Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Quaternary International
journal homepage: www.elsevier.com/locate/quaint
Sensitivity analysis and impact quantication of the main factors affecting peak
discharge in the SCS curve number method: An analysis of Iranian watersheds
Mohammad Reza Kousari a, *, Hossein Malekinezhad b, Hossein Ahani a, Mohammad Amin Asadi Zarch b
a
b
a r t i c l e i n f o
a b s t r a c t
Article history:
Available online 4 June 2010
The SCS curve number method is the most commonly used method for the estimation of peak discharge
in a watershed. This method is used in numerous complex models such as SWAT, HEC-HMS, EPIC, and
AGNPS, but has never been analyzed using the sensitivity analysis, to the best of the authors knowledge.
The present study deals with the effects of the time of concentration time, watershed area, amount of
rainfall at different return periods and the a-coefcient on the nature of the SCS curve number method in
the estimation of peak discharge and its reaction to change in the input parameters (coefcient of
estimation for getting the time of effective rainfall, which is usually equal to 0.133). Results indicate the
effective role of CN on the input to the peak discharge model. The sensitivity of the model, during the
estimation of peak discharge, increases following the increase in the return period. The sensitivity
analysis of SCS curve number method was performed via the MATLAB program. Due to the increased
application of MATLAB program for general basin conditions, it can be applicable in special watersheds to
nd out the parameter(s) having a signicant impact on peak discharge.
2010 Elsevier Ltd and INQUA. All rights reserved.
1. Introduction
In watershed analysis, the hydrological models are more diverse
rather than their other counterparts. A careful study shows that
other effective models such as sediment delivery and erosion
estimation models are affected by hydrological models and play
a basic role in the algorithm of other models. It is notable that
simple models play the basic role for complex models such as SWAT
(Soil & Water Assessment Tool) (Arnold et al., 1996), HEC-HMS
(Hydrologic Modeling System), EPIC (Erosion Productivity Impact
Calculator), and AGNPS (Agricultural Non-Point Source Pollution
Model), which were developed after the SCS curve number (Soil
Conservation Service; now Natural Resources Conservation Service
{NRCS}) hydrograph method.
One of the most commonly used and widely applicable
methods is the SCS curve number method used in the estimation
of watershed ood hydrograph ordinates. Easy usage and the
availability of the models inputs and numerous outputs such as
peak discharge of ood, time to peak, lag time and ood time
make the SCS method more applicable. Many exact hydraulic
models derive their input data (i.e., the peak discharge for steady
and uniform ows while ood hydrograph for unsteady ows)
from the SCS curve number method. The SCS curve number
method is widely used for predicting the storm runoff volume
(Zhan and Huang, 2004).
As for every other model, the SCS curve number model also
needs the input for getting output, where the amount of
precipitation, time of concentration time, watershed area and CN
are its main inputs. However, a question arises: which input in
which conditions is effective for the main output (i.e. peak
discharge)? Sensitivity analysis can answer this question. Sensitivity analysis accurately compares the certainty and efciency of
the models and nds the sensitive conditions for the calibration
of these models. Sensitivity analysis is less time consuming,
economical, and effective. It is important to evaluate how
a model responds to changes in its inputs as part of the process of
model development, verication and evaluation. Furthermore,
a sensitivity analysis of the models input parameters can serve as
a guide to any further application of the model. Quantitative
sensitivity analysis is being increasingly used for corroboration,
quality assurance, and the defensibility of model-based analyses
(Ascough et al., 2004).
Thus, a sensitivity analysis is usually the rst step towards
model calibration because it answers several questions such as (a)
where the data collection efforts should focus; (b) what degree of
care should be taken for parameter estimation; and (c) the relative
importance of various parameters (Cho and Lee, 2001). Sensitivity
analysis can be used as an aid in identifying the important uncertainties for the purpose of prioritizing the additional data collection
or research (Frey and Patil, 2002). In addition, the sensitivity
analysis can play an important role in model verication and validation throughout the course of model development and renement (Kleijnen and Sargent, 2000). Sensitivity analysis can also be
used to provide insight into the robustness of the model results
when making decisions (Saltelli et al., 2000).
Different methods are available for carrying out sensitivity
analyses and expressing their results (Lenhart et al., 2002; Van
Griensven et al., 2002; Van Griensven, 2006; Kannan et al.,
2007). Some methods use a percent change in input and report
a corresponding change in output variables. This is not always
suitable for the parameters such as saturated hydraulic conductivity and curve number (CN). Hydraulic conductivity can vary
over several orders of magnitude, and a 10% variate of a CN value
in hydrologic soil group C can lead to a CN value in the soil group B
or D (Neitsch et al., 2001). Ascough et al. (2004) have reported the
methods of sensitivity analysis especially in Natural Resource
Management with the eligibility and limitations of each method.
They concentrated on the qualitative evaluation of four sensitivity
analysis methods: 1) Fourier Amplitude Sensitivity Test (FAST), 2)
Response Surface Method (RSM), 3) Mutual Information Index
(MII), and 4) the methods of Sobol; and have mentioned that in
Natural Resource Management the FAST and Sobol methods are
particularly attractive.
On the other hand, there are different ways of classifying
sensitivity analysis methods, which are broadly classied as
mathematical, statistical (or probabilistic), and graphical (Frey and
Patil, 2002). Alternatively, these methods can be classied as
screening, local, and global (Saltelli et al., 2000; Ascough et al.,
2004). FAST, RSM, MII and Sobol methods are classied as statistical (Ascough et al., 2004).
In sensitivity analysis, the use of each method depends on the
structure of model, research condition and the level of accuracy.
Some methods are good for complex models and their application
to the simple model can consume extra time and energy.
The objective of this paper is to survey the effect of main factors
on the peak discharge in the SCS curve number method. The
mathematical-graphical method that determines the impact of
each input on peak discharge in a step by step manner was used. For
more accuracy and also for saving time, the algorithm for peak
discharge estimation by SCS method and sensitivity analysis were
recorded with the help of MATLAB program.
Direct application of sensitivity analysis studies to the SCS
curve number method cannot be found. Therefore, the results
from complex models that use the SCS curve number method have
been mentioned in these cases (Bhuyan et al., 2002). For the EPIC
model, sand, silt, coarse fragment contents, and CN were found to
be the most sensitive parameters (Brath and Montanari, 2003).
Another study indicated that the inltration is one of the important factors affecting peak discharge. Mohammed et al. (2004)
showed that in the evaluation of AGNPS (Agricultural Non-Point
Source Pollution) model, the CN is the most sensitive input for the
model. Holvoet et al. (2005) concluded that, in the sensitivity
analysis of SWAT model, the dominant hydrological parameters
were the curve number (CN), the surface runoff lag time, the
recharge to deep aquifer and the threshold depth of water in the
shallow aquifer. Cryer and Havens (1999) showed that the runoff
curve number was the most sensitive input parameter for the
GLEAMS (Groundwater Loading Effects of Agricultural Management Systems) model.
67
P Ia 2
P Ia S
Where:
Q is runoff; mm, cm or inch)
P is rainfall; mm, cm or inch)
S is the potential maximum soil moisture retention after runoff
begins (mm, cm or inch) Ia is the initial abstraction (mm, cm or
inch), or the amount of water before runoff, such as inltration, or
rainfall interception by vegetation; and it is generally assumed that
Ia 0.2S.
So the above equation could be written as:
Qp
P 2S2
P :8S
68
Fig. 1. The algorithm for peak discharge estimation for different return periods. qp peak discharge of unit hydrograph (m3.s1), Tp Time to peak (h), Tc Time of concentration
(h), S The potential maximum soil moisture retention after runoff begins (here is cm), CN curve number, PRT precipitation in related to its return period (cm), P precipitation
(cm), QRT The runoff height in cm and different return period (cm), D duration time of effective precipitation (h), Qpn peak discharge in different return period (m3.s1),
a the coefcient for determination of duration time of effective precipitation from Tc usually equal to 0.133.
watershed areas and the SCS output may be erroneous outside this
range (Mahdavi, 2007). As shown in Table 1, out of the four
combinations, the watershed area and the time concentration have
a major effect on the determined homogeneous combinations.
Table 2 shows the amount of 24 hour precipitations with return
periods ranging from 2 to 100 y. These combinations are related to
the size of watersheds.
The range of each of the input parameter for each watershed
may be different from the others, but this program only derives the
inputs and for each special condition. It can also carry out the
sensitivity analysis within that specic condition.
Table 1
Range of inputs, means and repetition steps of watershed area, time of concentration, curve number (CN) and a-coefcient for combinations 1e4.
Combination
Area (km2)
Mean
Repetition step
1
1e10
10e30
30e60
60e100
5.5
0.1
20
0.1
45
0.1
80
0.1
Mean
Curve number CN
Repetition step
0.25e1.0
1e3
3e6
6e10
0.625
0.1
2
0.1
4.5
0.1
8
0.1
Mean
Coefcient a
Repetition step
60e90
60e90
60e90
60e90
75
0.1
75
0.1
75
0.1
75
0.1
Mean
Repetition step
0.133e0.15
0.133e0.15
0.133e0.15
0.133e0.15
0.1415
0.001
0.1415
0.001
0.1415
0.001
0.1415
0.001
24eh
precipitation (cm)
Mean
precipitation (cm)
Repetition
step
2
5
10
25
50
100
0.2e0.5
0.5e1.0
1e2
2e4
3e6
5e10
0.35
0.75
1.5
3
4.5
7.5
0.001
0.1
0.1
0.1
0.1
0.1
69
Table 3
The systematic program runs used to estimate discharges. The CN is changed but the
other input parameters are xed at their averages.
Name of Input
Range of inputs and outputs
parameters and
also output discharge
CN
60
60.1
60.2
60.3
80
80
80
80
Area (km2)
concentration time (h)
8
8
8
8
Alpha coefcients
0.14
0.14
0.14
0.14
Precipitation in
2 years
0.35
0.35
0.35
0.35
different return
5 years
0.75
0.75
0.75
0.75
periods (cm)
10 years
1.5
1.5
1.5
1.5
25 years
3
3
3
3
50 years
4.5
4.5
4.5
4.5
100 years
7.5
7.5
7.5
7.5
2 years
0
0
0
0
Discharge in
5 years
0
0
0
0
different return
3
10 years
0
0
0
0
periods (m /2)
25 years
0
0
0
0
50 years
1.63
1.68
1.72
1.77
100 years 19.08 19.27 19.45 19.63
.
90
80
80
8
8
0.14
0.14
0.35
0.35
0.75
0.75
1.5
1.5
3
3
4.5
4.5
7.5
7.5
.
0
.
0.27
.
5.53
.
26.78
.
54.41
.
117.02
Xi X
st
Fig. 2. The change in peak discharge due to the change in standard normal variates of
input parameters for combination 1 given in Table 1.
70
Fig. 3. The change in peak discharge due to the change in standard normal variates of
input parameters for combination 2 given in Table 1.
Fig. 5. The change in peak discharge due to the change in standard normal variates of
input parameters for combination 4 given in Table 1.
3. Results
Fig. 4. The change in peak discharge due to the change in standard normal variates of
input parameters for combination 3 given in Table 1.
Fig. 6. The slope change graph in peak discharge for 2-year return period for combination 1 of Table 1 for different standard normal variate values of input parameters.
Fig. 7. The slope change graph in peak discharge for 5-year return period for combination 1 of Table 1 for different standard normal variate values of input parameters.
Fig. 8. The slope change graph in peak discharge for 10-year return period for
combination 1 of Table 1 for different standard normal variate values of input
parameters.
71
Fig. 9. The slope change graph in peak discharge for 25-year return period for
combination 1 of Table 1 for different standard normal variate values of input
parameters.
4. Discussion
The results derived from the present study are in agreement
with previous work (Cryer and Havens, 1999; Bhuyan et al., 2002;
Brath and Montanari, 2003; Mohammed et al., 2004). When the
time of concentration is short, this indicates small watershed area,
low potential for inltration, low loss of precipitation, more slope,
circularity of watershed, and short stream length. Any one of these
factors or a combination of them causes increased peak discharge
in the outlet. In other words, the evacuation of the ood-volume is
more that there is no storage routing in small and sloped watersheds. From Fig. 2, by increasing the time of concentration, peak
discharge decreases as compared to CN or precipitation. An
increasing concentration time is related to an increase in the stream
length, lesser slope of watershed, less slope of main stream,
increased loss of rainfall, more watershed area, more watershed
Fig. 10. The slope change graph in peak discharge for 50-year return period for
combination 1 of Table 1 for different standard normal variate values of input
parameters.
72
Fig. 11. The slope change graph in peak discharge for 100-year return period for
combination 1 of Table 1 for different standard normal variate values of input
parameters.
Fig. 13. The slope change graph in peak discharge for 5-year return period for
combination 4 of Table 1 for different standard normal variate values of input
parameters.
Fig. 12. The slope change graph in peak discharge for 2-year return period for
combination 4 of Table 1 for different standard normal variate values of input
parameters.
Fig. 14. The slope change graph in peak discharge for 10-year return period for
combination 4 of Table 1 for different standard normal variate values of input
parameters.
73
Fig. 15. The slope change graph in peak discharge for 25-year return period for
combination 4 of Table 1 for different standard normal variate values of input
parameters.
Fig. 17. The slope change graph in peak discharge for 100-year return period for
combination 4 of Table 1 for different standard normal variate values of input
parameters.
Fig. 16. The slope change graph in peak discharge for 50-year return period for
combination 4 of Table 1 for different standard normal variate values of input
parameters.
74
Fohrer, N., Haverkamp, S., Eckhardt, K., Frede, H.G., 2001. Hydrologic response to
land use changes on the catchment scale. Physics and Chemistry of Earth 26,
577e582.
Frey, H.C., Patil, R., 2002. Identication and review of sensitivity analysis methods.
Risk Analysis 22, 553e577.
Holvoet, K., van Griensven, A., Seuntjens, P., Vanrolleghem, P.A., 2005. Sensitivity
analysis for hydrology and pesticide supply towards the river in SWAT. Physics
and Chemistry of the Earth, 518e526.
Huang, M.B., Zhang, L., 2004. Hydrological responses to conservation practices in
a catchment of the Loess Plateau, China. Hydrological Processes 18, 1885e1898.
Kannan, N., White, S.M., Worrall, F., Whelan, M.J., 2007. Sensitivity analysis and
identication of the best evapotranspiration and runoff options for hydrological
modelling in SWAT-2000. Journal of Hydrology 332, 456e466.
Kleijnen, J.P.C., Sargent, R.G., 2000. A methodology for tting and validating metamodels in simulation. European Journal of Operational Research 120 (1),
14e29.
Laurance, W.F., 1998. A crisis in the making: responses of Amazonian forests to land
use and climate change. Trends in Ecology and Evolution 13, 411e415.
Lenhart, T., Eckhardt, K., Fohrer, N., Frede, H.G., 2002. Comparison of two different
approaches of sensitivity analysis. Physics and Chemistry of Earth 27 (27),
645e654.
Mahdavi, M., 2007. Applied Hydrology. Tehran University.
Mohammed, H., Yohannes, F., Zeleke, G., 2004. Validation of agricultural non-point
source (AGNPS) pollutionmodel in Kori watershed South Wollo, Ethiopia.
International Journal of Applied Earth Observationand Geoinformation,
97e109.
Neitsch, S.L., Arnold, J.G., Kiniry, J.R., Williams, J.R., 2001. Soil and Water Assessment
Tool-Version 2000-Users Manual Temple, Texas, USA.
Richey, J.E., Nobre, C., Deser, C., 1989. Amazon River discharge and climate variability: 1903e1985. Science 246, 101e103.
Saltelli, A., Chan, K., Scott, M., 2000. Sensitivity Analysis, Probability and Statistics
Series. Wiley, New York.
Van Griensven, A., Francos, A., Bauwens, W., 2002. Sensitivity analysis and autocalibration of an integral dynamic model for river water quality. Water Science
and Technology 45, 325e332.
Van Griensven, A., 2006. A global sensitivity analysis tool for the parameters of
multivariable catchment models. Journal of Hydrology 324 (1e4), 10e23.
Zhan, X., Huang, M.-L., 2004. ArcCN-Runoff: an ArcGIS tool for generating curve
number and runoff maps. Environmental Modelling and Software 19 (10),
875e879.