Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Pedosphere 21(6): 738–749, 2011

ISSN 1002-0160/CN 32-1315/P


c 2011 Soil Science Society of China
Published by Elsevier B.V. and Science Press

Application of the SCS-CN Model to Runoff Estimation in


a Small Watershed with High Spatial Heterogeneity∗1

XIAO Bo1,2,∗2 , WANG Qing-Hai1 , FAN Jun2 , HAN Feng-Peng2 and DAI Quan-Hou3
1
Beijing Research & Development Center for Grass and Environment, Beijing Academy of Agriculture and Forestry Sciences,
Beijing 100097 (China)
2
State Key Laboratory of Soil Erosion and Dryland Farming on the Loess Plateau, Institute of Soil and Water Conservation,
Chinese Academy of Sciences, Yangling 712100 (China)
3
College of Forest, Guizhou University, Guiyang 550025 (China)
(Received April 19, 2011; revised September 19, 2011)

ABSTRACT
For reasons of simplicity, the most commonly used hydrological models are based on the Soil Conservation Service
Curve Number (SCS-CN) model, which is probably a good choice for the estimation of runoff on the Loess Plateau of
China; however, the high spatial heterogeneity, mainly caused by a fragmented landform and variations in soil type, may
limit its applicability to this region. Therefore, applicability of the SCS-CN model to a small watershed, Liudaogou on the
plateau, was evaluated and the most appropriate initial abstraction ratio (Ia /S) value in the model was quantified by the
inverse method. The results showed that the standard SCS-CN model was applicable to the estimation of runoff in the
Liudaogou watershed and the model performance was acceptable according to the values of relative error and Nash-Sutcliffe
efficiency. The most appropriate Ia /S value for the watershed was 0.22 because with this modified Ia /S value, the model
performance was slightly improved. The model performance was not sensitive to the modification of the Ia /S value when
one heavy rainfall event (50.1 mm) was not considered, which implied that the model, using a standard Ia /S value, can be
recommended for the Liudaogou watershed because single rainfall events exceeding 50 mm seldom occurred in that region.
The runoff amount predicted for the Liudaogou watershed by the SCS-CN model, using the modified Ia /S value, increased
gradually with increasing rainfall when rainfall values were lower than 50 mm, whereas the predicted amount increased
rapidly when the rainfall exceeded 50 mm. These findings may be helpful in solving the problem of serious soil and water
loss on the Loess Plateau of China.
Key Words: curve number, initial abstraction ratio, model performance, rainfall, soil and water loss

Citation: Xiao, B., Wang, Q. H., Fan, J., Han, F. P. and Dai, Q. H. 2011. Application of the SCS-CN model to runoff
estimation in a small watershed with high spatial heterogeneity. Pedosphere. 21(6): 738–749.

INTRODUCTION reaches of the Yellow River (Milliman et al., 1987; Saito


et al., 2001). An estimation of surface runoff is essen-
The Loess Plateau of China is well known in the tial for reducing sediments and consequent hazards be-
world for its rapid rate of soil erosion (15 000–20 000 cause runoff is the driving force behind soil erosion.
t km−2 year−1 ), which is mainly due to the variable Although many hydrologic models are available for the
weather (drought followed by concentrated rainfalls), estimation of runoff, most physically based models are
sparse vegetation, loose soils, a complex landform, and limited because of their large number of input para-
the long-term improper use of land (Cha and Tang, meters and complicated calibration requirements (Wu
2000; Shi and Shao, 2000). The large amount of deli- et al., 1993; Kothyari and Jain, 1997). On the Loess
vered sediment seriously damages the middle and lower Plateau of China, the acquisition of these parameters

∗1
Supported by the National Natural Science Foundation of China (No. 41001156), the Beijing Novel Program, China
(No. 2009B25), the Beijing Municipal Natural Science Foundation, China (No. 8102015), and the Open Fund of the State
Key Laboratory of Soil Erosion and Dryland Farming on the Loess Plateau of China (No. 10501-295).
∗2
Corresponding author. E-mail: xiaoboxb@gmail.com.
SCS-CN MODEL AND RUNOFF ESTIMATION 739

at different scales (from single slopes to catchments) frequently questioned for its validity and applicability
and model calibration are difficult because of the to other regions (e.g., Ponce and Hawkins, 1996; Shi
scarcity of related research. Additionally, the tradi- et al., 2009). For example, some researchers indicated
tional hydrological models used for most small water- that the CN value should be verified before using this
sheds on the Loess Plateau can not represent the dis- model to estimate runoff in China, where the rainfalls
tributed nature of watershed-scale hydrological prop- are concentrated and intensive, much different from
erties like soil type, slope, and land use (Hellweger and those in America (Wang, 2005; Zhou et al., 2008).
Maidment, 1999). In this situation, the Soil Conser- The SCS-CN model was introduced in China in the
vation Service Curve Number (SCS-CN) model devel- 1990s and has been applied by many researchers, espe-
oped by the USDA-Soil Conservation Service in 1972 cially on the Loess Plateau (Pan, 1996; He, 2003; Liu
(USDA SCS, 1985) is probably a good choice for the et al., 2005; Huang et al., 2007; Liu and Li, 2008; Zhou
estimation of runoff. et al., 2008; Shi et al., 2009). The results of these stu-
The SCS-CN model is a simple, empirical model dies indicate that it is a valuable tool for the prediction
with clearly stated assumptions and few data require- of runoff from small watersheds on the Loess Plateau,
ments. Therefore, it has been widely used for wa- but its applicability should be locally validated and
ter resource management, storm water modeling and calibrated (Liu et al., 2005; Huang et al., 2007; Liu
runoff estimation for single rainfall events in small and Li, 2008). Additionally, the high spatial hetero-
agricultural or urban watersheds (Greene and Cruise, geneity of soil properties across the Loess Plateau, in-
1995; Tsihrintzis and Hamid, 1997; Lewis et al., cluding texture and depth (Wang et al., 2001; Li and
2000; Chandrmohan and Durbude, 2001; He, 2003; Shao, 2006), bulk density (Lian et al., 2006; Xiao et
Liu et al., 2005; Mishra et al., 2006; Liu and Li, al., 2009), moisture content (Qiu et al., 2001; Fu et al.,
2008; Sahu et al., 2010a). The model has also been 2003; Zhu and Shao, 2008), saturated hydraulic con-
adopted by many hydrological and ecological mo- ductivity (Stolte et al., 2003; Liu and Shao, 2009), and
dels to determine runoff, including CREAMS (Knisel, infiltrability (Jiang et al., 2005), limits the applicability
1980), ANSWERS (Beasley et al., 1980), AGNPS of the model to this region. High spatial heterogeneity
(Young et al., 1989), EPIC (Sharpley and Williams, makes runoff estimation become much more complex
1990), SWAT (Arnold et al., 1995), and QUALHYMO and difficult. The main objectives of this research were:
(http://www.waterbalance.ca/). Recently, the SCS- 1) to evaluate the applicability of SCS-CN model to a
CN model was extended to estimate sediment yield small watershed on the Loess Plateau of China with
and to model soil moisture (Reshmidevi et al., 2008; high spatial heterogeneity; 2) to accurately quantify
Singh et al., 2008). Some researchers also integrated the Ia /S value in order to increase the precision of es-
the SCS-CN model into the GIS/RS system to extend timation; and 3) to evaluate the relationship between
the model applicability to complex watersheds with rainfall and runoff resulting from the modified SCS-CN
high temporal and spatial variability in soil and land model.
use (Zhan and Huang, 2004; Geetha et al., 2007). Al-
THEORY OF THE SCS-CN MODEL
though the model is versatile and has been widely used
in the world, many researchers question its applica- The theory underlying the SCS-CN model is that
bility. Johnson (1998) pointed out that the SCS-CN runoff can be related to soil-cover complexes and rain-
model has been overused and applied to situations and fall through a curve number. The model assumes that
conditions for which it was not designed. Indeed, the the ratio of actual soil retention after runoff begins to
SCS-CN model was developed using small-sized test potential maximum retention is equal to the ratio of
plots and was originally designed for use in midsize ru- direct runoff to available rainfall. This relationship, af-
ral watersheds. Therefore, its extension to large basins ter algebraic manipulation and inclusion of simplifying
requires validation and calibration. Ponce (1989) con- assumptions, results in the following equation (USDA
cluded that this model should not be used for catch- SCS, 1985):
ments larger than 250 km2 without catchment subdi-
Q/(P − Ia ) = F/S (1)
vision. Apart from scale problems, the initial abstrac-
tion ratio (the ratio of initial abstraction to maximum where Q = runoff (mm), P = storm rainfall (mm), Ia
potential retention, Ia /S) in the SCS-CN model is of- = initial abstraction, F = cumulative infiltration, and
ten set equal to 0.20, according to experimental data S = potential maximum retention or infiltration (mm).
obtained in North America, but this value has been Traditionally, the value of Ia is often set equal to 0.2S
740 B. XIAO et al.

in Eq. 1. Therefore, the SCS-CN model can be ex- area of 7.03 km2 was selected. The watershed (38◦
pressed as follows: 46 –38◦ 51 N latitude, 110◦ 21 –110◦ 23 E longitude,
 1 094.0–1 273.9 m altitude) belongs to the loess hilly-
(P − 0.2S)2 /(P + 0.8S) P ≥ 0.2S
Q= (2) gully region of the Loess Plateau, located in Shenmu
0 P < 0.2S County of Shaanxi Province in northern China. The
For convenience and standardization application of Liudaogou watershed is located at approximately the
Eq. 2, S is expressed in the form of a dimensionless center of the wind-water erosion crisscross region on
runoff curve number (CN) which represents the runoff the Loess Plateau, which suffers the most serious water
potential of the land cover-soil complex characteris- erosion in summer and the most serious wind erosion
tics governed by soil antecedent moisture condition in winter (Cha and Tang, 2000). This region experi-
(AMC), soil type, and land use and treatment (USDA ences intense soil erosion and is the main source for
SCS, 1985): coarse sediment to the Yellow River (Cha and Tang,
2000). The mean annual precipitation is 409 mm, with
S = 25400/CN − 254 (3) a minimum of 110 mm and a maximum of 891 mm, and
Three AMCs were defined as dry (lower limit of rainfall from June to September can account for 80% of
moisture or upper limit of S), moderate (normal or the annual total (Cha and Tang, 2000). According to
average soil moisture condition), and wet (upper limit the water deficit index by Wang and Takahashi (1999)
of moisture or lower limit of S), and denoted as AMC I, and the relative dryness index, the region can be classi-
AMC II, and AMC III, respectively (Mishra and Singh, fied as semiarid (Kimura et al., 2006). Drought followed
2003). The CN value of AMC II (CNII ) was provided by concentrated rainfalls, sparse vegetation, loose soils,
by the SCS-CN manual and the CN value of AMC I a complex landform, and long-term improper use of
(CNI ) and CN value of AMC III (CNIII ) can be calcu- land (overgrazing, excessive deforestation, and culti-
lated by applying the following equations (USDA SCS, vation on steep sloping land) since the 17th century
1985): together have resulted in serious soil erosion and wa-
ter shortage consequent to water loss from runoff (Cha
CNI = 4.2CNII /(10 − 0.058CNII ) (4) and Tang, 2000).
The long-term serious soil loss has developed deep
CNIII = 23CNII /(10 + 0.13CNII ) (5) (up to 20 m) and dense (3–7 km km−2 ) gully networks
and has led to a fragmented landform digital eleva-
MATERIALS AND METHODS
tion model (DEM) and slope gradients (Fig. 1). Most
Study area importantly, the soils in the watershed are erosion-
vulnerable loess with areas of coarse sand (soil type
A representative watershed, Liudaogou, with an map in Fig. 2). There are three types of soils in the

Fig. 1 Digital elevation model (DEM) data and slope gradients of the Liudaogou watershed.
SCS-CN MODEL AND RUNOFF ESTIMATION 741

Fig. 2 Soil types and land use in the Liudaogou watershed.

TABLE I

Hydrologic and related properties of the soils in the Liudaogou watershed


Property Aeolian sandy Chestnut Loessial Rock, river,
soil soil soil or reservoir
Texture Sandy soil Loamy sand soil Sandy loam soil –
Saturated water content (mL L−1 ) 213.5±5.5 261.4±12.9 317.8±16.5 –
Water content at field capacity (mL L−1 ) 125.5±4.4 150.3±7.1 231.7±12.6 –
Water content at wilting point (mL L−1 ) 8.4 14.7 25.9 –
Saturated hydraulic conductivitya) (cm h−1 ) 71.34±7.19 19.40±2.10 10.52±0.25 –
Bulk density (g cm−3 ) 1.71±0.01 1.63±0.13 1.54±0.02 –
Soil hydraulic groupb) A A A D
Area (km2 ) 0.88 1.10 4.61 0.44
Area percentage (%) 12.5 15.6 65.6 6.3
a)
Measured by a tension infiltrometer.
b)
Soil hydraulic groups are mainly determined by infiltration rate: A > 7.26 mm h−1 , 3.81 mm h−1 < B < 7.26 mm h−1 ,
1.27 mm h−1 < C < 3.81 mm h−1 , and D < 1.27 mm h−1 .

Liudaogou watershed, aeolian sandy soil, chestnut soil, and land use map came from the Institute of Soil
and loessial soil (Fig. 2). The hydrologic and related and Water Conservation, Chinese Academy of Sciences
properties of these three soils are presented in Table I. (Liu, 1993). The topographic, soil, and land use maps
The fragmented landform and variations in soil type were digitized in the ArcGIS Desktop 9.3 and veri-
result in high spatial heterogeneity in soil texture and fied/corrected by field investigation using a Trimble
depth, bulk density, moisture, saturated hydraulic con- 5700 GPS system (Trimble Navigation Limited, USA)
ductivity, infiltrability, nutrients, and land use (Fig. 2) in 2004. The verified topographic map was then im-
(Jiang et al., 2005; Zhu and Shao, 2008; Liu and Shao, ported into ArcGIS, and the DEM map was generated
2009; Wang et al., 2009). from a contour line by the ArcGIS extension 3D Ana-
lyst for creating a TIN model from features and then
Data sources and processing
converted TIN to rasters. The slope map was com-
The standard 1:10 000 topographic map (contour pleted from surface analysis of the DEM using the 3D
line interval of 5 m) of the Liudaogou watershed, Analyst tools. The DEM and slope maps of the wa-
which was created in 1985, was obtained from the tershed are presented in Fig. 1. The corrected soil map
Shaanxi Bureau of Surveying and Mapping. The soil and land use map were reclassified according to the
742 B. XIAO et al.

manual of the SCS-CN model (USDA SCS, 1985), and RESULTS


the results are displayed in Fig. 2.
Determination of CN values
The ArcCN-Runoff tool, an extension of the Ar-
cGIS software for generating curve number and runoff The runoff generation mainly relied on the CN va-
maps (Zhan and Huang, 2004), was added to the Ar- lues, which are a function of AMC, slope, soil type, and
cGIS desktop software and activated. Then the soil land use. The CN values reflected the runoff potential.
map and land use map were imported, and a new Low CN values mean that the surface has a high poten-
land-soil file was generated. Finally, the CN value was tial to retain water, whereas high values indicate that
obtained, and the runoff depth was calculated in the the rainfall could only be stored to a limited extent.
ArcCN-Runoff tool. The total runoff volume of the Li- From Table I, we can see that the soil types changed
udaogou watershed after any rainfall event was com- from aeolian sandy soil to loessial soil, but the soils all
puted by multiplying the runoff depth by the area. The belonged to the same soil hydraulic group, A (infiltra-
detailed, stepwise operation of the runoff calculation tion rate > 7.26 mm h−1 ), according to the SCS-CN
by the ArcCN-Runoff tool in ArcGIS can be found in manual. This indicates that the SCS-CN model is a re-
Zhan and Huang (2004). latively coarse model in which the soil type is not clas-
During the period of 2004–2005, natural rainfall sified in detail. The bare rock land and river/lake land
was measured by an auto-recording rain gauge, and in the Liudaogou watershed were attributed to the soil
the actual runoff at the outlet of the watershed was hydraulic group D (infiltration rate < 1.27 mm h−1 ).
monitored by means of a V-shaped weir. The land use in the watershed was sorted into 14
types; these land use types and their corresponding CN
Model validation and runoff predication
values from the ArcCN-Runoff tool are listed in Ta-
The measured runoff was compared with the value ble II. There were three types of land use, rangeland
calculated by the SCS-CN model. Subsequently, the (42%), cropland (31%), and brushland (26%) (Table
applicability of the model to the Liudaogou watershed II). In the SCS-CN model, the CN value was influenced
was evaluated by testing the relative error (RE) and by the AMC, which was derived from rainfall measure-
Nash-Sutcliffe efficiency (NSE) (Nash and Sutcliffe, ments from the preceding five days. The CN value for
1970), both of which have been widely used in evalua- AMC II conditions can be converted into CN values for
tion of model performance (Moriasi et al., 2007). AMC I and AMC III conditions by using Eqs. 4 and 5.
However, in this study, only the CN value for AMC
RE = (Qcal obs obs
i − Qi )/Qi × 100% (6) II was used in the estimation of runoff; the depth of
n
 
n soil in this region can be up to 100 m and the infiltra-
NSE = 1 − (Qobs
i − Qcal
i )
2
(Qcal obs 2
i − Qmean ) (7) tion rate can be up to 71.34 cm h−1 . Thus, both the
i=1 i=1
AMC I and AMC III conditions did not appear. Fi-
where Qobs
i is the ith observation of runoff, Qcali is the nally, the CN value for the Liudaogou watershed was
obs
ith calculated runoff by the SCS-CN model, Qmean is obtained, and runoff depth was calculated using the
the mean value of observed runoff, and n is the total ArcCN-Runoff tool (Fig. 3).
number of observations. NSE ranges between −∞ and
SCS-CN model verification
1, with NSE = 1 being the optimal value. Values of
NSE between 0 and 1 are generally viewed as accept- During the period of 2004–2005, approximately 10
able levels of performance, whereas a value of NSE natural rainfall events were recorded. The rainfall for
< 0 indicates that the mean observed value is a bet- each event was between 5.3 and 50.1 mm, and the rain-
ter predictor than the simulated value, which indicates fall intensity was between 2.5 and 24.4 mm h−1 . The
unacceptable performance (Moriasi et al., 2007). runoff measured and that estimated by the SCS-CN
According to the results, the value of Ia /S, which model are presented in Table III. The relative error in
was traditionally set equal to 0.2 in Eq. 2, was modified Table III showed that the runoff was generally (8 out
until the model yielded the best performance as evalu- of 10) overestimated, being on the average 20% higher
ated by relative error and NSE in Eqs. 6 and 7. Finally, than the actual runoff.
this value of the initial abstraction ratio was assumed The NSE value calculated by Eq. 7 was 0.90 and
for the runoff estimation. the relative error ranged from 2.0% to 77.7%, with an
SCS-CN MODEL AND RUNOFF ESTIMATION 743

TABLE II

Land use and curve number (CN) in the Liudaogou watershed

Land use Hydraulic Area Area CNb)


conditiona) percen-
tage A (AMC II) D (AMC II) A (AMC I) D (AMC I)
km2 %
Row cropland (contoured, terraced) Good 0.34 4.84 62 81 41 64
Small grain land (contoured, terraced) Good 0.36 5.10 59 81 38 64
Close-seeded cropland (contoured) Good 1.48 21.00 55 83 34 67
Rangeland (herbaceous) Poor 2.03 28.94 30 93 15 85
Rangeland (herbaceous) Fair 0.39 5.59 30 89 15 77
Rangeland (herbaceous) Good 0.42 6.02 30 85 15 70
Rangeland (sagebrush) Fair 0.07 1.00 30 70 15 49
Brushland Poor 0.07 0.96 48 83 28 67
Brushland Fair 1.50 21.36 35 77 18 58
Deciduous forest 0.25 3.52 36 79 19 61
Orchard land 0.01 0.17 45 83 26 67
Residential land 0.00 0.07 61 86 40 72
Open space land Good 0.03 0.48 39 80 21 63
Bare rock land 0.07 0.99 100 100 100 100
a)
According to the manual of the SCS-CN model, the hydraulic condition is based on a combination of factors that affect
infiltration and runoff, including density and canopy of vegetative areas, amount of year-round cover, amount of grass or
close-seeded legumes, percentage of residue cover on the land surface, and degree of surface roughness. The good hydraulic
condition means that these factors encourage average and better than average infiltration and tend to decrease runoff; the
poor hydraulic condition means that these factors impair infiltration and tend to increase runoff; and the fair hydraulic
condition is in-between the former two.
b)
A and D are the soil hydraulic group and AMC is the soil antecedent moisture condition, dry (I) and moderate (II),
according to the manual of the SCS-CN model.

Fig. 3 Curve number (CN) values and the corresponding runoff in the case of 100 mm precipitation in the Liudaogou
watershed.

average of 27.0% (Table III), indicating that the perfor- Modification of Ia /S value in the SCS-CN model
mance of the standard SCS-CN model was acceptable
according to the explanation of NSE by Moriasi et al. Because the history and documentation are ob-
(2007). scure, the assumption that Ia = 0.2S has been freque-
744 B. XIAO et al.

TABLE III

Comparison of measured runoff and runoff estimated by the SCS-CN model in the Liudaogou watershed
Date Rainfall Rainfall intensity Measured Estimated Relative error
runoff runoff
Maximum Average
−1
mm mm h m3 %
Aug. 2, 2004 5.3 24.4 4.0 326.5 370.5 13.5
Aug. 10, 2004 7.3 4.6 1.5 369.2 511.2 38.5
Aug. 11, 2004 10.2 18.3 6.8 1 105.9 707.6 −36.0
Aug. 14, 2004 14.1 6.1 1.3 567.5 982.1 73.0
Aug. 15, 2004 15.2 4.6 1.3 625.0 1 055.9 68.9
Sep. 29, 2004 6.9 2.5 1.6 468.6 477.8 2.0
May 29, 2005 7.1 2.5 1.8 279.1 495.9 77.7
Jul. 19, 2005 50.1 10.7 1.3 4 139.2 5 296.4 28.0
Aug. 23, 2005 16.0 11.2 3.2 1 276.2 1 114.4 −12.7
Sep. 15, 2005 9.9 2.8 0.8 590.8 688.8 16.6

ntly questioned for its validity and applicability, in- of Ia /S for moderate rains. Because the watershed is
voking a critical examination of the Ia -S relationship located at the center of the Loess Plateau, where the
for pragmatic applications. Mishra et al. (2006) em- annual precipitation is 409 mm and a single rainfall
ployed a large dataset from 84 small watersheds (0.17– event exceeding 50 mm is very rare, we can thus in-
71.99 ha in area) in the USA to investigate a number of fer that the modification of the value of Ia /S will not
Ia -S relationships that incorporated antecedent mois- significantly increase the performance of the SCS-CN
ture as a function of antecedent precipitation. Jain et model. Therefore, the default value (0.20) of Ia /S was
al. (2006) reviewed the Ia -S relationship and proposed
a new non-linear relationship that incorporated storm
rainfall and S. Ponce and Hawkins (1996) suggested
that the fixed initial abstraction ratio 0.20 may not be
the most appropriate number and that it should be in-
terpreted as a regional parameter.
In our case (Fig. 4a), we found that the relative
error rapidly decreased with increasing Ia /S when it
was less than 0.15, while it was almost stable when
the Ia /S value was between 0.15 and 0.30. However,
the NSE value of the model increased initially with in-
creasing Ia /S and then decreased gradually when the
Ia /S value exceeded 0.20 (Fig. 4b). Thus, we can con-
clude that the SCS-CN model achieved the best pre-
cision (RE = 25.1%, NSE = 0.95) in the estimation
of runoff when Ia /S was 0.22, but the value 0.20 was
also acceptable (RE = 27.0%, NSE = 0.90). However,
it should be noted that the variation in model perfor-
mance presented in Fig. 4 was mainly affected by the
special rainfall event on July 19, 2005 of 50.1 mm rain-
fall. If this rainfall event was removed from the model
evaluation, the relative error and NSE of the SCS-CN
model were maintained at 26.8% and 0.50, respectively,
Fig. 4 Relationships between the initial abstraction ratio
with the Ia /S values from 0.10 to 0.30 (Fig. 5). This (Ia /S) and the performance of the SCS-CN model in the
means that the performance of the SCS-CN model in Liudaogou watershed, evaluated by the relative error (RE)
the Liudaogou watershed was not sensitive to the value (a) and the Nash-Sutcliffe efficiency (NSE) (b).
SCS-CN MODEL AND RUNOFF ESTIMATION 745

that produced the highest precision was 0.22. There-


fore, the SCS-CN model with this modified Ia /S value
was used to predict the runoff and the results are pre-
sented in Fig. 7. The runoff predicted increased gradu-
ally with increasing rainfall when the rainfall was less

Fig. 5 Relationships between the initial abstraction ratio


(Ia /S) and the performance of the SCS-CN model in the
Liudaogou watershed, evaluated by the relative error (RE)
(a) and the Nash-Sutcliffe efficiency (NSE) (b) when the
heavy rainfall of July 19, 2005 (50.1 mm) was removed from
the evaluation.

applicable. However, 0.22 was recommended under Fig. 6 Relationships between curve number (CN) values
most conditions. and estimated runoff in the case of 50 mm rainfall and an
initial abstraction ratio of 0.20 by the SCS-CN model (a)
Sensitivity analysis of the SCS-CN model and the relationships between initial abstraction ratio
(Ia /S) values and estimated runoff in the case of 50 mm
From the description of the theory of the SCS-CN rainfall and a CN value of 80 (b) in the Liudaogou water-
model, it can be appreciated that the model only re- shed.
quires two input factors, CN and Ia /S. We evaluated
the model sensitivity to these two factors, and the re-
sults are displayed in Fig. 6. At Ia /S = 0.20, the es-
timated runoff was zero when the CN values were less
than 50 and then rapidly increased with increasing CN
values (Fig. 6a). Conversely, the estimated runoff de-
creased almost linearly with Ia /S increasing from 0
to 0.50 (CN = 80) (Fig. 6b). These comparative data
confirmed that the SCS-CN model was much more sen-
sitive to changes in CN values and less sensitive to
Ia /S values, as was already pointed out by Ponce and
Hawkins (1996) and Wang and Huang (2008a).

Runoff predication in the Liudaogou watershed


Fig. 7 Relationships between rainfall and runoff predicted
The results of the SCS-CN model application to by the SCS-CN model with a modified Ia /S value (0.22)
the Liudaogou watershed indicated that the Ia /S value in the Liudaogou watershed.
746 B. XIAO et al.

than 50 mm, while it increased rapidly when the the Ia /S value only increased the model performance
rainfall exceeded 50 mm. This indicated that a large to a limited extent. However, some studies also re-
amount of runoff causing serious soil erosion can be ported that modifying the Ia /S value significantly im-
expected when a single rainfall event exceeded 50 mm. proved the performance of the model. Wang and Huang
(2008b) documented that the optimized Ia /S value
DISCUSSION should be 0.01 for the Xingdiangou watershed in Suide
County of Shaanxi Province, and Shi et al. (2009)
The SCS-CN model is a lumped model used to noted that the Ia /S value should be 0.053 in the Three
simulate rainfall-runoff; i.e., there is an input value Gorges area of China because this modification im-
of rainfall and an output value of runoff. The model proved the model NSE to 0.768, compared to 0.482 for
does not take into account the time, slope gradient, the traditional Ia /S value. He et al. (2001) reported
spatial and temporal variation of infiltration, or the that the best Ia /S value is 0.05 when the model is
cumulative processes of runoff during the calculation. applied to the suburbs of Shanghai. From these refe-
Additionally, the division of soils into hydraulic groups rences, we can see that the standard value for Ia /S
is very coarse and the definition of antecedent moisture (0.20) and the lower (down to 0.01) and higher values
condition is not quantitative. These disadvantages of (up to 0.24) of Ia /S have all been recommended for use
the SCS-CN model exclude its application for some in China and even in the Loess Plateau. At present, it
process-based purposes, but guarantee the model sim- is difficult to determine a universal Ia /S value for re-
plicity and stability in the estimation of runoff (Ponce gional or national scales in China. Therefore, the stan-
and Hawkins, 1996). In this study, considerable vari- dard Ia /S value is recommended whenever no specific
ations in the soil types in the Liudaogou watershed research has been conducted.
were neglected although the differences in infiltrabili- Some researchers tried to establish new CN va-
ty among soil types can be remarkable (Table I). Ad- lues for each land use type in China because they
ditionally, the fragmented landform of the watershed believed that the CN values from the United States
was not taken into account during the calculation pro- could not be exported to other countries. Zhang et
cedure. Although spatial heterogeneity was not consi- al. (2003) established a new approach to get CN va-
dered, the relative error and NSE values for the runoff lues from a particular rainfall event and for the rainfall
estimation, 27% and 0.90, respectively, confirmed the over the previous five days using a statistical method
applicability of the model to this watershed. in the Anjiagou watershed of Gansu Province. They
The most appropriate value for Ia /S was 0.22 for claimed that they obtained better performance of the
this particular watershed. However, if the single heavy SCS-CN model with these new CN values. Wang and
rainfall event recorded was removed from the evalua- Huang (2008b) optimized the CN value for millet (Pan-
tion, the model performance did not appear to be sen- icum miliaceum L.), alfalfa (Medicago sativa L.), and
sitive to the Ia /S value. Similar results were obtained sorghum (Sorghum bicolor (Linn.) Moench) using the
in other studies. The SCS-CN model, using standard inverse method (using measured runoff to calculate CN
Ia /S and CN values, performed well with a predic- values). Similar work has also been conducted in other
tion precision in excess of 75% and relative error be- watersheds (Jin et al., 2003; Zhang et al., 2004; Gao
tween 6.7% and 23.3% when applied to the Wang- et al., 2006), but it seems that the results are not con-
donggou watershed in Changwu County of Shaanxi sistent. The results of these studies indicate that the
Province (Liu et al., 2005), the Wangjiagou water- standard CN value in the SCS-CN model is an overes-
shed in Lishi County of Shanxi Province (Zhang et al., timate for some land use types and an underestimate
2008), and the Wenyuhe watershed in Beijing (Zhou for the others.
et al., 2008). Additionally, Wang (2005) pointed out Some researchers tried to add further parameters to
that the most appropriate Ia /S value is 0.204 for the the SCS-CN model to reflect the effects of some factors
Mengjiagou watershed in Shenmu County of Shaanxi (e.g., slope gradient, rainfall intensity, and soil mois-
Province. Wang and Huang (2008a) concluded that ture conditions) in order to increase its precision in
the optimized Ia /S value was 0.24 for the Feijiamao specific areas. For example, Wang and Huang (2008a)
watershed in Suide County and the Yangdaogou wa- modified the SCS-CN model with an additional rainfall
tershed in Lishi County, both of which are in Shanxi intensity factor for the Loess Plateau, and the model
Province. They concluded that the modification of performance evaluated by NSE was significantly incre-
SCS-CN MODEL AND RUNOFF ESTIMATION 747

ased from 0.351 to 0.812. In addition, Sahu et al. Beasley, D. B., Huggins, L. F. and Monke, E. J. 1980.
(2010b) recently incorporated a continuous function for ANSWERS: a model for watershed planning. Tran.
antecedent soil moisture to avoid unreasonable sudden ASABE. 23(4): 938–944.
jumps in curve number, which resulted in correspon- Cha, X. and Tang, K. L. 2000. Study on comprehensive
control model of small watershed eco-environment in
ding jumps in the estimated runoff. These attempts
water and wind crisscrossed erosion zone. J. Nat. Res.
may increase the model performance but also result in
(in Chinese). 15(1): 97–100.
a simultaneous loss in the model simplicity and stabi- Chandrmohan, T. and Durbude, D. G. 2001. Estimation of
lity. runoff using small watershed models. Hydrol. J. 24(2):
45–53.
CONCLUSIONS Fu, B. J., Wang, J., Chen, L. D. and Qiu, Y. 2003.
The effects of land use on soil moisture variation in
The applicability of the SCS-CN model to a small the Danangou catchment of the Loess Plateau, China.
watershed on the Loess Plateau of China with high Catena. 54(1–2): 197–213.
spatial heterogeneity was evaluated in this study. The Gao, Y., Zhu, B., Miao, C. Y. and Zhang, J. Z. 2006. Appli-
most appropriate Ia /S value was quantified by the in- cation of SCS model to estimate the volume of rainfall
verse method. The standard model was applicable to runoff in sloping field of purple soil. Chin. Agr. Sci.
the estimation of runoff in the Liudaogou watershed Bull. (in Chinese). 22(11): 396–400.
and the model performance was acceptable according Geetha, K., Mishra, S. K., Eldho, T. I., Rastogi, A. K. and
Pandey, R. P. 2007. Modifications to SCS-CN method
to relative error (2.0%–77.7%, averaging 27.0%) and
for long-term hydrologic simulation. J. Irrig. Drain.
NSE (0.90) values. The most appropriate Ia /S value
Eng. 133(5): 475–486.
for the Liudaogou watershed was equal to 0.22 because Greene, R. G. and Cruise, J. F. 1995. Urban watershed
with this modified Ia /S value, the model performance modeling using geographic information system. J. Wa-
increased slightly, as evaluated by the relative error ter Resour. Plann. Manage. 121(4): 318–325.
(averaging 25.1%) and NSE (0.95) values. The model He, B. G., Zhou, N. S., Gao, X. J., Xu, S. Y. and Yuan,
performance was not sensitive to the Ia /S value when Y. K. 2001. Precipitation-runoff relationship in farm-
a heavy rainfall (50.1 mm) event was omitted from the land nonpoint source pollution research: amending co-
dataset, which implied that the SCS-CN model with effcient of SCS hydrologic method. Res. Environ. Sci.
(in Chinese). 14(3): 49–51.
a standard Ia /S value could be recommended for use
He, C. 2003. Integration of geographic information systems
in the watershed studied because single rainfall events
and simulation model for watershed management. En-
exceeding 50 mm were very scarce in this region. The viron. Model. Softw. 18(8–9): 809–813.
runoff in the Liudaogou watershed predicted by the Hellweger, F. L. and Maidment, D. R. 1999. Definition
SCS-CN model with the modified Ia /S value increased and connection of hydrologic elements using geographic
gradually with increasing rainfall when the rainfall was data. J. Hydrol. Eng. 4: 10–18.
less than 50 mm and increased rapidly when the rain- Huang, M. B., Gallichand, J., Dong, C. Y., Wang, Z. L.
fall exceeded 50 mm. These findings may be helpful in and Shao, M. A. 2007. Use of soil moisture data and
solving the problem of serious soil and water loss on curve number method for estimating runoff in the Loess
Plateau of China. Hydrol. Proc. 21(11): 1471–1481.
the Loess Plateau of China.
Jain, M. K., Mishra, S. K., Suresh, B. P. and Venugopal,
K. 2006. On the Ia -S relation of the SCS-CN method.
ACKNOWLEDGEMENT
Nord. Hydrol. 37(3): 261–275.
The authors are grateful to Dr. Wen Bin, Institute Jiang, N., Shao, M. A., Lei, T. W. and Zhang, X. C. 2005.
Spatial variability of soil infiltration properties on na-
of Geographical Sciences and Natural Resources, Chi-
tural slope in Liudaogou catchment on Loess Plateau.
nese Academy of Sciences, for providing runoff data of
J. Soil Water Conserv. (in Chinese). 19(1): 14–17.
the Liudaogou watershed. Jin, H., Sun, X. H. and Li, Y. B. 2003. Application of SCS
model in Lanhe watersheds. J. Taiyuan Univ. Technol.
REFERENCES (in Chinese). 34(6): 735–736, 752.
Johnson, R. R. 1998. An investigation of curve number
Arnold, J. G., Williams, J. R., Srinivasan, R., King, K. W. applicability to watersheds in excess of 25 000 hectares
and Griggs, R. H. 1995. SWAT: Soil Water Assessment (250 km2 ). J. Environ. Hydrol. 6: 1–10.
Tool. Texas A&M University, Texas Agricultural Ex- Kimura, R., Fan, J., Zhang, X. C., Takayama, N.,
perimental Station, Blackland Research Center, Tem- Kamichika, M. and Matsuoka, N. 2006. Evapotranspi-
ple. ration over the grassland field in the Liudaogou Basin
748 B. XIAO et al.

of the Loess Plateau, China. Acta Oecol. 29(1): 45–53. Nash, J. E. and Sutcliffe, J. V. 1970. River flow forecas-
Knisel, W. G. 1980. CREAMS: A Field Scale Model ting through conceptual models. Part I. A discussion
for Chemicals, Runoff, and Erosion from Agricul- of principles. J. Hydrol. 10(3): 282–290.
tural Management Systems. United States Department Pan, J. J. 1996. Study on hydrological property and runoff
of Agriculture Conservation Research Report No. 26. of catchments by remote sensing. J. Soil Erosion Soil
United States Department of Agriculture, Washington Water Conserv. (in Chinese). 2(3): 88–92.
D.C. Ponce, V. M. 1989. Engineering Hydrology—Principles and
Kothyari, U. C. and Jain, S. K. 1997. Sediment yield esti- Practices. San Diego State University, San Diego.
mation using GIS. Hydrol. Sci. J. 42(6): 833–843. Ponce, V. M. and Hawkins, R. H. 1996. Runoff curve num-
Lewis, D., Singer, M. J. and Tate, K. W. 2000. Applicabi- ber: has it reached maturity? J. Hydrol. Eng. 1(1):
lity of SCS curve number method for a California oak 11–19.
woodlands watershed. J. Soil Water Conserv. 55(2): Qiu, Y., Fu, B. J., Wang, J. and Chen, L. D. 2001. Spatial
226–230. variability of soil moisture content and its relation to
Li, Y. Y. and Shao, M. A. 2006. Change of soil physical environmental indices in a semi-arid gully catchment
properties under long-term natural vegetation restora- of the Loess Plateau, China. J. Arid Environ. 49(4):
tion in the Loess Plateau of China. J. Arid. Environ. 723–750.
64(1): 77–96. Reshmidevi, T. V., Jana, R. and Eldho, T. I. 2008.
Lian, G., Guo, X. D., Fu, B. J. and Hu, C. X. 2006. Spa- Geospatial estimation of soil moisture in rain-fed paddy
tial variability of bulk density and soil water in a small fields using SCS-CN-based model. Agr. Water Manage.
catchment of the Loess Plateau. Acta Ecol. Sin. (in 95(4): 447–457.
Chinese). 26(3): 647–654. Sahu, R. K., Mishra, S. K. and Eldho, T. I. 2010a. Com-
Liu, C. L. and Shao, M. A. 2009. Spatial variation of sa- parative evaluation of SCS-CN-inspired models in ap-
turated hydraulic conductivity and soil water of the plications to classified datasets. Agr. Water Manage.
surface layer of a slope on the Loess Plateau. Sci. Soil 97(5): 749–756.
Water Conserv. (in Chinese). 7(1): 13–18. Sahu, R. K., Mishra, S. K. and Eldho, T. I. 2010b. An im-
Liu, X. Z., Kang, S. Z., Liu, L. D. and Zhang, X. P. 2005. proved AMC-coupled runoff curve number model. Hy-
SCS model based on geographic information and its drol. Proc. 24(20): 2834–2839.
application to simulate rainfall-runoff relationship at Saito, Y., Yang, Z. and Hori, K. 2001. The Huanghe (Yel-
typical small watershed level in Loess Plateau. Trans. low River) and Changjiang (Yangtze River) deltas: a
CSAE (in Chinese). 21(5): 93–97. review on their characteristics, evolution and sediment
Liu, X. Z. and Li, J. Z. 2008. Application of SCS model discharge during the Holocene. Geomorphology. 41(2–
in estimation of runoff from small watershed in Loess 3): 219–231.
Plateau of China. Chin. Geogra. Sci. 18(3): 235–241. Sharpley, A. N. and Williams, J. R. 1990. EPIC, Ero-
Liu, Y. M. 1993. Special Issue For the Study of Ecolo- sion/Productivity Impact Calculator: 1. Model Docu-
gical and Environmental Remediation and Experimen- mentation. US Department of Agriculture Technical
tal Demonstration on the Wind Water Crisscross Re- Bulletin, Washington, D.C.
gion on the Loess Plateau of China: Memoir of North- Shi, H. and Shao, M. A. 2000. Soil and water loss from the
western Institute of Soil and Water Conservation, Loess Plateau in China. J. Arid Environ. 45(1): 9–20.
Academia Sinica and Ministry of Water Resources (in Shi, Z. H., Chen, L. D., Fang, N. F., Qin, D. F. and Cai, C.
Chinese). Institute of Soil and Water Conservation, F. 2009. Research on the SCS-CN initial abstraction
Chinese Academy of Sciences, Xi’an. ratio using rainfall-runoff event analysis in the Three
Milliman, J. D., Qin, Y. S., Ren, M. E. and Yoshiki, S. Gorges Area, China. Catena. 77(1): 1–7.
1987. Man’s influence on the erosion and transport of Singh, P. K., Bhunya, P. K., Mishra, S. K. and Chaube, U.
sediment by Asian rivers: the Yellow River (Huanghe) C. 2008. A sediment graph model based on SCS-CN
example. J. Geol. 95(6): 751–762. method. J. Hydrol. 349(1–2): 244–255.
Mishra, S. K. and Singh, V. P. 2003. Soil Conservation Ser- Stolte, J., van Venrooij, B., Zhang, G. H., Trouwborst,
vice Curve Number (SCS-CN) Methodology. Springer, K. O., Liu, G. B., Ritsema, C. J. and Hessel, R. 2003.
Dordrecht. Land-use induced spatial heterogeneity of soil hydraulic
Mishra, S. K., Tyagi, J. V., Singh, V. P. and Singh, R. properties on the Loess Plateau in China. Catena.
2006. SCS-CN-based modeling of sediment yield. J. 54(1–2): 59–75.
Hydrol. 324(1–4): 301–322. Tsihrintzis, V. A. and Hamid, R. 1997. Urban stormwa-
Moriasi, D. N., Arnold, J. G., Van Liew, M. W., Bingner, ter quantity/quality modeling using the SCS method
R. L., Harmel, R. D. and Veith, T. L. 2007. Model eva- and empirical equations. J. Am. Water Resour. As-
luation guidelines for systematic quantification of accu- soc. 33(1): 163–176.
racy in watershed simulations. Tran. ASABE. 50(3): USDA-Soil Conservation Service (USDA SCS). 1985. Na-
885–900. tional Engineering Handbook. Section 4. Hydrology.
SCS-CN MODEL AND RUNOFF ESTIMATION 749

USDA-SCS, Washington, D.C. Young, R. A., Onstad, C. A., Bosch, D. D. and Ander-
Wang, B. L. 2005. The revised SCS model in China. Yellow son, W. P. 1989. AGNPS: a nonpoint-source pollution
River (in Chinese). 27(5): 24–26. model for evaluating agricultural watersheds. J. Soil
Wang, Q. and Takahashi, H. 1999. A land surface water Water Conserv. 44(2): 168–173.
deficit model for an arid and semiarid region: impact of Zhan, X. and Huang, M. L. 2004. ArcCN-Runoff: an Ar-
desertification on the water deficit status in the Loess cGIS tool for generating curve number and runoff maps.
Plateau, China. J. Climate. 12: 244–257. Environ. Model. Softw. 19(10): 875–879.
Wang, S. Q., Zhu, S. L. and Zhou, C. H. 2001. Charac- Zhang, M. H., Wang, X. Y. and Qin, F. L. 2004. Applica-
teristics of spatial variability of soil thickness in China. tion of SCS model to estimate the quantity of rainfall
Geogr. Res. (in Chinese). 20(2): 161–169. runoff of small watershed in Shixia, Miyun County.J.
Wang, Y. and Huang, M. B. 2008a. Application of the SCS- Capital Norm. Univ. (Nat. Sci. Ed.) (in Chinese). 25
CN method on runoff estimation in small watershed on (S1): 155–158.
Loess Plateau. Sci. Soil Water Conserv. (in Chinese).
Zhang, X. Y., Meng, F. and Ding, N. 2003. Application
6(6): 87–91.
of SCS model to estimating the quantity of runoff of
Wang, Y. and Huang, M. B. 2008b. Optimizing parameters
small watershed in semi-arid or arid region. Res. Soil
of SCS-CN method for application on the Loess Pla-
Water Conserv. (in Chinese). 10(4): 172–174, 249.
teau. Bull. Soil Water Conserv. (in Chinese). 28(1):
54–58. Zhang, Y. X., Mu, X. M. and Wang, F. 2008. Calibra-
Wang, Y. Q., Zhang, X. C. and Huang, C. Q. 2009. Spatial tion and validation to parameter λ of soil conserva-
variability of soil total nitrogen and soil total phospho- tion service curve number method in hilly region of
rus under different land uses in a small watershed on the Loess Plateau. Agr. Res. Arid Area. (in Chinese).
the Loess Plateau, China. Geoderma. 150(1–2): 141– 26(5): 124– 128.
149. Zhou, C. N., Ren, S. M. and Yan, M. J. 2008. Application
Wu, T. H., Hall, J. A. and Bonta, J. V. 1993. Evaluation of of SCS model to simulate rainfall-runoff relationship in
runoff and erosion models. J. Irrig. Drain. Eng. 119 Wenyu river basin in Beijing. Trans. CSAE (in Chi-
(2): 364–382. nese). 24(3): 87–90.
Xiao, B., Wang, Q. H., Que, X. E. and Cao, Z. D. 2009. Zhu, Y. J. and Shao, M. A. 2008. Variability and pattern
Spatial variation of soil nutrients and bulk density in of surface moisture on a small-scale hillslope in Liudao-
rehabilitated slope land on northeast of Loess Plateau. gou catchment on the northern Loess Plateau of China.
J. Soil Water Conserv. (in Chinese). 23(3): 92–96. Geoderma. 147(3–4): 185–191.

You might also like