KSCE Journal of Civil Engineering Water Resources and Hydrologic Engineering

Copyright ⓒ2019 Korean Society of Civil Engineers

DOI 10.1007/s12205-019-0532-1 pISSN 1226-7988, eISSN 1976-3808
www.springer.com/12205
TECHNICAL NOTE

Curve Number Estimation of Ungauged Catchments considering

Characteristics of Rainfall and Catchment
Nam Won Kim* and Mun-Ju Shin**
Received March 16, 2018/Revised June 5, 2018/Accepted December 18, 2018/Published Online March 4, 2019

··································································································································································································································

Abstract
The Soil Conservation Service Curve Number (CN) method is widely used to calculate the flood runoff in ungauged catchments.
However, the existing CN calculation method has a disadvantage in that it cannot calculate the CNs considering the spatiotemporal
variability of rainfall in the ungauged catchments. In this study, the authors used a distributed rainfall-runoff model and a simple
runoff generation method to generate the hourly runoffs of the ungauged catchments considering the spatiotemporal variability of
rainfall, and estimate the CNs of the ungauged catchments using the generated runoff and the CN back calculation method. As a
result of calculating the CNs of the ungauged catchments for past 20 rainfall events, the CNs had a large variability even in the same
catchment. In addition, the mean CNs for the 20 rainfall events of independent ungauged catchments differed even if the ungauged
catchments were close to each other. The method of estimating the CNs of the ungauged catchments presented in this study properly
reflects the characteristics of actual rainfall and catchment, and thus can be usefully used to estimate valid CN.
Keywords: ungauged catchments, runoff data generation method, curve number, flood events, spatiotemporal variability of rainfall
··································································································································································································································

1. Introduction is a bare soil and a hydrologic soil group is A that has low runoff
potential and high infiltration rates. However, if the hydrological
The Soil Conservation Service Curve Number (CN) method modeler determines that the hydrologic soil group is B that has a
(U.S. Soil Conservation Service, 1972; U.S. Department of moderate infiltration rate, then the modeler will use CN of 86
Agriculture, Natural Resources Conservation Service, 2004) is a instead of 77 (refer to Table 9–1 in Chapter 9 of the Natural
widely used method for calculating flood runoff for the design of Resources Conservation Service (www.nrcs.usda.gov)) and
hydraulic structures in ungauged catchments due to the simplicity of consequently produce more simulated runoff. However, this
use and the ease of obtaining the required data (Ponce and Hawkins, handbook method has a problem that there is a large difference
1996; Shaw and Walter, 2009; Nagarajan and Poongothai, 2012). from the CN calculated using observed rainfall and runoff
This method determines the CN using antecedent moisture (Titmarsh et al., 1995; Stewart et al., 2012; Tedela et al., 2012).
condition, soil type, land use, and hydrologic surface condition A geographic information system can be used as an alternative
presented in the SCS-CN handbook. Then the empirical parameter, method for estimating the CNs of ungauged catchments (Jena et
CN, is used to predict the direct runoff of ungauged catchment. al., 2012; El-Hames, 2012). However, the CN calculated by this
The CN ranges from 0 to 100, where it is 100 for impervious method has a disadvantage in that it does not consider the
areas or water surfaces and less than 100 for natural surfaces. spatiotemporal variability of rainfall in the ungauged catchments. It
The Soil Conservation Service has tabulated the CN for various means that the result of the flood runoff calculation of the
conditions that can be found in Chapter 9 (Hydrologic Soil- ungauged catchment using the estimated CN can be significantly
Cover Complexes) of the Natural Resources Conservation different from the actual flood runoff.
Service (www.nrcs.usda.gov). Therefore, the CN of the target In this study, the authors propose a method to estimate the CNs
catchment is read from the CN table and used. Hydrological of ungauged catchments in the Andongdam catchment in South
modelers determine the CN by considering the aforementioned Korea considering the spatiotemporal variability of rainfall and
catchment conditions, and the subject of the modeler is involved the spatial variability of the physical characteristics of catchments.
during the CN determination process. Therefore, the CN difference This method calculates the CN by applying the rainfall and
due to subjective intervention affects the runoff simulation runoff of the ungauged catchment to the back calculation method
results. For example, suppose a cover treatment in a fallow area (Hawkins, 1973). The back calculation method has been applied

*Senior Research Fellow, Water Resources and River Research Institute, Korea Institute of Civil Engineering and Building Technology, Goyang 10223,
Korea (E-mail: [email protected])
**Member, Postdoctoral Research Fellow, Water Resources and River Research Institute, Korea Institute of Civil Engineering and Building Technology,
Goyang 10223, Korea (Corresponding Author, E-mail: [email protected])

−1−
 Nam Won Kim and Mun-Ju Shin

to various studies (Hawkins et al., 1985; Hawkins, 1993; Simanton catchments. Therefore, three gauged catchments were used for
et al., 1996; Bonta, 1997; Stewart et al., 2012; Tedela et al., 2012). calibration and validation of parameters. In order to analyze the
The runoff of the ungauged catchment is generated using the effect of storm on the CNs of ungauged catchments, 47 ungauged
Grid-based Rainfall-runoff Model (GRM) (Choi et al., 2015) and catchments located upstream of the Andongdam catchment
a simple runoff data generation method (Kim and Shin, 2017). (Watch Points in Fig. 1 and WPs in Table 1) were additionally
The purpose of this study is to 1) generate the runoff of ungauged selected. In Table 1, the “Class” column represents the catchment
catchments for past 20 rainfall events and calculate the CNs classification, and “Main” represents a catchment with an outlet
using the generated runoff, 2) investigate the variability of the on the main stream, Nakdong River. In addition, “Sub” represents a
CNs by the different rainfall events in the same catchment, and catchment with an outlet on the sub-stream of the main stream,
3) investigate the variability of mean CNs by comparing the and “Sub-upper” represents the catchment located upstream of
mean CNs of the 20 rainfall events for completely independent the sub-stream. The sub-upper catchments are completely
ungauged catchments. The mean CN represents the average independent ungauged catchments.
condition of the catchment, so this value can be used for design
of structures and as conservation measures (Hawkins et al., 2.2 Data
2009). In order to simulate the flood runoff, the hourly rainfall data of
27 rainfall stations located in and nearby the Andongdam catchment
2. Material were used (Fig. 1). These rainfall stations are operated by the
Korea Meteorological Administration (http://www.kma.go.kr),
2.1 Study Area K-water (http://www.kwater.or.kr) and Ministry of Land,
Study area is the Andongdam catchment (1,584 km2), which is Infrastructure and Transport (http://www.molit.go.kr). Twenty
a mountainous catchment in South Korea (Fig. 1). One multipurpose rainfall events (Table 2) were selected from 1998 to 2009, where
dam is located at the outlet of this catchment and it plays an the quality of the data was relatively good. The gauge rainfall
important role in mitigating flood damage in downstream areas. data for each rainfall event was converted to grid rainfall
The water produced in this catchment plays an important role in (resolution 200 m × 200 m) using ordinary kriging method, and
domestic use and in the water supply of industry and agriculture. the grid rainfall was used to simulate GRM runoff. Therefore, the
Two water level gauge stations (Dosan and Socheon) are located runoff simulation results reflect the characteristics of the
upstream of the Andongdam catchment, and these water level spatiotemporal distribution of rainfall.
gauge stations are also selected as the outlets of these gauged In order to generate hourly observed runoff data for calibration
and validation of GRM parameters, water level and rating data
provided by the Water Resources Management Information
System (http://www.wamis.go.kr) were used. The observed
runoff data of the three gauged catchments were generated for
the 20 rainfall events.
In order to generate physical input data of the GRM, grid data
(resolution of 200 m × 200 m) was generated using digital elevation
model (National Geographic Information Institute, http://www.ngii.
go.kr), soil map (National Academy of Agricultural Science,
http://www.naas.go.kr), and land use map (Ministry of Environment,
http://www.me.go.kr) with scales of 1:25000. Therefore, the
runoff simulation results take into account the spatial variability
of the physical characteristics of the catchment.

2.3 Rainfall-runoff Model

The GRM is a physically-based distributed rainfall-runoff
model that can simulate hourly flood runoff for rainfall events.
This model simulates surface, channel and subsurface flow using
the kinematic wave model (Beven, 1981) and calculates the
infiltration using the Green-Ampt model. In addition, it computes
the baseflow using the Darcy's law (Freeze and Cherry, 1979).
Since this model is a distributed rainfall-runoff model, it calculates
all these hydrological elements on a grid-by-grid basis. This
means that the flood runoff can be simulated at any target point.
Using these advantages of the distributed model, flood runoff
Fig. 1. Study Area data of ungauged catchment was generated (see Section 3.1).

−2− KSCE Journal of Civil Engineering

 Curve Number Estimation of Ungauged Catchments considering Characteristics of Rainfall and catchment

Table 1. Catchment Characteristics

Name of Area Channel length Average [a] Name of Area Channel length Average [a]
catchment (km2)
Class (km2)
Class
(km) slope catchment (km) slope
Andongdam 1,584.2 147.2 0.21 Main WP26 751.4 82.5 0.24 Main
WP2 14.7 5.6 0.19 Sub-upper WP27 39.1 10.6 0.24 Sub-upper
WP3 1,459.7 132.4 0.21 Main WP28 28.4 11.1 0.27 Sub-upper
WP4 51.6 12.6 0.10 Sub Socheon 645.0 62.5 0.25 Main
WP5 13.8 5.3 0.13 Sub-upper WP30 104.5 20.8 0.28 Sub
WP6 1,327.7 119.6 0.23 Main WP31 32.7 7.4 0.30 Sub-upper
WP7 89.6 19.9 0.20 Sub WP32 540.0 61.7 0.24 Main
WP8 81.2 16.4 0.21 Sub WP33 88.5 23.3 0.21 Sub
WP9 17.6 5.5 0.20 Sub-upper WP34 37.3 10.1 0.21 Sub-upper
WP10 1,235.9 118.4 0.23 Main WP35 402.1 49.9 0.25 Main
WP11 1,221.5 114.0 0.23 Main WP36 33.9 9.3 0.22 Sub
WP12 36.6 12.6 0.14 Sub-upper WP37 13.0 4.6 0.24 Sub-upper
Dosan 1,157.4 101.6 0.23 Main WP38 60.5 14.2 0.24 Sub
WP14 8.9 5.8 0.23 Sub-upper WP39 23.2 7.4 0.24 Sub-upper
WP15 1,142.6 99.5 0.23 Main WP40 251.2 29.2 0.25 Main
WP16 11.2 5.6 0.25 Sub-upper WP41 49.2 12.8 0.26 Sub
WP17 231.7 33.1 0.19 Sub WP42 22.2 9.0 0.28 Sub-upper
WP18 34.0 6.5 0.10 Sub-upper WP43 14.2 4.2 0.22 Sub-upper
WP19 5.6 4.3 0.13 Sub-upper WP44 189.9 25.2 0.25 Main
WP20 28.6 7.5 0.29 Sub-upper WP45 58.6 15.1 0.24 Sub-upper
WP21 47.0 10.8 0.24 Sub-upper WP46 131.2 24.6 0.25 Sub
WP22 850.0 89.5 0.24 Main WP47 41.6 10.9 0.27 Sub
WP23 838.8 83.1 0.24 Main WP48 32.6 7.7 0.28 Sub-upper
WP24 86.4 20.3 0.18 Sub WP49 53.6 12.5 0.23 Sub
WP25 39.2 10.2 0.18 Sub-upper WP50 19.6 7.0 0.25 Sub-upper
Note: [a] “Main” represents the catchment with the outlet on the main stream, Nakdong River, “Sub” represents the catchment with the outlet on the
main stream tributary, and “Sub-upper” represents the catchment located upstream of the sub-stream.

Table 2. The 20 Rainfall Events

Event name Period (year/month/day:hour) Event name Period (year/month/day:hour)
1998_Evt1 1998/08/14:19-1998/08/17:23 2004_Evt1 2004/06/19:04-2004/06/23:03
1999_Evt1 1999/08/02:13-1999/08/05:04 2004_Evt2 2004/07/15:09-2004/07/19:02
1999_Evt2 1999/09/19:20-1999/09/22:01 2004_Evt3 2004/08/17:19-2004/08/20:22
1999_Evt3 1999/09/23:04-1999/09/25:19 2006_Evt1 2006/07/10:07-2006/07/12:08
2000_Evt1 2000/09/13:01-2000/09/15:18 2006_Evt2 2006/07/15:14-2006/07/17:10
2000_Evt2 2000/09/16:04-2000/09/18:04 2006_Evt3 2006/07/28:04-2006/07/31:01
2002_Evt1 2002/08/05:20-2002/08/09:16 2007_Evt1 2007/09/01:02-2007/09/03:17
2002_Evt2 2002/08/30:23-2002/09/02:04 2008_Evt1 2008/07/24:19-2008/07/27:03
2003_Evt1 2003/06/27:09-2003/06/29:18 2009_Evt1 2009/07/11:21-2009/07/14:08
2003_Evt2 2003/09/11:24-2003/09/15:01 2009_Evt2 2009/07/14:15-2009/07/16:22

Table 3. Parameters of the GRM for Calibration

Number. Name of parameter Unit Parameter range Description of parameter
1 IniSaturation - 0−1 Initial soil saturation ratio
2 MinSlopeLdSrf - 0.0001−0.04 Minimum slope of land surface
3 MinSlopeChBed - 0.001−0.04 Minimum slope of channel bed
4 MinChBaseWidth m - Minimum channel width
5 ChRoughness m(−1/3)s 0.025-0.15 Channel roughness coefficient

Among the five parameters used for calibration in this study the characteristics of the storm. On the other hand, parameter No.
(Table 3), parameters Nos. 2 to 5 are related to the physical 1 is related to the initial soil saturation ratio and has different
characteristics of the catchment and are therefore independent of values for each rainfall event because the preceding soil moisture

−3−
 Nam Won Kim and Mun-Ju Shin

is different for each rainfall event. A description of how to rainfall-runoff model (herein GRM) to generate hourly hydrographs
calibrate and validate these parameters is given in Section 3.1. of ungauged catchments. It generates hydrographs of ungauged
catchments by applying calibrated parameters of a downstream
2.4 Objective Function and Model Performance Evaluation gauged catchment (calibration catchment, the Andongdam in
Statistics Fig. 1) to upstream ungauged catchments. Therefore, this method
The objective function used for parameter calibration is the assumes that the simulated flow at each upstream point is valid if
Nash-Sutcliffe Efficiency (NSE) (Nash and Sutcliffe, 1970). the simulated flow at the catchment outlet is valid. This
NSE, which emphasizes high flow fitting (Legates and McCabe, assumption is valid because the simulated flows from the grids at
1999), is suitable for use as an objective function for flood the upstream points join the catchment outlet through the flow
simulation as in this study. NSE is as follows: routing process. This method is a special case of regionalization
because it transfers downstream gauged catchment information to
∑ (Q − Qsim ,i )
n 2
upstream ungauged catchments. In addition, this method is more
obs ,i
NSE = 1 - i =1
(1) efficient and practical than the DMIP method (Smith et al., 2012)
∑ (Q )
n 2

i =1 obs ,i − Qobs for flood data generation in ungauged catchment because it can
generate flow data only for flood events of interest.
where Qobs i is the observed flow of time step i (herein hourly),
, This method simultaneously calibrates the physical parameters
Qobs is the mean of the observed flow, Qsim i is a simulated flow,
,
(Nos. 2 to 5 in Table 3) for all rainfall events. This is because the
and n is the total number of time steps. The range of NSE is from physical characteristics of the catchment are not related to the
−∞ to 1 and 1 means that the rainfall-runoff model completely storm characteristics, so these parameters must have a consistent
simulates the observed flow. This objective function was also value regardless of the magnitude and distribution of the rainfall.
used as a model performance evaluation statistic because the Since the GRM uses physical input data, it reflects the spatial and
NSE provides overall information about the simulated flow physical characteristics of the catchment in the runoff simulation,
(Moriasi et al., 2007). and these calibrated parameter values are used as minimum
The modified correlation coefficient (rmod) (McCuen and thresholds. In this study, these parameters were manually
Snyder, 1975) and percent bias (PBIAS) were used as additional calibrated using the NSE objective function. The simultaneous
model performance evaluation statistics. The rmod solves the calibration of parameters for multiple events has a key idea
problem that outliers have a serious impact on the correlation similar to the simultaneous calibration of parameters for multiple
coefficient and the difference between two hydrographs does not catchments (Parajka et al., 2007). The parameter values of a
affect the correlation coefficient (McCuen and Snyder, 1975). distributed rainfall-runoff model can in theory be calculated
rmod is as follows: using the physical characteristics of the catchment, but due to
min{σ sim , σ obs } problems such as scale differences between the measurement
rmod = r × (2) data and the model grid, insufficient data to calculate the parameter
max{σ sim , σ obs }
values, and measurement errors in the data, the parameters are
where r is the correlation coefficient, and σsim and σobs are the generally calibrated (Abbott et al., 1986; Refsgaard and Storm,
standard deviations for the simulated and observed flow, respectively. 1996; Eckhardt and Arnold, 2001; Lee et al., 2005; Sahoo et al.,
PBIAS is a statistic that can clearly show the poor model 2006). After calibration of the physical parameters, the parameter
simulation of the flow volume (Gupta et al., 1999). It is the for the antecedent soil moisture condition (No. 1 in Table 3) was
percentage difference between the volume of simulated and manually calibrated for each rainfall event. This is because the
observed flow. The best PBIAS value is 0 and a negative value initial soil saturation ratio is different for each rainfall event.
indicates an underestimation of the simulated runoff. The SDE method does not use the split-sample test (Klemeš,
1986), which uses independent historical events for parameter
∑ (Q − Qobs ,i )
n

× (100 )
sim ,i validation, because this method uses all historical data for
PBIAS = i =1
(3)
∑
n parameter calibration. Instead, this method validates the parameters
Q
i =1 obs ,i
using all historical data in upstream gauged catchments
The range of satisfactory model performance evaluation statistic (validation catchments, herein the Dosan and Socheon catchments
values selected in this study are NSE > 0.5 (Moriasi et al., 2007), in Fig. 1), which are assumed to be ungauged catchments.
rmod > 0.6 (Smith et al., 2012), and PBIAS = ± 30% (Smith et al., Good model performance evaluation statistics (NSE, PBIAS,
2012). and rmod) for validation catchments indicate that the assumption
of the SDE method is appropriate. Therefore, hourly flood data
3. Method for the ungauged catchments (herein 47 catchments in Table 1)
can be generated using the SDE method. This parameter
3.1 Spatial Data Extension Method calibration and validation method is a special method for
The Spatial Data Extension (SDE) method (Kim and Shin, regionalization.
2017) is a simple data generation method that uses a distributed

−4− KSCE Journal of Civil Engineering

 Curve Number Estimation of Ungauged Catchments considering Characteristics of Rainfall and catchment

3.2 Curve Number Estimation using Rainfall and Simulated Dosan and Socheon catchments) over the 20 rainfall events. The
Flow Data bold and underlined values are the performance statistic values
CN can be calculated by Eq. (4) as follows (USDA-NRCS, placed outside of the acceptance boundaries. The Andongdam
2004): catchment had satisfactory NSE, rmod, and PBIAS values for all
rainfall events (the maximum values of NSE, rmod and PBIAS
1000
CN = were 0.99, 0.97, and 28.4, respectively, and the minimum values
S (4)
10 + were 0.70, 0.66, and 0.7, respectively). The Dosan catchment
25.4
had satisfactory model performance evaluation statistics for the
where S is potential maximum retention in mm. most rainfall events, but had not for one rainfall event (the
S can be calculated by Eq. (5), which is suggested by Hawkins maximum values of NSE, rmod and PBIAS were 0.98, 0.97, and
(1973), if initial loss is 20% of S. The initial loss value can be 34.3, respectively, and the minimum values were 0.48, 0.70, and
estimated using the observed rainfall and flow data in gauged 1.8, respectively). The simulated flow of the 2000_Evt1 event
catchments. For ungauged catchments, however, there is no for the Dosan catchment was slightly overestimated with respect
observed data to estimate S, so the commonly used 20% of S was to the PBIAS value (34.3) however, the NSE value could be
applied as follows (Grimaldi et al., 2013): acceptable because it is very close to the acceptance boundary
(0.48). In addition, the rmod value was placed within the acceptance
S = 5 ⎡ ( P + 2Q ) − Q ( 4Q + 5 P ) ⎤ (5)
⎣ ⎦ boundary; therefore, the authors accepted the simulated flow in
where P is the total volume of rainfall (mm) and Q is the total this rainfall event. In the same way, the rainfall events with the
volume of the simulated flow (mm) in a rainfall event. Given unsatisfactory model performances for the Socheon catchment
Eqs. (4) and (5), CN of the ungauged catchment can be were screened out (the maximum values of NSE, rmod and
calculated. PBIAS were 0.98, 0.96, and 40.0, respectively, and the minimum
values were 0.00, 0.52, and 1.0, respectively), and one event of
4. Results and Discussion 2004_Evt1 had unsatisfactory results with respect to all three
performance evaluation statistics. The main reason could be
4.1 Model Performance error in rating formula used to calculate the observed flow rather
Table 4 shows the three model performance evaluation statistics than error in the climate data for input or the GRM model
(NSE, rmod, and PBIAS) for the one calibration catchment (the structure in consideration of the acceptable results for the other
Andongdam catchment) and the two validation catchments (the catchments and the acceptable results for the other events for the

Table 4. Model Performance Evaluation Statistics for the Three Gauged Catchments Over the 20 Rainfall Events [a]

Andongdam Dosan Socheon

Event
NSE rmod PBIAS NSE rmod PBIAS NSE rmod PBIAS
1998_Evt1 0.86 0.90 −15.0 0.80 0.76 14.2 0.88 0.86 −3.0
1999_Evt1 0.72 0.88 −21.7 0.83 0.94 −18.9 0.53 0.65 −14.1
1999_Evt2 0.70 0.66 −5.9 0.87 0.86 −10.7 0.78 0.79 −3.5
1999_Evt3 0.90 0.89 −9.2 0.93 0.89 −8.3 0.75 0.68 −4.9
2000_Evt1 0.82 0.89 17.9 0.48 0.71 34.3 0.93 0.96 14.0
2000_Evt2 0.89 0.93 −8.4 0.90 0.93 12.1 0.70 0.79 −21.5
2002_Evt1 0.86 0.88 0.7 0.81 0.70 −21.5 0.87 0.93 13.4
2002_Evt2 0.99 0.96 0.7 0.85 0.76 −29.7 0.98 0.85 −1.0
2003_Evt1 0.93 0.91 −12.0 0.83 0.73 −23.5 0.76 0.81 −16.8
2003_Evt2 0.89 0.86 22.0 0.92 0.86 −8.4 0.07 0.52 20.0
2004_Evt1 0.78 0.86 28.4 0.75 0.90 29.0 0.05 0.56 40.0
2004_Evt2 0.74 0.82 −2.7 0.84 0.89 −2.7 0.00 0.52 −7.2
2004_Evt3 0.85 0.93 18.9 0.96 0.97 15.0 0.45 0.62 24.7
2006_Evt1 0.92 0.90 −5.1 0.95 0.89 −2.5 0.79 0.73 12.5
2006_Evt2 0.95 0.96 −7.3 0.92 0.88 −1.8 0.55 0.70 20.7
2006_Evt3 0.92 0.97 −9.2 0.98 0.97 −2.5 0.68 0.65 4.8
2007_Evt1 0.88 0.84 −2.3 0.91 0.82 −8.8 0.26 0.56 14.9
2008_Evt1 0.89 0.93 −2.3 0.95 0.96 −10.3 0.83 0.79 6.2
2009_Evt1 0.93 0.88 8.9 0.96 0.89 −7.6 0.87 0.91 −7.2
2009_Evt2 0.90 0.90 −8.4 0.81 0.90 −26.8 0.75 0.76 −23.7
Note: [a]Dosan and Socheon catchments are assumed as ungauged catchments. The bold and underlined values represent the performance statistic
values placed outside of the acceptance boundaries (NSE > 0.5, rmod > 0.6, and PBIAS ± 30 %).

−5−
 Nam Won Kim and Mun-Ju Shin

Table 5. The CNs of the Gauged and Ungauged Catchments for the 20 Rainfall Events
1998_ 1999_ 1999_ 1999_ 2000_ 2000_ 2002_ 2002_ 2003_ 2003_
Catchment
Evt1 Evt1 Evt2 Evt3 Evt1 Evt2 Evt1 Evt2 Evt1 Evt2
Andongdam 82 85 80 78 78 89 65 74 76 92
WP2 75 73 71 67 69 88 54 61 65 87
WP3 82 85 81 78 79 89 64 74 76 92
WP4 82 81 84 77 79 92 64 80 78 93
WP5 82 79 80 77 79 93 63 80 76 93
WP6 83 85 80 79 80 90 64 73 76 92
WP7 77 80 76 74 80 90 59 72 75 89
WP8 76 79 74 73 78 89 58 71 73 88
WP9 68 69 67 63 64 88 45 60 62 82
WP10 83 85 81 79 80 90 64 73 76 92
WP11 83 85 81 79 80 90 64 73 76 92
WP12 85 83 84 83 89 94 71 85 80 95
Dosan 83 85 81 78 80 90 63 73 76 77
WP14 71 69 70 68 67 89 51 67 64 85
WP15 82 85 81 78 80 90 63 73 76 69
WP16 73 72 76 71 70 89 54 73 66 84
WP17 72 74 71 70 69 88 52 69 68 85
WP18 68 69 68 65 61 86 47 67 66 74
WP19 63 68 65 60 58 83 41 61 65 82
WP20 74 75 75 75 77 92 54 72 70 90
WP21 71 72 68 71 68 90 51 68 67 90
WP22 84 87 83 80 82 91 66 73 77 90
WP23 84 87 83 79 82 91 65 73 77 89
WP24 78 80 80 76 79 91 60 77 73 85
WP25 79 82 81 76 79 91 60 76 73 84
WP26 85 87 83 80 82 91 66 72 77 90
WP27 74 78 71 70 72 89 52 64 64 85
WP28 70 73 64 66 67 86 48 58 60 80
Socheon 83 85 84 78 82 91 64 73 78 87
WP30 80 80 83 78 83 91 60 76 76 89
WP31 80 80 87 82 89 93 62 81 81 89
WP32 84 86 84 79 82 91 65 72 78 87
WP33 67 80 74 67 74 87 53 59 68 79
WP34 60 72 66 65 65 86 46 54 62 77
WP35 83 84 86 79 82 91 65 73 79 87
WP36 67 69 59 62 25 85 45 53 59 74
WP37 70 67 57 63 61 85 45 54 59 76
WP38 80 82 82 75 82 91 60 64 75 82
WP39 75 79 75 71 76 88 41 57 70 80
WP40 82 79 85 79 81 90 63 75 79 86
WP41 73 71 77 71 72 86 51 69 73 79
WP42 77 73 81 75 78 88 53 73 77 80
WP43 70 68 71 66 63 85 46 65 68 78
WP44 85 81 87 81 84 91 65 77 80 87
WP45 82 79 82 78 79 90 62 72 76 83
WP46 86 82 89 83 86 92 67 80 82 89
WP47 79 74 79 76 76 89 57 73 75 85
WP48 78 74 78 76 76 89 55 72 74 85
WP49 84 81 89 84 89 93 67 81 83 89
WP50 81 77 83 79 81 91 61 76 78 87

−6− KSCE Journal of Civil Engineering

 Curve Number Estimation of Ungauged Catchments considering Characteristics of Rainfall and catchment

Table 5. (continue)
2004_ 2004_ 2004_ 2006_ 2006_ 2006_ 2007_ 2008_ 2009_ 2009_
Catchment
Evt1 Evt2 Evt3 Evt1 Evt2 Evt3 Evt1 Evt1 Evt1 Evt2
Andongdam 68 84 79 77 71 87 81 73 81 81
WP2 52 78 67 62 62 73 70 67 70 76
WP3 68 84 80 77 72 87 82 72 81 81
WP4 64 84 73 77 73 88 80 77 83 81
WP5 63 85 73 78 76 84 82 75 83 82
WP6 68 85 80 77 73 88 82 72 81 81
WP7 63 82 74 72 69 84 77 73 77 80
WP8 61 81 72 71 68 82 76 71 76 80
WP9 48 73 62 61 59 69 66 63 67 75
WP10 68 85 81 78 73 88 83 72 82 81
WP11 68 85 81 78 73 88 83 71 81 81
WP12 71 89 81 81 79 92 87 80 86 97
Dosan 68 84 81 78 73 87 83 70 81 81
WP14 54 78 68 66 66 73 71 63 73 77
WP15 68 84 81 78 74 87 83 70 81 81
WP16 58 80 71 72 69 76 77 70 75 81
WP17 54 76 68 71 69 76 74 63 72 67
WP18 48 71 63 70 63 69 70 61 71 68
WP19 43 67 58 68 58 63 68 57 65 75
WP20 61 82 74 72 75 79 82 64 75 76
WP21 54 76 68 68 70 73 73 62 70 75
WP22 70 86 83 79 75 89 86 71 83 82
WP23 70 86 83 80 76 88 86 71 83 82
WP24 65 81 77 76 74 83 81 75 82 82
WP25 65 83 76 76 76 82 81 75 81 83
WP26 70 86 83 80 76 89 87 71 83 82
WP27 54 78 70 66 70 74 75 64 73 78
WP28 48 73 65 62 74 69 69 57 67 74
Socheon 69 84 82 81 72 88 87 70 82 81
WP30 67 81 82 79 79 85 86 67 80 79
WP31 74 87 85 84 81 88 93 72 84 82
WP32 69 85 82 81 69 88 88 71 83 82
WP33 54 76 70 68 58 76 76 60 75 76
WP34 46 72 61 59 63 69 70 56 68 74
WP35 71 85 83 82 78 88 89 71 83 82
WP36 45 67 61 63 62 68 69 51 64 71
WP37 44 68 60 64 63 67 70 51 64 74
WP38 64 84 79 76 76 84 88 61 77 78
WP39 58 79 71 71 71 79 83 57 73 76
WP40 70 84 83 81 78 88 88 75 83 82
WP41 61 77 74 75 70 79 80 63 76 75
WP42 67 80 77 78 73 84 85 67 80 81
WP43 53 72 66 70 65 72 72 60 71 76
WP44 72 85 84 83 80 90 89 79 85 85
WP45 68 83 80 79 77 87 87 73 81 83
WP46 74 86 86 85 82 92 90 82 87 86
WP47 64 81 77 78 75 84 82 74 80 82
WP48 63 81 76 77 75 83 82 73 79 81
WP49 75 88 87 86 83 93 90 85 89 88
WP50 68 84 80 83 77 87 85 80 84 86

−7−
 Nam Won Kim and Mun-Ju Shin

Socheon catchment, respectively. In addition, the observed flow Table 6. Mean Value of the CNs of the 20 Rainfall Events
did not include outliers or anomaly. Therefore, the simulated Catchment Mean CN Catchment Mean CN Catchment Mean CN
flow for the 2004_Evt1 event was also accepted. Andongdam 79 WP18 66 WP35 81
WP2 69 WP19 63 WP36 61
4.2 Curve Number Estimation of Ungauged Catchments WP3 79 WP20 75 WP37 63
Table 5 shows the CNs calculated using the simulated flows of WP4 79 WP21 70 WP38 77
all catchments for the 20 rainfall events. For the same ungauged WP5 79 WP22 81 WP39 71
catchment, very different CNs were shown depending on the WP6 79 WP23 81 WP40 81
characteristics of the rainfall events such as magnitude and WP7 76 WP24 78 WP41 73
spatiotemporal variation. This result supports McCuen (2002)'s WP8 75 WP25 78 WP42 76
study that CNs are different for different rainfall events. WP9 66 WP26 81 WP43 68
Figure 2 shows the distribution of CNs versus peak flows for WP10 80 WP27 71 WP44 83
the 20 rainfall events in six ungauged catchments. The catchments in WP11 80 WP28 66 WP45 79
this figure are independent ungauged catchments located upstream, WP12 85 Socheon 80 WP46 84
midstream, and downstream of the Andongdam catchment (Fig. Dosan 79 WP30 79 WP47 77
1). The CNs were not correlated with the peak flows and were WP14 69 WP31 83 WP48 76
widely distributed. Therefore, CNs of the 20 rainfall events were WP15 78 WP32 80 WP49 85
WP16 73 WP33 70 WP50 80
averaged to estimate the representative CN that can be used for
WP17 70 WP34 64
design flood estimation of the ungauged catchment (Table 6).
The mean CNs of the ungauged catchments ranged from 61 to

Fig. 2. Distribution of the CNs corresponding to the Peak Flows of Fig. 3. The Mean CNs for 22 Independent Ungauged Catchments
the 20 Rainfall Events for Six Ungauged Catchments (The (The numbers in the Catchments Represent the Catchment
Horizontal Axis Represents the Peak Flow Rate in mm/h.) Number in Fig. 1.)

−8− KSCE Journal of Civil Engineering

 Curve Number Estimation of Ungauged Catchments considering Characteristics of Rainfall and catchment

85. The different spatiotemporal characteristics of the 20 rainfall (86)90114-9.

events and different physical characteristics of the ungauged Beven, K. (1981). “Kinematic subsurface stormflow.” Water Resources
catchments such as height, steepness, soil type, soil depth, and Research, Vol. 17, No. 5, pp. 1419-1424, DOI: 10.1029/
WR017i005p01419.
vegetation can cause the wide distribution of the mean CNs.
Bonta, J. V. (1997). “Determination of watershed curve number using
Figure 3 shows the mean CNs of 22 completely independent derived distributions.” Journal of Irrigation and Drainage Engineering,
ungauged catchments (“Sub-upper” catchments in Table 1). The Vol. 123, No. 1, pp. 28-36, DOI: 10.1061/(ASCE)0733-9437(1997)
mean CNs ranged from 63 to 85, and they were significantly 123:1(28).
different even in nearby catchments. It implies that the CNs of Choi, Y. S., Choi, C. K., Kim, H. S., Kim, K. T., and Kim, S. (2015).
ungauged catchments should be estimated considering the “Multi-site calibration using a grid-based event rainfall–runoff
characteristics of storms rather than by the SCS handbook model: A case study of the upstream areas of the Nakdong River
method, which considers antecedent moisture condition, soil basin in Korea.” Hydrological Processes, Vol. 29, No. 9, pp. 2089-
2099, DOI: 10.1002/hyp.10355.
type, land use, and hydrologic surface condition only. Therefore,
Eckhardt, K. and Arnold, J. G. (2001). “Automatic calibration of a
the CN estimated by the proposed method is an improved value distributed catchment model.” Journal of Hydrology, Vol. 251,
compared to the CN estimated by the conventional SCS Nos. 1-2, pp. 103-109, DOI: 10.1016/S0022-1694(01)00429-2.
handbook method. El-Hames, A. S. (2012). “An empirical method for peak discharge
prediction in ungauged arid and semi-arid region catchments based
5. Conclusions on morphological parameters and SCS curve number.” Journal of
Hydrology, Vol. 456, pp. 94-100, DOI: 10.1016/j.jhydrol. 2012.06.016.
This study proposed a method to estimate CNs of ungauged Freeze, R. A. and Cherry, J. A. (1979). Groundwater, Prentice Hall Inc.,
NJ, USA.
catchments using simulated runoff considering spatiotemporal
Grimaldi, S., Petroselli, A., and Romano, N. (2013). “Green-Ampt
variability of the rainfall and spatial variability of the physical Curve-Number mixed procedure as an empirical tool for rainfall–
characteristics of the catchments. Firstly, the NSE, rmod, and runoff modelling in small and ungauged basins.” Hydrological
PBIAS statistics were used to evaluate the appropriateness of the Processes, Vol. 27, No. 8, pp. 1253-1264, DOI: 10.1002/hyp.9303.
simulated runoff by the GRM for the three gauged catchments. Gupta, H. V., Sorooshian, S., and Yapo, P. O. (1999). “Status of automatic
As a result, proper simulation results were obtained for the 20 calibration for hydrologic models: Comparison with multilevel
rainfall events in the calibration and validation catchments. Since expert calibration.” Journal of Hydrologic Engineering, Vol. 4,
the estimated parameters of the GRM were appropriate, the SDE No. 2, pp. 135-143, DOI: 10.1061/(ASCE)1084-0699(1999)4:2(135).
Hawkins, R. H. (1973). “Improved prediction of storm runoff in mountain
method was applied to generate the runoff of the 47 ungauged
watersheds.” Journal of the Irrigation and Drainage Division, Vol. 99,
catchments for the 20 rainfall events, and the CNs of the No. 4, pp. 519-523, DOI: 10.1061/(ASCE)0733-9437(1993)119:2(334).
ungauged catchments were estimated using the generated runoff. Hawkins, R. H. (1993). “Asymptotic determination of runoff curve
As a result, even the same catchment had different CNs for numbers from data.” Journal of Irrigation and Drainage Engineering,
different rainfall events. The mean CNs of the ungauged catchments Vol. 119, No. 2, pp. 334-345, DOI: 10.1061/(ASCE)0733-9437
were different, and even the neighboring ungauged catchments (1993)119:2(334).
had different mean CNs. Therefore, it is necessary to use the CN Hawkins, R. H., Hjelmfelt Jr, A. T., and Zevenbergen, A. W. (1985).
estimation method which can consider the characteristics of actual “Runoff probability, storm depth, and curve numbers.” Journal of
Irrigation and Drainage Engineering, Vol. 111, No. 4, pp. 330-340,
storms in the ungauged catchments proposed in this study for
DOI: 10.1061/(ASCE)0733-9437(1985)111:4(330).
better CN estimation. Hawkins, R. H., Ward, T. J., Woodward, D. E., and Van Mullem, J. A.
(2009). Curve number hydrology-state of practice, The ASCE/
Acknowledgements EWRI Publication, p. 37.
Jena, S. K., Tiwari, K. N., Pandey, A., and Mishra, S. K. (2012). “RS
This research was supported by a grant from a Strategic Research and Geographical Information System–based evaluation of distributed
Project (Development of Hydrological Safety Assessment System and composite curve number techniques.” Journal of Hydrologic
for Hydraulic Structures), that was funded by the Korea Institute Engineering, Vol. 17, No. 11, pp. 1278-1286, DOI: 10.1061/(ASCE)
HE.1943-5584.0000651.
of Civil Engineering and Building Technology (project number:
Kim, N. W. and Shin, M. J. (2017). “Experimental study of rainfall
20180374-001). We gratefully acknowledge Yun Seok Choi for spatial variability effect on peak flow variability using a data generation
valuable discussions regarding the study. method.” Journal of Korea Water Resources Association, Vol. 50,
No. 6, pp. 359-371, DOI: 10.3741/JKWRA.2017.50.6.359 (in Korean).
References Klemeš, V. (1986). “Operational testing of hydrological simulation
models.” Hydrological Sciences Journal, Vol. 31, No. 1, pp. 13-24,
Abbott, M. B., Bathurst, J. C., Cunge, J. A., O'Connell, P. E., and Rasmussen, DOI: 10.1080/02626668609491024.
J. (1986). “An introduction to the european hydrological system— Lee, H., McIntyre, N., Wheater, H., and Young, A. (2005). “Selection of
systeme hydrologique europeen,“SHE”, 1: History and philosophy conceptual models for regionalisation of the rainfall-runoff relationship.”
of a physically-based, distributed modelling system.” Journal of Journal of Hydrology, Vol. 312, Nos. 1-4, pp. 125-147, DOI:
Hydrology, Vol. 87, Nos. 1-2, pp. 45-59, DOI: 10.1016/0022-1694 10.1016/j.jhydrol.2005.02.016.

−9−
 Nam Won Kim and Mun-Ju Shin

Legates, D. R. and McCabe, G. J. (1999). “Evaluating the use of SHE, to predict streamflow at high frequency in a flashy mountainous
“goodness-of-fit” measures in hydrologic and hydroclimatic model Hawaii stream.” Journal of Hydrology, Vol. 327, Nos. 1-2, pp. 94-
validation.” Water Resources Research, Vol. 35, No. 1, pp. 233-241, 109, DOI: 10.1016/j.jhydrol.2005.11.012.
DOI: 10.1029/1998WR900018. Shaw, S. B. and Walter, M. T. (2009). “Improving runoff risk estimates:
McCuen, R. H. (2002). “Approach to confidence interval estimation for Formulating runoff as a bivariate process using the SCS curve
curve numbers.” Journal of Hydrologic Engineering, Vol. 7, No. 1, number method.” Water Resources Research, Vol. 45, No. 3,
pp. 43-48, DOI: 10.1061/(ASCE)1084-0699(2002)7:1(43). pp. W03404, DOI: 10.1029/2008WR006900.
McCuen, R. H. and Snyder, W. M. (1975). “A proposed index for Simanton, J. R., Hawkins, R. H., Mohseni-Saravi, M., and Renard, K. G.
comparing hydrographs.” Water Resources Research, Vol. 11, No. 6, (1996). “Runoff curve number variation with drainage area, Walnut
pp. 1021-1024, DOI: 10.1029/WR011i006p01021. Gulch, Arizona.” Transactions of the ASAE, Vol. 39, No. 4, pp. 1391-
Moriasi, D. N., Arnold, J. G., Van Liew, M. W., Bingner, R. L., Harmel, 1394, DOI: 10.13031/2013.27630.
R. D., and Veith, T. L. (2007). “Model evaluation guidelines for Smith, M. B., Koren, V., Zhang, Z., Zhang, Y., Reed, S. M., Cui, Z., and
systematic quantification of accuracy in watershed simulations.” Anderson, E. A. (2012). “Results of the DMIP 2 Oklahoma
Transactions of the ASABE, Vol. 50, No. 3, pp. 885-900, DOI: experiments.” Journal of Hydrology, Vol. 418, pp. 17-48, DOI:
10.13031/2013.23153. 10.1016/j.jhydrol.2011.08.056.
Nagarajan, N. and Poongothai, S. (2012). “Spatial mapping of runoff Stewart, D., Canfield, E., and Hawkins, R. (2012). “Curve number
from a watershed using SCS-CN method with remote sensing and determination methods and uncertainty in hydrologic soil groups
GIS.” Journal of Hydrologic Engineering, Vol. 17, No. 11, pp. 1268- from semiarid watershed data.” Journal of Hydrologic Engineering,
1277, DOI: 10.1061/(ASCE)HE.1943-5584.0000520. Vol. 17, No. 11, pp. 1180-1187, DOI: 10.1061/(ASCE)HE.1943-
Nash, J. E. and Sutcliffe, J. V. (1970). “River flow forecasting through 5584.0000452.
conceptual models part I—A discussion of principles.” Journal of Tedela, N. H., McCutcheon, S. C., Rasmussen, T. C., Hawkins, R. H.,
Hydrology, Vol. 10, No. 3, pp. 282-290, DOI: 10.1016/0022- Swank, W. T., Campbell, J. L., and Tollner, E. W. (2012). “Runoff
1694(70)90255-6. curve numbers for 10 small forested watersheds in the mountains of
Parajka, J., Blöschl, G., and Merz, R. (2007). “Regional calibration of the eastern United States.” Journal of Hydrologic Engineering, Vol. 17,
catchment models: Potential for ungauged catchments.” Water No. 11, pp. 1188-1198, DOI: 10.1061/(ASCE)HE.1943-5584.0000436.
Resources Research, Vol. 43, No. 6, pp. W06406, DOI: 10.1029/ Titmarsh, G. W., Cordery, I., and Pilgrim, D. H. (1995). “Calibration
2006WR005271. procedures for rational and USSCS design flood methods.” Journal
Ponce, V. M. and Hawkins, R. H. (1996). “Runoff curve number: Has it of Hydraulic Engineering, Vol. 121, No. 1, pp. 61-70, DOI: 10.1061/
reached maturity?” Journal of Hydrologic Engineering, Vol. 1, No. 1, (ASCE)0733-9429(1995)121:1(61).
pp. 11-19, DOI: 10.1061/(ASCE)1084-0699(1996)1:1(11). US Department of Agriculture, Natural Resources Conservation Service
Refsgaard, J. C. and Storm, B. (1996). “Construction, calibration and (USDA–NRCS) (2004). “Part 630: Hydrology.” National Engineering
validation of hydrological models.” Distributed Hydrological Modelling, Handbook, US Government Printing Office, Washington, DC, USA.
M. B. Abbott and J. C. Refsgaard, Eds., Kluwer Academic, pp. 41- US Soil Conservation Service (SCS) (1972). “Hydrology.” National
54. Engineering Handbook, US Government Printing Office, Washington,
Sahoo, G. B., Ray, C., and De Carlo, E. H. (2006). “Calibration and DC, USA.
validation of a physically distributed hydrological model, MIKE

− 10 − KSCE Journal of Civil Engineering