
(2004). Reburial of fossil organic carbon in marine sediments. Nature, 427,
336-339.
(67) DNR (1995). Methods for Determining Gasoline Range Organics. Wisconsin
DNR, Publication no. PUBL-SW-140.
(68) Draper, W.M., Dhaliwal, J. S., Perera, S., K., Bauman, F., J. (1996).
Determination of diesel fuel and motor oil in water and wastes by a modified
diesel-range organics total petroleum hydrocarbon method. Journal of AOAC
International, 79, 508-519.
(69) Drapper, D., Tomlinson, R., Williams, P. (2000). Pollutant concentration in
road runoff: SouthEast Queensland case study. Journal of Environmental
Engineering, 126, 313-320.
(70) Driver, N. E., Troutman, B. M. (1989). Regression models for estimating
urban storm-runoff quality and quantity in the United States. Journal of
Hydrology, 109, 221-236.
(71) Duncan, H. P. (1995). A review of urban stormwater quality processes.
Cooperative Research Centre for Catchment Hydrology: Report no 95/9.
269
(72) Duncan, M.J. (1972). The performance of a rainfall simulator and an
investigation of plot hydrology. MAgrSc Thesis, University of Canterbury,
New Zealand.
(73) EEA (2003). Mapping the impacts of recent natural disasters and
technological accidents in Europe. E. E. Agency. Copenhagen, Denmark.
(74) Egodawatta, P. K. (2007). Translation of small-plot scale pollutant build-up
and wash-off measurements to urban catchment scale. Queensland
University of Technology, Brisbane. PhD Thesis.
(75) Egodawatta, P., Thomas, E., Goonetilleke, A. (2007). Mathematical
interpretation of pollutant wash-Off from urban road surfaces using
simulated rainfall. Water Research, 41, 3025-3031.
(76) Egodawatta, P., Thomas, E., Goonetilleke, A. (2009). Understanding the
physical processes of pollutant build-up and wash-off on roof surfaces.
(77) Ellis, J. B., Revitt, D. M. (1982). Incidence of heavy metals in street surface
sediments: solubility and grainsize studies. Water Air Soil Pollution, 17, 87-
100.
(78) Ellis, J.B., Mitchell, G. (2006). Urban diffuse pollution: key data information
approaches for the Water Framework Directive. Water and Environment
Journal, 20, 19-26.
(79) EPA (2008). Test methods for evaluating solid waste, Physical/Chemical
Methods: SW-846. 3rd Edition. United States Environmental protection
Agency.
http://www.epa.gov/epawaste/hazard/testmethods/sw846/online/index.htm.
270
(80) EPA (2010). Water quality criteria for nitrogen and phosphorus pollution.
United States Environmental Protection Agency,
http://water.epa.gov/scitech/swguidance/waterquality/sandards/criteria/
(81) EPA(1994). Determination of trace elements in waters and wastes by
inductively coupled plasma – mass spectrometry. US Environmental
Protection Agency, Method 200.8.
(82) Eriksson, E., Baun, A., Mikkelsen, P. S., Ledin, A. (2005). Chemical hazard
identification and assessment tool for evaluation of stormwater priority
pollutants. Water Science & Technology, 51, 47-55.
(83) FHWA (2005). Federal Highway Administration Technical Advisory Report
(84) Filella, I., Peñuelas, J. (2006). Daily, weekly, and seasonal time courses of
VOC concentrations in a semi-urban area near Barcelona. Atmospheric
Environment, 40, 7752-7769.
(85) Floyd, C.N. (1981). Mobile rainfall simulator for small plot field
experiments. Journal of Agricultural Engineering Research, 26, 307.
(86) Fujiwara, F., Rebagliati, R. J., Marrero, J., Gómez, D., Smichowski, P.
(2010). Antimony as a traffic-related element in size-fractionated road dust
samples collected in Buenos Aires. Microchemical Journal,
doi:10.1016/j.microc.2010.05.006.
(87) Garcia-Ochoaa, F., Gomeza, E., Santosa, V. E., Jose C. Merchuk, J. C.
(2010). Oxygen uptake rate in microbial processes: an overview.
Biochemical Engineering Journal, 49, 289-307.
271
(88) Gardiner, L. R., Armstrong, W. (2007). Identifying sensitive receiving
environments at risk from road runoff. Land Transport New Zealand:
Research report No 315. pp. 68.
(89) GCCC (2001). The Gold Coast '2010' cities for climate protection program
2001 action plan. Gold Coast, QLD, Gold Coast City Council, pp. 41.
(90) GCCC (2007). Carbon neutral by 2020: Gold Coast city council responding
to climate change, Gold Coast City Council, pp.17.
(91) GCCC. (2006). Gold Coast Priority Infrastructure Plan- Derivation of Trunk
Road Infrastructure Network Charges: FinalReport. Veitch Lister
Consulting,pp:8.http://www.goldcoast.qld.gov.au/gcplanningscheme_0509/at
tachments/planning_scheme_documents/part8_infrastructure/reference_docu
mentation.pdf . Accessed 2 September 2010.
(92) Gilli, G., Scursatone, E., Bono R., Palin, L. (1989). Use of leaded gasoline
and volatile halogenated hydrocarbon emission from automotive exhaust.
(93) Goonetilleke, A., Egodawatta, P., Kitchen, B. (2009). Evaluation of pollutant
build-up and wash-off from selected land uses at the port of Brisbane,
Australia. Marine Pollution Bulletin, 58, 213-221.
(94) Goonetilleke, A., Thomas, E. (2003). Water quality impacts of urbanisation,
Queensland University of Technology: pp.93.
(95) Goonetilleke, A., Thomas, E., Ginn, S., Gilbert, D. (2005). Understanding
the role of land use in urban stormwater quality management. Journal of
Environmental Management, 74, 31-42.
272
(96) Gouider, M., Bouzid, J., Sayadi, S., Montiel, A. (2009). Impact of
orthophosphate addition on biofilm development in drinking water
distribution systems. Journal of Hazardous Materials, 167, 1198-1202.
(97) Goyer, R.A., Clarsksom, W.T. (2001). Toxic effects of metals, in Casarett
and Doull’s Toxicology: The Basic Science of Poisons. Klaassen, C.D.
(Ed.), McGraw-Hill, New York, pp. 811-867.
(98) Grieshop, A. P., Lipsky, E. M., Pekney, N. J., Takahama, S., Robinson, A. L.
(2006). Fine particle emission factors from vehicles in a highway tunnel:
effects of fleet composition and season. Atmospheric. Environment, 40,
S287-S298.
(99) Gromaire-Mertz, M. C., Garnaud, S., Gonzalez, A., Chebbo, G. (1999).
Characterisation of urban runoff pollution in Paris. Water Science and
(100) Grottker, M. (1987). Runoff quality from a street with medium traffic
loading. The Science of the Total Environment, 59, 457-466.
(101) Gupta, K., Saul, A. J. (1996). Specific relationships for the first flush load in
combined sewer flows. Water Research, 30, 1244-1252.
(102) Gupta, M. K., Agnew, R. W., Gruber, D., Kreutzberger, W. (1981).
Constituents of highway runoff Vol. 4: characteristics of runoff from
operating highways. Federal Highway Administration, Washington, D. C.:
FHWD-RD-81-045.
(103) Gurran, N., Squires, C., Blakley, E. (2006). Meeting the sea change
challenge: Best Practice Models of Local and Regional Planning for Sea
Change Communities in Coastal Australia. Report No. 2 for the National
273
Seachange Taskforce. The University of Sydney Planning Research Centre,
Sydney.
(104) Haiping, Z., Yamada, K. (1998). Simulation of nonpoint source pollutant
loadings from urban area during rainfall: an application of a physically-based
distributed model. Water Science and Technology, 38, 199-206.
(105) Hall, M. J., Ellis, J. B. (1985). Water quality problems of urban areas.
GeoJournal, 11, 265-275.
(106) Harrison, R. M. (1979). Toxic metals in street and household dusts. Science
of the Total Environment, 11, 89-97.
(107) Harrison, R. M., Wilson, S. J. (1985). The chemical composition of high-
way drainage waters. The Science of the Total Environment, 43, 63-77.
(108) Heinonen-Tanski, H., Wijk-Sijbesma, C. V. (2005). Human excreta for plant
production. Bioresource Technology, 96, 403-411.
(109) Hering, J. G., Morel, F. M. M. (1988). Humic acid complexation of calcium
and copper. Environmental Science & technology, 22, 1234-1237.
(110) Herngren, L. (2005). Build-up and wash-off process kinetics of PAHs and
heavy metals on paved surfaces using simulated rainfall. Queensland
University of Technology, Brisbane. PhD Thesis.
(111) Herngren, L., Goonetilleke, A., Ayoko, G. A. (2005a). Understanding heavy
metal and suspended solids relationships in urban stormwater using
simulated rainfall. Journal of Environmental Management, 76, 149-158.
(112) Herngren, L., Goonetilleke, A., Sukpum, R., De Silva, D. Y. (2005b).
Rainfall simulation as a tool for urban water quality research. Environmental
Engineering Science, 22, 378-383.
274
(113) Herngren, L., Goonetilleke, A., Ayoko, G. A. (2006). Analysis of heavy
metals in road-deposited sediments. Analytica Chimica Acta, 571, 270-278.
(114) Hoffman, E. J., Latimer, J. S., Mills, G. L., Quinn, J. G. (1982). Petroleum
hydrocarbons in urban runoff from a commercial land use area. Journal
Water Pollution Control Federation, 54, 1517-25.
(115) Hoffman, E. J., Mills, G. L., Latimer, J. S., Quinn, J. G. (1984). Urban runoff
as a source of polycyclic aromatic hydrocarbons to coastal waters.
Environmental Science & Technology, 18, 580-587.
(116) Hoffman, E. J., Mills, G. L., Latimer, J. S., Quinn, J. G. (1985). Stormwater
runoff from highways. Water Air Soil Pollution, 25, 349-364.
(117) Hogland, W., Berndtsson, R., Larson, M. (1984). Estimation of quality and
pollution load of combined sewer overflow discharge. Proceedings of 3rd
International Conference on Urban Storm Drainage, Göteborg, Sweden, pp.
841-850.
(118) Hojae, S., Wei, M., Aijun, L., Kaicho, C. (2009). Bio-removal of mixture of
benzene, toluene, ethylbenzene, and xylenes/total petroleum
hydrocarbons/trichloroethylene from contaminated water. Journal of
Environmental Sciences, 21, 758-763.
(119) Horwitz, W. (1982). Evaluation of analytical methods used for regulation of
foods and drugs. Analytical Chemistry, 54, 67A-76A.
(120) Huber, W. C. (1986). Deterministic modeling of urban runoff quality. In
Urban runoff pollution. Torno, H. C., Desbordes, M., Marsalek, J. (Eds.),
NATO ASI series, Springer-Verlag, Berlin, New York; pp. 167-242.
(121) Hung, C., Gong, G., Chen, H., Hsieh, H., Santschi, P. H., Wade, T. L.,
Sericano, J, L. (2007). Relationships between pesticides and organic carbon
275
fractions in sediments of the Danshui River estuary and adjacent coastal
areas of Taiwan. Environmental Pollution, 148, 546-554.
(122) Hunter, J. V., Sabatino, T., Gomperts, R., MacKenzie, M. J. (1979).
Contribution of urban runoff to hydrocarbon pollution. Journal Water
Pollution Control Federation, 51, 2129-38.
(123) Hutcheson, M. S., Pedersen, D., Anastas, N. D., Fitgerald, J., Silverman, D.
(1996). Beyond TPH: health-based evaluation of petroleum hydrocarbon
exposures. Regulatory Toxicology & Pharmacology. 24, 85-101.
(124) Hvitved-Jacobsen, T.; Vollertsen, J.; Nielsen, A. H. (2010). Urban and
highway stormwater pollution: concepts and engineering. CRC Press,
Florida.
(125) Hwang, D. F.; Lu, Y. H. (2000). Influence of environmental and nutritional
factors on growth, toxicity, and toxin profile of dinoflagellate Alexandrium
minutum. Toxicon, 38, 1491- 1503.
(126) Hwang, H. -M., Foster, G. D. (2006). Characterization of polycyclic
aromatic hydrocarbons in urban stormwater runoff flowing into the tidal
Anacostia River, Washington, DC, USA. Environmental Pollution, 140, 416-
426.
(127) IARC (2009). IARC Monographs on the Evaluation of Carcinogenic Risks to
Humans. http://monographs.iarc.fr/ENG/Classification/crthgr01.php. Access
date 20 March 2010.
(128) IEAUST (2001). Australian rainfall and runoff: A guide to flood estimation.
Reprinted edition 2001, Institute of Engineers Australia, Barton, ACT.
(129) IPCC (2000). Special report on emissions scenarios. A special report of
Working Group III of the Intergovernmental Panel on Climate Change.
276
Nakićenović, N., Swart, R. (Eds.). Inter-Governmental Panel on Climate
Change, Cambridge University Press, UK. 570pp.
(130) IPCC (2001). Climate change 2001: the scientific basis. Houghton, J.T.,
Ding, Y., Griggs, D.J., Noguer, M., Van der Linden, P. J., Xiaosu, D.
(Eds.). Inter-Governmental Panel on Climate Change, Cambridge University
Press, UK: pp. 944.
(131) IPCC (2007). Climate change 2007: the physical science basis. Solomon, S.,
Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K.B., Tignor, M.,
Miller, H.L (Eds.). Inter-Governmental Panel on Climate Change,
Cambridge, United Kingdom and New York, USA, Cambridge University
Press: pp. 996.
(132) Irish, L. B., Barrett, M. E., Malina, J. F., Charbeneau, R. J. (1998). Use of
regression models for analyzing highway storm-water loads. Journal of
Environmental Engineering, 124, 987-993.
(133) Jackson, D. A. (1993). Stopping rules in principal component analysis: A
comparison of heuristical and statistical approach. Ecology, 74, 2204-2214.
(134) Jartun, M., Ottesen, R. T., Steinnes, E., Volden, T. (2008). Runoff of particle
bound pollutants from urban impervious surfaces studied by analysis of
sediments from stormwater traps. Science of the Total Environment, 396,
147-163.
(135) Jo, M., Rene, E. R., Kim, S., Park, H. (2008). Removal of BTEX compounds
by industrial sludge microbes in batch systems: statistical analysis of main
and interaction effects. World Journal of Microbiology and Biotechnology,
24, 73-78.
277
(136) Johansson, C., Norman, M., Burman, L. (2009). Road traffic emission
factors for heavy metals. Atmospheric Environment, 43, 4681-4688.
(137) Johnson, T. D., Belitz, K. (2009). Assigning land use to supply wells for the
statistical characterization of regional groundwater quality: Correlating urban
land use and VOC occurrence. Journal of Hydrology, 370, 100-108.
(138) Karlsson, K., Viklander, M., Scholes, L., Revitt, M. (2010). Heavy metal
concentrations and toxicity in water and sediment from stormwater ponds
and sedimentation tanks. Journal of Hazardous Material, 178, 612-618.
(139) Kay, P., Edwards, A. C., Foulger, M., (2009). A review of the efficacy of
contemporary agricultural stewardship measures for ameliorating water
pollution problems of key concern to the UK water industry. Agricultural
Systems, 99, 67-75.
(140) Kayhanian, M., Singh, A., Suverkropp, C., Borroum, S. (2003). Impact of
annual average daily traffic on highway runoff pollutant concentrations.
Journal of Environmental Engineering, 129, 975-990.
(141) Kayhanian, M., Suverkropp, C., Ruby, A., Tsay, K. (2007). Characterization
and prediction of highway runoff constituent event mean concentration.
(142) Keller, H. R., Massart, D. L., Brans, J. P. (1991). Multicriteria decision
making: A case study. Chemometrics and Intelligent laboratory Systems, 11,
175-189.
(143) Kimura, I. (1988). Aquatic pollutions problems in Japan. Aquatic
Toxicology, 11, 287- 301.
278
(144) Ko, J.-Y., Day, J. W. (2004). A review of ecological impacts of oil and gas
development on coastal ecosystems in the Mississippi Delta. Ocean and
Coastal Management, 47, 597-623.
(145) Kokot, S., Grigg, M., Panayiotuo, H., Dong Phuong, T. (1998). Data
interpretation by some common chemometrics methods. Electroanalysis, 10,
1081-1088.
(146) Kreider, M. L., Panko, J. M., McAtee, B.,L., Sweet, L. I., Finley, B. L.
(2010). Physical and chemical characterization of tire-related particles:
comparison of particles generated using different methodologies. Science of
the Total Environment, 408, 652-659.
(147) Kucklick, J. R., Siversten, S. K., Sanders, M., & Scott, G. I. (1997). Factors
influencing polycylic aromatic hydrocarbon distributions in South Carolina
estuarine sediments. Journal of Experimental Marine Biology and Ecology,
213, 13-29.
(148) Kudchadker, A. P., Zwolinski, B. J. (1966). Vapour pressures and boiling
points of normal alkanes, C21 to C100. Journal of Chemical and Engineering
Data, 11, 253-255.
(149) Kummer, U.; Pacyna, J.; Pacyna, E.; Friedrich, R. (2009). Assessment of
heavy metal releases from the use phase of road transport in Europe.
(150) Kuo, T.-T., Yan, Y-L., & Li, K-C. (1993). A simplified computer model for
nonpoint source pollution in a small urban area. Water Science &
(151) Lan, W. G., Wong, M. K., Chen, N., Sin, Y. M. (1994). Orthogonal array
design as a chemometric method for optimisation of analytical procedures
279
part2.* Four -level design and its application in microwave dissolution of
biological samples. Analyst, 119, 1669-1675.
(152) Laxen, D. P. H., Harrison, R. M. (1977). Highway as a source of water
pollution: an appraisal with the heavy metal lead. Water Research, 11, 1-11.
(153) Lee, B., Dong, T. T. T. (2010). Effects of road characteristics on distribution
and toxicity of polycyclic aromatic hydrocarbons in urban road dust of
Ulsan, Korea. Journal of Hazardous Materials, 175, 540-550.
(154) Legret, M., Colandini, V. (1999). Effects of a porous pavement with
reservoir structure on runoff water: water quality and fate of heavy metals.
Water Science & Technology, 39, 111-117.
(155) Lim, M. C. H., Ayoko, G. A., Morawska, L., Ristovski, Z. D., Jayaratne, E.
R., Kokot, S. (2006). A comparative study of the elemental composition of
the exhaust emissions of cars powered by liquefied petroleum gas and
unleaded petrol. Atmospheric Environment, 40, 3111-3122.
(156) Loch, R.J., Robotham, B.G., Zeller, L., Masterman, N., Orange, D.N.,
Bridge, B., Sheridan, G., Bourke, J.J. (2001). A multi-purpose rainfall
simulator for field infiltration and erosion studies. Australian Journal of Soil
Research, 39, 599.
(157) Mahbub, P., Ayoko, G. A., Goonetilleke, A., Egodawatta, P., Kokot, S.
(2010a). Impacts of traffic and rainfall characteristics on heavy metals build-
up and wash-off from urban roads. Environmental Science & Technology,
44, 8904-8910.
(158) Mahbub, P., Ayoko, G., Egodawatta, P., Yigitcanlar, T., Goonetilleke, A.
(2010b). Traffic and climate change impacts on water quality: measuring
build-up and wash-off of heavy metals and petroleum hydrocarbons. In
280
Rethinking Sustainable Development: Urban management, Engineering and
Design. Yigitcanlar, T., (Ed.). Engineering Science Reference, New York,
pp. 147-167.
(159) Mahbub, P., Goonetilleke, A., Ayoko, G. A. (2010c). A prediction model of
the build-up of volatile organic carbons on urban roads. Environmental
Science and Technology, (Under Review).
(160) Main Roads (2009). Austraods Test Manual, Method AGPT/T250,
http://www.austroads.com.au/pavement/testmethods.html, Accessed 05
January 2009.
(161) Makepeace, D. K., Smith, D.W., Stanley, S.J. (1995). Urban stormwater
quality: summary of contaminant data. Critical Reviews in Environmental
Science & Technology, 25, 93-139.
(162) Makropoulos, C. K., Natsis, K., Liu, S., Mittas, K., Butler, D. (2008).
Decision support for sustainable option selection in integrated urban water
management. Environmental Modelling & Software, 23, 1448-1460.
(163) Malvern (1994). Mastersizer S Long Bed Version 2.19 © Malvern
Instruments Ltd. 1992-1994.
(164) Manoli, E., Voutsa, D., Samara, C. (2002). Chemical characterization and
source identification/ apportionment of fine and coarse air particles in
Thessalonoki, Greece. Atmospheric Environment, 36, 949-961.
(165) Mareschal, B., Brans, J. P. (1988). Geometrical representations for MCDA.
European Journal of Operational Research, 34, 69-77.
(166) Marsalek, J., Brownlee, B., Mayer, T., Lawal, S., Larkin, G. A. (1997).
Heavy metals and PAHs in stormwater runoff from the skyway bridge,
281
Burlington, Ontario. Water Quality Research Journal of Canada, 32, 815-
827.
(167) Massart, D. L., Vandeginste, B. G. M., Buydens, L. M. C., De Jong, S.,
Lewi, P. J. Smeyers-Verbeke, J. (1997). Handbook of chemometrics and
qualimetrics Part A, Elsevier: 771-804.
(168) Méndez-Armenta, M., Ríos, C. (2007). Cadmium neurotoxicity,
Environmental Toxicology & Pharmacology, 23, 350-358.
(169) Meyers, L. S., Gamst, G., Guarino, A. J. (2006). Applied multivariate
research: design and interpretation. Sage Publications Inc. pp.481-482.
(170) Mohamed, M. F., Kang, D., Aneja, V. P. (2002). Volatile organic
compounds in some urban locations in United States. Chemosphere, 47, 863-
882.
(171) Moilleron, R., Gonzalez, A., Chebbo, G., Thévenot, D. R. (2002).
Determination of aliphatic hydrocarbons in urban runoff samples from the
‘‘Le Marais’’ experimental catchment in Paris centre. Water Research, 36,
1275-1285.
(172) Morgan, W. K. C., Reger, R. B., Tucker, D. M. (1997). Health effects of
diesel emissions. Annals of Occupational Hygiene, 41, 643-658.
(173) Morrison, R.T., Boyd, R. N. (1992). Organic Chemistry 6th edition.
Prentice-Hall Inc. pp.94-97.
(174) Murakami, M., Fujita, M., Furumai, H., Kasuga, I., Kurisu, F. (2009).
Sorption behavior of heavy metal species by soakaway sediment receiving
urban road runoff from residential and heavily trafficked areas. Journal of
Hazardous Material, 164, 707-712.
282
(175) Neeft, J. P. A., Makkee, M., Moulijn, J. A. (1996). Diesel particulate
emission control. Fuel Processing Technology, 47, 1-69.
(176) Ni, Y., Wang, L., Kokot, S. (2001). Simultaneous determination of
nitrobenzene and nitro-substituted phenols by differential pulse voltammetry
and chemometrics. Analytica Chimica Acta, 431, 101-113.
(177) Niemczynowicz, J. (1999). Urban hydrology and water management –
present and future challenges. Urban Water, 1, 1-14.
(178) Ning, Z., Cheung, C. S., Lu, Y., Liu, M. A., Hung, W. T. (2005).
Experimental and numerical study of the dispersion of motor vehicle
pollutants under idle condition. Atmospheric Environment, 39, 7880-7893.
(179) Nixon, H., Saphores, J. –D. (2007). Impacts of motor vehicle operation on
water quality in the US – Cleanup costs and policies. Transportation
Research Part D, 12 , 564-576.
(180) Novotny, V., Goodrich-Mahoney, J. (1978). Comparative assessment of
pollution loadings from non-point sources in urban land use. Progress in
Water Technology, 10, 775-785.
(181) Novotny, V., Olem, H. (1994). Water quality prevention, identification, and
management of diffuse pollution. New York, Van Nostrand Reinhold.
(182) Novotny, V., Sung, H. M., Bannerman, H., Baum, K. (1985). Estimating
nonpoint pollution from small urban watersheds. Journal Water Pollution
Control Federation, 57, 744-748.
(183) Ogden, K. W., Taylor, S. Y. (1999). Traffic engineering and management.
Institute of Transport Studies, Monash University, Australia. pp. 592- 594.
(184) Öko-Test (2002). Testbericht Bremsbeläge. Öko-Test Verlag, Frankfurt,
January 2002 (in German).
283
(185) Oliva, S. R., Espinosa, A. J. F. (2007). Monitoring of heavy metals in
topsoils, atmospheric particles and plant leaves to identify possible
contamination sources. Microchemical Journal, 86, 131-139.
(186) Olson, D. A., Hammond, D. M., Seila, R. L., Burke, J. M., Norris, G. A.
(2009). Spatial gradients and source apportionment of volatile organic
compounds near roadways. Atmospheric Environment, 43, 5647-5653.
(187) Otto, M. (1988). Fuzzy theory explained. Chemometrics and Intelligent
laboratory Systems, 4, 101-120.
(188) Otto, M., Wegscheider, W. (1985). Spectrophotometric multicomponent
analysis applied to trace metal determinations. Analytical Chemistry, 57, 63-
69.
(189) Overton, D. E., Meadows, M. E. (1976). Stormwater modeling. New York,
Academic Press.
(190) Patra, A., Colvile, R., Arnold, S., Bowen, E., Shallcross, D., Martin, D.,
Price, C., Tate, J., Apsimon, H., Robbins, A. (2008). On street observations
of particulate matter movement and dispersion due to traffic on an urban
road. Atmospheric Environment, 42, 3911-3926.
(191) Peña-Méndez, E. M., Astorga-España, M. S., García-Montelongo, F. J.
(1999). Interpretation of analytical data on n-alkanes and polynuclear
aromatic hydrocarbons in Arbacia lixula from the coasts of Tenerife (Canary
Islands, Spain) by multivariate data analysis. Chemosphere, 39, 2259-2270.
(192) Peterson, S. M., Batley, G. E. (1992). Road runoff and its impacts on the
aquatic environment: a review. CSIRO Investigation Report:
CET/LH/IR076.
284
(193) Pitt, R. (1979). Demonstration of nonpoint pollution abatement through
improved street cleaning practices. U.S. Environmental Protection Agency:
Report No. EPA/600/2-79-161.
(194) Pitt, R., Field, R., Lalor, M., Brown, M. (1995). Urban stormwater toxic
pollutants: assessment, sources, and treatability. Water Environment
Research, 67, 260-275.
(195) Prévôt, A. S. H., Dommen, J., Bäumle, M. (2000). Influence of road traffic
on volatile organic compound concentrations in and above a deep Alpine
valley. Atmospheric Environment, 34, 4719-4726.
(196) Richard, F.C., Bourg, A.C.M., (1991). Aqueous geochemistry of chromium:
a review. Water Research, 25, 807-816.
(197) Rogge, W.F., Hildemann, L.M., Mazurek, M.A., Cass, G.R., Simoneit,
B.R.T. (1993). Sources of fine organic aerosol. 3. Road dust, tire debris, and
organometallic brake lining dust: roads as sources and sinks. Environmental
Science & Technology, 27, 1892-1904.
(198) Sansalone, J. J., Buchberger, S. G. (1997a). Characterization of solid and
metal element distributions in urban highway stormwater. Water Science &
(199) Sansalone, J. J., Buchberger, S. G. (1997b). Partitioning and first flush of
metals in urban roadway storm water. Journal of Environmental Engineering
123, 134-143.
(200) Sansalone, J. J., Christina, C. M. (2004). First flush concepts for suspended
and dissolved solids in small impervious watersheds. Journal of
285
(201) Sartor, J. D., Boyd, G. B. (1972). Water pollution aspects of street surface
contaminants. Washington, D. C., U.S. Environmental Protection Agency:
pp. 170.
(202) Sartor, J. D., Boyd, G.B., Agardy, F.J. (1974). Water pollution aspects of
street surface contaminants. Water Pollution Control Federation, 46, 458-
467.
(203) Sawyer, C. N., McCarty, P.L., Parkin, G.F. (1994). Solids. Chemistry for
Environmental Engineering, McGraw-Hill: pp. 567-576.
(204) Schaarup-Jensen, K.; Hvitved-Jacobsen, T. (1991). Similation of Dissolved
Oxygen Depletion in Streams Receiving Combined Sewer Overflows. In
New Technologies in Urban Drainage, Maksimovic, C. (Ed), Elsevier
Applied Science, New York: 273 - 282.
(205) Senhorst, H. A. J., Zwolsman, J. J. G. (2005). Climate change and effects on
water quality: a first impression. Water Science & Technology, 51, 53-59.
(206) Sericano, J.L., Wade, T.L., Atlas, E.L., Brooks, J.M., (1990). Historical
perspective on the environmental bioavailability of DDT and its derivatives
to Gulf of Mexico oysters. Environmental Science & Technology, 24, 1541-
1548.
(207) Settle, S., Goonetilleke,A., Ayoko, G. A. (2007). Determination of surrogate
indicators for phosphorus and solids in urban stormwater: application of
multivariate data analysis techniques. Water Air Soil Pollution, 182, 149-
161.
(208) Shahin, M. Y. (2005). Pavement management for airports, roads, and parking
lots. 2nd Edition.Springer Science Inc. pp. 93-94.
286
(209) Shaw, S. B., Parlange, J.-Y., Lebowitz, M., Walter, M. T. (2009).
Accounting for surface roughness in a physically-based urban wash-off
model. Journal of Hydrology, 367, 79-85.
(210) Shaw, S. B., Walter, M. T., Steenhuis, T. S. (2006). A physical model of
particulate wash-off from rough impervious surfaces. Journal of Hydrology,
327, 618-626.
(211) Shepherd, K. A.; Ellis, P. A.; Rivett, M. O. (2006). Integrated understanding
of urban land, groundwater, baseflow and surface-water quality—The City of
Birmingham, UK. Science of the Total Environment, 360, 180-195.
(212) Shinya, M., Tsuchinaga, T., Kitano, M., Yamada, y., Ishikawa, M. (2000).
Characterization of heavy metals and polycyclic aromatic hydrocarbons in
urban highway runoff. Water Science & Technology, 42, 201-208.
(213) Shivalingaiah, B., James, W. (1984). Algorithms for build-up and wash-off
and routing pollutants in urban runoff. Proceedings of the Third International
Conference on Urban Storm Drainage, Chalmers University, Goteborg,
Sweden, June 1984, pp.1445-1456.
(214) Silburn, D.M., Hargreaves, P., Budd, N., and Glanville, S.G. (1996).
Endosulfan on cotton plants— persistence and wash-off during rain.
INTERSECT 96—International Symposium on Environmental Chemistry
and Toxicology. Sydney: RACI, Abstract O58.
(215) Sirius (2008). SIRIUS version 7.1© copyright 1987-2008, Pattern
Recognition Systems AS, http://www.prs.no, Help Topics, 2008.
(216) Sivakumar, T., Manavalan, R., Muralidharan, C., Valliappan, K. (2007).
Multi-criteria decision making approach and experimental design as
chemometric tools to optimize HPLC separation of domperidone and
287
pantoprazole. Journal of Pharmaceutical and Biomedical Analysis, 43, 1842-
1848.
(217) Smit, R., Brown, A. L., Chan, Y. C. (2008). Do air pollution emissions and
fuel consumption models for roadways include the effects of congestion in
the roadway traffic flow? Environmental Modelling & Software, 23, 1262-
1270.
(218) Smith, L., Mukerjee S., Gonzales M., Stallings C., Neas L., Norris, G.,
Özkaynak, H. (2006). Use of GIS and ancillary variables to predict volatile
organic compound and nitrogen dioxide levels at unmonitored locations.
(219) Soh, Y. C., Roddick, F., Leeuwen, J. V. (2008). The future of water in
Australia: the potential effects of climate change and ozone depletion on
Australian water quality, quantity and treatability. Environmentalist, 28, 158-
165.
(220) Soonthornnonda, P., Christensen, E. R., Liu, Y., Li, J. (2008). A wash-off
model for stormwater pollutants. Science of the Total Environment, 402,
248-256.
(221) Sotelo, M. F.; Tailer, R.; Vives, I.; Grimalt, J. O. (2008). Assessment of the
environmental and physiological processes determining the accumulation of
organochlorine compounds in European mountain lake fish through
multivariate analysis (PCA and PLS). Science of the Total Environment,
404, 148-161.
288
(223) Staehelin, J., Schläpfer, K., Bürgin, T., Steinemann, U., Schneider, S.,
Brunner, D., Bäumle, M., Meier, M., Zahner, C., Keiser, S., Stahel, W.,
Keller, C. (1995). Emission factors from road traffic from a tunnel study
(Gubrist tunnel, Switzerland). Part I: concept and first results. Science of the
Total Environment, 169, 141-147.
(224) Sternbeck, J., Sjödin, Å., Andréasson, K. (2002). Metal emissions from road
traffic and the influence of resuspension – results from two tunnel studies.
(225) Tarley, C. R. T., Silveira, G., dos Santos, W. N. L., Matos, G. D., da Silva,
E. G. P., Bezerra, M. A., Miró, M., Ferreira, S. L. C. (2009). Chemometric
tools in electroanalytical chemistry: Methods for optimization based on
factorial design and response surface methodology. Michrochemical Journal,
92, 58-67.
(226) Thurman, E. M. (1985). Organic Geochemistry of Natural Waters. Martinus
Nijhoff/ Dr W. Junk Publishers, Dodrecht.
(227) Tomanovic, A., Maksimovic, C. (1996). Improved modelling of suspended
solids discharge from asphalt surface during storm event. Water Science &
(228) Tomerini, D. M. (1997). Predicting the impacts of road transport on urban
water quality. Brisbane, Griffith University, Nathan. Honours Thesis.
(229) Tomerini, D. M., Brown, A. L. (1998). Predicting the impacts of road
transport on urban water quality. The International Conference on Integrated
Modelling of the Urban Environment, Sydney, Australia, July 1998.
289
(230) UNEG (1997). United Nations Environment Glossary: Glossary of
Environment Statistics. Series F Number 67. United Nations, New York,
1997. http://unstats.un.org/unsd/environmentgl/
(231) USEPA (1996). Indicators of the environmental impacts of transportation.
US Environmental Protection Agency, Publication number: EPA 230-R-96-
009, Washington, DC, 1996.
(232) Vaze, J., Chiew, F. H. S. (2002). Experimental study of pollutant
accumulation on an urban road surface. Urban Water, 4, 379-389.
(233) Von Uexküll, O., Skerfving, S., Doyle, R., Braungart, M. (2005). Antimony
in brake pads – a carcinogenic component? Journal of Cleaner Production,
13, 19-31.
(234) Walker, C. H., Hopkin, S. P., Sibly, R. M., Peakall, D. B. (2006). Principles
of ecotoxicolgy. CRC Press, Boca Raton, Florida.
(235) Wang, H., Liu, Q., Tu, Y. (2005). Interpretation of partial least-squares
regression models with VARIMAX rotation. Computational Statistics &
Data Analysis, 48, 207-219.
(236) Wang, P., Zhao, W. (2008). Assessment of ambient volatile organic
compounds (VOCs) near major roads in urban Nanjing, China. Atmospheric
Research, 89, 289-297.
(237) Wanielista, M. P., Yousef, Y. A. (1992). Stormwater management. New
York, Wiley.
(238) Watson, C.J., Billingsley, P., Croft, D.J., Huntsberger, D.V. (1993). Statistics
for Management and Economics. 5th Edition. ©1993, 1990, Allyn and
Bacon, Boston, USA, pp. 343-348.
290
(239) Whetton, P.H., McInnes, K.L., Jones, R. N., Hennessy, K.J., Suppiah, R.,
Page, C. M., Bathols, J., Durack, P. J. (2005). Australian climate change
projections for impact assessment and policy application: a review. Climate
Impact Group, CSIRO Marine and Atmospheric Research: Aspendale,
Victoria, Australia.
(240) Whipple, W., Hunter, J. V., Yu, S. L. (1974). Unrecorded pollution from
urban runoff. Journal Water Pollution Control Federation, 46, 873-885.
(241) Wold, S. (1978). Cross-Validatory Estimation of the Number of Components
in Factor and Principal Components Models. Technometrics, 20, 397.
(242) Woolgar, L. (2008). Assessing the increasing risk of marine oil pollution
spills in China. Conference Proceedings, International Oil Spill Conference,
Savannah, GA, USA, May 4–8, 2008, pp. 711-715.
(243) Wu, J. S., Allan, C. J., Saunders, W. L., Evett, J. B. (1998). Characterization
and pollutant loading estimation for highway runoff. Journal of
(244) Xun, L., Feng-Min, L., Da-Qian, L., Guo-Jun, S. (2010). Soil organic carbon,
carbon fractions and nutrients as affected by land use in semi-arid region of
Loess Plateau of China. Pedosphere, 20, 146-152.
(245) Yidana, M. S., Ophori, D., Yakubo, B. (2008). A multivariate statistical
analysis of surface water chemistry data—The Ankobra Basin, Ghana.
(246) Yuan, Y., Hall, K., Oldham, C. (2001). A preliminary model for predicting
heavy metal contaminant loading from an urban catchment. Science of the
Total Environment, 266, 299-307.
291
(247) Zhao, H., Li, X., Wang, X., Tian, D. (2010). Grain size distribution of road-
deposited sediment and its contribution to heavy metal pollution in urban
runoff in Beijing, China. Journal of Hazardous Materials, doi:
10.1016/j.jhazmat.2010.07.012.
(248) Zheng, N., Liu, J., Wang, Q., Liang, Z. (2010). Health risk assessment of
heavy metal exposure to street dust in the zinc smelting district, northeast of
China. Science of the Total Environment, 408, 726-733.
(249) Zheng, N., Liu, J., Wang, Q., Liang, Z. (2010). Health risk assessment of
heavy metal exposure to street dust in the zinc smelting district, northeast of
China. Science of the Total Environment, 408, 726-733.
(250) Zhu, X., Zhang, L., Che, X., Wang, L. (1999). The classification of
hydrocarbons with factor analysis and the PONA analysis of gasoline.
Chemometrics and Intelligent Laboratory Systems, 45, 147-155.
292
APPENDIX A.1
INTENSITY-FREQUENCY-DURATION TABLE
FOR STATION 40584
293
294
Table A.1.1 IFD data for rain gauge Station 40584 including (28.05°S, 153.29°E)
ARI 1* ARI 2 ARI 5 ARI 10 ARI 20 ARI 50 ARI 100
Duration
5 125 158 193 213 240 275 302
5.5 121 153 187 206 233 267 292
6 117 148 182 200 226 259 284
6.5 114 144 176 194 219 252 276
7 111 140 172 189 214 245 269
7.5 108 137 167 185 208 239 262
8 105 133 163 180 203 233 256
8.5 103 130 160 176 199 228 250
9 100 127 156 172 194 223 245
9.5 98 124 153 168 190 219 240
10 96 122 150 165 186 214 235
11 92 117 144 159 179 206 226
12 89 113 139 153 173 199 218
13 86 109 134 148 167 192 211
14 83 105 130 143 162 186 205
15 81 102 126 139 157 181 199
16 78 99 122 135 153 176 193
17 76 96 119 131 149 171 188
18 74 94 116 128 145 167 183
19 72 91 113 125 141 163 179
20 70 89 110 122 138 159 175
21 69 87 108 119 135 155 171
22 67 85 105 116 132 152 167
23 66 83 103 114 129 149 163
24 64 82 101 112 126 146 160
25 63 80 99 109 124 143 157
26 62 78 97 107 122 140 154
27 60 77 95 105 119 138 151
28 59 75 93 103 117 135 149
29 58 74 92 102 115 133 146
30 57 73 90 100 113 131 144
32 55 70 87 97 110 126 139
34 54 68 84 94 106 123 135
36 52 66 82 91 103 119 131
38 50 64 80 88 100 116 127
40 49.1 62 77 86 98 113 124
45 46 59 73 81 92 106 117
50 43.4 55 69 76 87 100 110
55 41.1 52 65 72 82 95 105
60 39.2 49.9 62 69 79 91 100
75 34.3 43.7 54 61 69 80 88
90 30.8 39.2 48.8 54 62 72 79
105 28 35.7 44.5 49.5 56 65 72
120 25.8 32.9 41.1 45.7 52 60 66
295
Table A.1.1 IFD data for rain gauge Station 40584 (Contd.)
ARI 1* ARI 2 ARI 5 ARI 10 ARI 20 ARI 50 ARI 100
Duration
135 24.1 30.6 38.2 42.5 48.4 56 62
150 22.5 28.7 35.8 39.9 45.4 53 58
165 21.3 27.1 33.8 37.6 42.8 49.6 55
180 20.2 25.7 32 35.7 40.6 47 52
195 19.2 24.4 30.5 34 38.7 44.8 49.4
210 18.3 23.4 29.2 32.4 36.9 42.8 47.2
225 17.6 22.4 27.9 31.1 35.4 41 45.3
240 16.9 21.5 26.9 29.9 34 39.4 43.5
270 15.7 20 25 27.8 31.7 36.7 40.5
300 14.7 18.8 23.4 26.1 29.7 34.4 38
360 13.2 16.8 20.9 23.3 26.6 30.8 34
420 12 15.3 19.1 21.2 24.2 28 30.9
480 11 14.1 17.6 19.6 22.3 25.8 28.5
540 10.3 13.1 16.3 18.2 20.7 24 26.5
600 9.62 12.3 15.3 17.1 19.5 22.6 24.9
660 9.07 11.6 14.5 16.1 18.4 21.3 23.5
720 8.6 11 13.7 15.3 17.4 20.2 22.3
840 7.83 10 12.5 14 16 18.6 20.5
960 7.22 9.23 11.6 13 14.8 17.2 19.1
1080 6.72 8.6 10.8 12.1 13.9 16.2 17.9
1200 6.3 8.07 10.2 11.4 13.1 15.2 16.9
1320 5.94 7.61 9.64 10.8 12.4 14.5 16
1440 5.63 7.22 9.16 10.3 11.8 13.8 15.3
1800 4.89 6.28 8.01 9.02 10.4 12.1 13.5
2160 4.35 5.6 7.17 8.08 9.3 10.9 12.1
2520 3.93 5.07 6.51 7.35 8.48 9.96 11.1
2880 3.6 4.64 5.97 6.76 7.8 9.18 10.2
3240 3.32 4.28 5.53 6.27 7.24 8.53 9.52
3600 3.08 3.98 5.15 5.85 6.76 7.98 8.91
3960 2.87 3.72 4.82 5.48 6.35 7.5 8.38
4320 2.69 3.49 4.54 5.16 5.98 7.07 7.91
* ARI 1 stands for average recurrence interval of 1 year and so on for other ARIs.
296
APPENDIX A.2 SUPPLEMENTARY
INFORMATION FOR CHAPTER 4
Manuscript Title
Impacts of traffic and rainfall characteristics on heavy
metals build-up and wash-off from urban roads
Parvez Mahbub, Godwin A. Ayoko, Ashantha Goonetilleke,
Prasanna Egodawatta, Serge Kokot
Supporting Information
Number of Pages: 19
Number of Figures: 3
Number of Tables: 6
297
298
Build-up Sample Collection
This study used ‘Wet and Dry Vacuum System’ (1) for the collection of build-up
samples from the urban road surfaces. A domestic vacuum cleaner with a water
filtration system was used to collect the road dust from a 2 x 1.5 m plot area in the
middle of the traffic lane in 25L plastic containers. Immediately afterwards de-
ionised water was sprayed at 2 bar pressure on the collection plots and vacuuming
was undertaken again to collect any remaining dust into the plastic containers. In
terms of collecting samples from a road surface subject to natural and traffic related
degradation, calibration studies confirmed that this method achieved a collection
efficiency of over 90%, same as described in earlier studies performed on synthetic
surfaces (2, 3). Samples were collected from the road surfaces after an antecedent
dry period of 7 days. This was in conformity with the findings of Egodawatta (4),
who noted that pollutant build-up on road surfaces asymptote to an almost constant
value after a seven day antecedent dry period. Homogeneous 500 mL subsamples
were transferred to high density 1 L polyethylene bottles using a churn splitter. The
particle size distributions of the suspended solids in the subsamples were obtained
using a Malvern Mastersizer S Particle Size Analyser capable of analysing particles
between 0.05 and 900 µm. Based on the particle size distributions, the total
particulate analytes were fractioned into four size ranges, namely, 300 µm, 150-299
µm, 75-149 µm, 1-74 µm using wet sieving. The filtrate that passed through a 1 µm
membrane filter was considered as the potential total dissolved fraction.
Wash-off Sample Collection
The study used a specially designed rainfall simulator to replicate the design rainfall
events common to the study region. The rainfall simulator was based on the design
299
of simulators used in agricultural research as described by Floyd (5) and Silburn et
al. (6). It consisted of an A-frame structure made of aluminium tubing of 40-mm
diameter. Three Veejet 80100 nozzles, spaced 1 m apart, were mounted on a
stainless steel boom at a height of 2.4 m to obtain rainfall drop size and terminal
velocity similar to natural rainfall. Further details on the design of the rainfall
simulator can be found in Herngren et al. (7).
Based on the detailed study by CSIRO (8) and the regional climate change study (9),
the predicted rainfall characteristics for year 2030 can be identified for the Gold
Coast region. The wash-off simulation methodology is described in detail by
Mahbub et al. (1). In this paper, a total of 22 wash-off events were simulated. The
runoff water was vacuumed continuously into 25L plastic containers using the same
vacuum cleaner as for the build-up sample collection. Subsequently, 500mL event
mean concentration (EMC) samples were extracted using a churn splitter. The EMC
represented a flow weighted average concentration of the pollutants computed based
on the total pollutant mass divided by the total runoff volume for a given rainfall
event. As pollutant concentrations may vary by orders of magnitude during a runoff
event, the EMC samples (representing single indices) were found to be appropriate
for evaluating the impacts of stormwater runoff on receiving waters (10). The
particulate and dissolved fractions in wash-off were separated in similar manner as
described for build-up.
Quality control measures in Sample testing
For quality control, calibration standards, internal standards, blanks and certified
reference materials were used. The calibration standard supplied by Accustandard®
contained each of the target heavy metal analytes at a concentration of 100 mg/L.
300
Six different calibration standards at concentrations of 20, 10, 5, 1, 0.1 and 0.01
mg/L were prepared. The internal standard containing Indium (In), Bismuth (Bi),
Terbium (Tb), Scandium (Sc) and Yttrium (Y) was prepared at a concentration of 1
mg/L.
The certified reference material (TraceCERT, Sigma-Aldrich®) contained 10 mg/L
of each target analyte except iron (100 mg/L). The volumes of samples, standards
and blanks were all kept at 50 mL after digestion while the concentration of internal
standards was kept at 0.02 mg/L for analysis. The laboratory fortified blanks were
prepared by adding the certified reference materials to the deionised water to obtain
a concentration of 0.1 mg/L for each target analyte except iron (1 mg/L) and were
treated exactly as samples for analysis. The analytical technique used for the
analysis of the metal analytes was Inductively Coupled Plasma/Mass Spectrometry
(ICP/MS). The percentage recoveries of the spiked blanks with known concentration
of analytes were estimated using the following equation:
R = ( LFB − LRB ) / C × 100 ---------------------------------------- (A.2.1)
where R= percent recovery, LFB= laboratory fortified blank, LRB= blank and C=
stated concentrations of analytes in the LFB. For different heavy metals, percent
recoveries were found to be 89% to 115%. The percentage recovery data was used to
verify the accuracy of the analytical methodology.
To determine the repeatability of the process, seven replicates of a randomly chosen
sample from each batch were analysed. The relative standard deviations (RSD) of
the above samples were measured using the equation:
RSD = ( S rsd / X rsd ) × 100 ------------------------------------------- (A.2.2)
301
where S rsd = Standard deviation of the replicate samples and X rsd = mean of the
replicate samples. The relative standard deviations were from 2.1% to 14.7%.
The reporting limits of the method were established by estimating the limits of
detection (LOD) using seven separate blank samples. The LOD is the lowest
concentration of an analyte measured by a method that could be reliably
distinguished from zero. The LOD was calculated using the equation:
LOD = X LOD + 3S LOD ----------------------------------------- (A.2.3)
where X LOD = mean of the seven blanks and S LOD = standard deviation of the seven
blanks. The LOD for different heavy metals were established from 0.008 mg/L to
0.053 mg/L. All of the test results were found within the specified limits of the test
methods described in USEPA 200.8 (11)
Data Analysis Techniques

PCA
PCA is a pattern recognition technique employed to investigate the correlations
contains the second largest variance and so on. The application of PCA to a data
matrix generates a loading for each variable and a score for each object on the
principal components. Consequently, the data can be presented diagrammatically by
plotting the loading of each variable in the form of a vector and the score of each
object in the form of a data point. This type of plot is referred to as a ‘Biplot’.
Detailed descriptions of PCA can be found elsewhere (12).
302
Fuzzy Clustering
Fuzzy clustering is an object classification method that assigns a degree of class
membership for a given object over several classes (13, 14). The classification is
performed with a user specified membership function which, in the case of ‘SIRIUS’
software (15) used for the analysis is similar to that described by Bezdek (13). An
P
example of a membership function is m( x) = 1 − c x − a , where a and c are
constants and p is called cluster exponent with a suggested value between 1 to 3
(15).
Values closer to 1 result in hard clustering where the objects are placed into their
most preferred classes while values closer to 3 result in soft clustering where the
objects are allowed to spread over as many classes as possible. The sum of the
membership values of each object is 1. The main advantage of fuzzy clustering is
that it facilitates the distinction between the objects that clearly belong to one cluster
and those that are members of several clusters. A class membership threshold is
defined as 1/ n (n = number of clusters).
PROMETHEE and GAIA

PROMETHEE (preference ranking organisation method for enrichment evaluation)
is a method that is designed to rank a number of objects in terms of the data criteria.
The ranking for each variable or criterion is performed by a user specified
preference function. The positive and negative partial outranking flows, φ + and
φ − are calculated from the preference functions for each object or action. The
φ + values indicate how each action outranks all the others, while the φ − values
indicate how each action is outranked by all the others. This procedure is known as
PROMETHEE I. In some instances, objects may perform equally well for a different
303
set of variables. To eliminate such outcomes, the net outranking flow φ , which is the
difference between φ + and φ − , for each action is calculated. This process is termed
as PROMETHEE II. Further details can be found in Keller et al. (16).
The GAIA (Geometrical Analysis for Interactive Aid) is essentially a principal
component analysis (PCA) biplot (16). It is generated from the matrix derived from
the decomposition of φ net outranking flow values from PROMETHEE II. GAIA
provides a graphical display of the relationships between objects, variables and each
other. An additional feature of GAIA is the inclusion of the decision axis, π , which
gives an indication of the degree of decision power (length of the π vector) as well
as the quality of the preferred object or action.
Additional Figures and Tables
The additional Figure A.2.1 shows the individual PCA biplots for five size fractions
as mentioned in the original paper in order to better understand the the relationship
between the object scores and variable loadings. Figures A.2.2 and A.2.3 show the
results of the particle size distributions in the build-up and wash-off samples
respectively from the Mastersizer S Particle Size Analyser. The additional Table
A.2.1 shows the traffic related attributes of the study sites, Table A.2.2 provides the
rainfall simulation plan based on the climate change studies (8, 9), Table A.2.3
provides individual results of the quality control parameters for each target heavy
metals, Table A.2.4 illustrates the possible sources of elemental emissions from
vehicles based on literatures (17-26), and finally Tables A.2.5 and A.2.6 show the
chemical compositions of build-up and wash-off respectively.
304
(a)
(b) (c)
(d) (e)
Figure A.2.1 PCA biplots of particulate and potential dissolved fraction for heavy metals build-
up for (a) >300µm, (b)150-300 µm, (c)75-150 µm, (d)1-75 µm and (e) <1 µm; objects are
indicated by labels with the prefix I, C or R starting for industrial, commercial and residential
sites, respectively
305
100
90
80
70
Shipper
Cummulative Volume, %
Lindfield
60
Billinghurst
Peanba Park
50 Discovery
Dalley Park
40 Reserve
Beattie
30 Abraham
Town Centre
20 Hope IsL
10
0
0.04 0.4 4 40 400
Particle size, µm
Figure A.2.2 Results from Mastersizer S particle size distribution analysis showing
that fractions <1µm constituted 2%-10% whilst fractions up to 300µm constituted
almost 68% to 90% volume of the total build-up particles in samples collected from
the 11 sites
306
100
1
2
90
3
4
80 5
6
70 7
8
Cummulative Volume, %
60 9
10
11
50
12
13
40 14
15
30 16
17
18
20
19
20
10
21
22
0
0.03 0.3 3 30 300
Particle Size, µm
Figure A.2.3 Results from Mastersizer S particle size distribution analysis showing that
fractions <1µm constituted around 2% whilst fractions up to 300µm constituted almost 82% to
96% volume of the total wash-off particles in samples collected from the 22 simulated rain
events
307
Table A.2.1 Traffic data for the study sites
Average Pavement
Site Name Daily Surface type
Volume to Lane
Land Use Coordinate Traffic Texture
Capacity Width, m % of
Site Label Location (ADT), Depth
Ratio (V/C) Aggregate
vehicles/day (STD), mm
Binder
Abraham Road 27.865°S DG14a
Residential 13028 1.11 0.6467 3.5
RA 153.307°E 5.1
Reserve Road 27.870°S DG14a

Residential 6339 0.45 0.7505 3.5
RR 153.301°E 5.1
Peanba Park road 27.851°S DG10b

Residential 581 0.15 0.6844 2.8
RP 153.281°E 5.3
Billinghurst Cres 27.856°S DG10b

Residential 5936 0.74 0.7015 2.9
RB 153.298°E 5.3
Beattie Road 27.868°S DG14a

Industrial 2670 0.24 0.7074 3.5
IBT 153.324°E 5.1
Shipper Drive 27.861°S DG14a

Industrial 7530 0.55 0.6788 3.5
IS 155.332°E 5.1
Hope Island Road 27.882°S DG14a

Commercial 7534 0.57 0.7254 3.4
CH 153.328°E 5.1
Lindfield Road 27.922°S DG10b

Commercial 2312 0.33 0.9417 3.3
CL 153.334°E 5.3
Town Centre Drive 27.929°S DG14a

Commercial 24506 0.62 0.6416 3.5
CT 153.337°E 5.1
Dalley Park Drive 27.887°S DG10b

Residential 3534 0.42 0.8342 2.9
RD 153.346°E 5.3
Discovery Drive 27.899°S DG14a

Residential 9116 0.25 0.6957 2.9
RDS 153.327°E 5.1
a
b
308
Table A.2.2 Simulation rain events for the Gold Coast region at present and as predicted for
year 2030
Current events Predicted events for 2030*
Duration Intensity Object Duration Intensity Object

ARI (year) ARI (year)
(min) (mm/hr) Identifiers (min) (mm/hr) Identifiers
1 120 24.6 4 1 65 37.39 2
10 300 24 - 1 120 24.6 4
100 45 125 12 1 5 125 7
10 160 39.3 6 10 85 58.3 22
1 60 39.3 1 1 25 63 19
2 90 39.3 3 2 42.5 61.2 20
5 133 39.3 5 5 69 59.2 21
10 52.5 77 14 5 16 125 10
20 67.5 77 15 10 21 122 11
50 86.7 77 16 2 10.5 120 9
100 101.25 77 17 1 5.75 119 8
100 105 75 18 100 49 115 13
* Based on the climate change studies (8, 9)
309
Table A.2.3 Limits of detection, percent recovery and relative standard deviation percentage
found in the heavy metal analysis with corresponding molecular weights shown alongside each
element
Relative Standard
Heavy metal elements LOD (mg/L) Recovery, %
Deviation, %
Al / 27 0.010 105.939 11.297
Cr / 53 0.020 100.767 9.449
Mn / 55 0.026 89.914 12.567
Fe / 56 0.007 106.537 2.098
Fe / 57 0.005 97.355 2.433
Ni / 60 0.053 99.805 4.946
Cu / 63 0.019 110.402 7.460
Zn / 66 0.012 115.707 14.657
Cd / 111 0.078 114.481 -
Pb / 206 0.026 111.988 4.573
Pb / 207 0.045 110.841 4.991
Pb / 208 0.008 102.777 3.874
Table A.2.4 Possible sources of elements frequently found in exhaust and non-exhaust emissions
of motor vehicle
Elements Possible Sources
Cu, Sb Bushing, thurstbearing, brake (17-21)
Zn, Cu Lubricants, engine oil (22, 23)
Zn, Cd Tyre (19, 24, 25)
Cr Alloy wheel plate, crankshaft, metal plating, yellow
paint of pavement (19, 24)
Pb, Ni Exhaust emission (18)
Cu, Sb, Ba Rush hour stop-start (26)
310
1 Table A.2.5 Chemical compositions (mean±standard deviations) of the Build-up of Heavy metals in the selected sites
Site Name
Al, mg/m2 Cr, mg/m2 Mn, mg/m2 Fe, mg/m2 Ni, mg/m2 Cu, mg/m2 Zn, mg/m2 Cd, mg/m2 Pb, mg/m2 TSS, mg/m2 TOC, mg/m2
Site Label
Hope Island
Rd 13.18±4.8 0.06±0.05 0.48±0.11 39.58±4.3 0.06±0.03 1.20±0.09 2.53±0.21 0.52±0.48 0.34±0.02 260.27±19.2 39.49±8.1
CH
Shipper Dr
7.26±3.2 0.04±0.02 0.40±0.31 15.67±4.5 0.02±0.01 2.46±1.21 3.22±1.4 0.52±0.48 0.93±0.09 186.16±31.8 14.85±6.3
IS
Billinghurst
Cres 13.17±7.2 0.04±0.11 0.40±0.08 33.75±5.9 0.02±0.01 1.44±0.04 3.57±0.9 0.52±0.48 0.52±0.05 469.07±33.8 39.78±17.2
RB
Peanba Park
3.33±0.69 0.11±0.05 0.11±0.09 9.51±2.8 0.14±0.11 0.76±0.05 1.12±0.7 0.52±0.48 0.22±0.1 62.43±5.7 12.18±5.9
RP
Dalley Park
36.74±10.1 0.07±0.01 2.63±0.98 110.58±3.6 0.10±0.03 2.86±0.94 4.01±0.29 0.42±0.29 1.21±0.61 1097.87±140.8 52.63±6.7
RD
Lindfield Rd
CL 1.96±0.38 0.04±0.02 0.05±0.01 4.91±3.1 0.03±0.01 2.15±0.25 2.12±1.1 0.52±0.48 0.59±0.07 55.01±12.1 15.03±5.4
Town Centre
Dr 1.74±0.92 0.11±0.09 0.03±0.01 2.89±2.1 0.06±0.01 0.76±0.08 1.43±0.52 0.52±0.48 0.29±0.07 71.20±20.2 11.05±3.8
CT
Abraham Rd
RA 6.91±1.2 0.09±0.03 0.26±0.11 14.61±2.6 0.08±0.01 1.01±0.04 2.36±0.22 0.52±0.48 0.29±0.06 212.00±33.9 19.70±3.7
Discovery Dr
RDS 2.66±0.94 0.03±0.01 0.06±0.01 4.96±1.4 0.20±0.08 0.83±0.06 2.25±0.49 0.52±0.48 0.48±0.11 67.20±12.5 17.69±4.9
Beattie Rd
1.35±0.32 0.11±0.09 0.06±0.04 3.71±1.1 0.27±0.01 0.89±0.71 1.61±1.1 0.52±0.48 0.38±0.08 63.20±15.7 11.06±5.1
IBT
Reserve Rd
13.11±3.2 0.24±0.12 0.42±0.15 23.98±2.8 0.48±0.09 1.56±0.91 2.59±0.71 0.52±0.48 0.34±0.09 644.27±51.1 28.62±1.8
RR
311
1 Table A.2.6 Chemical compositions (mean±standard deviations) of the wash-off of Heavy metals for the simulated rain events
Rain Al, Cr, Mn, Fe, Pb, Ni, Cu, Zn, Cd, TSS, TOC, EC,
pH
Events* mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L µs/cm
1 4.13±1.6 0.02±0.01 0.02±0.01 1.01±0.21 0.09±0.02 0.02±0.01 0.34±0.1 1.65±0.82 0.20±0.06 30.60±8.1 9.56±3.6 7.08±0.03 56.80±0.01
2 0.78±0.11 0.02±0.01 0.01±0.005 0.75±0.1 0.10±0.05 0.03±0.01 0.14±0.05 1.11±0.32 0.20±0.06 31.10±3.8 9.69±1.5 7.08±0.02 53.70±0.05
3 0.70±0.13 0.03±0.01 0.02±0.005 0.75±0.09 0.09±0.01 0.04±0.01 0.21±0.08 2.03±0.12 0.20±0.02 35.00±10.2 8.78±1.8 7.01±0.01 43.20±0.04
4 0.63±0.12 0.02±0.01 0.02±0.005 0.80±0.06 0.05±0.01 0.09±0.02 0.38±0.11 10.46±2.1 0.16±0.05 36.10±11.3 7.64±4.2 7.06±0.02 42.00±0.06
5 1.48±0.05 0.02±0.01 0.04±0.01 0.83±0.07 0.05±0.01 0.04±0.01 0.40±0.1 11.17±1.32 0.16±0.09 20.40±9.2 7.69±0.98 7.06±0.01 39.30±0.02
6 0.33±0.09 0.01±0.005 0.02±0.01 0.74±0.12 0.08±0.02 0.01±0.005 0.30±0.05 9.81±2.2 0.16±0.04 16.70±7.1 7.27±2.4 7.07±0.01 38.30±0.02
7 1.16±0.25 0.02±0.01 0.02±0.01 1.44±0.12 0.05±0.01 0.01±0.005 0.26±0.09 10.21±2.4 0.16±0.04 60.70±18.2 28.60±5.4 7.01±0.01 74.80±0.02
8 0.81±0.21 0.02±0.01 0.02±0.01 1.58±0.1 0.04±0.01 0.01±0.005 0.29±0.06 7.80±2.1 0.16±0.04 69.80±18.1 26.22±5.9 7.01±0.01 63.40±0.02
9 4.33±1.2 0.02±0.01 0.02±0.01 1.33±0.09 0.06±0.02 0.01±0.005 0.25±0.12 8.62±3.2 0.16±0.01 24.86±3.6 21.18±6.4 7.10±0.01 44.80±0.01
10 0.80±0.38 0.03±0.01 0.02±0.01 1.53±0.1 0.04±0.01 0.04±0.01 0.17±0.04 7.82±1.05 0.16±0.04 30.40±12.1 18.17±8.1 7.09±0.02 38.70±0.01
11 0.53±0.07 0.03±0.01 0.05±0.01 1.02±0.3 0.08±0.04 0.03±0.01 0.24±0.06 2.50±1.3 0.20±0.03 22.10±10.7 16.25±8.6 7.09±0.03 33.80±0.01
12 0.48±0.1 0.09±0.03 0.05±0.01 0.96±0.04 0.13±0.01 0.06±0.01 0.25±0.08 2.22±0.95 0.16±0.01 15.80±6.5 12.92±6.4 7.07±0.01 28.00±0.01
13 0.43±0.2 0.05±0.02 0.03±0.01 0.66±0.4 0.10±0.06 0.06±0.02 0.24±0.09 1.93±0.98 0.20±0.01 20.00±7.2 8.71±1.1 7.06±0.01 28.10±0.01
14 0.18±0.05 0.03±0.01 0.04±0.01 0.54±0.23 0.13±0.06 0.03±0.01 0.16±0.06 1.52±0.31 0.20±0.06 20.20±7.8 17.57±4.3 7.04±0.01 23.20±0.02
15 0.56±0.1 0.03±0.01 0.03±0.01 0.62±0.21 0.11±0.02 0.06±0.02 0.17±0.1 1.42±0.12 0.20±0.05 28.20±3.4 15.98±3.5 7.00±0.01 23.80±0.02
16 0.40±0.08 0.03±0.01 0.01±0.005 0.52±0.2 0.10±0.05 0.09±0.01 0.16±0.03 1.49±0.15 0.20±0.05 26.30±2.5 16.44±4.2 7.02±0.01 23.70±0.02
17 0.45±0.05 0.03±0.01 0.01±0.005 0.64±0.12 0.07±0.02 0.03±0.01 0.17±0.06 1.11±0.14 0.20±0.05 16.10±9.3 14.42±2.6 7.01±0.01 22.80±0.01
18 0.26±0.06 0.02±0.01 0.01±0.005 0.55±0.41 0.14±0.05 0.06±0.02 0.19±0.01 1.29±0.24 0.20±0.02 21.80±1.4 13.70±2.8 6.99±0.01 23.50±0.02
19 0.27±0.06 0.04±0.02 0.02±0.01 0.60±0.31 0.07±0.01 0.08±0.02 0.17±0.09 1.68±0.31 0.20±0.04 35.10±1.8 15.82±3.2 7.16±0.01 33.40±0.03
20 0.31±0.01 0.03±0.01 0.01±0.005 0.51±0.12 0.13±0.04 0.08±0.02 0.17±0.06 1.30±0.34 0.20±0.04 20.10±6.4 12.20±6.7 7.20±0.02 29.40±0.06
21 0.26±0.05 0.05±0.02 0.02±0.01 0.46±0.13 0.09±0.01 0.11±0.03 0.19±0.08 1.31±0.25 0.20±0.07 12.50±6.7 9.36±2.8 7.19±0.02 25.90±0.02
22 0.55±0.1 0.03±0.01 0.03±0.005 0.42±0.11 0.13±0.05 0.11±0.05 0.18±0.06 1.16±0.24 0.20±0.07 18.60±3.1 10.39±4.1 7.17±0.03 25.40±0.02
*
2 Numerical object identifiers used elsewhere are same as rain events numbers
312
References
Traffic and climate change impacts on water quality: measuring build-up and
wash-off of heavy metals and petroleum hydrocarbons. In Rethinking
Sustainable Development: Urban management, Engineering and Design;
Yigitcanlar, T. Ed.; Engineering Science Reference, New York, 2010.
(2) Herngren, L. Build-up and wash-off process kinetics of PAHs and heavy
metals on paved surfaces using simulated rainfall. PhD Dissertation,
Queensland University of Technology, Brisbane, 2005.
(3) Deletic, A.; Orr, D., W. Pollution buildup on road surfaces. J. Environ. Eng.
2005, 131, 49-59.
(4) Egodawatta, P. K. Translation of small-plot scale pollutant build-up and
wash-off measurements to urban catchment scale. PhD Dissertation,
Queensland University of Technology, Brisbane, 2007.
(5) Floyd, C.N. Mobile rainfall simulator for small plot field experiments. J.
Agric. Eng. Res. 1981, 26, 307-314.
(6) Silburn, D.M., Hargreaves, P., Budd, N., and Glanville, S.G. Endosulfan on
cotton plants— Persistence and wash-off during rain. INTERSECT 96—
International Symposium on Environmental Chemistry and Toxicology.
Sydney: RACI, Abstract O58, 1996.
313
(7) Herngren, L.; Goonetilleke, A.; Sukpum, R.; De Silva, D. Y. Rainfall
simulation as a tool for urban water quality research. Environ. Eng. Sci.
2005, 22, 378-383.
(8) CSIRO: Climate change in Australia. Technical Report 2007,
http://www.csiro.au/resources/Publications.html.
(9) CSIRO Division of Marine and Atmospheric Research: The impact of
climate change on extreme rainfall and coastal sea levels over south-east
Queensland. Part 2: a high-resolution modelling study of the effect of climate
change on the intensity of extreme rainfall events. 2007, Aspendale, Victoria.
http://www.csiro.au/resources/Publications.html
(10) Sansalone, J. J.; Buchberger, S. G. Partitioning and first flush of metals in
urban roadway storm water. J. Environ. Eng.1997, 123, 134-143.
(11) US Environmental Protection Agency: Determination of trace elements in
waters and wastes by Inductively Coupled Plasma – Mass Spectrometry.
Method 200.8, 1994.
(12) Massart, D. L.; Vandeginste, B. G. M.; Buydens, L. M. C.; De Jong, S.;
Lewi, P. J.; Smeyers-Verbeke, J. Handbook of Chemometrics and
Qualimetrics Part A; Elsevier: Amsterdam, 1997.
(13) Bezdek, J. C. Pattern Recognition with Fuzzy Objective Function
Algorithms; Plenum Press: New York, 1982.
314
(14) Otto, M. Fuzzy theory explained. Chemom. Intell. Lab. Syst.1988, 4, 101-
120.
(15) SIRIUS version 7.1© copyright 1987-2008, Pattern Recognition Systems
AS, http://www.prs.no, Help Topics, 2008.
(16) Keller, H. R.; Massart, D. L.; Brans, J. P. Multicriteria decision making: a
case study. Chemom. Intell. Lab. Syst.1991, 11, 175-189.
(17) Sternbeck, J.; Sjödin, Å.; Andréasson, K. Metal emissions from road traffic
and the influence of resuspension – results from two tunnel studies. Atmos.
Environ. 2002, 36, 4735-4744.
(18) Johansson, C.; Norman, M.; Burman, L. Road traffic emission factors for
heavy metals. Atmos. Environ.2009, 43, 4681-4688.
(19) Harrison, R. M. Toxic metals in street and household dusts. Sci. Total
Environ.1979, 11, 89-97.
(20) Fujiwara, F.; Rebagliati, R. J.; Marrero, J.; Gómez, D.; Smichowski, P.
Antimony as a traffic-related element in size-fractionated road dust samples
collected in Buenos Aires. Microchem. J. 2010,
doi:10.1016/j.microc.2010.05.006.
(21) Von Uexküll, O.; Skerfving, S.; Doyle, R.; Braungart, M. Antimony in brake
pads – a carcinogenic component? J. Cleaner Prod. 2005, 13, 19-31.
315
(22) Lim, M. C. H.; Ayoko, G. A.; Morawska, L.; Ristovski, Z. D.; Jayaratne, E.
R.; Kokot, S. A comparative study of the elemental composition of the
exhaust emissions of cars powered by liquefied petroleum gas and unleaded
petrol. Atmos. Environ.2006, 40, 3111-3122.
(23) Cadle, S.H.; Mulawa, P.A.; Hunsanger, E.C. Composition of light duty
motor vehicle exhaust particulate matter in the Denver, Colorado Area.
Environ. Sci. Technol. 1999, 33, 2328–2339.
(24) Adachi, K.; Tainosho, Y. Characterization of heavy metal particles
embedded in tire dust. Environ. Int. 2004, 30, 1009– 1017.
(25) Kummer, U.; Pacyna, J.; Pacyna, E.; Friedrich, R. Assessment of heavy
metal releases from the use phase of road transport in Europe. Atmos.
Environ.2009, 43, 640-647.
(26) Grieshop, A. P.; Lipsky, E. M.; Pekney, N. J.; Takahama, S.; Robinson, A.
L. Fine particle emission factors from vehicles in a highway tunnel: effects
of fleet composition and season. Atmos. Environ.2006, 40, S287–S298.
316
Manuscript Title
Analysis of the Build-up of Semi and Non Volatile Organic Compounds
on Urban Roads

Egodawatta1
1

2
s.mahbub@qut.edu.au; g.ayoko@qut.edu.au; a.goonetilleke@qut.edu.au;
*Corresponding Author: Parvez Mahbub;Tel: 61 7 3138 9945;Fax: 61 7 3138
1170;
Number of Pages: 11
Number of Supplementary Tables 10
317
318
Table A.3.1 Test results for the build-up of SVOCs and NVOCs along with physico-chemical
parameters for the > 300 µm particle size fraction
Target variables for Site Identifiera
chemical analysis CH RDS RR CT RD IBT CA RP IS RB CL
OCT 0.27 0.27 0.27 0.27 0.27 1.23 0.27 0.27 2.30 0.27 0.27
DEC 0.16 0.16 0.16 0.16 0.16 2.21 0.16 0.16 1.30 0.16 0.16
DOD 7.60 0.81 2.23 38.96 5.67 22.11 0.22 9.17 0.22 0.22 0.22
TED 0.36 0.38 0.19 6.14 3.70 0.60 0.61 1.33 1.98 0.19 1.74
HXD 0.83 0.48 1.45 7.49 1.13 0.61 1.11 0.77 3.70 0.73 2.36
b
SVOC ,
OCD 0.61 0.34 0.55 12.67 2.07 1.55 0.34 4.87 0.98 0.83 14.23
mg/L
EIC 0.17 3.63 0.17 21.58 3.43 0.17 1.52 6.74 21.22 0.40 17.21
DOC 0.69 4.58 0.27 23.89 3.43 0.27 2.97 1.35 24.57 2.32 21.49
TTC 0.03 6.41 0.98 28.70 4.37 0.03 1.78 4.64 27.49 0.03 22.77
HXC 0.54 3.51 3.32 16.74 2.66 0.54 3.74 3.37 8.98 2.01 7.88
OCC 1.06 2.33 1.93 13.82 3.41 0.58 2.54 3.79 5.97 0.58 5.63
TCT 0.51 2.20 1.98 3.61 2.32 0.51 1.77 2.72 4.04 1.33 0.51
DTT 0.66 2.09 1.93 1.94 2.73 1.58 1.99 2.70 1.84 1.58 0.66
NVOCc, TRT 0.62 1.48 1.58 0.62 1.62 1.38 1.44 1.20 0.62 0.62 0.62
mg/L HXT 0.53 35.83 0.35 1.21 1.35 0.68 17.28 0.94 0.53 1.10 0.53
OTT 0.30 0.30 0.73 0.73 0.30 0.61 0.30 0.78 2.81 0.30 0.30
TTT 0.43 0.43 0.43 0.43 1.32 0.87 0.43 0.43 3.53 1.15 0.43
PSD*(%) 20.27 17.43 17.69 17.64 15.42 38.04 37.63 25.80 23.31 36.58 16.01
pH 7.25 7.17 7.48 7.39 7.17 7.31 7.26 7.30 6.72 7.01 7.65
Physico-
EC, micro-
Chemical 66.10 32.20 38.40 21.80 39.30 23.40 36.20 18.77 18.23 30.90 16.84
siemens/cm
Parameters
TSS, mg/L 1.60 8.00 16.27 2.40 45.07 0.27 1.87 5.33 3.73 64.80 1.87
TOC, mg/L 2.17 1.52 1.87 1.98 2.38 1.95 2.56 2.06 2.06 2.67 2.34
a
Site identifiers are same as the site identifiers described in Table 1 of the
manuscript
b
Target SVOCs are given in Section 2.5 of the manuscript
c
Target NVOCs are given in Section 2.5 of the manuscript
*Particle Size Distribution (% of the total volume of the corresponding particle
fraction)
319
parameters for the 150- 300 µm particle size fraction
Target variables Site Identifiera
for chemical
CH RDS RR CT RD IBT CA RP IS RB CL
analysis
OCT 0.27 0.27 0.27 0.27 0.27 2.31 0.27 0.27 1.20 0.27 0.27
DEC 0.16 0.16 0.16 0.16 0.16 1.32 0.16 0.16 1.50 0.16 0.16
DOD 5.76 14.68 6.87 0.22 3.67 5.94 12.93 0.22 5.46 0.22 0.22
TED 0.19 4.63 9.64 1.06 1.00 3.66 0.19 1.01 1.31 0.48 4.00
HXD 0.43 1.33 1.77 1.82 1.16 0.90 0.87 1.03 0.43 0.80 0.43
SVOCb,
OCD 0.74 1.01 6.13 7.31 0.35 0.35 0.35 0.35 2.90 0.79 9.43
mg/L
EIC 0.17 5.29 12.64 16.36 0.89 0.17 0.69 0.81 0.55 6.89 25.11
DOC 1.01 6.04 8.21 12.47 3.07 0.27 4.42 2.42 1.21 6.56 25.01
TTC 0.03 5.56 14.10 15.14 3.34 0.03 3.59 0.03 0.03 6.47 25.75
HXC 1.55 4.52 9.01 2.40 2.69 48.17 2.25 2.28 5.45 4.43 3.76
OCC 0.58 3.33 7.14 3.59 1.39 0.58 3.06 0.58 0.58 3.06 3.07
TCT 1.61 2.13 6.10 1.67 0.51 0.87 0.51 0.51 1.26 2.40 2.21
DTT 1.38 1.57 2.67 2.18 27.53 1.13 0.66 1.52 1.99 2.12 1.62
TRT 0.62 1.31 2.52 1.35 0.62 1.48 0.62 0.62 0.62 1.44 0.62
NVOCc,
mg/L
HXT 0.53 0.53 1.17 0.53 0.53 0.53 0.53 7.66 20.92 24.56 0.85
OTT 0.30 0.30 0.30 0.97 0.30 0.30 0.30 0.30 0.30 0.30 1.01
TTT 0.43 0.43 2.26 1.03 0.43 0.43 0.43 1.41 0.43 0.43 0.43
PSD*(%) 14.21 3.79 8.01 8.78 10.10 11.84 10.69 16.60 8.84 14.24 8.56
pH 7.25 7.17 7.48 7.39 7.17 7.31 7.26 7.30 6.72 7.01 7.65
Physico-
EC, micro-
Chemical 66.10 32.20 38.40 21.80 39.30 23.40 36.20 18.77 18.23 30.90 16.84
siemens/cm
Parameters
TSS, mg/L 10.40 4.27 26.67 8.00 123.73 1.07 0.27 5.60 6.40 12.53 3.73
TOC, mg/L 3.68 2.19 2.66 2.00 32.24 1.85 2.03 1.83 1.56 2.00 2.87
a
manuscript
b
c
fraction)
320
parameters for the 75-150 µm particle size fraction
Target variables Site Identifiera
for chemical
CH RDS RR CT RD IBT CA RP IS RB CL
analysis
OCT 0.27 0.27 0.27 0.27 0.27 3.21 0.61 0.27 3.10 0.27 0.27
DEC 0.16 0.16 0.16 0.32 0.16 0.95 0.16 0.16 0.88 0.16 0.16
DOD 8.47 0.22 23.18 7.67 51.86 0.22 47.71 22.04 0.22 11.11 0.22
TED 7.31 0.86 4.06 0.19 2.01 0.19 4.50 0.19 0.19 3.13 0.79
HXD 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 1.40 4.81 0.43
SVOCb,
OCD 0.35 0.35 0.35 0.35 1.43 7.26 1.44 0.35 3.84 11.79 0.71
mg/L
EIC 0.99 0.33 0.17 0.17 0.89 23.70 0.61 0.17 11.07 25.02 14.76
DOC 0.95 0.27 0.27 0.59 6.26 29.17 2.79 0.27 1.98 24.63 14.13
TTC 0.03 0.03 0.36 0.03 8.29 27.28 2.00 0.03 15.46 32.35 22.19
HXC 26.18 21.45 4.16 1.25 5.73 3.59 0.54 0.54 0.54 7.16 7.99
OCC 0.58 0.58 2.30 1.55 5.88 6.28 0.58 0.58 0.58 5.79 4.71
TCT 1.17 0.51 0.51 1.18 3.67 3.01 0.51 0.51 1.40 4.36 3.48
DTT 1.41 0.66 0.66 1.35 1.92 1.87 0.66 1.63 2.02 1.55 2.77
TRT 0.62 0.62 1.36 0.62 1.35 0.62 0.62 1.52 0.62 1.37 1.76
NVOCc,
mg/L
HXT 1.13 0.53 0.53 0.53 0.53 1.32 0.53 10.75 43.38 1.19 0.53
OTT 0.30 0.30 0.30 1.52 0.30 1.03 0.30 0.30 0.30 0.91 0.74
TTT 0.43 0.43 0.43 1.12 0.93 0.43 0.43 1.85 0.43 0.43 1.81
PSD*(%) 10.10 8.36 5.77 12.04 11.67 11.18 7.07 16.86 8.67 10.98 13.76
pH 7.25 7.17 7.48 7.39 7.17 7.31 7.26 7.30 6.72 7.01 7.65
Physico-
EC, micro-
Chemical 66.10 32.20 38.40 21.80 39.30 23.40 36.20 18.77 18.23 30.90 16.84
siemens/cm
Parameters
TSS, mg/L 21.07 2.67 16.80 1.87 102.67 8.27 20.80 7.47 30.93 50.40 6.93
TOC, mg/L 4.02 3.29 3.46 1.85 3.05 1.75 2.19 2.49 1.86 5.16 2.44
a
manuscript
b
c
fraction)
321
parameters for the 1-75 µm particle size fraction
OCT 0.27 0.27 0.66 1.32 0.27 5.32 3.65 0.27 6.50 0.27 0.98
DEC 2.32 1.25 0.98 3.21 0.16 2.37 0.89 0.16 4.21 0.16 0.32
DOD 5.99 0.22 0.22 2.53 41.94 12.91 21.62 0.22 0.22 14.51 0.22
TED 7.76 71.20 5.08 4.81 4.38 21.77 1.49 2.31 1.26 2.57 2.33
HXD 0.43 1.52 0.99 1.13 0.43 1.76 0.43 0.43 0.43 0.43 0.43
SVOCb,
OCD 2.74 2.22 1.12 0.35 0.92 6.88 3.97 0.35 1.33 6.49 0.35
mg/L
EIC 1.60 2.60 0.66 0.39 2.62 11.75 8.31 0.17 4.70 13.90 0.17
DOC 3.47 2.55 2.57 2.19 5.37 14.19 11.44 4.62 7.62 18.95 0.27
TTC 4.59 2.65 4.61 2.52 2.70 14.05 14.43 0.03 0.40 29.15 0.03
HXC 3.35 2.09 3.22 1.37 3.13 6.86 5.06 4.00 0.54 5.73 0.54
OCC 2.52 2.18 1.94 0.58 3.58 6.98 5.40 3.18 0.58 6.59 0.58
TCT 1.92 1.82 1.09 0.89 2.57 0.51 2.36 1.68 1.92 3.32 0.51
DTT 1.82 1.82 1.94 1.82 1.98 0.66 1.81 1.49 2.06 1.73 0.66
TRT 1.54 0.62 0.62 1.63 0.62 0.62 1.66 0.62 1.44 1.41 0.62
NVOCc,
mg/L
HXT 0.53 0.53 0.53 1.17 0.53 0.53 0.53 11.55 0.53 1.06 0.53
OTT 0.30 0.30 0.57 0.30 1.83 0.30 0.30 0.30 0.30 0.30 0.30
TTT 0.43 1.10 1.11 1.27 1.66 0.43 0.43 1.59 1.11 2.31 0.43
PSD*(%) 51.52 64.83 59.29 56.47 53.88 36.05 41.36 37.94 52.53 35.07 58.22
pH 7.25 7.17 7.48 7.39 7.17 7.31 7.26 7.30 6.72 7.01 7.65
Physico-
Chemical EC, 66.10 32.20 38.40 21.80 39.30 23.40 36.20 18.77 18.23 30.90 16.84
Parameters microsiemens/cm
TSS, mg/L 157.60 15.20 546.67 33.60 785.07 28.27 152.53 23.47 124.53 307.20 23.47
TOC, mg/L 11.50 2.98 14.54 2.86 5.67 2.22 5.25 2.99 5.33 19.88 3.60
a
manuscript
b
c
fraction)
322
parameters for the <1 µm particle size fraction
OCT 0.88 0.65 1.31 2.38 0.62 2.65 1.32 0.72 1.58 0.27 0.85
DEC 0.61 0.16 2.31 1.32 0.16 3.10 0.64 0.16 0.97 0.16 0.16
DOD 19.07 1.75 1.54 0.22 23.00 6.58 0.22 0.22 0.22 0.22 0.22
TED 29.45 13.54 0.19 61.18 3.07 7.31 1.50 0.97 0.98 1.77 0.71
HXD 0.43 8.46 0.43 1.10 0.43 0.43 2.41 0.43 0.43 0.43 0.43
b
SVOC ,
mg/L OCD 0.35 12.25 0.84 1.05 0.35 0.35 1.38 2.39 0.98 4.54 0.79
EIC 0.17 16.77 0.40 0.17 0.66 2.03 16.01 3.64 0.53 7.73 0.34
DOC 1.10 16.09 0.27 3.09 5.55 2.73 21.14 4.36 0.27 11.69 0.69
TTC 0.03 22.45 1.46 0.03 0.03 1.98 33.38 3.02 0.36 13.65 0.66
HXC 2.15 10.85 3.35 3.28 4.53 2.65 3.53 3.92 1.23 4.82 0.54
OCC 2.12 9.43 2.84 2.38 2.29 1.57 8.92 2.88 1.25 4.46 1.50
TCT 1.95 5.14 1.93 1.25 2.22 1.48 0.51 1.43 0.51 3.75 1.19
DTT 2.38 3.16 1.91 2.51 1.54 1.41 0.66 1.74 0.66 2.42 1.65
TRT 1.46 2.29 1.35 1.17 1.80 0.62 0.62 1.50 0.62 1.82 0.62
NVOCc,
mg/L
HXT 2.05 1.26 0.53 0.53 17.76 1.07 0.53 0.53 0.53 0.53 0.53
OTT 0.30 0.30 1.33 0.30 0.30 0.65 0.30 0.87 0.30 2.19 0.30
TTT 0.43 0.95 1.19 0.43 0.43 1.69 0.43 0.43 0.43 3.27 0.43
PSD*(%) 3.90 5.59 9.24 5.07 8.94 2.90 3.24 2.80 6.65 3.14 3.45
pH 7.25 7.17 7.48 7.39 7.17 7.31 7.26 7.30 6.72 7.01 7.65
Physico-
EC,
Chemical microsiemens/cm
66.10 32.20 38.40 21.80 39.30 23.40 36.20 18.77 18.23 30.90 16.84
Parameters
TDS, mg/L 69.60 37.07 37.87 25.33 41.33 25.33 36.53 20.56 20.56 34.13 19.01
DOC, mg/L 18.11 7.71 6.08 2.36 9.28 3.28 7.67 2.81 4.03 10.06 3.78
a
manuscript
b
c
fraction)
323
Table A.3.6 Simple bi-variate correlation matrix between target variables for the >300 µm particle size fraction from original data given in
supplementary Table A.3.1
OCT DEC DOD TED HXD OCD EIC DOC TTC HXC OCC TCT DTT TRT HXT OTT TTT PSD pH EC TSS TOC
OCT 1.00 0.78 -0.03 -0.01 0.18 -0.22 0.40 0.41 0.39 0.14 0.05 0.36 -0.02 -0.21 -0.20 0.89 0.88 0.02 -0.04 -0.16 -0.24 -0.16
DEC 0.78 1.00 0.24 -0.12 -0.04 -0.20 0.04 0.05 0.04 -0.13 -0.16 -0.07 -0.08 0.04 -0.20 0.46 0.47 0.15 -0.08 -0.12 -0.27 -0.21
DOD -0.03 0.24 1.00 0.68 0.65 0.46 0.31 0.25 0.31 0.53 0.63 0.21 0.08 -0.16 -0.26 -0.04 -0.22 -0.03 0.25 -0.18 -0.28 -0.27
TED -0.01 -0.12 0.68 1.00 0.83 0.63 0.68 0.63 0.68 0.80 0.89 0.57 0.27 -0.23 -0.26 0.15 0.10 -0.43 0.08 -0.01 -0.04 -0.01
HXD 0.18 -0.04 0.65 0.83 1.00 0.64 0.83 0.82 0.84 0.97 0.96 0.60 -0.02 -0.49 -0.26 0.38 0.18 -0.34 0.04 0.17 -0.26 -0.11
OCD -0.22 -0.20 0.46 0.63 0.64 1.00 0.70 0.68 0.69 0.72 0.75 0.05 -0.26 -0.46 -0.29 -0.10 -0.26 -0.04 0.58 0.30 -0.27 0.04
EIC 0.40 0.04 0.31 0.68 0.83 0.70 1.00 0.98 0.99 0.90 0.86 0.58 -0.08 -0.58 -0.21 0.58 0.41 -0.40 -0.07 0.11 -0.34 -0.10
DOC 0.41 0.05 0.25 0.63 0.82 0.68 0.98 1.00 0.99 0.88 0.81 0.49 -0.21 -0.62 -0.17 0.54 0.43 -0.18 -0.08 -0.05 -0.29 -0.04
TTC 0.39 0.04 0.31 0.68 0.84 0.69 0.99 0.99 1.00 0.91 0.86 0.56 -0.12 -0.56 -0.16 0.56 0.41 -0.14 -0.06 -0.17 -0.34 -0.14
HXC 0.14 -0.13 0.53 0.80 0.97 0.72 0.90 0.88 0.91 1.00 0.97 0.63 0.02 -0.46 -0.12 0.37 0.16 -0.08 0.08 -0.50 -0.28 -0.13
OCC 0.05 -0.16 0.63 0.89 0.96 0.75 0.86 0.81 0.86 0.97 1.00 0.63 0.10 -0.41 -0.16 0.29 0.08 -0.13 0.14 -0.05 -0.29 -0.14
TCT 0.36 -0.07 0.21 0.57 0.60 0.05 0.58 0.49 0.56 0.63 0.63 1.00 0.63 -0.04 0.05 0.66 0.51 -0.27 -0.18 -0.19 -0.04 -0.26
DTT -0.02 -0.08 0.08 0.27 -0.02 -0.26 -0.08 -0.21 -0.12 0.02 0.10 0.63 1.00 0.62 0.20 0.13 0.12 0.00 -0.16 -0.16 0.24 -0.16
TRT -0.21 0.04 -0.16 -0.23 -0.49 -0.46 -0.58 -0.62 -0.56 -0.46 -0.41 -0.04 0.62 1.00 0.41 -0.29 -0.27 0.04 0.17 0.01 0.04 -0.28
HXT -0.20 -0.20 -0.26 -0.26 -0.26 -0.29 -0.21 -0.17 -0.16 -0.12 -0.16 0.05 0.20 0.41 1.00 -0.24 -0.23 -0.01 -0.09 0.07 -0.15 -0.39
OTT 0.89 0.46 -0.04 0.15 0.38 -0.10 0.58 0.54 0.56 0.37 0.29 0.66 0.13 -0.29 -0.24 1.00 0.87 -0.04 -0.62 -0.29 -0.24 -0.20
TTT 0.88 0.47 -0.22 0.10 0.18 -0.26 0.41 0.43 0.41 0.16 0.08 0.51 0.12 -0.27 -0.23 0.87 1.00 0.05 -0.81 -0.27 0.14 0.09
PSD 0.20 0.45 -0.03 -0.43 -0.34 -0.40 -0.40 -0.38 -0.43 -0.38 -0.43 -0.27 0.00 0.04 -0.01 -0.04 0.05 1.00 -0.26 -0.10 0.13 0.42
pH -0.64 -0.28 0.25 0.08 0.04 0.58 -0.07 -0.08 -0.06 0.08 0.14 -0.48 -0.26 0.17 -0.09 -0.62 -0.81 -0.26 1.00 -0.01 -0.31 -0.09
EC -0.36 -0.32 -0.18 -0.31 -0.37 -0.50 -0.61 -0.55 -0.57 -0.50 -0.45 -0.39 -0.26 0.11 0.07 -0.39 -0.27 -0.10 -0.01 1.00 0.13 0.10
TSS -0.24 -0.27 -0.28 -0.04 -0.26 -0.27 -0.34 -0.29 -0.34 -0.28 -0.29 -0.04 0.24 0.04 -0.15 -0.24 0.14 0.13 -0.31 0.13 1.00 0.50
TOC -0.16 -0.21 -0.27 -0.01 -0.11 0.04 -0.10 -0.04 -0.14 -0.13 -0.14 -0.26 -0.16 -0.28 -0.39 -0.20 0.09 0.42 -0.09 0.10 0.50 1.00
324
Table A.3.7 Simple bi-variate correlation matrix between target variables for the 150-300 µm particle size fraction from original data given in
OCT 1.00 0.89 0.06 0.07 -0.23 -0.21 -0.33 -0.37 -0.37 0.91 -0.43 -0.23 -0.15 0.10 0.08 -0.21 -0.24 0.05 -0.02 -0.30 -0.20 -0.16
DEC 0.89 1.00 0.06 -0.01 -0.35 -0.15 -0.35 -0.39 -0.40 0.63 -0.46 -0.22 -0.15 -0.04 0.34 -0.22 -0.25 -0.04 -0.11 -0.06 -0.20 -0.18
DOD 0.06 0.06 1.00 0.23 0.09 -0.37 -0.39 -0.34 -0.31 0.08 0.17 0.04 -0.12 0.10 -0.32 -0.48 -0.15 -0.52 -0.16 0.13 -0.12 -0.09
TED 0.07 -0.01 0.23 1.00 0.46 0.44 0.41 0.28 0.44 0.27 0.71 0.83 -0.14 0.77 -0.30 0.02 0.63 -0.53 0.04 -0.28 -0.05 -0.16
HXD -0.23 -0.35 0.09 0.46 1.00 0.21 0.22 0.05 0.23 -0.01 0.64 0.43 0.14 0.68 -0.36 0.11 0.65 -0.36 0.12 -0.07 0.20 0.10
OCD -0.21 -0.15 -0.37 0.44 0.21 1.00 0.93 0.86 0.92 -0.18 0.54 0.50 -0.20 0.25 -0.18 0.85 0.34 -0.42 0.06 -0.16 -0.17 -0.22
EIC -0.33 -0.35 -0.39 0.41 0.22 0.93 1.00 0.96 0.99 -0.20 0.59 0.49 -0.18 0.27 -0.18 0.86 0.24 -0.38 0.25 -0.15 -0.17 -0.20
DOC -0.37 -0.39 -0.34 0.28 0.05 0.86 0.96 1.00 0.97 -0.27 0.47 0.32 -0.14 0.05 -0.20 0.86 0.07 -0.36 0.34 -0.06 -0.14 -0.14
TTC -0.37 -0.40 -0.31 0.44 0.23 0.92 0.99 0.97 1.00 -0.22 0.64 0.50 -0.11 0.26 -0.24 0.83 0.24 -0.42 0.28 -0.23 -0.08 -0.12
HXC 0.91 0.63 0.08 0.27 -0.01 -0.18 -0.20 -0.27 -0.22 1.00 -0.20 -0.06 -0.14 0.33 -0.14 -0.17 -0.08 0.05 0.09 -0.22 -0.17 -0.14
OCC -0.43 -0.46 0.17 0.71 0.64 0.54 0.59 0.47 0.64 -0.20 1.00 0.85 -0.14 0.75 -0.19 0.21 0.62 -0.50 0.34 0.12 -0.04 -0.16
TCT -0.23 -0.22 0.04 0.83 0.43 0.50 0.49 0.32 0.50 -0.06 0.85 1.00 -0.22 0.82 -0.02 0.05 0.68 -0.35 0.31 0.15 -0.09 -0.25
DTT -0.15 -0.15 -0.12 -0.14 0.14 -0.20 -0.18 -0.14 -0.11 -0.14 -0.14 -0.22 1.00 -0.20 -0.15 -0.14 -0.13 -0.05 -0.10 0.08 0.99 1.00
TRT 0.10 -0.04 0.10 0.77 0.68 0.25 0.27 0.05 0.26 0.33 0.75 0.82 -0.20 1.00 -0.05 -0.08 0.64 -0.30 0.24 0.01 -0.10 -0.25
HXT 0.08 0.34 -0.32 -0.30 -0.36 -0.18 -0.18 -0.20 -0.24 -0.14 -0.19 -0.02 -0.15 -0.05 1.00 -0.26 -0.15 0.30 -0.03 -0.28 -0.15 -0.21
OTT -0.21 -0.22 -0.48 0.02 0.11 0.85 0.86 0.86 0.83 -0.17 0.21 0.05 -0.14 -0.08 -0.26 1.00 -0.01 -0.26 0.26 -0.11 -0.17 -0.14
TTT -0.24 -0.25 -0.15 0.63 0.65 0.34 0.24 0.07 0.24 -0.08 0.62 0.68 -0.13 0.64 -0.15 -0.01 1.00 0.02 0.39 -0.25 -0.02 -0.17
PSD 0.05 -0.04 -0.52 -0.53 -0.36 -0.42 -0.38 -0.36 -0.42 0.05 -0.50 -0.35 -0.05 -0.30 0.30 -0.26 0.02 1.00 -0.08 0.16 -0.05 -0.04
pH -0.22 -0.51 -0.16 0.44 0.31 0.56 0.65 0.64 0.68 0.09 0.44 0.31 -0.10 0.24 -0.73 0.56 0.39 -0.08 1.00 -0.01 -0.08 -0.07
EC -0.30 -0.36 0.33 -0.08 -0.07 -0.36 -0.35 -0.36 -0.30 -0.20 0.02 0.15 0.18 0.01 -0.28 -0.41 -0.05 0.16 -0.01 1.00 0.25 0.24
TSS -0.20 -0.20 -0.12 -0.05 0.20 -0.17 -0.17 -0.14 -0.08 -0.17 -0.04 -0.09 0.99 -0.10 -0.15 -0.17 -0.02 -0.05 -0.08 0.25 1.00 0.98
TOC -0.16 -0.18 -0.09 -0.16 0.10 -0.22 -0.20 -0.14 -0.12 -0.14 -0.16 -0.25 1.00 -0.25 -0.21 -0.14 -0.17 -0.04 -0.07 0.24 0.98 1.00
325
OCT 1.00 0.98 -0.36 -0.38 -0.01 0.39 0.51 0.39 0.45 -0.32 0.14 0.10 0.30 -0.45 0.62 0.10 -0.34 -0.33 -0.46 -0.15 -0.09 -0.52
DEC 0.98 1.00 -0.43 -0.45 -0.03 0.38 0.50 0.38 0.43 -0.33 0.16 0.11 0.32 -0.47 0.59 0.25 -0.29 -0.57 -0.41 -0.19 -0.13 -0.55
DOD -0.36 -0.43 1.00 0.33 -0.14 -0.21 -0.45 -0.26 -0.34 -0.30 0.04 -0.04 -0.29 0.19 -0.26 -0.39 0.00 -0.13 0.64 0.29 0.61 0.65
TED -0.38 -0.45 0.33 1.00 0.08 -0.08 -0.24 -0.18 -0.26 0.50 -0.18 -0.14 -0.41 -0.11 -0.33 -0.38 -0.45 -0.46 0.75 0.01 0.18 0.59
HXD -0.01 -0.03 -0.14 0.08 1.00 0.84 0.63 0.50 0.64 -0.06 0.36 0.54 0.08 0.20 0.10 0.22 -0.26 0.40 -0.48 -0.07 0.31 0.64
OCD 0.39 0.38 -0.21 -0.08 0.84 1.00 0.87 0.83 0.85 -0.18 0.60 0.66 0.20 0.02 0.08 0.35 -0.37 0.50 -0.44 -0.17 0.25 0.36
EIC 0.51 0.50 -0.45 -0.24 0.63 0.87 1.00 0.93 0.98 -0.15 0.69 0.74 0.52 0.14 0.10 0.43 -0.13 0.15 -0.18 -0.35 0.44 0.73
DOC 0.39 0.38 -0.26 -0.18 0.50 0.83 0.93 1.00 0.92 -0.13 0.84 0.79 0.44 0.17 -0.21 0.48 -0.10 0.20 0.02 -0.24 0.12 0.54
TTC 0.45 0.43 -0.34 -0.26 0.64 0.85 0.98 0.92 1.00 -0.20 0.77 0.84 0.60 0.25 0.10 0.38 -0.06 0.18 -0.18 -0.37 0.20 0.34
HXC -0.32 -0.33 -0.30 0.50 -0.06 -0.18 -0.15 -0.13 -0.20 1.00 -0.17 -0.07 -0.17 -0.21 -0.31 -0.26 -0.27 -055 0.03 0.09 -0.48 0.53
OCC 0.14 0.16 0.04 -0.18 0.36 0.60 0.69 0.84 0.77 -0.17 1.00 0.91 0.51 0.43 -0.34 0.39 0.05 0.53 0.16 -0.16 0.48 0.18
TCT 0.10 0.11 -0.04 -0.14 0.54 0.66 0.74 0.79 0.84 -0.07 0.91 1.00 0.67 0.43 -0.16 0.37 0.10 0.82 -0.04 -0.15 0.55 0.29
DTT 0.30 0.32 -0.29 -0.41 0.08 0.20 0.52 0.44 0.60 -0.17 0.51 0.67 1.00 0.42 0.28 0.24 0.55 0.64 0.04 -0.21 0.90 -0.22
TRT -0.45 -0.47 0.19 -0.11 0.20 0.02 0.14 0.17 0.25 -0.21 0.43 0.43 0.42 1.00 -0.20 -0.10 0.64 0.47 0.40 -0.25 0.86 0.30
HXT 0.62 0.59 -0.26 -0.33 0.10 0.08 0.10 -0.21 0.10 -0.31 -0.34 -0.16 0.28 -0.20 1.00 -0.26 -0.08 -0.04 -0.71 -0.26 0.03 -0.33
OTT 0.10 0.25 -0.39 -0.38 0.22 0.35 0.43 0.48 0.38 -0.26 0.39 0.37 0.24 -0.10 -0.26 1.00 0.15 0.29 0.25 -0.39 -0.25 -0.19
TTT -0.34 -0.29 0.00 -0.45 -0.26 -0.37 -0.13 -0.10 -0.06 -0.27 0.05 0.10 0.55 0.64 -0.08 0.15 1.00 0.84 0.51 -0.19 -0.15 -0.29
PSD -0.13 -0.07 -0.13 -0.46 0.00 0.00 0.15 0.20 0.18 -0.15 0.23 0.32 0.64 0.47 -0.04 0.29 0.84 1.00 0.23 -0.40 -0.02 -0.15
pH -0.46 -0.41 0.04 0.05 -0.48 -0.44 -0.18 0.02 -0.18 0.03 0.16 -0.04 0.04 0.40 -0.71 0.25 0.51 0.23 1.00 -0.01 -0.39 -0.13
EC -0.35 -0.39 0.29 0.91 -0.07 -0.17 -0.35 -0.24 -0.37 0.69 -0.16 -0.15 -0.41 -0.25 -0.36 -0.39 -0.49 -0.40 -0.01 1.00 0.27 0.56
TSS -0.09 -0.13 0.61 0.18 0.31 0.25 0.04 0.12 0.20 -0.08 0.48 0.55 0.20 0.26 0.03 -0.25 -0.15 -0.02 -0.39 0.27 1.00 0.34
TOC -0.52 -0.55 0.05 0.59 0.64 0.36 0.13 0.14 0.14 0.53 0.18 0.29 -0.22 0.30 -0.33 -0.19 -0.29 -0.15 -0.13 0.56 0.34 1.00
326
OCT 1.00 0.66 -0.04 -0.13 0.18 0.31 0.41 0.37 0.06 0.07 0.13 -0.21 -0.16 0.24 -0.24 -0.25 -0.39 -0.27 -0.43 -0.07 -0.29 -0.33
DEC 0.66 1.00 -0.35 0.02 0.28 -0.02 -0.02 -0.07 -0.24 -0.32 -0.35 -0.30 0.18 0.48 -0.30 -0.33 -0.31 0.17 -0.41 -0.07 -0.36 -0.25
DOD -0.04 -0.35 1.00 -0.20 -0.22 0.26 0.35 0.38 0.32 0.41 0.52 0.51 0.16 0.00 -0.23 0.79 0.20 -0.27 -0.16 0.08 0.67 0.56
TED -0.13 0.02 -0.20 1.00 0.68 0.13 -0.01 -0.15 -0.10 -0.03 0.00 -0.08 -0.02 -0.35 -0.16 -0.13 -0.09 0.39 -0.35 0.05 -0.28 -0.28
HXD 0.18 0.28 -0.22 0.68 1.00 0.29 0.14 0.00 0.02 0.21 0.15 -0.52 -0.29 -0.34 -0.22 -0.19 -0.22 0.16 0.52 -0.16 -0.28 -0.32
OCD 0.31 -0.02 0.26 0.13 0.29 1.00 0.94 0.89 0.88 0.82 0.88 0.27 -0.26 0.18 -0.28 -0.24 -0.01 -0.67 -0.26 0.15 -0.49 0.35
EIC 0.41 -0.02 0.35 -0.01 0.14 0.94 1.00 0.97 0.90 0.74 0.85 0.41 -0.17 0.25 -0.26 -0.15 0.16 -0.71 0.42 -0.05 -0.21 0.34
DOC 0.37 -0.07 0.38 -0.15 0.00 0.89 0.97 1.00 0.90 0.77 0.86 0.50 -0.08 0.28 -0.10 -0.11 0.29 -0.81 0.47 -0.06 0.35 0.40
TTC 0.06 -0.24 0.32 -0.10 0.02 0.88 0.90 0.90 1.00 0.75 0.83 0.50 -0.06 0.32 -0.22 -0.17 0.30 -0.67 -0.23 0.02 0.69 0.61
HXC 0.07 -0.32 0.41 -0.03 0.21 0.82 0.74 0.77 0.75 1.00 0.96 0.27 -0.21 -0.02 0.12 -0.02 0.08 -0.80 -0.42 0.12 0.41 0.29
OCC 0.13 -0.35 0.52 0.00 0.15 0.88 0.85 0.86 0.83 0.96 1.00 0.39 -0.23 0.00 0.01 0.04 0.14 -0.80 -0.14 0.14 0.12 0.27
TCT -0.21 -0.30 0.51 -0.08 -0.52 0.27 0.41 0.50 0.50 0.27 0.39 1.00 0.62 0.37 0.01 0.29 0.60 -0.30 -0.64 0.15 0.42 0.53
DTT -0.16 0.18 0.16 -0.02 -0.29 -0.26 -0.17 -0.08 -0.06 -0.21 -0.23 0.62 1.00 0.44 -0.07 0.28 0.44 0.27 -0.54 0.04 0.46 0.35
TRT 0.24 0.48 0.00 -0.35 -0.34 0.18 0.25 0.28 0.32 -0.02 0.00 0.37 0.44 1.00 -0.25 -0.34 -0.03 -0.19 -0.39 0.25 -0.59 0.25
HXT -0.24 -0.30 -0.23 -0.16 -0.22 -0.28 -0.26 -0.10 -0.22 0.12 0.01 0.01 -0.07 -0.25 1.00 -0.13 0.31 -0.39 0.37 -0.13 -0.24 -0.21
OTT -0.25 -0.33 0.79 -0.13 -0.19 -0.24 -0.15 -0.11 -0.17 -0.02 0.04 0.29 0.28 -0.34 -0.13 1.00 0.31 0.69 -0.55 0.12 0.85 0.00
TTT -0.39 -0.31 0.20 -0.09 -0.22 -0.01 0.16 0.29 0.30 0.08 0.14 0.60 0.44 -0.03 0.31 0.31 1.00 -0.19 -0.40 -0.18 0.41 0.44
PSD -0.27 0.17 -0.27 0.39 0.16 -0.67 -0.71 -0.81 -0.67 -0.80 -0.80 -0.30 0.27 -0.19 -0.39 0.19 -0.19 1.00 0.22 0.12 0.13 -0.18
pH -0.43 -0.41 -0.16 -0.05 0.20 -0.26 -0.42 -0.47 -0.23 -0.02 -0.14 -0.64 -0.54 -0.39 0.07 -0.05 -0.40 0.22 1.00 -0.01 -0.08 -0.18
EC -0.37 -0.07 0.28 0.05 -0.16 0.15 -0.05 -0.06 0.12 0.22 0.14 0.35 0.40 0.25 -0.30 0.22 -0.18 0.12 -0.01 1.00 0.39 0.46
TSS -0.29 -0.36 0.67 -0.28 -0.28 -0.09 -0.01 0.05 0.09 0.11 0.12 0.42 0.46 -0.19 -0.24 0.85 0.41 0.13 -0.08 0.39 1.00 0.47
TOC -0.33 -0.25 0.06 -0.28 -0.32 0.35 0.34 0.40 0.61 0.29 0.27 0.53 0.35 0.25 -0.21 0.00 0.44 -0.18 -0.18 0.46 0.47 1.00
327
Table A.3.10 Simple bi-variate correlation matrix between target variables for the <1 µm particle size fraction from original data given in supplementary
Table A.3.5
OCT DEC DOD TED HXD OCD EIC DOC TTC HXC OCC TCT DTT TRT HXT OTT TTT PSD pH EC TDS DOC
OCT 1.00 0.81 -0.17 0.44 -0.19 -0.39 -0.28 -0.29 -0.23 -0.34 -0.30 -0.51 -0.27 -0.62 -0.26 -0.31 -0.16 -0.06 0.13 -0.29 -0.30 -0.49
DEC 0.81 1.00 -0.08 0.11 -0.25 -0.36 -0.32 -0.36 -0.27 -0.27 -0.32 -0.29 -0.19 -0.44 -0.24 0.08 0.16 0.13 0.18 -0.07 -0.11 -0.31
DOD -0.17 -0.08 1.00 0.09 -0.19 -0.27 -0.32 -0.20 -0.33 -0.03 -0.26 0.07 0.06 0.28 0.78 -0.28 -0.21 0.32 -0.05 0.08 0.09 0.05
TED 0.44 0.11 0.09 1.00 0.08 -0.03 -0.19 -0.15 -0.21 0.03 -0.10 -0.01 0.49 0.10 -0.11 -0.31 -0.23 -0.07 0.16 0.16 0.01 0.06
HXD -0.19 -0.25 -0.19 0.08 1.00 0.91 0.78 0.63 0.63 0.87 0.81 0.67 0.48 0.49 -0.11 -0.25 -0.05 0.02 -0.08 0.04 0.09 0.05
OCD -0.39 -0.36 -0.27 -0.03 0.91 1.00 0.73 0.57 0.54 0.91 0.72 0.85 0.62 0.67 -0.17 0.10 0.26 -0.05 -0.20 -0.05 0.01 0.06
EIC -0.28 -0.32 -0.32 -0.19 0.78 0.73 1.00 0.96 0.97 0.71 0.98 0.47 0.14 0.29 -0.19 0.02 0.18 -0.27 -0.17 0.04 0.06 0.11
DOC -0.29 -0.36 -0.20 -0.15 0.63 0.57 0.96 1.00 0.97 0.63 0.94 0.38 0.06 0.27 -0.03 0.06 0.20 -0.26 -0.17 0.09 0.10 0.16
TTC -0.23 -0.27 -0.33 -0.21 0.63 0.54 0.97 0.97 1.00 0.54 0.95 0.29 -0.02 0.12 -0.21 0.04 0.16 -0.28 -0.13 0.08 0.08 0.13
HXC -0.34 -0.27 -0.03 0.03 0.87 0.91 0.71 0.63 0.54 1.00 0.74 0.85 0.62 0.80 0.12 0.09 0.22 0.13 -0.16 0.11 0.16 0.13
OCC -0.30 -0.32 -0.26 -0.10 0.81 0.72 0.98 0.94 0.95 0.74 1.00 0.48 0.21 0.36 -0.15 -0.02 0.09 -0.13 -0.08 0.15 0.16 0.16
TCT -0.51 -0.29 0.07 -0.01 0.67 0.85 0.47 0.38 0.29 0.85 0.48 1.00 0.80 0.86 0.10 0.35 0.51 0.09 -0.12 0.22 0.28 0.33
DTT -0.27 -0.19 0.06 0.49 0.48 0.62 0.14 0.06 -0.02 0.62 0.21 0.80 1.00 0.77 -0.08 0.22 0.28 -0.01 0.21 0.26 0.13 0.09
TRT -0.62 -0.44 0.28 0.10 0.49 0.67 0.29 0.27 0.12 0.80 0.36 0.86 0.77 1.00 0.33 0.30 0.27 0.27 -0.14 0.06 0.14 0.14
HXT -0.26 -0.24 0.78 -0.11 -0.11 -0.17 -0.19 -0.03 -0.21 0.12 -0.15 0.10 -0.08 0.33 1.00 -0.22 -0.19 0.54 -0.10 0.26 0.26 0.25
OTT -0.31 0.08 -0.28 -0.31 -0.25 0.10 0.02 0.06 0.04 0.09 -0.02 0.35 0.22 0.30 -0.22 1.00 0.87 -0.08 -0.10 -0.03 -0.05 0.06
TTT -0.16 0.16 -0.21 -0.23 -0.05 0.26 0.18 0.20 0.16 0.22 0.09 0.51 0.28 0.27 -0.19 0.87 1.00 -0.21 -0.21 -0.03 -0.02 0.11
PSD -0.06 0.13 0.32 -0.07 0.02 -0.05 -0.27 -0.26 -0.28 0.13 -0.13 0.09 -0.01 0.27 0.54 -0.08 -0.21 1.00 -0.12 0.18 0.15 0.03
pH 0.13 0.18 -0.05 0.16 -0.08 -0.20 -0.17 -0.17 -0.13 -0.16 -0.08 -0.12 0.21 -0.14 -0.10 -0.10 -0.21 -0.12 1.00 -0.01 -0.04 -0.18
EC -0.29 -0.07 0.68 0.16 0.04 -0.05 0.04 0.09 0.08 0.11 0.15 0.22 0.26 0.36 0.26 -0.03 -0.03 0.18 -0.01 1.00 0.99 0.93
TDS -0.30 -0.11 0.69 0.21 0.09 0.01 0.06 0.10 0.08 0.16 0.16 0.28 0.33 0.41 0.26 -0.05 -0.02 0.15 -0.04 0.99 1.00 0.95
DOC -0.49 -0.31 0.65 0.06 0.05 0.06 0.11 0.16 0.13 0.13 0.16 0.33 0.29 0.41 0.25 0.06 0.11 0.03 -0.18 0.93 0.95 1.00
328
Manuscript Title
Prediction Model of the Build-up of Volatile Organic Compounds
on Urban Roads
Parvez Mahbub, Ashantha Goonetilleke, Godwin A. Ayoko
Number of Pages: 12
Number of Tables: 4
329
330
Quality Control Measures in Sample Testing
For quality control, calibration standards, internal standards, spikes and blanks were
used. Ten different calibration standards (Chemservice® THM501 – 1RPM) at 1, 2,
5, 10, 20, 50, 100, 150, 200 and 250 µg/L concentrations were prepared for each
target analyte. Volatile internal standards (Chemservice® IS-8260ARPM) consisting
of flourobenzene, chlorobenzene-d5 and 1, 4- dichlorobenzene-d4 were added to
each sample and standards at 50 µg/L concentration.
The detection limit was found to be 0.01 µg/L for the test method. Three quality
control standards at 10, 50 and 100 µg/L concentrations were prepared
with the quality control standards at a concentration of 20 µg/L. The percentage
R = (C1 − C 2) / C1×100 ------------------------------------------------------------------- (A.4.1)
final spike concentration. The percentage recoveries were found to be within 90%-
95%.
Principal Component Analysis (PCA)
The principal component analysis (PCA) is a data extraction method that extracts
information from a data matrix by the projection of objects and variables to the
principal components. The principal components (PCs) and are considered as the
latent variables which are linear combinations of the original variables of the dataset
331
(1). The PCs arise from the measured variables in the data matrix and hence,
regarded as dependent variables. The analysis is done such that once the first PC is
extracted, all succeeding PCs must be orthogonal to each other. Also, the first PC
accounts for most data variance and the remaining variance is described by
successive PCs in decreasing order.
This type of approach simplifies a complex data structure and reveals the latent
variables that are easily explicable in terms of their correlations with the objects.
Graphically, the outcome of PCA is presented as scores and loadings plots, which
reveal the patterns in the objects and variables, respectively. While the scores plot
describes the relationships among the objects in terms of their projections on the
newly extracted PCs, the loadings plot describes the contributions of original
variables on the extracted PCs. When scores and loadings vectors are displayed on
the same visual representation, a biplot is obtained, which provides information
about the degree of similarity between objects and correlation between variables.
Detailed descriptions of PCA can be found elsewhere (2).
Two Phase Factor Analysis (FA)
Factor analysis (FA) is another type of data extraction method that has been used in
prediction models for spatially different data observations (3) including
hydrocarbons classifications (4). As opposed to PCA, the factors in the FA method
are extracted as independent variables that account only for the shared variance,
namely, the covariance of the variables in the data matrix.
332
several extraction processes described in Meyers et al. (1), the Principal Component
Extraction (PCE) is the most efficient extraction process that could be performed by
microprocessors. Factor rotation, on the other hand, is used to achieve simple
structures where the measured variables could be associated with each factor in
terms of their strong correlations. Amongst several factor rotation methods, the
several environmental studies (e.g., 5). This rotation method allows a factor to
correlate quite strongly with some variables (correlations closer to 1), but more
weakly with the other variables (correlations closer to 0) in the data matrix. The total
variance explained by the factors before and after rotation remains the same. This
study adopted the FA method including the PCE followed by varimax with a view to
strengthen a validation strategy for the PCA components through the factor
extraction process.
Partial Least Squares Regression (PLS)

Partial least squares regression (PLS) is a multivariate regression method that is used
to predict response variables (R) from predictors or independent variables (C). This
method simultaneously estimates the underlying factors in both the response and the
predictor data matrix. The matrices are decomposed as follows:
R = TP + E ------------------------------------------------------------------------- (A.4.2)
C = UQ + F ------------------------------------------------------------------------- (A.4.3)
underlying factors of R and C in equations A.4.2 and A.4.3 respectively.
333
Detailed discussions of PLS estimation of responses from predictors can be found in
Beebe and Kowalski (6). Several environmental studies (e. g., 7) and analytical
chemistry studies (e. g., 8) have applied PLS along with other prediction methods
such as PCR (principal component regression) and established PLS as a better tool
than PCR when there are independently varying major response components. In this
study SIRIUS (9) and Statistical Excel (10) was used for PCA data analyses, PLS
regression and plotting purposes. This study also used SPSS (11) for factor analyses.
334
Additional Tables
Table A.4.1 Traffic and pavement characteristics of eleven study sites

Site Name Average Surface
Land Use Daily Traffic Volume to Texture Lane Top Coat Age of the
Suburb Capacity
Label (ADT), Depth Width, m Material section, yr
vehicles/day Ratio (V/C) (STD), mm
Abraham Road Upper

Residential 13028 1.11 0.6467 3.5 DG14a 3
C6 Coomera
Reserve Road Upper

Residential 6339 0.45 0.7505 3.5 DG14a 3
R3 Coomera
Peanba Park road Upper

Residential 581 0.15 0.6844 2.8 DG10b 4
R2 Coomera
Billinghurst Cres Upper

Residential 5936 0.74 0.7015 2.9 DG10b 10
R1 Coomera
Beattie Road
Industrial Coomera 2670 0.24 0.7074 3.5 DG14a 2
I4
Shipper Drive
Industrial Coomera 7530 0.55 0.6788 3.5 DG14a 6
I5
Hope Island Road

Commercial Helensvale 7534 0.57 0.7254 3.4 DG14a 3
C10
Lindfield Road
Commercial Helensvale 2312 0.33 0.9417 3.3 DG10b 10
C9
Town Centre Drive

Commercial Helensvale 24506 0.62 0.6416 3.5 DG14a 4
C7
Dalley Park Drive

Residential Helensvale 3534 0.42 0.8342 2.9 DG10b 10
R11
Discovery Drive
Residential Helensvale 9116 0.25 0.6957 2.9 DG14a 2
R8
a
b
335
Table A.4.2 Chemical composition (mean±standard deviation) of VOCs at five size fractions; inter-site percentage coefficient of variations ranging from
29%-44% for eleven sites
Size Fractions
Site
>300µm 150-300 µm 75-150 µm 1-75 µm <1 µm
Name
TOL, ETB, MPX, OX, TOL, ETB, MPX, OX, TOL, ETB, MPX, OX, TOL, ETB, MPX, OX, TOL, ETB, MPX, OX,
Label
µg/m2 µg/m2 µg/m2 µg/m2 µg/m2 µg/m2 µg/m2 µg/m2 µg/m2 µg/m2 µg/m2 µg/m2 µg/m2 µg/m2 µg/m2 µg/m2 µg/m2 µg/m2 µg/m2 µg/m2
Hope
0.071 0.090 0.030 0.018
Island 0.154 0.103 0.161± 0.055± 0.136 0.133 0.043± 0.123± 0.061± 0.102 0.031± 0.066± 0.029± 0.060± 0.018± 0.061±
±0.00 ±0.00 ±0.00 ±0.00
Road ±0.03 ±0.01 0.02 0.001 ±0.03 ±0.04 0.01 0.01 0.02 ±0.03 0.007 0.005 0.009 0.006 0.007 0.004
2 8 32 6
C10
Shipper 0.034 0.032 0.098 0.095 0.024 0.013
0.156 0.109± 0.043± 0.165 0.045± 0.185± 0.033± 0.122 0.042± 0.058± 0.029± 0.022± 0.016± 0.013±
Drive ±0.00 ±0.00 ±0.00 ±0.00 ±0.00 ±0.00
±0.02 0.04 0.005 ±0.07 0.005 0.06 0.009 ±0.06 0.004 0.002 0.004 0.001 0.005 0.006
I5 5 3 4 2 5 5
Billinghur 0.032 0.141 0.035 0.100 0.098 0.024 0.012
0.175 0.116± 0.048± 0.040± 0.140± 0.029± 0.046± 0.109 0.060± 0.028± 0.075± 0.020± 0.035±
s Cres ±0.00 ±0.00 ±0.00 ±0.00 ±0.00 ±0.00 ±0.00
±0.05 0.005 0.004 0.002 0.06 0.001 0.009 ±0.02 0.003 0.002 0.001 0.003 0.001
R1 6 2 2 5 3 1 5
Peanba 0.071 0.091 0.028 0.084 0.078 0.061
0.122 0.085± 0.026± 0.032± 0.145± 0.044± 0.110 0.036± 0.135 0.153± 0.045± 0.179± 0.055± 0.141±
Park road ±0.00 ±0.00 ±0.00 ±0.00 ±0.00 ±0.00
±0.06 0.003 0.001 0.008 0.07 0.008 ±0.05 0.003 ±0.08 0.05 0.008 0.02 0.001 0.02
R2 2 3 3 3 1 1
Dalley
0.014 0.067 0.022 0.057 0.050 0.013 0.018 0.013
Park 0.031 0.031± 0.013± 0.013± 0.065± 0.022± 0.013± 0.016± 0.013± 0.013± 0.013± 0.013±
±0.00 ±0.00 ±0.00 ±0.00 ±0.00 ±0.00 ±0.00 ±0.00
Drive ±0.01 0.001 0.005 0.001 0.002 0.006 0.001 0.002 0.001 0.001 0.001 0.002
2 2 3 1 4 1 3 3
R11
Lindfield 0.028 0.025 0.099 0.027 0.025
0.147 0.097± 0.038± 0.165 0.103 0.042± 0.144± 0.018± 0.041± 0.142 0.091± 0.048± 0.127± 0.019± 0.060±
Road ±0.00 ±0.00 ±0.00 ±0.00 ±0.00
±0.02 0.005 0.005 ±0.05 ±0.08 0.005 0.06 0.004 0.001 ±0.02 0.006 0.004 0.05 0.002 0.004
C9 1 3 3 3 3
Town
0.076 0.043 0.064 0.059
Centre 0.134 0.124± 0.048± 0.149 0.118 0.048± 0.149± 0.044± 0.113 0.048± 0.153 0.136± 0.051± 0.172± 0.087± 0.175±
±0.00 ±0.00 ±0.00 ±0.00
Drive ±0.04 0.04 0.002 ±0.05 ±0.04 0.001 0.03 0.005 ±0.04 0.002 ±0.05 0.06 0.001 0.04 0.001 0.03
8 4 2 3
C7
Abraham 0.040 0.080 0.099 0.065 0.042
0.189 0.094 0.177± 0.062± 0.159 0.110 0.037± 0.099± 0.026± 0.026± 0.097± 0.029± 0.130± 0.027± 0.085±
Road ±0.00 ±0.00 ±0.00 ±0.00 ±0.00
±0.03 ±0.06 0.09 0.003 ±0.07 ±0.05 0.003 0.003 0.001 0.003 0.004 0.006 0.06 0.002 0.002
C6 2 5 3 6 5
Discovery 0.050 0.050 0.050
0.152 0.137 0.171± 0.064± 0.183 0.128 0.055± 0.150± 0.102± 0.178 0.047± 0.124 0.077± 0.040± 0.160± 0.074± 0.137±
Drive ±0.00 ±0.00 ±0.00
±0.05 ±0.05 0.08 0.008 ±0.07 ±0.05 0.005 0.003 0.06 ±0.05 0.004 ±0.07 0.003 0.005 0.05 0.003 0.05
R8 1 3 1
Beattie 0.086 0.024 0.060 0.076 0.039 0.014 0.030
0.120 0.044 0.098± 0.036± 0.020± 0.097± 0.016± 0.029± 0.035± 0.013± 0.086± 0.059± 0.073±
Road ±0.00 ±0.00 ±0.00 ±0.00 ±0.00 ±0.00 ±0.00
±0.04 ±0.01 0.007 0.003 0.006 0.005 0.003 0.001 0.001 0.001 0.004 0.004 0.005
I4 3 3 4 9 1 5 7
Reserve 0.048 0.093 0.027 0.013
0.141 0.056 0.107± 0.040± 0.149 0.105 0.042± 0.174± 0.031± 0.123 0.041± 0.099± 0.037± 0.036± 0.013± 0.013±
Road ±0.00 ±0.00 ±0.00 ±0.00
±0.04 ±0.02 0.003 0.001 ±0.05 ±0.04 0.005 0.06 0.004 ±0.05 0.003 0.005 0.008 0.006 0.005 0.002
R3 4 6 6 4
336
Table A.4.3 Correlation coefficients matrix achieved through VARIMAX factor rotation
including pH and EC as new variables with bold coefficients showing variables strongly
associated with corresponding factors
Correlation coefficients after rotation Correlation coefficients after rotation for
Measured
variables for total particulate fraction all fractions taken together
Factor 1 Factor 2 Factor 1 Factor 2
TSS .984 .119 .199 .904
ADT -.922 -.257 -.359 -.820
TOC .881 .443 .418 .906
STD .729 .602 .667 .664
pH .746 .451 .389 .807
EC .726 .501 .486 .740
V/C .042 .891 .881 .052
OX .490 .871 .863 .498
MPX .418 .831 .801 .448
TOL .529 .789 .843 .475
ETB .508 .797 .818 .487
Table A.4.4 PLS calibration set (X1*, X2*: Two independent factors found in FA method;
TOL= Toluene, ETB=Ethylbenzene, MPX= Meta and Para Xylene, OX= Ortho Xylene; E1-E12
represents the twelve individual experiments; C1-C3 represents the replicate experiments at
centre)
Experiments X1 X2 TOL ETB MPX OX TSS TOC ADT VC STD
E1 -1.000 -1.000 0.031 0.014 0.031 0.001 45.07 2.38 2670 0.24 0.834
E2 1.000 -1.000 0.189 0.094 0.177 0.062 1.87 2.56 7534 0.57 0.647
E3 -1.000 1.000 0.120 0.044 0.098 0.036 0.27 1.95 3534 0.42 0.707
E4 1.000 1.000 0.067 0.022 0.057 0.001 123.73 32.24 2670 0.24 0.834
E5 -1.414 0.000 0.159 0.040 0.110 0.037 0.27 2.03 7534 0.57 0.647
E6 1.414 0.000 0.130 0.027 0.085 0.042 36.53 7.67 7534 0.57 0.647
E7 0.000 -1.414 0.065 0.022 0.050 0.001 102.67 3.05 2670 0.24 0.834
E8 0.000 1.414 0.099 0.026 0.080 0.026 20.80 2.19 7534 0.57 0.647
E9 0.000 0.000 0.001 0.001 0.001 0.001 41.33 9.28 2670 0.24 0.834
E10 0.000 0.000 0.001 0.018 0.016 0.001 785.07 5.67 2670 0.24 0.834
E11 0.000 0.000 0.099 0.065 0.097 0.029 152.53 5.25 7534 0.57 0.647
E12 0.000 0.000 0.039 0.014 0.035 0.001 28.27 2.22 3534 0.42 0.707
C1 0.000 0.000 0.097 0.016 0.076 0.029 8.27 1.75 3534 0.42 0.707
C2 0.000 0.000 0.086 0.024 0.060 0.020 1.07 1.85 3534 0.42 0.707
C3 0.000 0.000 0.086 0.059 0.073 0.030 25.33 3.28 3534 0.42 0.707
* As the values of independent factors X1 and X2 are unknown, the coded values
(9) for a two factor orthogonal design were used in the optimisation process.
337
Additional Figures
Figure A.4.1 Eleven study sites are shown on the map of selected suburbs of Coomera, Upper
Coomera and Helensvale; Corresponding traffic parameters and labels of each site are shown
alongside the site name
338
1.0
I5 5
V/C >300 µm
0.8 R2 R11 150-300 µm
4
0.6 I4
0.4
ADT 3
PC 2 (3.5%)
0.2
PC 2 (28.0%)
C9
C6 ETSTD
B 2
0.0
TOL
OX
MPX
R3 TOC TOC
-0.2 R8 R1 TSS
1
R11 TSS
-0.4
C7 R3
0 C9 STD
-0.6 R8 I5 MPEXTB
C10 R2 TOL
V/C
C7 C6 I4 OX R1
-0.8 -1
C10 ADT
-1.0
-5 0 5 10 -2
-4 1 6
PC 1 (93.6%)
PC 1 (69.5%)
(a) 1.5
(b) 2.5
R11 R11
75-150 µm 1-75 µm
2.0
1.0 1.5
TSS
1.0
C10 TSS C10
PC 2 (9.9%)
0.5
R3
PC2 (3.8%)
C7 0.5
STD R1
C7 TOC
TOC R1 0.0 C6
STD R8
C6
0.0 R3 C9
TOL
C9 -0.5 ETB
ETB I4
R8
TOL I5 MO X
PX
MPX
OX -1.0
I4 R2 V/C
-0.5
ADT
R2 I5 -1.5
ADT
V/C
-2.0
-5 0 5 10
-1.0
-5 0 5 10
PC 1 (83.3%)
PC 1 (92.9%)
(c) (d)
2.0
<1 µm
R11
1.5
I5
1.0
R3
STD
TDS
DOC
PC2(9.2%)
0.5 R1
C10 V/C
C9
0.0
C6
TOL
-0.5
ADT
ETB
R8 I4
-1.0 C7 MPX
R2 OX
-1.5
-5 0 5 10
PC 1 (86.6%)
(e)
Figure A.4.2 PCA biplots for (a-d) particulate fractions and (e) potential dissolved fraction with
eleven land use scores shown with initials C, I and R for commercial, industrial and residential
respectively
339
R11
1.0 I5
R2 V/C 4 150-300 µm
>300 µm
0.8
I4
0.6 3
PC 2 (18.7%)
0.4
PC 2 (3.2%)
ADT 2
0.2 C9 TOC
ETB TSS
C6 MPX
TOL
0.0 OD
ST X
pH 1
R3 R8 TOC
-0.2 R1
EC
R11 TSS R3 EC
-0.4 0 C7 STD
C10 R2 pH
V/C
C9 R8 TOL
C6 I4 I5 MP
OX
EX
TB
R1
-0.6
C7 ADT
-1
-0.8 C10
-1.0 -2
-3 2 7
-5 0 5
PC 1 (94.0%)
PC 1 (78.8%)
(a) (b)
2.5 R11
1-75 µm
1.0 I5
R2 V/C 2.0
75-150 µm
I4
1.5 TSS
0.5
ADT
1.0 C10
C9 MPX
TOL
PC 2 (8.5%)
R8 OX R3
ETB EC
R3
PC 2 (3.4%)
0.0 pH R1 0.5 C7 R1
STD
C6 TOC TOC
STD
EC 0.0 C9C6 pH
C7 R8
-0.5 TSS OX
-0.5 I4 TOL
ETB
C10 I5 MPX
-1.0
-1.0 R11 R2
ADT
-1.5 V/C
-1.5 -2.0
-5 0 5 10 -5 0 5
PC1 (93.8%) PC 1 (85.8%)
(c) (d)
2.5 R11
1.5
R11 Total Particulate
<1µm 2.0
1.0 R3
C10 1.5
I5 E
TCDS
DOC C10 TSS
C9 STD 1.0
0.5 pH R1
C7R3 TOC
PC 2 (6.8%)
PC 2 (9.9%)
0.5 EC
R1
C6
V/C C6
0.0 C9
0.0 STD
pH
MPX
-0.5 R8 TOL
OX
-0.5 TOL I4 ETB
ADT
C7
R8I4
R2
-1.0 ADT
-1.0 ETB I5V/C

R2 OX -1.5
MPX
-1.5 -2.0
-5 0 5 -5 0 5
PC 1 (89.1%) PC 1 (85.1%)
(e) (f)
Figure A.4.3 PCA biplots for (a-d) particulate fractions, (e) dissolved fraction and (e) total
particulates (1->300 µm) incorporating pH and EC as new variables
340
References
(1) Meyers, L. S.; Gamst, G.; Guarino, A. J. Applied Multivariate Research:
Design and Interpretation. Sage Publications Inc. 2006, 481-482.
(2) Massart, D. L.; Vandeginste, B. G. M.; Buydens, L. M. C.; De Jong, S.;

Lewi, P. J.; Smeyers-Verbeke, J. Handbook of Chemometrics and
Qualimetrics Part A, Elsevier, 1997, 771-804.
(3) Christensen, W. F.; Amemiya, Y. Modeling and prediction for multivariate

spatial factor analysis. J. Stat. Plan. Infer. 2003, 115, 543-564.
(4) Zhu, X.; Zhang, L.; Che, X.; Wang, L. The classification of hydrocarbons
with factor analysis and the PONA analysis of gasoline. Chemom. Intell.
Lab. Syst. 1999, 45, 147-155.
(5) Yidana, M. S.; Ophori, D.; Yakubo, B. A multivariate statistical analysis of

surface water chemistry data—The Ankobra Basin, Ghana. J. Environ.
Manage. 2008, 86, 80-87.
(6) Beebe, K. R.; Kowalski, B. R. An introduction to multivariate calibration

and analysis. Anal. Chem. 1987, 59, 1007A-1017A.
(7) Sotelo, M. F.; Tailer, R.; Vives, I.; Grimalt, J. O. Assessment of the
environmental and physiological processes determining the accumulation of
organochlorine compounds in European mountain lake fish through
multivariate analysis (PCA and PLS). Sci. Total Environ. 2008, 404, 148-
161.
(8) Ni, Y.; Wang, L.; Kokot, S. Simultaneous determination of nitrobenzene and
nitro-substituted phenols by differential pulse voltammetry and
chemometrics. Anal. Chim. Acta, 2001, 431, 101-113.
(9) SIRIUS (2008). SIRIUS version 7.1© copyright 1987-2008, Pattern

Recognition Systems AS, Help Topics, http://www.prs.no.
(10) Statistical Excel. Version 1.8 © StatisXL 2007.

341
342
Manuscript Title
Prediction of the Wash-off of Traffic Related Semi and Non Volatile Organic
Compounds from Urban Roads under Changed Rainfall Characteristics
Parvez Mahbub, Ashantha Goonetilleke, Godwin A. Ayoko
Number of Pages: 14
Number of Tables: 3
343
344
Testing of SVOCs and NVOCs
Calibration standards (17 component FTRPH calibration standards from
Accustandard®) were prepared at 0.1, 0.5, 0.7, 1, 1.4, 7, 10, 28, 50 mg/L
concentrations for each target analyte. The prescribed DRO internal standard
(Sigma-Aldrich®) consisting of acenaphthene-d10, chrysene-d12, naphthalene-d8,
perylene-d12, phenanthrene-d10, 1, 4-dichlorobenzene-d4 were added to each
sample and standards at 5 mg/L concentration. Field blanks were used during each
field sample collection episode and all results were blank corrected. Seven field
blanks were used to establish the limits of detection for each analyte.
Three quality control standards (TPH Mix-1-DRO certified reference materials from
Sigma-Aldrich®) at 1, 10 and 50 mg/L concentrations were included in each batch
for comparison with the calibration standards. One sample from each batch was
spiked with another quality control standard at a concentration of 35 mg/L.
Surrogate standards (Accustandard®) consisting of 10 mg/L of n-triacontane-d62
were added to seven randomly chosen samples. The spike or surrogate recoveries
were calculated using the following equation:
R = (C1 − C 2) / C1× 100 ----------------------------------------- (A.5.1)
where R= percent recovery, C1= initial spike/surrogate concentration before
extraction, C2= final spike/surrogate concentration.
The extractions of SVOCs and NVOCs were performed by separatory funnel liquid-
liquid extraction with 250 mL hexane as the exchange solvent according to USEPA
method 3510C (18). The extracted samples were then cleaned using standard
column cleanup protocol with 5 cm silica gel and 5 cm pyrex® glass wool topped
345
with 5 cm anhydrous Na2SO4 (18). Sample concentration was then carried out using
the Kuderna-Danish apparatus followed by the nitrogen blowdown technique (18).
The sample concentration was continued until a final concentrated volume of 1 mL
was achieved for Gas Chromatographic (GC) analyses.
A temperature programmed GC capillary column (HP5MS Agilent®) of 30 m
length, 0.32 mm internal diameter and 0.25 µm film thickness was used in the study
to separate the analytes, which were then detected by a mass spectrometer interfaced
to the GC. A splitless sample injection of 2 µL at an inlet temperature of 280°C,
inlet pressure of 35.58 kN/m2 (5.16 psi) and a flowrate of 2.4 mL/min was used. The
initial oven temperature was set at 40°C with intial temperature holding time of 12
min., followed by an increase of 10°C per min. until the oven temperature reached
300°C with final temperature holding time of 20 min.

PCA
The PCA technique is used to transform the inter-correlated original variables to a
new orthogonal (uncorrelated) set of Principal Components (PCs) such that the first
PC contains most of the data variance and the second PC contains the second largest
variance and so on. Though PCA produces the same number of PCs as the original
variables, the first few contain most of the variance. Therefore, the first few PCs are
often selected for interpretation. Hence, the PCA technique reduces the number of
variables without losing useful information contained in the original data set.
The PCs are often regarded as the latent variables that can be easily explained in
terms of their correlations with the objects. Graphically, the outcome of PCA is
346
presented as scores and loadings plots, which reveal the patterns in the objects and
variables, respectively. The visual representations of scores and loadings vectors of
corresponding objects and variables are known as a biplot. It provides information
about the degree of similarity between objects and correlation between variables.
Detailed descriptions of the PCA technique can be found in Massart et al. (28). This
study used Sirius software (22) for PCA.
FA
Factor analysis is applied to explain the correlation between a set of variables in
terms of a small number of underlying factors on the basis of the shared variance-
covariance of the variables in the analysis (29). As opposed to the PCA technique
where the measured variables are analogous to independent variables and the PCs
are analogous to dependent variables, factors in factor analysis are analogous to
independent variables and measured variables are analogous to dependent variables.
several extraction processes described in Meyers et al. (29), the principal component
extraction (PCE) is the most efficient extraction process that could be performed by
microprocessors. Factor rotation, on the other hand, is used to achieve simple
structures where the measured variables could be associated with each factor in
terms of their strong correlations. Amongst several factor rotation methods, the
several environmental studies (e.g., 30, 31). The FA technique has also been
successfully used in hydrocarbon classification (32) and source apportionment (33).
This study adopted the FA method including the PCE followed by varimax with a
view to strengthen a validation strategy for the PCA components. FA was
undertaken using PASW statistics software (21).
347
Experimental Design
Experimental design is a chemometric approach that deals with optimisation and
understanding of a system’s performance. Various experimental designs are
discussed in detail by Deming and Morgan (34). Experimental designs are used to
reduce the number of experiments, incorporate the interactions between variables as
well as to select the optimal experimental conditions (35). The orthogonal rotatable
central composite design was successfully used by Sivakumar et al. (26) for
optimising the chromatographic separations in relation to the complex experimental
conditions during various commercial pharmaceutical preparations. As the total
number of experiments in this study are quite high (110 for the five size fractions)
and the interactions between the measured variables are complex under the changed
climatic conditions, this study also adopted the orthogonal rotatable central
composite design to optimise three calibration matrices for light SVOCs, heavy
SVOCs and NVOCs. Sirius software (22) was used in this study for experimental
design.
PLS
PLS is a multivariate regression method that is used to predict response variables
(R) from predictor variables (C). This method simultaneously estimates the
underlying factors in both the response and the predictor data matrix (36). The
matrices are decomposed as follows:
R = TP + E ------------------------------------------------------------------------- (A.5.2)
C = UQ + F ------------------------------------------------------------------------- (A.5.3)
underlying factors of R and C in equations A.5.2 and A.5.3 respectively. While
348
estimating the factors using both R and C matrices, PLS technique assumes that the
factors have the following relationship:
u = bt + ε --------------------------------------------------------------------------- (A.5.4)
where u and t matrices are the column vectors of U and T respectively and b is the
inner relationship between u and t and is used to calculate subsequent factors if
necessary.
Several studies have used the PLS technique with varimax rotation to improve the
explanatory abilities of prediction models (e.g., 37). The PLS was also suggested as
the preferred prediction methodology by Ni et al. (27) in analytical chemistry
studies. In this study, PLS was performed using the Sirius software (22).
349
Additional Figures
Figure A.5.1 Four wash-off road sites located within 5 km radius of the meteorological station
40166 (adapted from the Google Earth map services)
350
Concentration, ppm
0.5
1.5
2.5
3.5
4.5
5.5
6.5
-0.5
OCT(predicted)
OCT(observed)
DEC(predicted)
median
minimum
DEC(observed)
maximum
75th qua rtile
25th qua rtile
DOD(predicted)
DOD(observed)
TED(predicted)
TED(observed)
HXD(predicted)
HXD(observed)
OCD(predicted)
OCD(observed)
EIC(predicted)
observed and predicted target compounds

EIC(observed)
DOC(predicted)
DOC(observed)
351
TTC(predicted)
TTC(observed)
HXC(predicted)
HXC(observed)
OCC(predicted)
OCC(observed)
TCT(predicted)
TCT(observed)
DTT(predicted)
DTT(observed)
150-300µm
TRT(predicted)
TRT(observed)
HXT(predicted)
HXT(observed)
OTT(predicted)
OTT(observed)
Figure A.5.2 Distributions of the box plot statistics at 150-300 µm particulate fraction for
TTT(predicted)
TTT(observed)
Concentration, ppm
0
1
2
3
4
0.5
1.5
2.5
3.5
4.5
-0.5
OCT(predicted)
OCT(observed)
DEC(predicted)
DEC(observed)
median
minimum
DOD(predicted)
maximum
75th quartile
25th quartile
DOD(observed)
TED(predicted)
TED(observed)
HXD(predicted)
HXD(observed)
OCD(predicted)
OCD(observed)
EIC(predicted)
observed and predicted target compounds

EIC(observed)
DOC(predicted)
DOC(observed)
352
TTC(predicted)
TTC(observed)
HXC(predicted)
HXC(observed)
OCC(predicted)
OCC(observed)
TCT(predicted)
TCT(observed)
DTT(predicted)
DTT(observed)
75-150 µm
TRT(predicted)
TRT(observed)
HXT(predicted)
HXT(observed)
OTT(predicted)
OTT(observed)
Figure A.5.3 Distributions of the box plot statistics at 75-150 µm particulate fraction for
TTT(predicted)
TTT(observed)
Concentration, ppm
0.5
1.5
2.5
3.5
4.5
5.5
6.5
7.5
8.5
-0.5
OCT(predicted)
OCT(observed)
DEC(predicted)
DEC(observed)
DOD(predicted)
media n
minimum
ma ximum
DOD(observed)
75th qua rtile
25th qua rtile
TED(predicted)
TED(observed)
HXD(predicted)
and predicted target compounds

HXD(observed)
OCD(predicted)
OCD(observed)
EIC(predicted)
EIC(observed)
DOC(predicted)
DOC(observed)
353
TTC(predicted)
TTC(observed)
HXC(predicted)
HXC(observed)
OCC(predicted)
OCC(observed)
TCT(predicted)
TCT(observed)
DTT(predicted)
DTT(observed)
TRT(predicted)
TRT(observed)
HXT(predicted)
HXT(observed)
1-75 µm
OTT(predicted)
OTT(observed)
TTT(predicted)
TTT(observed)
Figure A.5.4 Distributions of the box plot statistics at 1-75 µm particulate fraction for observed
Additional Tables
Table A.5.1 PLS calibration set for light SVOC: E1-E28 represent the 28 individual experiments; C1-C7 represent the replicate experiments at centre;
L1-L4 represent underlying factors in the light SVOC data matrix
Experiments L1 L2 L3 L4 OCT DEC DOD TED pH EC TSS TOC Intensity Duration
E1 -1 -1 -1 -1 2.12 0.30 1.82 0.12 3.48 3.55 0.15 0.22 0.97 3.06
E2 1 -1 -1 -1 2.08 0.37 2.31 0.14 4.05 4.58 0.57 0.31 1.13 2.43
E3 -1 1 -1 -1 1.45 0.27 1.65 0.11 3.17 3.15 0.02 0.18 0.88 3.35
E4 1 1 -1 -1 0.35 0.03 0.05 1.69 6.32 5.40 0.25 0.35 2.78 1.03
E5 -1 -1 1 -1 2.89 0.38 2.31 0.15 4.56 6.34 0.10 0.25 1.20 1.95
E6 1 -1 1 -1 1.22 0.28 1.76 0.12 3.53 3.60 0.03 0.21 0.98 3.10
E7 -1 1 1 -1 2.29 0.32 2.01 0.14 4.15 4.69 0.30 0.23 1.16 2.48
E8 1 1 1 -1 0.17 0.41 2.75 0.11 3.05 3.02 0.09 0.16 0.85 3.21
E9 -1 -1 -1 1 0.36 0.03 0.05 0.12 6.37 5.45 0.81 0.32 2.80 1.04
E10 1 -1 -1 1 0.34 0.03 0.05 0.08 6.03 4.52 0.11 0.24 2.56 1.66
E11 -1 1 -1 1 0.24 0.57 4.04 0.15 4.29 5.96 0.21 0.21 1.13 1.84
E12 1 1 -1 1 1.54 0.20 1.28 0.11 3.59 3.66 0.10 0.18 1.00 3.15
E13 -1 -1 1 1 2.41 0.29 1.79 0.14 4.17 4.71 0.35 0.31 1.17 2.50
E14 1 -1 1 1 0.18 0.30 2.00 0.10 3.14 3.12 0.13 0.23 0.87 3.31
E15 -1 1 1 1 0.33 0.03 0.05 0.38 5.81 4.98 2.98 0.82 2.56 0.95
E16 1 1 1 1 0.31 0.03 0.04 4.71 5.54 4.14 0.17 0.27 2.35 1.52
E17 -2 0 0 0 3.41 0.25 1.61 0.15 4.61 6.40 0.53 0.40 1.21 1.97
E18 2 0 0 0 2.16 0.23 1.43 0.12 3.55 3.63 0.05 0.20 0.99 3.12
E19 0 -2 0 0 2.74 0.16 0.78 0.15 4.20 4.21 0.39 0.44 3.70 0.44
E20 0 2 0 0 1.73 0.22 1.39 0.10 3.20 3.18 0.03 0.18 0.89 3.38
E21 0 0 -2 0 0.35 0.03 0.05 1.18 6.29 5.38 0.58 0.47 2.76 1.02
E22 0 0 2 0 2.30 0.34 2.05 0.15 4.63 6.43 0.12 0.27 1.22 1.98
E23 0 0 0 -2 1.43 0.22 1.36 0.12 3.27 3.33 2.08 1.07 0.91 2.87
E24 0 0 0 2 0.02 0.36 2.07 0.12 3.67 4.15 3.10 1.40 1.03 2.20
E25 0 0 0 0 0.22 0.02 0.03 0.35 3.85 3.85 2.27 3.90 3.39 0.41
E26 0 0 0 0 0.17 0.28 1.79 0.10 2.97 2.95 1.51 0.85 0.82 3.13
E27 0 0 0 0 0.27 0.02 2.49 2.49 4.85 4.15 2.28 3.91 2.13 0.79
E28 0 0 0 0 0.28 0.03 0.04 3.13 5.03 3.77 2.00 3.24 2.14 1.39
C1 0 0 0 0 0.86 0.31 1.82 0.13 4.01 4.01 0.90 0.28 3.53 0.42
C2 0 0 0 0 0.32 0.03 0.05 0.48 5.84 4.37 0.82 0.35 2.48 1.61
C3 0 0 0 0 0.29 0.06 1.10 0.16 4.30 4.31 0.03 0.20 3.79 0.45
C4 0 0 0 0 0.28 0.23 0.48 0.20 4.25 4.26 0.73 0.31 3.75 0.45
C5 0 0 0 0 1.61 0.33 0.68 0.15 4.33 4.90 0.53 0.22 1.21 2.60
C6 0 0 0 0 0.25 6.07 0.04 0.51 4.52 3.39 0.54 0.31 1.92 1.25
C7 0 0 0 0 2.20 0.27 1.66 0.13 3.99 5.54 3.59 1.78 1.05 1.71
354
Table A.5.2 PLS calibration set for heavy SVOC: E1-E28 represent the 28 individual experiments; C1-C7 represent the replicate experiments at centre;
H1-H4 represent underlying factors in the heavy SVOC data matrix
Experiments H1 H2 H3 H4 HXD OCD EIC DOC TTC HXC OCC pH EC TSS TOC Intensity Duration
E1 -1 -1 -1 -1 2.70 1.74 2.46 2.54 2.14 0.05 1.49 9.29 4.48 0.38 0.29 1.65 2.30
E2 1 -1 -1 -1 1.89 0.54 1.33 1.06 1.56 0.08 0.83 6.40 2.08 0.03 0.19 1.14 4.23
E3 -1 1 -1 -1 2.46 0.69 1.48 1.04 1.28 0.05 1.11 8.47 2.42 0.20 0.33 4.66 0.73
E4 1 1 -1 -1 2.49 2.31 2.61 2.61 2.38 0.50 1.17 8.31 1.70 0.32 0.24 2.92 2.35
E5 -1 -1 1 -1 2.22 2.98 3.18 2.88 3.25 0.54 1.44 7.45 1.51 0.02 0.39 2.61 2.69
E6 1 -1 1 -1 3.97 0.84 1.83 1.03 0.03 0.07 0.73 8.69 1.85 1.24 0.57 2.26 3.01
E7 -1 1 1 -1 2.65 0.52 2.31 1.33 1.28 0.06 1.07 9.29 4.48 1.05 0.27 1.65 2.30
E8 1 1 1 -1 1.87 0.15 1.15 0.50 1.00 0.04 0.79 6.44 2.10 0.12 0.17 1.15 4.26
E9 -1 -1 -1 1 1.46 2.29 1.73 4.45 1.45 0.05 1.67 8.22 2.35 0.40 0.20 4.52 0.71
E10 1 -1 -1 1 2.03 1.72 2.25 2.12 2.23 3.36 1.05 6.91 1.41 0.10 0.18 2.43 1.95
E11 -1 1 -1 1 2.20 3.15 4.40 4.01 4.16 0.52 1.42 7.38 1.50 0.02 0.23 2.59 2.67
E12 1 1 -1 1 1.84 1.11 1.48 1.05 1.35 2.44 1.30 6.24 1.26 0.12 0.18 2.14 2.74
E13 -1 -1 1 1 7.06 5.00 4.57 4.21 2.22 0.17 1.05 8.46 1.80 0.12 0.19 2.20 2.93
E14 1 -1 1 1 3.10 1.19 1.39 1.71 2.36 0.05 1.93 9.29 4.48 0.54 0.30 1.65 2.30
E15 -1 1 1 1 1.90 0.50 0.75 0.64 0.69 0.04 0.91 6.42 2.09 0.19 0.24 1.14 4.25
E16 1 1 1 1 3.92 1.64 0.50 2.15 0.03 0.31 1.87 8.19 2.34 0.53 0.37 4.50 0.71
E17 -2 0 0 0 1.97 1.85 2.35 1.86 2.63 3.42 1.68 6.72 1.37 0.54 0.23 2.36 1.90
E18 2 0 0 0 2.72 0.81 1.53 1.05 1.27 0.10 1.18 9.53 4.59 0.05 0.30 1.69 2.36
E19 0 -2 0 0 1.86 0.33 1.52 1.41 1.72 0.04 0.72 6.40 2.08 0.05 0.19 1.14 4.24
E20 0 2 0 0 2.95 0.66 1.65 1.42 0.94 0.10 0.60 8.44 2.42 0.18 0.52 4.64 0.73
E21 0 0 -2 0 2.38 1.97 3.36 3.06 2.53 0.15 1.88 7.97 1.63 2.31 0.22 2.80 2.25
E22 0 0 2 0 1.83 1.59 1.53 1.33 1.63 3.00 0.87 6.36 1.29 0.76 0.23 2.23 2.30
E23 0 0 0 -2 2.01 1.00 1.65 1.77 1.67 0.04 6.01 6.83 1.38 1.23 0.23 2.34 3.00
E24 0 0 0 2 2.42 0.73 0.89 1.10 1.26 0.11 1.38 8.19 3.95 4.27 1.92 1.45 2.03
E25 0 0 0 0 1.84 1.03 1.30 1.37 1.47 0.03 0.71 5.98 1.94 2.05 0.89 1.06 3.95
E26 0 0 0 0 3.76 2.68 2.18 2.90 0.03 0.07 0.37 7.37 2.11 2.75 3.23 4.05 0.64
E27 0 0 0 0 2.12 1.54 2.50 2.05 3.45 0.39 3.55 7.37 1.50 1.94 3.59 2.59 2.08
E28 0 0 0 0 2.93 0.59 0.21 0.97 0.12 0.07 0.67 8.36 1.78 2.03 1.95 2.17 2.90
C1 0 0 0 0 2.17 0.31 2.27 1.92 1.85 0.04 1.86 7.23 1.46 0.21 0.21 2.48 3.17
C2 0 0 0 0 2.19 1.91 1.75 1.76 2.45 0.35 0.65 7.34 1.49 1.93 0.24 2.57 2.65
C3 0 0 0 0 2.12 0.92 1.54 1.36 1.46 0.19 1.07 7.21 1.45 0.32 0.21 2.47 3.16
C4 0 0 0 0 3.41 0.59 0.43 0.56 2.95 0.92 0.96 8.44 1.80 0.34 0.21 2.19 2.93
C5 0 0 0 0 2.98 0.64 0.15 0.95 0.03 2.26 0.68 7.90 1.68 0.02 0.31 2.05 2.74
C6 0 0 0 0 2.07 1.51 1.75 1.47 1.96 0.18 0.86 7.06 1.43 1.80 3.32 2.48 2.55
C7 0 0 0 0 2.12 1.86 2.28 2.36 2.23 0.04 1.02 6.60 1.33 1.71 2.63 2.26 2.90
355
Table A.5.3 PLS calibration set for NVOC: E1-E46 represent the 46 individual experiments; C1-C4 represent the replicate experiments at centre; N1-N5
represent underlying factors in the NVOC data matrix
Experiments N1 N2 N3 N4 N5 TCT DTT TRT HXT OTT TTT TSS TOC pH EC Intensity Duration
E1 -1 -1 -1 -1 -1 3.56 0.12 0.20 0.26 0.72 0.17 6.29 5.00 0.32 0.28 1.42 1.96
E2 1 -1 -1 -1 -1 1.76 5.13 2.23 0.25 0.96 1.20 6.20 4.66 0.11 0.25 1.33 2.10
E3 -1 1 -1 -1 -1 1.01 0.82 1.02 0.19 0.62 0.84 4.72 2.61 0.17 0.23 1.07 3.28
E4 1 1 -1 -1 -1 3.15 0.31 6.17 0.36 0.26 0.03 4.17 2.24 0.03 0.18 0.94 3.48
E5 -1 -1 1 -1 -1 1.31 0.91 2.83 0.22 0.69 1.61 5.24 1.75 0.02 0.40 2.33 2.39
E6 1 -1 1 -1 -1 2.38 2.29 2.34 0.29 1.15 0.89 7.97 3.69 0.26 0.33 2.85 1.03
E7 -1 1 1 -1 -1 1.00 2.30 2.28 0.96 0.93 0.57 7.20 2.91 0.85 0.33 2.48 1.57
E8 1 1 1 -1 -1 1.30 1.79 1.86 0.23 0.47 0.87 5.72 2.01 1.02 0.54 1.89 2.50
E9 -1 -1 -1 1 -1 0.84 1.09 0.04 0.26 2.04 0.68 6.33 4.75 0.27 0.23 1.35 2.14
E10 1 -1 -1 1 -1 0.92 0.82 0.90 0.13 0.63 0.51 4.76 2.63 0.03 0.21 1.07 3.31
E11 -1 1 -1 1 -1 1.16 0.73 0.92 0.17 0.56 0.46 4.25 2.28 0.10 0.17 0.96 3.55
E12 1 1 -1 1 -1 5.32 3.41 2.83 0.23 0.93 0.58 5.41 2.93 0.03 0.19 3.87 0.45
E13 -1 -1 1 1 -1 4.24 0.95 1.06 0.26 0.73 0.71 5.54 2.62 0.33 0.19 3.86 0.60
E14 1 -1 1 1 -1 3.05 0.15 2.84 0.21 5.70 2.19 5.17 1.73 0.02 0.23 2.30 2.35
E15 -1 1 1 1 -1 1.51 3.33 1.75 2.20 1.40 0.90 7.64 3.53 0.81 0.29 2.73 0.98
E16 1 1 1 1 -1 2.63 3.10 2.56 0.66 3.23 5.59 7.16 2.90 0.11 0.22 2.47 1.56
E17 -1 -1 -1 -1 1 1.43 2.07 1.82 0.26 0.77 0.52 5.93 2.08 0.11 0.20 1.96 2.59
E18 1 -1 -1 -1 1 3.62 2.75 1.33 0.18 0.81 0.66 6.17 4.91 0.44 0.29 1.39 1.93
E19 -1 1 -1 -1 1 1.14 0.41 1.35 0.26 0.82 0.67 6.24 4.69 0.60 0.41 1.34 2.11
E20 1 1 -1 -1 1 0.95 0.82 1.03 0.19 0.62 0.51 4.74 2.62 0.11 0.18 1.07 3.29
E21 -1 -1 1 -1 1 0.96 0.73 0.92 0.17 0.56 0.46 4.24 2.28 0.15 0.23 0.96 3.54
E22 1 -1 1 -1 1 0.84 1.71 1.19 0.23 0.71 1.28 5.41 2.93 0.78 0.29 3.87 0.45
E23 -1 1 1 -1 1 0.80 1.77 1.24 0.24 0.73 0.88 5.55 2.62 0.45 0.36 3.88 0.61
E24 1 1 1 -1 1 0.99 1.13 1.09 0.21 0.51 0.55 5.03 1.68 1.64 0.24 2.24 2.29
E25 -1 -1 -1 1 1 0.96 2.16 3.48 5.72 0.87 2.19 6.67 3.08 2.86 0.71 2.38 0.86
E26 1 -1 -1 1 1 2.54 2.65 2.78 0.31 1.27 1.00 7.34 2.97 0.19 0.28 2.53 1.60
E27 -1 1 -1 1 1 1.27 1.51 1.85 0.29 0.51 1.03 5.90 2.07 0.30 0.21 1.95 2.58
E28 1 1 -1 1 1 1.29 1.08 1.35 0.94 2.21 0.68 6.27 4.99 0.04 0.29 1.41 1.96
E29 -1 -1 1 1 1 2.71 1.16 0.29 1.71 0.61 0.50 4.63 2.56 0.06 0.20 1.05 3.22
E30 1 -1 1 1 1 0.53 0.74 0.92 0.33 0.56 0.46 4.26 2.29 0.04 0.18 0.96 3.55
E31 -1 1 1 1 1 1.31 1.25 1.48 0.25 0.71 2.69 5.42 2.93 0.42 0.43 3.88 0.45
E32 1 1 1 1 1 1.11 1.22 1.49 0.28 2.98 2.09 5.54 2.62 0.15 0.49 3.87 0.60
E33 -2.37841 0 0 0 0 2.55 0.89 1.12 0.21 0.68 0.56 5.15 1.72 0.76 0.27 2.29 2.34
E34 2.37841 0 0 0 0 1.13 2.77 3.04 1.83 4.19 4.02 7.57 3.50 0.58 0.43 2.70 0.97
E35 0 -2.37841 0 0 0 2.17 2.37 1.82 0.37 1.26 0.59 7.24 2.93 0.72 0.38 2.50 1.58
E36 0 2.37841 0 0 0 1.25 2.11 1.58 0.33 0.49 0.62 5.94 2.08 0.02 0.33 1.96 2.59
E37 0 0 -2.37841 0 0 1.07 0.94 1.18 0.27 0.72 0.59 5.47 4.35 3.54 1.86 1.23 1.71
E38 0 0 2.37841 0 0 0.94 0.93 1.16 0.22 0.71 0.58 5.38 4.04 4.05 1.82 1.15 1.82
E39 0 0 0 -2.37841 0 1.92 1.12 0.53 0.20 1.13 0.47 4.31 2.38 2.31 1.07 0.97 3.00
E40 0 0 0 2.37841 0 0.80 0.65 0.86 0.71 0.52 0.16 3.97 2.13 1.69 0.86 0.89 3.31
E41 0 0 0 0 -2.37841 0.67 1.40 1.47 3.45 0.61 0.27 4.62 2.50 2.28 3.54 3.31 0.38
E42 0 0 0 0 2.37841 1.32 1.56 2.21 0.22 0.64 1.43 4.91 2.32 2.28 3.13 3.43 0.54
E43 0 0 0 0 0 1.72 0.40 1.03 0.21 0.63 2.75 4.73 1.58 1.50 3.22 2.10 2.15
E44 0 0 0 0 0 1.31 2.69 2.07 1.53 0.99 0.92 6.71 3.10 2.64 4.08 2.39 0.86
E45 0 0 0 0 0 1.30 1.89 2.48 0.54 0.84 0.28 6.54 2.65 2.18 3.18 2.26 1.42
E46 0 0 0 0 0 1.15 1.51 1.74 0.24 0.17 1.44 5.47 1.92 1.65 1.84 1.80 2.39
C1 0 0 0 0 0 1.24 0.14 1.16 0.29 0.71 0.58 5.40 2.92 1.02 0.28 3.86 0.45
C2 0 0 0 0 0 1.59 0.97 1.33 0.35 0.74 0.60 5.61 2.65 0.17 0.32 3.91 0.61
C3 0 0 0 0 0 1.59 1.06 1.33 0.32 0.81 0.85 6.16 4.90 0.87 0.26 1.39 1.93
C4 0 0 0 0 0 0.78 0.79 1.33 2.50 0.95 0.66 6.17 4.64 0.13 0.28 1.32 2.09
356
APPENDIX A.6
REGRESSION EQUATIONS OF THE
PREDICTION FRAMEWORK
357
358
Chapter 8 and 9 presented a prediction framework based on multivariate
chemometric tools. The partial least square regression parameters for the prediction
of the build-up of VOCs and the wash-off and SVCOs and NVOCs were given in
Table 8.2, Chapter 8 and Table 9.3, Chapter 9, respectively. A set of equations was
derived for the study sites in the Gold Coast region and the object classification
system described in Chapter 4 determined the applicable ranges of the traffic and
rainfall characteristics for these equations. The coefficients of determination (r2) of
these equations were given in Table 8.2, Chapter 8 and Table 9.3, Chapter 9. Also
the limit of detections and highest concentrations of calibration standards of the
VOCs, SVOCs and NVOCs as described in the sample testing sections of Chapters 6
and 7 determined the applicable lower and upper ranges of target compound
concentrations for these regression equations. The number of data points was 132 for
equations A.6.1 to A.6.4, 392 for equations A.6.5 to A.6.8, 476 for equations A.6.9
to A.6.15 and 782 for equations A.6.16 to A.6.21. However, in order to apply the
prediction framework in areas other than the study sites, the dynamic changes in
traffic and rainfall characteristics need to be established using the object
classification system decribed in Chapter 4. Then the prdiction framework described
in Chapters 8 and 9 can be used to derive a new set of regression equations to predict
the build-up and wash-off of traffic generated pollutants on urban roads. The
regression equations for the study sites in Gold Coast region of South-East
Queensland Australia are given below:
TOL = 0.14299 − 0.00004 × TSS − 0.00041 × TOC + 0.00001 × ADT + 0.09551 × V / C − 0.16459 × STD
------------------------------------------------------------------------------------------(A.6.1)
ETB = 0.04342 − 0.00001 × TSS − 0.00025 × TOC + 0.03352 × V / C − 0.05634 × STD -----------(A.6.2)
MPX = 0.1151 − 0.00003 × TSS − 0.00034 × TOC + 0.07334 × V / C − 0.12632 × STD ------------(A.6.3)
OX = 0.04195 − 0.00001 × TSS − 0.00024 × TOC + 0.03529 × V / C − 0.06083 × STD ------------(A.6.4)
359
OCT = 1.43637 − 0.20623 × X 1 − 0.17806 × pH + 0.21926 × EC − 0.29119 × TSS − 0.28215 × TOC
------------------------------------------------------------------------------------------(A.6.5)
DEC = 0.39725 − 0.03939 × pH − 0.02799 × TOC − 0.0387 × Intensity + 0.03409 × Duration ----(A.6.6)
DOD = 2.59367 − 0.05784 × X1 − 0.27143 × pH + 0.01027 × EC − 0.08502 × TOC − 0.26865 × Intensity + 0.20442 × Duration
------------------------------------------------------------------------------------------(A.6.7)
TED = −0.3599 + 0.1293 × X 2 + 0.0259 × X 3 + 0.2922 × pH + 0.2583 × TOC − 0.23967 × Duration

------------------------------------------------------------------------------------------(A.6.8)
HXD = −1.30769 + 0.28855 × X 3 + 0.50387 × pH + 0.03698 × TSS ----------------------------(A.6.9)
OCD = 1.23907 − 0.2156 × X1 − 0.12769 × X 2 + 0.08085 × pH − 0.19386 × EC + 0.21585 × Intensity − 0.19178 × Duration
------------------------------------------------------------------------------------------(A.6.10)
EIC = 1.18368 − 0.46395 × X 1 + 0.17913 × pH − 0.27463 × EC -------------------------------(A.6.11)
DOC = −0.38802 − 0.23081 × X 2 + 0.20858 × pH − 0.18596 × EC + 0.46536 × Intensity --------- (A.6.12)
TTC = 2.14959 − 0.40283 × X 1 − 0.36492 × X 3 − 0.20721 × Intensity --------------------------(A.6.13)
HXC = 3.64724 − 0.27293 × pH − 0.2694 × EC − 0.12746 × TSS − 0.10119 × TOC − 0.08157 × Duration
------------------------------------------------------------------------------------------(A.6.14)
OCC = 1.87262 − 0.30651 × X 4 − 0.16803 × EC + 0.16691 × TSS − 0.09037 × Duration ----------(A.6.15)
TCT = 1.71 − 0.1275 × X 2 − 0.179 × X 5 + 0.081× TOC − 0.234 × pH − 0.177 × EC + 0.134 × Intensity − 0.1124 × Duration
------------------------------------------------------------------------------------------(A.6.16)
DTT = −1.7895 − 0.0754 × X 5 + 0.59567 × TSS − 0.04034 × pH ------------------------------(A.6.17)
TRT = 0.43675 + 0.36919 × X 1 − 0.15933 × X 5 + 0.30537 × TSS − 0.16648 × TOC --------------(A.6.18)
HXT = −0.1178 − 0.0703 × X 1 + 0.16575 × X 4 + 0.1705 × TSS + 0.2484 × pH − 0.2223 × Duration

------------------------------------------------------------------------------------------(A.6.19)
OTT = 0.00878 + 0.29912 × X 1 + 0.26974 × TSS − 0.1637 × pH − 0.16256 × Duration -----------(A.6.20)
TTT = 0.2411 + 0.1852 × X 1 + 0.1556 × X 3 + 0.1458 × X 4 + 0.23 × TSS − 0.0533 × pH − 0.2274 × Duration
------------------------------------------------------------------------------------------(A.6.21)
360

Parvez Mahbub Thesis

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Parvez Mahbub Thesis

Caricato da

Copyright:

Formati disponibili

Impact of Urban Traffic and

Climate Change on Water Quality

S. M. Parvez Bin Mahbub

A Thesis Submitted in Partial Fulfilment of the Requirements

School of Urban Development

Climate change, heavy metals, hydrocarbons, multivariate data analyses, pollutant

Table 2.1 Heavy Metals sources from exhaust and non-exhaust

Figure 2.1 Illustration of pollutant build-up on road surfaces .............. 23

Appendix A.1 Intensity-Frequency-Duration Table for Station 40584

I would like to sincerely thank my principal supervisor, Professor Ashantha

research culminating in this thesis.

I would also thank my co-supervisors, Associate Professor Godwin A. Ayoko, Dr

supervision for this research.

Without the generous financial assistance provided by the Queensland University of

to undertake this postgraduate research. I also wish to extend my gratitude towards

the laboratory technicians of School of Urban Development and Chemistry

(3) Mahbub, P., Ayoko, G. A., Goonetilleke, A., Egodawatta, P. (2011).

important sources of pollutants to urban receiving waters. This situation is further

urbanisation in the continent’s major cities (BITRE 2008).

Additionally, climate change is expected to result in changed rainfall characteristics

throughout Australia (CSIRO 2003, 2007). The changes in the rainfall

authorities to undertake adaptive mitigation strategies to safeguard the water quality

in urban areas and to maintain ecosystem functions.

and wash-off processes under static environmental conditions (for example

traffic related pollutant build-up on urban roads. Likewise, rainfall characteristics

petroleum hydrocarbons are regarded as pollutants that are generated by traffic.

and predicted changes in traffic characteristics as well as pollutant wash-off and

predicted changes in rainfall characteristics difficult to determine. A better

understanding of the pollutant build-up and wash-off processes due to changing

stormwater quality mitigation under such changed situation.

1.3.Aims and Objectives

1. The impact of traffic characteristics on pollutant build-up on urban roads.

2. The impact of rainfall characteristics on the pollutant wash-off from urban

The objectives of this research study were as follows:

generated pollutants with changing urban traffic and climate change,

2. Characterisation of the build-up and wash-off processes of stormwater

pollutants generated from urban traffic.

3. Development of prediction frameworks for traffic generated stormwater

4. Provide guidance to stormwater quality mitigation strategies under dynamic

urban traffic and climate change scenarios.

1. Traffic generated pollutant build-up on urban roads is influenced by changes

in urban traffic characteristics such as average daily traffic, congestion and

road surface texture depth.

2. The pollutant wash-off from urban roads is influenced by changes in rainfall

characteristics such as intensity, frequency and duration due to climate

1.5.Scope of the Research

Southeast Queensland region of Australia, the methodologies, prediction

frameworks as well as the pollutant build-up and wash-off characterisations have

generic applicability to urban areas in other geographical locations.

relation to pollutant wash-off studies. The changes in rainfall characteristics due to

characteristics for this year.

In relation to pollutant build-up and wash-off, heavy metals, volatile, semi-volatile

Characterisation of the build-up and wash-off of stormwater pollutants from other

described in section 1.3.

1.6.Justification for the Research

water bodies, adaptive strategies need to be undertaken. However, limited

conceptualisation of appropriate adaptive strategies. Research studies have been

undertaken in the past on the process kinetics of several stormwater pollutants.

Unfortunately, these studies have commonly focussed on the characterisation of

traffic characteristics were not specifically investigated in these studies in relation to

pollutant build-up. Similarly, there is lack of knowledge on the impacts of climate

change induced rainfall characteristics on pollutant wash-off. Even though some

studies have investigated rainfall characteristics whilst characterising pollutant