Journal of Environmental Sciences 2010, 22(6) 846–850

Evaluation of the relationship between two different methods for enumeration

fecal indicator bacteria: Colony-forming unit and most probable number
Kyung Hwa Cho1 , Dukki Han1 , Yongeun Park1 , Seung Won Lee1 ,
Sung Min Cha1 , Joo-Hyon Kang2 , Joon Ha Kim1,3,∗

1. Department of Environmental Science and Engineering, Gwangju Institute of Science and Technology, 261 Cheomdan-gwagiro, Buk-gu,
Gwangju, 500-712, Korea. E-mail: firstkh@gist.ac.kr
2. Deparment of Civil and Environmental Engineering, Dongguk University-Seoul, Seoul 100-715, Korea
3. Sustainable Water Resource Technology Center, Gwangju Institute of Science and Technology,
261 Cheomdan-gwagiro, Buk-gu, Gwangju, 500-712, Korea

Received 18 October 2009; revised 29 December 2009; accepted 16 January 2010

Abstract
Most probable number (MPN) and colony-forming unit (CFU) estimates of fecal indicator bacteria (FIB) concentration are common
measures of water quality in aquatic environments. Thus, FIB intensively monitored in Yeongsan Watershed in an attempt to compare
two different methods and to develop a statistical model to convert from CFU to MPN estimates or vice versa. As a result, the significant
difference was found in the MPN and CFU estimates. The enumerated Escherichia coli concentrations in MPN are greater than those in
CFU, except for the measurement in winter. Especially in fall, E. coli concentrations in MPN are one order of magnitude greater than
that in CFU. Contrarily, enterococci bacteria in MPN are lower than those in CFU. However, in general, a strongly positive relationship
are found between MPN and CFU estimates. Therefore, the statistical models were developed, and showed the reasonable converting
FIB concentrations from CFU estimates to MPN estimates. We expect this study will provide preliminary information towards future
research on whether different analysis methods may result in different water quality standard violation frequencies for the same water
sample.

Key words: most probable number; colony-forming unit; fecal indicator bacteria; Escherichia coli; enterococci
DOI: 10.1016/S1001-0742(09)60187-X

Introduction Multiple-tube fermentation and membrane filtration fecal

Water-borne diseases are caused by fecal bacteria propa-
gated through contaminated drinking water or recreational
surface water. The 2000 National Water Quality Inventory
(NWQI) reported that approximately 93,000 river and
stream miles in the United States contain high levels of
fecal indicator bacteria (FIB) (US EPA, 2002). Therefore,
mandatory water quality improvement programs, such
as the US EPA Total Maximum Daily Load (TMDL),
have established in an attempt to minimize the fecal
contaminations in water resources. The Europe Union has
also published several standards regarding guideline and
mandatory concentration levels of the pathogenic bacteria
in coastal waters (CDC, 2004). For the effective manage-
ment of public health, it is necessary to further investigate
the diversities of transport and fate characteristics of
fecal-borne microorganisms, which are generally quanti-
fied using Escherichia coli. However, different kinds of
methods in quantification of FIB may result in complexity
and uncertainty in environmental assessment and design.
* Corresponding author. E-mail: Joonkim@gist.ac.kr
are common procedures for estimating fecal coliform
concentrations (Eckner, 1998) and are evaluated as most
probable number (MPN) and colony-forming unit (CFU)
estimates of FIB concentration. Several studies recently
investigated the empirical relationships between FIB con-
centrations resulted from different testing procedures, and
have revealed that the two different test procedures are
likely to result in different fecal coliform concentrations
because they assess different metabolic process endpoints
(Noble et al., 2003). Therefore, the relationship between
MPN and CFU needs to be investigated by assessing the
samples from various waterbodies. Furthermore, empiri-
cal modeling approaches can provide the alternatives for
reducing the inconvenience associated with two different
methods in quantification of FIB.
As such, the primary objectives of this article are: (1) to
spatiotemporally enumerate FIB concentration from same
water body using two different procedures (i.e., MPN
and CFU); (2) to investigate the relationship between
two procedures; (3) to develop the empirical equation
for converting FIB concentration from MPN and CFU
estimates or vice versa.
No. 6 Evaluation of the relationship between two different methods for enumeration fecal indicator bacteria······
1 Materials and methods
1.1 Study area and sampling design
As shown in Fig. 1, the Yeongsan Watershed (YSW)
located in the southwestern Korea is one of the four largest
river basins in Korea (3455 km2 ) and goes through the
city of Gwangju and finally flows into the Yellow Sea. It
is encompassed by a substantial agricultural area (1000
km2 ) with 13 tributaries merging with the mainstream.
Throughout the watershed, there are several water storage
systems for supplying the water resources for agricultural
development including four upstream and one downstream
reservoir.
Although it features a relatively short mainstream length
(130 km) and slow urbanization, it shows the worst water
quality out of the four largest watersheds. In the upland
region of YSW, the Gwangju City, a highly urbanized
metropolitan area, is located and has been recognized
as the most polluted area in YSW (Kang et al., 2009),
while the other regions are mainly rural areas (over 50%
is agricultural areas, forest or grass land). To this end,
many research works and management schemes have been
undertaken and suggested to improve the water quality
in YSW. However, these have not been as successful as
desired; rather, the water quality in the region has been
getting worse (YRBEO, 2005). Especially, according to
Cha et al. (2009), the FIB concentration in receiving water
body of YSW (i.e., Yeongsan estuarine dam) significantly
increased since 2006.
Thus, two FIBs i.e., E. coli and enterococci (ENT)
were monthly measured at each monitoring site from
January to November 2008. At each sampling event, water
samples were collected using a polycarbonate sampling
bar at approximately 1 m below the water surface. The
collected samples were stored in ice chests and transported
to the laboratory in the Gwangju Institute of Science and
Technology. All the analyses were conducted within 6 hr
Fig. 1 Map of the Yeongson Watershed (YSW) showing the monitoring
stations, ranging from the Damyang and the Yeongsan estuarine dam.
847
after sample collection. For CFU procedures, improved
enumeration method was applied to estimate ENT and E.
coli.
1.2 Experiment
In determining FIB concentrations, the specified sub-
strate tests using Coliert and EnterolertTM (IDEXX Labo-
ratories, Inc., Westbrook, Maine) were conducted to quan-
tify E. coli and ENT using 97 well Quanti-trayTM /2000
trays (Olstadt et al., 2007). This simplified test for counting
FIB population is a presence-absence test using specified
medium containing o-nitrophenyl-β-D-galactopyranoside
and 4-methylumbelliteryl-β-D-glucuronide as nutrients.
The sealed Quanti-traysTM /2000 were incubated for 24 hr
at (41 ± 0.5)°C for ENT and (35 ± 0.5)°C for E. coli.
For measuring E. coli and ENT concentrations, fluorescent
wells of each Quanti-trayTM /2000 were counted under 365
nm UV light (Spectroline CM-10, Spectro Corporation,
USA). The number of counted wells could be finally
converted to most probable number per 100 mL (MPN/100
mL) according to the conversion table provided by IDEXX
Laboratories, Inc. (USA). For CFU estimates, the im-
proved enumeration method was applied to estimate ENT,
and E. coli; this method provides step-by-step instructions
for faster and easier enumeration using the improved two
media (i.e., mE Agar for ENT and mTEC Agar for E. coli)
(US EPA, 2000).
1.3 Regression models
Regression models were created using CFU and MPN
estimates in the YSW as the explanatory and dependent
variables, respectively. The basic form of regression equa-
tion is
MPNestimates = a + b × CFUestimates (1)
The other form of regression can be defined as the follow-
ing:
MPNestimates = a × CFUbestimates (2)
where, MPNestimates and CFUestimates estimates are the FIB
concentrations in MPN and CFU, a and b are regression
coefficients.
2 Results
Table 1 shows the basic statistics of FIB measurement
in both CFU and estimates in YSW. On the basis on
the averaged FIB concentrations, the significant difference
was found in the MPN and CFU estimates. In general,
it can be seen that the enumerated E. coli in MPN is
greater than those in CFU, except for the measurement
in winter. Especially in fall, E. coli in MPN is a one
order of magnitude greater than that in CFU. Contrarily,
ENT in MPN is lower than that in CFU. It implies
that the each enumeration method for FIB can lead to a
different concentrations, and thereby result in the difficulty
in environmental management schemes.
Table 2 indicates the Pearson correlations between MPN
and CFU estimates for the YSW. Strongly positive rela-
848 Kyung Hwa Cho et al. Vol. 22

Table 1 Basic statistics of FIB measurements for the YSW

Spring Summer Fall Winter

E. coli Average 265 (626)* 1978.27 (2696.71) 297.54 (3452.18) 171.94 (93.33)
Std 724.37 (2493.61) 3095.88 (5796.49) 462.18 (6403.91) 546.66 (372.29)
Max 4300 (25000) 17200 (42196) 2800 (25000) 3933 (3448)
Min 0 (9) 0 (9) 0 (9) 0 (9)
N 134 129 89 103
ENT Average 77 (61) 775.03 (188.63) 506.85 (343.69) 120.75 (26.93)
Std 208.76 (186.75) 1395 (533) 722. 58 (588.37) 408.03 (54.41)
Max 1933 (1722) 9900 (3654) 3180 (2143) 3200 (487)
Min 0 (9) 0 (9) 0 (9) 0 (9)
N 134 129 89 103
* Values in parenthesis indicate FIB concentration in MPN. ENT: enterococci.

Table 3 Coefficients of regression models for converting FIB

tionships were found in between MPN and CFU estimates concentration from CFU to MPN estimates
of E. coli in both summer and fall, and between MPN and
CFU estimates of ENT in both spring and winter. Even Spring Summer Fall Winter
though there is the significant difference on magnitude of E. coli∗,∗∗,∗∗∗ lna 1.27 –0.51 –1.23 2.09
FIB between two methods, the positive correlation was (std. error) (0.18) (0.33) 0.47 (0.31)
found between MPN and CFU procedures for enumerating b 0.80 1.04 1.36 0.36
(std. error) (0.04) (0.05) 0.09 (0.07)
E. coli and ENT (Table 2). Overall, while the measured E. ENT** a 0 7.90 0 11.47
coli concentrations between MPN and CFU estimates are (std. error) (0) (39.4) (0) (3.86)
highly correlated during summer and fall, the measured b 0.72 0.19 0.42 0.13
ENT concentrations by two methods are well associated (std. error) (0.04) (0.02) (0.04) (0.01)
during spring and winter. Thus, these correlations led to * Regression model: lnMPNestimates = lna + b × lnCFUestimates .
develop the regression for converting E. coli and ENT ** Regression model: MPNestimates = a + b × CFUestimates .
*** Values in bold indicate the significant parameter that their magnitude
concentrations from MPN and CFU estimates or vice of coefficient is greater than their std. error.
versa; the MPN and CFU estimates are independent and
dependent variables, respectively. For E. coli, as mentioned spring, summer, and fall) might lead to a high turbidity in
before, E. coli in MPN is a one order of magnitude water body, and thereby results in an increase in sediment-
greater than that in CFU, except for the winter. Thus, attached FIB. In general, FIB can be easily attached on
logarithmic transformations were applied into the depen- the surface of suspended particles in the turbid water body.
dent and independent variables. Contrarily, the data on Consequently, these sediment-attached in water body can
ENT was used in development of regression model without cause different concentrations from two analyses.
the logarithmic transformation because all variables are Figure 2 shows the comparison between the observed
normally distributed. and predicted E. coli in four seasons. In general, the pre-
The regression coefficients and standard error are given dicted E. coli concentrations by the regression model are in
in Table 3. If the absolute value of a given regression good agreement with the observed E. coli concentrations.
coefficient in an regression model is greater than twice its As well, the goodness of fit (r2 ) and Nash-Sutcliff model
standard error, that parameter can be defined as significant efficiency coefficient (NSE) were used for evaluating the
(Rawlings et al., 1998). On the basis of this criterion, all performance of regression models. Table 4 presents the
regression models, except for two models for E. coli and calculated r2 and NSE of regression models for both E.
ENT for summer, have a regression coefficient greater than coli and ENT. The models demonstrate that it is useful
twice their standard error. The variation of FIB concentra- for converting E. coli concentrations from CFU to MPN
tions in summer is greater than those in the other seasons, estimates; in particular, regression model for summer is the
and thereby results in a difficulty in the determination most accurate.
of the robust and reliable coefficients in the regression Figure 3 presents the comparison between the observed
model. For regression models for E. coli, the magnitudes of and predicted ENT in four seasons. The predicted ENT
slope coefficient (b) are relatively larger in spring, summer, concentrations by the regression model well reproduce
and fall, and lower in winter. Similarly, the magnitude the variations of the observed ENT concentrations. Even
of slope coefficient for ENT is the lowest in winter. It though the accuracies of models for ENT is generally
can be explained that the runoff in three seasons (i.e.,
Table 4 Goodness of fit (r2 ) and Nash-Sutcliff model efficiency
Table 2 Pearson correlation between MPN and CFU estimates coefficient (NSE) of regression models
Spring Summer Fall Winter
Spring Summer Fall Winter
E. coli Pearson correlation 0.38 0.68 0.67 0.38
E. coli r2 0.81 0.83 0.67 0.55
p-value (0.00) (0.00) (0.00) (0.00)
NSE 0.81 0.83 0.76 0.30
ENT Pearson correlation 0.79 0.47 0.68 0.75
ENT r2 0.47 0.52 0.49 0.88
p-value (0.00) (0.00) (0.00) (0.00)
NSE 0.76 0.52 0.49 0.85
MPN: most probable number; CFU: colony-forming unit.
No. 6 Evaluation of the relationship between two different methods for enumeration fecal indicator bacteria······ 849

Fig. 2 Measured vs. predicted E. coli in four seasons.

Fig. 3 Measured vs. predicted ENT in four seasons.

850 Kyung Hwa Cho et al.
lower than those of model for E. coli, the calculated r2
and NSE of regression models for ENT demonstrate that
the developed model can be useful for converting ENT
concentrations. The relative poor prediction on ENT can be
explained by the factor that the linear regression model is
not enough to capture the non-linear relationship between
the MPN and the CFU estimates. As shown in Table 4, r2
of the regression models for ENT are the greatest in winter
and the lowest in spring.
3 Conclusions
MPN and CFU are the common procedures for enu-
merating E. coli and ENT. The purposes of this study
is to investigate the relationship between two different
methods using the sample from the same water body, and
to develop the statistical model for converting E. coli and
ENT concentrations from CFU to MPN estimates, or vice
versa. Here, the results of this study can be summarized as
follow:
(1) In general, the enumerated E. coli in MPN is greater
than that in CFU, except for the measurement in winter.
Contrarily, ENT in MPN is lower than that in CFU.
(2) Even though there is the significant difference on
magnitude of FIB between two methods, the positive
correlation was found between MPN and CFU procedures
for enumerating E. coli and ENT.
(3) While nonlinear regression model between E. coli
concentrations in MPN and CFU resulted in the reasonable
model, the linear regression demonstrated that it can be
useful for application of ENT concentration.
We expect this study will provide preliminary informa-
tion towards future research on whether different analysis
methods may result in different water quality standard
violation frequencies for the same water sample.
Acknowledgments
This work was supported by the Korean Ministry of
Environment as “The Eco-technopia 21 Project” (No. 019-
071-053).
Vol. 22
References
CDC (Centers for Disease Control), 2004. Healthy swim-
ming. Center for Disease Control, Atlanta, Georgia.
http://www.cdc.gov/healthyswimming/index.htm, accessed
December 14, 2007.
Cha S M, Ham Y S, Ki S J, Lee S W, Cho K H, Park Y E
et al., 2009. Evaluation of pollutants removal efficiency
to achieve successful urban river restoration project at the
Gwangju Stream, Korea. Water Science and Technology,
59(11): 2101–2109.
Eckner K F, 1998. Comparison of membrane filtration and
multiple-tube fermentation by the colilert and enterolert
methods for detection of waterborne coliform bacteria,
Escherichia coli, and enterococci used in drinking and
bathing water quality monitoring in southern sweden. Ap-
plied Environmental Microbiology, 64(8): 3079–3083.
Kang J H, Lee Y S, Ki S J, Lee Y G, Cha S M, Cho K H et al.,
2009. Characteristics of wet and dry weather heavy metal
discharges in the Yeongsan Watershed, Korea. Science of
the Total Environment, 407(11): 3482–3493.
Noble R T, Weisberg S B, Leecaster M K, McGee C D, Ritter K J,
Walker K O et al., 2003. Comparison of beach bacterial wa-
ter quality indicator measurement methods. Environmental
Monitoring Assessment. 81(1): 301–312.
Olstadt J, Schauer J J, Standridge J, Kluender S, 2007. A
comparison of ten USEPA approved total coliform/E. coli
tests. Journal of Water and Health, 5(2): 267–282.
Rawlings J O, Pantula S G, Dickey D A , 1998. Ap-
plied regression analysis. In: A Research Tool (2nd ed.).
Springer-Verlag, New York.
US EPA (United States Environmental Protection Agency), 2002.
The 2000 national water quality inventory. EPA-841-R-02-
001. US EPA Office of Water, Washington DC.
US EPA (United States Environmental Protection Agency),
2000. Method 1600: improved enumeration methods for
the recreational water quality indicators: enterococci and
Escherichia coli. EPA/821/R-97/004. Office of Science and
Technology, Washington DC.
YRBEO (Yeongsan River Basin Environmental Office), 2005.
A special plan to improve water quality of the Yeongsan
watershed, Gwangju, Korea.