CN-China: Revised runoff curve number by using rainfall-runoff events data in China

Water Research 177 (2020) 115767

Contents lists available at ScienceDirect

Water Research journal homepage: www.elsevier.com/locate/watres

CN-China: Revised runoff curve number by using rainfall-runoff events data in China Huishu Lian a, b, Haw Yen c, Jr-Chuan Huang d, Qingyu Feng e, Lihuan Qin a, Muhammad Amjad Bashir a, Shuxia Wu a, A-Xing Zhu f, g, Jiafa Luo h, Hongjie Di i, Qiuliang Lei a, *, Hongbin Liu a, ** a

Key Laboratory of Nonpoint Source Pollution Control, Ministry of Agriculture, Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing, 10081, China School of Environmental Science and Engineering, Southern University of Science and Technology, Shenzhen, Guangdong, 518055, China c Blackland Research and Extension Center, Texas A&M Agrilife Research, Texas A&M University, Temple, TX, 76502, USA d Department of Geography, National Taiwan University, Taipei, 10617, Taiwan e Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing, 100085, China f Key Laboratory of Virtual Geographic Environment, Nanjing Normal University, Ministry of Education, Nanjing, 210023, China g Department of Geography, University of Wisconsin-Madison, Madison, WI, 53706, USA h AgResearch Limited, Ruakura Research Centre, Hamilton, 3240, New Zealand i Centre for Soil and Environmental Research, Lincoln University, Lincoln, Christchurch, 7647, New Zealand b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 22 September 2019 Received in revised form 2 March 2020 Accepted 28 March 2020 Available online 2 April 2020

The curve number (CN) method developed by the United States Department of Agriculture (USDA) in 1954 is the most common adopted method to estimate surface runoff. For years, applicability of the CN method is a conundrum when implementing to other countries. Specifically, countries with more complex natural environment may require more dedicated adjustments. Therefore, the current CN lookup table provided by USDA might not be appropriate and could be questionable to be applied directly to regions elsewhere. Some studies have been conducted to modify CN values according to specified natural characteristics in scattered regions of mainland China. However, an integral and representative work is still not available to address potential concerns in general matters. In this study, a large set of rainfallrunoff monitoring data were collected to adjust CN values in 55 study sites across China. The results showed that the revised CN values are largely different from CN look-up table provided by USDA, which would lead to huge errors in runoff estimation. In this study, the revised CN (dubbed CN-China) provides better reference guidelines that are suitable for most natural conditions in China. In addition, scientists and engineers from other parts of the world can take advantage of the proposed work to enhance the quality of future programs related to surface runoff estimation. © 2020 Elsevier Ltd. All rights reserved.

Keywords: Hydrology Curve number Surface runoff estimation Modeling

1. Introduction The high-density construction of meteorological stations around the world makes it easy to acquire historical and real-time rainfall data, but there are few stations for runoff monitoring. Runoff monitoring is often carried out for short-term and limited area in order to accomplish specific research projects (Fauvel et al., 2016; Gwynne and Glover, 1961; Hvitvedjacobsen and Yousef, 1988; Kim and Sansalone,

* Corresponding author. ** Corresponding author. E-mail addresses: [email protected] (Q. Lei), [email protected] (H. Liu). https://doi.org/10.1016/j.watres.2020.115767 0043-1354/© 2020 Elsevier Ltd. All rights reserved.

2008). Surface runoff estimation serves as the most important role in hydrology related research (Hawkins,1993; Jiang et al., 2012; Kim and Sansalone, 2008; Kirchner et al., 2000; Muche et al., 2019; Steenhuis et al.,1995; Tyagi et al., 2008; Wang and Wang, 2018). Fortunately, the relationship between runoff and rainfall allows us to use mathematical methods to estimate runoff based on readily available rainfall data (Clyde and Work, 1943; Guo et al., 2017). The Green-Ampt infiltration curve (Freyberg et al., 1980; Li et al., 2015; Stewart, 2018), the Philip infiltration curve (Triadis and Broadbridge, 2012), and the Horton infiltration curve methods are all for calculating runoff (Esen, 1987; Grimaldi et al., 2013). Applications of these methods are very limited because they require many parameters and detailed soil

2

H. Lian et al. / Water Research 177 (2020) 115767

attribute data that are difficult to acquire. In 1949, Mockus proposed a framework to predict runoff based on land use, soil, antecedent precipitation, storm duration, and annual mean temperature based on mechanism of storage-excess flow (Mockus, 1949). The USDA Soil Conservation Service (SCS) now the Natural Resources Conservation Service (NRCS) proposed CN method in 1954 according to this framework (Boughton, 1989; Hawkins et al., 2019), which was published in the National Engineering Handbook (USDA,1954) in order to meet the design needs of the Watershed Protection and Flood Prevention Act (Public Law 83-566) (Hawkins et al., 2019). The CN method has been widely used since then. The conceptual and empirical CN method is simple with only one parameter CN, making it the most widely used method for calculating runoff based on rainfall events (D’Asaro et al., 2014; Garen and Moore, 2005; Gaume et al., 2004; Lal et al., 2019; Wilson et al., 2017). Besides, the CN method has been widely adopted in many hydrological models (Hawkins et al., 2019; Soulis et al., 2017). The CN method will be performed well if model users obtain CN values from measured rainfall-runoff data when available (Lal et al., 2017). CN parameters reflect the effect of soil infiltration capacity (Ponce and Hawkins, 1996; Zoure et al., 2019). Key impact factors of CN include soil properties, land use type, slope, antecedent moisture conditions, vegetation coverage, land management practices (Boughton, 1989, Chin, 2017, Hawkins, 1978a, b, Lal et al., 2019, Shi et al., 2017, Zoure et al., 2019). CN values have been derived experimentally from rainfall-runoff events measurement over a wide range of geographic, soil, and land management conditions. Based on the monitoring data of 150 watersheds in the U.S., scientists organized a CN look-up table (USDA, 1954), which is widely used around the world. Subsequently, Hawkins found that the runoff calculated by this method was very sensitive to CN values even more than rainfall depth (Hawkins, 1975). Previous study has shown that a variation of ±10% of CN value leads to a variation from
45% to þ55% runoff (Boughton, 1989). Therefore, the accuracy of the CN value plays a crucial role in the estimation of runoff. Various applications and modifications of CN method have been implemented in China, including finding the local initial abstraction (Ia) and CN values based on rainfall-runoff monitoring data (Fu et al., 2011; Jiao et al., 2015; Shi et al., 2009), the effects of different soil, land use, slope on CN values (Dong et al., 2015; Huang et al., 2007; Huang et al., 2006), relationship between CN value and the rainfall depth under long-term sequence and using remote sensing to study the spatial heterogeneity of CN values (Chen et al., 2017; Zhang et al., 2019). There is a common conundrum that scientists could not find the corresponding CN values from CN look-up table. The primary reasons are the complexity and interactions among land use type, soil infiltration capacity, and climate characteristics which could result in an enormous error between CN from look-up table and reality. For instance, nearly 70% of the rainfall in the U.S. will seep into the soil, while farmlands are fragmented and densely drained in China. Infiltration rate is low due to the dense surface irrigation and drainage system (He et al., 2001). In addition, topographic variations in China are substantially more complicated than in the U.S. Therefore, parameters are being over-calibrated in order to achieve better simulation performance, especially when users adopt CN values from look-up table to hydrological models. The embedded uncertainty with potentially large errors in surface runoff estimation is going to affect the corresponding scientific credibility in the following analytical processes such as the simulation of nitrogen and phosphorus in watershed scale (Yen et al., 2014). Thus, a national-scale research is urgently needed to alleviate the tremendous differences in CN value and the corresponding predominant controlling factors. The primary goal of this study was to adjust CN values based on the monitored rainfall-runoff data in China. In order to provide

revised CN values that can be better representative to reflect actual hydrological conditions in China, three objectives were defined to: (i) calculate CN values using the monitored rainfall-runoff data under different climate, soil, land use, and slope conditions across mainland China (contiguous China); (ii) compare CN values obtained from CN look-up table with the calculated CN values under corresponding field conditions (soil, land use and slope); and, (iii) identify the key factors differentiating CN values under Chinese field conditions from those obtained by NRCS. 2. Materials and methods 2.1. Overview of SCS-CN method The CN method was developed by USDA in 1950s originally intended for estimating the depth of surface runoff in an ungauged small catchment and later implemented also to non-agricultural watersheds with other extended applications (Boughton, 1989; Hawkins et al., 2019; Ponce and Hawkins, 1996). The simple structure approach relies on the parameter CN, a lumped expression of a watershed potential hydrologic response. The CN method is based on a water balance function (Eq. (1)) and two fundamental hypotheses. Eq. (1) shows the rainfall depth equals to the summery of Ia, cumulative infiltration (F), and direct runoff (Q). The first hypothesis is that the ratio of direct runoff to the maximum potential runoff is equal to the ratio of the amount of actual infiltration to the potential maximum retention (Eq. (2)). The second hypothesis states that the amount of initial abstraction is some fraction of the potential maximum retention (Eq. (3)) (Mishra and Singh, 1999). The method has interpreted empirical relationships between initial abstraction (Ia) and potential maximum retention or infiltration (S) through extensive rainfall-runoff experimental data: Ia ¼ 0.2S. Functions for S (Eq. (4)) and CN (Eq. (5)) are as follow:

P ¼ Ia þ F þ Q

(1)

Q F ¼ P
Ia S

(2)

Ia ¼ lS

(3)


 1=2  S ¼ 5 P þ 2Q
4Q 2 þ 5PQ

(4)

CN ¼ 25400=ðS þ 254Þ

(5)

where, P ¼ rainfall depth (mm); Ia ¼ initial abstraction of the rainfall (mm); F ¼ cumulative infiltration excluding Ia (mm); Q ¼ runoff depth (mm); S ¼ potential maximum retention or infiltration (mm); and, l is the initial abstraction coefficient, empirically l is treated as a constant 0.2, the relationship between Ia and S is Ia ¼ lS; CN ¼ curve number, which is a dimensionless parameter with the range from 0 to 100, and the higher the CN value, the greater the potential of surface runoff. The SCS model has no strict physical theory, and its runoff calculation formula is summarized by measured data from more than 150 small watersheds in the U.S. It reflects the empirical law of rainfall-runoff events. The key to affect the accuracy of results is whether the parameter CN value can accurately reflect the watershed characteristics. The range of CN values is 0e100. In this study, spatial variable CN values are calculated into a single value to represent the average condition of study sites. The arithmetic means is used to determine the CN value. When CN values are calculated from real storm data as outlined in the preceding

H. Lian et al. / Water Research 177 (2020) 115767

section, a secondary relationship always happened between decreased CN and increased rainfall depth, and finally CN approach a constant (Hawkins, 1993). There are many methods to determine CN values from observed rainfall-runoff data, while no agreement has been settled because no one method shows outstanding advantage (Lal et al., 2017). In addition, the fact that we use mean CN value otherwise asymptotic CN value is the lack of abundant rainfall-runoff data at each study site. What’s more, frequency matching is used to calculate the chosen return-period runoff from the same return-period rainfall depth (Hawkins, 1993). The rainfall and runoff depths are collected separately and then arranged on a rank-order to produce rainfall-runoff pairs during equal return periods (Hawkins et al., 2009). In this work, we used the actual rainfall-runoff events data to calculate CN values other than the rank-order data, because we don’t have enough rainfall-runoff events in some study sites to match rainfall-runoff data in the same return period. Therefore, no obvious difference identified between the two forms of rainfall-runoff data.

3

2.3. Validation of the revised CN In order to verify whether CN values calculated based on the monitored data can well simulate the runoff in the study area, four research sites were selected with abundant data and divided the rainfall event into two parts. The first part was used to determine CN values, and the second part was used to verify the validity of the CN value by using the calculated CN value to estimate the runoff depth and compare with the measured runoff depth. The NashSutcliffe Efficiency (NSE) (Nash and Sutcliffe, 1970) was used to evaluate the performance of the method. The NSE was used as indices of the agreement between the calculated and observed values of runoff depth (perfect fit when NSE equals to 1, and the worse value of NSE is -∞). When the Nash coefficient is above 0.6, the calculated values were deemed qualified.

PN ðQ
Qc Þ2i NSE ¼ 1
P i¼1 0 2 N i¼1 ðQ0
Q0 Þi

(6)

where, Qo ¼ observed runoff depth; Qc ¼ calculated runoff depth. 2.2. Data collection In this study, rainfall-runoff monitoring data were collected from 31 monitoring plots and 24 watersheds to calculate the CN value. These sites located in 21 provinces in China. The distribution of all study sites is shown in Fig. 1. Total 55 sites were selected depending on available monitoring rainfall-runoff event data. The land use, soil and plant type, slope, and other attributes for these sites are showed in Table 1. In addition, the detailed data sources for each study site are showed in Supplementary Material Table S2. The soil and land use data are derived from soil surveys conducted by Soil Testing and Formula Fertilization Program, while the slope and climatic data are collected from literature review. The rainfallrunoff data used to determine CN values were collected from literature and Chinese Ecosystem Research Network (CERN). The time scale of rainfall-runoff data is the day.

2.4. CN values obtained from the table CN entries The CN values for these 55 study sites also obtained from the lookup table according to land use type, management practice, hydrological condition (HC) and hydrologic soil groups (HSG). HSG (A/B/C/D) is classified according to soil saturated hydraulic conductivity Ks (mm/h). A (Ks >180), B (18< Ks 180), C (1.8< Ks 18), D (Ks <1.8). Ks is calculated according to the soil properties data (Table 1) from the Soil Testing and Formula Fertilization Program. The equation for calculating Ks is as follows:

Ks ¼ 0:056C þ 0:016s þ 0:231Om
0:693

(7)

where, C is the percentage of clay content in soil; s is the percentage of sand in soil, and Om is the percentage of organic matter in soil.

Fig. 1. The distribution of monitoring sites for rainfall and runoff. There are totally 55 sites, including 24 for watershed scale and 31 for plot scale.

ID Province

Watershed Area/ km2

Time

N

100.00

2011 e2014 2004 2011 e2014 2004 e2006 2014 2011 e2014 2003 2002 1982 e2011 2000 1993 e2006 2011 e2014 2006 2011 1989 e1996 1990 e2009 2011 e2014 1958 e1966 1958 e1966 2006 1982 e1987 1981 e1994 2013 2001 e2006 1995 1993 2015 2015 2016 1959 e2005 1997 2013 2004 e2005 2011 e2014 1982 2012 e2014 1989

371 2033

3850

12.82 56.15 2.84 158.58

B

42 669 115 1500

660 670

6.13 62.78 7.44 52.36 66.18 24.89 0.94 232.99

42

1030

50

60 1465 136 1613 5 5 1

1

Sichuan

2 3

Beijing Yunnan

4

Jiangsu

100.00

5 6

Jiangxi Guangdong

100.00 100.00

7 8 9

Guizhou Guangdong Beijing

3.22 100.00

1441.00 25.00 176.00

10 Heilongjiang 11 Beijing

100.00 50.00

12 Jilin

100.00

13 Beijing 14 Fujian 15 Hubei

59.10 16.70

16 Hunan

5600.00

100.00

17 Guangdong 18 Shaanxi

100.00 3.70

19 Shaanxi

2.00

20 Sichuan 21 Shanxi

1146.00

22 Beijing

158.00

23 Zhejiang 24 Beijing

1444.00

25 26 27 28 29 30

Guangdong Ningxia Anhui Shaanxi Hubei Shaanxi

31 Shandong 32 Sichuan 33 Shandong

100.00

50.00 100.00 100.00 50.00 0.50 0.14 3.39 10336.00 100.00 323.10

34 Beijing

100.00

35 Anhui 36 Guizhou

2670.00

37 Gansu 38 Jiangxi

8.48

100.00

100.00

Annual Rainfall/ mm

Altitude/ Clay m %

Sand SOM Ks (mm/ HSG CN % % h) Calculated

CN Looked up

Slope Lon

94.73

58

NA

101.96 29.99 alpine meadow

B A

97.45 77.14

63 45

NA NA

117.18 40.58 multiple land use type 100.68 21.59 broadleaf evergreen forest

43.34 19.91 2.04 162.75

B

90.54

60

NA

120.60 28.10 broadleaf evergreen forest

70 NA

14.64 55.84 2.86 93.67 56.89 17.89 2.34 184.31

B A

88.43 68.86

60 45

10 NA

114.92 25.85 forest 113.33 23.33 broadleaf evergreen & deciduous scrub

1220 1523 537

670 40 NA

21.89 58.03 1.27 121.01 8.60 84.78 2.40 80.03 13.20 63.34 2.14 93.26

B B B

82.12 87.34 77.60

60 66 58

NA 2 NA

107.29 26.09 multiple land use type 110.30 21.20 forest 117.07 40.53 multiple land use type

36 3

455 484

176 660

44.89 28.54 0.82 176.00 6.13 62.78 0.94 52.36

B B

85.41 97.20

66 78

10 14.6

123.48 47.90 pine forest 117.18 40.58 dryland

67

800

1090

26.06 30.94 4.06 132.00

B

84.89

66

NA

20 33 21

473 1478 1013

1320 5 250

22.21 48.50 5.13 150.72 15.12 31.58 1.79 64.37 27.04 49.45 0.96 109.98

B B B

78.24 83.73 76.71

60 66 60

8 15 20

128.10 42.40 broadleaf deciduous & needleleaf evergreen forest 116.34 40.28 multiple land use type ̊ 119.14 26.16 orchard 110.46 31.13 forest

28

1474

150

27.93 29.15 3.82 133.13

B

86.59

72

NA

112.72 25.73 multiple land use type

201 1737

NA

3.15

89.95 0.25 58.84

B

80.27

66

NA

112.52 23.10 broadleaf evergreen & deciduous scrub

8

500

1300

7.86

73.66 0.47 62.04

B

71.88

58

30

110.32 39.12 montane steppe

18

500

1200

13.47 55.65 1.81 82.24

B

94.60

81

15

109.30 36.80 dryland

42 4

863 510

1800 1525

41.03 31.75 1.23 143.82 7.79 78.86 0.36 65.27

B B

90.26 91.00

77 78

5 NA

102.18 27.54 dryland 112.00 39.03 multiple land use type

1

537

NA

13.20 63.34 2.14 93.26

B

72.89

60

NA

117.07 40.53 multiple land use type

5 9

1425 508

650 NA

18.98 57.59 0.66 108.21 24.71 43.52 1.40 92.42

B B

87.88 88.68

76 78

NA NA

121.16 30.04 multiple land use type 116.04 40.59 dryland

13 16 6 16 12 3

1720 428 995 635 850 458

NA 3940 30 580 35 477

7.24 9.21 36.35 21.98 28.78 19.66

60.15 90.73 116.06 69.01 90.92 70.35

B B B B B B

75.63 94.75 87.85 86.90 90.70 80.85

66 86 81 81 86 78

NA 10 NA 10 5 3

113.37 105.30 117.55 108.04 117.15 109.17

28 25 11

671 826 612

NA 570 NA

27.21 53.00 1.08 110.71 13.17 67.92 0.74 82.86 45.49 30.29 2.12 150.60

B B B

83.37 76.29 82.82

81 74 81

NA 6.5 NA

118.50 36.42 multiple land use type 105.32 31.49 cropland 117.05 36.42 multiple land use type

35

626

1300

20.73 41.92 0.61 97.75

B

67.71

66

NA

115.31 40.11 broadleaf deciduous scrub

17 5

1600 863

100 1380

20.49 47.11 3.36 81.00 47.34 20.06 0.78 183.36

B A

75.47 44.40

75 45

NA 5

118.40 29.76 multiple land use type 105.43 27.00 broadleaf evergreen & deciduous scrub

5 24

300 1881

1970 NA

12.15 53.92 0.28 61.86 18.75 68.25 1.97 90.81

B B

73.35 77.27

74 78

35 NA

104.26 35.62 dryland 116.77 28.34 Subtropical park woodland

60.44 62.06 26.11 26.59 24.40 38.10

0.75 0.81 0.89 2.22 3.02 0.72

Lat

23.07 36.00 31.75 34.17 31.70 34.76

Land use

orchard bare land dryland dryland bare land dryland

H. Lian et al. / Water Research 177 (2020) 115767

Plot Area/ m2

4

Table 1 Properties for all 55 study sites. Including the number of rainfall-runoff events N, annual rainfall (mm), altitude (m), soil clay content (%), soil organic matter (SOM) (%), Ks, HSG (hydrological soil group), land use type and slope ().

cropland dryland fallow multiple land use type dryland forest 30.07 38.26 30.07 30.84 40.97 26.98 107.38 110.33 107.38 116.36 122.27 112.39 10 NA 10 NA 10 NA 71 78 86 78 75 60 B B B B B B 102.78 58.47 102.78 94.07 115.27 124.49 23.35 77.00 23.35 68.83 50.03 44.13

0.34 1.10 1.10 1.42 1.76 1.25

3.2. Comparison of calculated and table CN entries CN values

60.00 100.00

100.00 984.00

Note: the references for these 55 study sites are added in Supplementary Material Table S2.

31.77 6.37 31.77 13.43 27.24 30.97 519 900 519 910 10 110 1071 440 1071 1400 480 1500 60 8 12 6 18 10 100.00 2.54

Chongqing Shaanxi Chongqing Anhui Liaoning Hunan

49 Shaanxi

50 51 52 53 54 55

100.00 48 Hunan

0.86

100.00 2.54 46 Shaanxi

47 Hunan

0.33

100.00 44 Sichuan

1.14

45 Shaanxi

100.00 42 Gansu 43 Yunnan

0.41

There is a linear correlation between rainfall and runoff data from 401 events (Fig. 2). Distribution of scatters in the graph is highly concentrated. Slope of the trend line in the graph reflects the runoff coefficient, which is around 0.62. The intercept represents the minimum rainfall that produces runoff, which is 20.94 mm. It was found that the SCS-CN model has better simulation effect on study areas with runoff coefficient greater than 0.5 than areas with runoff coefficient less than 0.5 (Peng and You, 2006). In different study areas with same rainfall depth, the amount of runoff produced was substantially different. In the same study area, the same amount of rainfall at different times or seasons also had a large difference in runoff generation. These two phenomena indicated that influencing factors of the CN value are complex. Interactions in external natural geographical conditions will cause the difference in regional runoff generation ability, while the internal seasonal variation and rainfall duration will also cause the difference in the runoff. Relationships between CN values and runoff depth are shown in Fig. 3 under different rainfall depths. According to the SCS-CN method, larger CN values may generate greater runoff under the same rainfall event. However, small rainfall events can still generate runoff by proper amount of initial abstraction and the soil infiltration ability. It indicates the research area has a strong ability to produce runoff. Therefore, the corresponding CN value could be large.

53.03 58.80 64.16 53.23 40.92 12.80

110.22 39.23 dryland NA 78 70.74 B 18.26 55.21 2.18 103.05 7

440

1250

B 51

1335

430

36.04 19.61 3.17 142.31

48.89

55

41

109.71 26.86 broadleaf deciduous & needleleaf evergreen forest 109.71 26.86 needleleaf evergreen forest 40 55 B

49.47

3.1. Relationships between rainfall and runoff events

36.04 19.61 3.17 142.31 430 1335 52

110.32 39.12 pasture 30 69 B 5

440

1300

7.86

73.66 0.47 62.04

64.07

109.73 34.91 dryland NA 78 74.03 B 20.10 34.33 0.67 72.92 5

561

410

B 38

909

965

35.74 28.60 1.01 127.23

63.06

66

NA

106.06 34.65 dryland 101.20 26.58 broadleaf deciduous & needleleaf evergreen forest 104.02 31.67 broadleaf evergreen & deciduous scrub NA NA 77 60 74.69 57.68 B B 20.71 35.72 2.93 75.15 13.10 46.77 1.53 87.93 14 90

522 1200

1625 2300

B B B 5.30 219.00

39 Yunnan 40 Sichuan 41 Shaanxi

5

3. Results

2004 e2006 2013 2013 1959 e1995 1987 2011 e2014 2011 e2014 1959 e2008 1959 e1964 2011 e2014 2011 e2014 1959 e1972 2006 1959 2006 1987 2008 2009

3 15 3

547 900 556

1997 NA 1230

21.32 35.27 1.24 91.27 13.39 63.71 1.03 81.75 28.85 24.94 0.93 93.59

58.78 76.29 76.01

60 78 78

NA 6.5 NA

100.18 25.35 multiple land use type 105.57 31.06 dryland 108.19 35.21 dryland

H. Lian et al. / Water Research 177 (2020) 115767

CN values of the 55 study sites were calculated from the measured data of rainfall and runoff events, and the average CN value of different events in the same study area represents the CN parameters of this site. The calculated CN values have showed clear differences compared to the CN values obtained from the table CN entries (USDA, 1954) for the study area (Fig. 4, Supplementary materials Table S1). Results revealed that the calculated CN value ranges 12.80e97.45, while the looked-up table indicates as 36e91. The difference between calculated and looked-up values ranges between
40.41e34.08. Comparisons of average value and error at each point indicated that the CN value acquired by table CN entries have reduced its variation. It could be substantially different by regions, which largely ignores the study area characteristics. Moreover, the CN table CN entries provided by USDA-SCS were summarized from a limited set of watershed observations (Bartlett et al., 2016; Ogden et al., 2017). Thus, these differences have supported our doubts on the applicability of the existing table CN entries in China, and for the areas with different characteristics in different regions, the existing looked-up table for CN is not suitable for runoff estimation. The blue line represents the calculated value of the monitoring point CN, and the orange line represents the CN value obtained by the existing look- up the table (Fig. 4). It can be found that CN value based on the measured data has a large variation range. Differences between the monitoring points can be clearly found, while the orange line has a small fluctuation range, floating between 50e80. In literature, it is nearly impossible to see that the CN value is less than 30 under actual environmental conditions, and the range of variation is between 30e100. Here, some study sites with CN values more than 80 and some are less than 30. After repeated inspections of the data, the occurrence of extreme CN values is because of the different forest type and local characters.

6

H. Lian et al. / Water Research 177 (2020) 115767

3.3. Model performance The purpose of validation conducted in the study was to ensure the driven results are reasonable in scientific measures. One should keep in mind that the derived results could be biased by changing validation standard (Yen et al., 2016). In this study, 0.6 is the value of NSE set to be the statistical guideline of satisfactory. To figure out the reliability of CN method, the simulated runoff was compared with observed runoff depths. For this purpose, four sites with large set of rainfall-runoff data were selected to validate the performance of this model. The freedom of rainfall-runoff events to validate the accuracy of CN parameters of these four sites is 7, 14, 8, and 14 respectively. The NSE results are 0.98, 0.76, 0.86, and 0.84 respectively, far more than 0.6, which indicates that the CN method is accurate enough to simulate runoff depth (Fig. 5). 4. Discussions 4.1. Differences of CN values between the United States and China Studies have shown that soil moisture contents, hydrological processes, infiltration, and other losses through rainfall have significant spatial and temporal variability (Ponce and Hawkins, 1996; Tramblay et al., 2010; Wang, 2018; Zeng et al., 2017). It indicated that the forestland has lowest runoff production capacity and the paddy field has highest (Choi et al., 2019). Other researchers also found that the land use types with high CN values are mainly urban area, farmland and water bodies, and the low CN values are in forests, orchards, shrubs and grasslands (Choi et al., 2019; Egodawatta et al., 2007; Li et al., 2018; Soulis et al., 2017; Zhang et al., 2018). This might be associated with less vegetation cover in urban area, farmland and water bodies compared to forests. Although CN-method may introduce large uncertainty in runoff estimation of paddy field, many scholars still treat paddy filed as one kind of land use type with low infiltration capacity in order to estimate runoff depth easily. The main character of curve number method is its base on accumulated experience in use, mainly in the U.S. There are many weaknesses and limitations when the original CN values were applied in China (Boughton, 1989). Furthermore, some watersheds performed quite differently from the basic CN runoff response patterns, leading to great differences between model and the actual processes. The major weakness and utmost potential source of error is the sensitivity of estimated runoff in the selection of the curve number.

Fig. 2. Relationships between rainfall and runoff of all monitoring sites. A total of 600 rainfall events were collected across the country.

Fig. 3. Relationships between CN values and runoff depth under different rainfall depths.

4.2. Impact factors of the revised CN Failure to evaluate the size of regional scale might affect the corresponding error between actual runoff and CN parameters simulated runoff. By summarizing the previous studies, it was found that this model is suitable for the study of small and medium-scale watersheds, with an area of about 0.25e1000 km2. Even the CN method was appropriate still the dominant behavior of CN value’s response to rainfall depth is important to calculate the runoff depth (Muche et al., 2019; Soulis et al., 2009). Results indicated a decrease in CN value with increase in rainfall but stable at high rainfall intensity (Fig. 6). It indicates that the rainfall depth is the primary factor to determine the CN value, and the asymptotic constant value is the best way to identify CN parameters for a watershed. And other studies have showed that CN values calculated from measured rainfall-runoff data vary systematically with the rainfall depth, the determination of a single asymptotic CN value observed for very high rainfall depths to characterize the watershed’s runoff response (Soulis and Valiantzas, 2012). Standard behavior of CN response for the rainfall-runoff relationship found in our study with a distinct bias for high CN at lower rainfall depths indicates that rainfall is another key factor in determining CN. Rainfall duration could be an important phenomenon to increase soil water content, which will be at maximum after certain rainfall period (Thorndahl and Willems, 2008). After attaining maximum soil water contents, CN values will be increased nearly up to 100 indicating runoff generation capacity is also increased. As the previous soil moisture condition (AMC) has a great influence on the results (Lal et al., 2019), the study sites with detailed monitoring data can be more reliable. For same test sites, runoff coefficient is different under different rain intensity. Therefore, CN value as the parameter of regional runoff needs to be able to represent the runoff property of the region, which needs to be calculated. Longterm monitoring of rainfall in different seasons and intensity can reflect the overall situation of the area (Yeh et al., 2018). Curve number varies spatially with the change in land use type, soil type and soil humidity in the early stage (Ross et al., 2018). Urban land and water bodies have higher CN value compared to

H. Lian et al. / Water Research 177 (2020) 115767

7

Fig. 4. The calculated CN value and looked-up CN value for 55 study sites. Only one third of total study sites can use table CN entries (USDA and Soil Conservation Service 1954) from USDA to gain satisfied (±5%) CN values.

Fig. 5. The results of SCS-CN method validation. Four study sites with enough data are selected to validate the accuracy of the model. Qob is the observed runoff depth, Qca is the calculated runoff depth, and n represent the number of rainfall-runoff events.

forests, mountains, and hills (Yang and Toor, 2017). The effect of soil properties on CN is divided into four categories according to the permeability of soil. American soil experts initially based on more than 14,000 soil data classified soils with similar runoff generation capacity called the hydrological soil group (HSG) (Ross et al., 2018; Stewart et al., 2012). The runoff generation capacities from low to high are divided into four categories: A, B, C, and D. Detailed classification criteria showed in the US National Engineering Handbook, 1972 (USDA, 1972). 4.3. Shortcomings and future perspectives The main limitation of our study is the ignorance of antecedent moisture condition (AMC). Because the rainfall-runoff event data we used did not have the rainfall data of the days before the event occurred, it was impossible to know the soil moisture conditions in the early stage of the event. Another limitation is the number of rainfall-runoff events is still not large enough to reflect the more

detailed hydrological conditions for some study sites. In this study, we have tried our best to collect more data in order to cover more regions of China and collect more rainfall-runoff events to calculate accurate results. The runoff calculated by the SCS-CN method is known to be sensitive to CN values, which is the largest potential source of error. Low frequency and poor resolution of monitoring data will result in poor estimation of runoff. Even with years of data available for calibration of the curve numbers, substantial errors may still occur (Boughton, 1989). Obviously, we are getting more and more accurate and high-resolution rainfall data through advanced observation devices. Thus, a strong runoff estimation method is urgently needed. While it is very difficult to establish the mechanism equation of rainfall and runoff because infiltration capacity is affected by many factors such as soil properties, vegetation coverage, land use type, climate, etc. Fortunately, we have such a simple empirical method to estimate runoff and meet the needs of massive simulations. There are several studies mapping the global

8

H. Lian et al. / Water Research 177 (2020) 115767

Acknowledgments This study was supported by funding from the National Natural Science Foundation of China (Grant No.: 31572208); the Special Fund for Agro-scientific Research in the Public Interest (Grant No.: 201303089); and the Newton Fund (Grant Ref: BB/N013484/1). And acknowledgment for the data support from Chinese Ecosystem Research Network (CERN). Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.watres.2020.115767. References Fig. 6. Response of CN values to rainfall intensity. Decreasing CN value with increasing rainfall but stable at high rainfall intensity.

CN values based on Moderate Resolution Imaging Spectroradiometer (MODIS), Harmonized World Soil Database (HWSD), and Normalized Difference Vegetation Index (NDVI) data sets (Lin et al., 2017; Zeng et al., 2017). Those methods are efficient to produce high-resolution CN values. While the huge error will be introduced by the original CN lookup table which may not suitable for other countries. Our work is going to be very useful to resolve this issue with relatively minor effort. And future research should focus on the establishment of a dynamic runoff estimation model. The input data of the model should include meteorological, topographical, soil type, land use, and vegetation cover data sets. For ungauged watersheds, CN parameters are usually estimated by well-known handbook tables (USDA, 1972), according to the hydrologic soil group (HSG), land use, surface condition and antecedent moisture condition (AMC). In the presence of variable conditions (soil type, land cover, and land use) within the basin, an area-weighted average CN is often used. According to our results, it is urgently needed a CN handbook table for different regions, which can be established based on local rainfall runoff monitoring network data.

5. Conclusion The Curve Number approach is an effective, simple and widely used method to estimate surface runoff volume. The revised CN values calculated by using monitoring data from 55 sites in China are considerably different from the CN values acquired from USbased CN value lookup-table. It has been shown that the monitoring data can provide valuable information to calculate the CN value. In this study, the modified CN values (CN-China) derived from the actual rainfall-runoff data play an important role and a very useful reference in future runoff estimation. The results of this study provide new insight into the establishment of the CN table that is more suitable for China and possibly many other countries around the world. One can take advantage of the given work to incorporate more details of the curve number approach to further improve the corresponding accuracy and applicability.

Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Bartlett, M.S., Parolari, A.J., McDonnell, J.J., Porporato, A., 2016. Beyond the SCS-CN method: a theoretical framework for spatially lumped rainfall-runoff response. Water Resour. Res. 52 (6), 4608e4627. Boughton, W.C., 1989. A review of the USDA SCS curve number method. Aust. J. Soil Res. 27 (3), 511e523. Chen, H.L., Liang, Z.Y., Liu, Y., Liang, Q.H., Xie, S.G., 2017. Integrated remote sensing imagery and two-dimensional hydraulic modeling approach for impact evaluation of flood on crop yields. J. Hydrol. 553, 262e275. Chin, D.A., 2017. Estimating the parameters of the curve number model. J. Hydrol. Eng. 22 (6), 06017001. Choi, D., Park, H., Kim, Y.J., Jung, J.W., Choi, W.J., Her, Y.G., Yoon, K.S., 2019. Curve numbers for rice paddies with different water management practices in korea. J. Irrigat. Drain. Eng. 145 (5), 06019003. Clyde, G.D., Work, R.A., 1943. Precipitation runoff relationships as a basis for watersupply forecasting. Trans. Am. Geophys. Union 24, 43e49. D’Asaro, F., Grillone, G., Hawkins, R.H., 2014. Curve number: empirical evaluation and comparison with curve number handbook tables in sicily. J. Hydrol. Eng. 19 (12), 04014035. Dong, L.H., Xiong, L.H., Lall, U., Wang, J.W., 2015. The effects of land use change and precipitation change on direct runoff in Wei River watershed, China. Water Sci. Technol. 71 (2), 289e295. Egodawatta, P., Thomas, E., Goonetilleke, A., 2007. Mathematical interpretation of pollutant wash-off from urban road surfaces using simulated rainfall. Water Res. 41 (13), 3025e3031. Esen, I.I., 1987. Least-squares estimates of the Horton infiltration parameters. Soil Sci. 144 (1), 6e10. Fauvel, B., Cauchie, H.M., Gantzer, C., Ogorzaly, L., 2016. Contribution of hydrological data to the understanding of the spatio-temporal dynamics of F-specific RNA bacteriophages in river water during rainfall-runoff events. Water Res. 94, 328e340. Freyberg, D.L., Reeder, J.W., Franzini, J.B., Remson, I., 1980. Application of the greenampt model to infiltration under time-dependent surface-water depths. Water Resour. Res. 16 (3), 517e528. Fu, S., Zhang, G., Wang, N., Luo, L., 2011. Initial abstraction ratio in the scs-cn method in the loess plateau of China. Transact. Asabe 54 (1), 163e169. Garen, D.C., Moore, D.S., 2005. Curve number hydrology in water quality modeling: uses, abuses, and future directions. J. Am. Water Resour. Assoc. 41 (2), 377e388. Gaume, E., Livet, M., Desbordes, M., Villeneuve, J.P., 2004. Hydrological analysis of the river Aude, France, flash flood on 12 and 13 November 1999. J. Hydrol. 286 (1e4), 135e154. Grimaldi, S., Petroselli, A., Romano, N., 2013. Green-Ampt Curve-Number mixed procedure as an empirical tool for rainfall-runoff modelling in small and ungauged basins. Hydrol. Process. 27 (8), 1253e1264. Guo, A.J., Chang, J.X., Liu, D.F., Wang, Y.M., Huang, Q., Li, Y.Y., 2017. Variations in the precipitation-runoff relationship of the weihe river basin. Nord. Hydrol 48 (1), 295e310. Gwynne, M.D., Glover, J., 1961. Light rainfall and plant survival - meaurement of stem flow run-off. Nature 191 (479), 1321. &. Hawkins, R.H., 1975. The importance of accurate curve numbers in the estimation of storm runoff. Water Resour. Bull. 11 (5), 887e891. Hawkins, R.H., 1978. Infiltration formula based on scs curve number. J. Irrigation and Drainage Division-Asce 104 (4), 464e467. Hawkins, R.H., 1978. Runoff curve numbers with varying site moisture. J. Irrigation and Drainage Division-Asce 104 (4), 389e398. Hawkins, R.H., 1993. Asymptotic determination of runoff curve numbers from data. J. Irrigat. Drain. Eng. 119 (2), 334e345. Hawkins, R.H., Theurer, F.D., Rezaeianzadeh, M., 2019. Understanding the basis of the curve number method for watershed models and TMDLs. J. Hydrol. Eng. 24 (7), 06019003. Hawkins T.W., R.H., Ward, T.J., Woodward, D.E., Van Mullem, J.A., 2009. Curve Number Hydrology State of the Practice. American Society of Civil Engineers, Reston Va, 106pp. He, B., Zhou, N., Gao, X., 2001. Precipitation-runoff relationship in farmland nonpoint source pollution research: amending coefficient of SCS hydrologic

H. Lian et al. / Water Research 177 (2020) 115767 method. Res. Environ. Sci. 14 (3), 49e51. Huang, M.B., Gallichand, J., Wang, Z.L., Goulet, M., 2006. A modification to the Soil Conservation Service curve number method for steep slopes in the Loess Plateau of China. Hydrol. Process. 20 (3), 579e589. Huang, M.B., Gallichand, J., Dong, C.Y., Wang, Z.L., Shao, M.G., 2007. Use of soil moisture data and curve number method for estimating runoff in the Loess Plateau of China. Hydrol. Process. 21 (11), 1471e1481. Hvitvedjacobsen, T., Yousef, Y.A., 1988. Analysis of rainfall series in the design of urban drainage control-systems. Water Res. 22 (4), 491e496. Jiang, W.Y., Haver, D., Rust, M., Gan, J., 2012. Runoff of pyrethroid insecticides from concrete surfaces following simulated and natural rainfalls. Water Res. 46 (3), 645e652. Jiao, P.J., Xu, D., Wang, S.L., Yu, Y.D., Han, S.J., 2015. Improved SCS-CN method based on storage and depletion of antecedent daily precipitation. Water Resour. Manag. 29 (13), 4753e4765. Kim, J.Y., Sansalone, J.J., 2008. Event-based size distributions of particulate matter transported during urban rainfall-runoff events. Water Res. 42 (10e11), 2756e2768. Kirchner, J.W., Feng, X.H., Neal, C., 2000. Fractal stream chemistry and its implications for contaminant transport in catchments. Nature 403 (6769), 524e527. Lal, M., Mishra, S.K., Pandey, A., Pandey, R.P., Meena, P.K., Chaudhary, A., Jha, R.K., Shreevastava, A.K., Kumar, Y., 2017. Evaluation of the Soil Conservation Service curve number methodology using data from agricultural plots. Hydrogeol. J. 25 (1), 151e167. Lal, M., Mishra, S.K., Kumar, M., 2019. Reverification of antecedent moisture condition dependent runoff curve number formulae using experimental data of Indian watersheds. Catena 173, 48e58. Li, J., Wang, Z.G., Liu, C.M., 2015. A combined rainfall infiltration model based on Green-Ampt and SCS-curve number. Hydrol. Process. 29 (11), 2628e2634. Li, C.L., Liu, M., Hu, Y.M., Shi, T., Zong, M., Walter, M.T., 2018. Assessing the impact of urbanization on direct runoff using improved composite CN method in a large urban area. Int. J. Environ. Res. Publ. Health 15 (4), 775. Lin, W., Yang, F., Zhou, L., Xu, J.G., Zhang, X.Q., 2017. Using modified Soil Conservation Service curve number method to simulate the role of forest in flood control in the upper reach of the Tingjiang River in China. J. Mt. Sci. 14 (1), 1e14. Mishra, S.K., Singh, V.P., 1999. Another look at scs-cn method. J. Hydrol. Eng. 4 (3), 257e264. Mockus, V., 1949. Estimation of total (and peak rates of) surface runoff for individual storms. Exhibit A of Appendix B, Interim Survey Report, Grand (Neosho) River Watershed. USDA Soil Conservation Service. Muche, M.E., Hutchinson, S.L., Hutchinson, J.M.S., Johnston, J.M., 2019. Phenologyadjusted dynamic curve number for improved hydrologic modeling. J. Environ. Manag. 235, 403e413. Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models part I- A discussion of principles. J. Hydrol. 10 (3), 282e290. Ogden, F.L., Hawkins, R., Walter, M.T., Goodrich, D.C., 2017. Comment on "Beyond the SCS-CN method: a theoretical framework for spatially lumped rainfallrunoff response. In: Bartlett, M.S. (Ed.), Water Resources Research vol. 53(7), pp. 6345e6350 et al. Peng, D., You, J., 2006. Application of modified SCS model into runoff simulation. J. Water Res. Water Eng. 17, 20e24. Ponce, V.M., Hawkins, R.H., 1996. Runoff curve number: has it reached maturity? J. Hydrol. Eng. 1 (1), 11e19. Ross, C.W., Prihodko, L., Anchang, J., Kumar, S., Ji, W.J., Hanan, N.P., 2018. HYSOGs250m, global gridded hydrologic soil groups for curve-number-based runoff modeling. Scientific Data 5, 150091. Shi, Z.H., Chen, L.D., Fang, N.F., Qin, D.F., Cai, C.F., 2009. Research on the SCS-CN initial abstraction ratio using rainfall-runoff event analysis in the Three Gorges Area, China. Catena 77 (1), 1e7. Shi, W.H., Huang, M.B., Gongadze, K., Wu, L.H., 2017. A modified SCS-CN method incorporating storm duration and antecedent soil moisture estimation for runoff prediction. Water Resour. Manag. 31 (5), 1713e1727. Soulis, K.X., Valiantzas, J.D., 2012. SCS-CN parameter determination using rainfallrunoff data in heterogeneous watersheds - the two-CN system approach. Hydrol. Earth Syst. Sci. 16 (3), 1001e1015.

9

Soulis, K.X., Valiantzas, J.D., Dercas, N., Londra, P.A., 2009. Investigation of the direct runoff generation mechanism for the analysis of the SCS-CN method applicability to a partial area experimental watershed. Hydrol. Earth Syst. Sci. 13 (5), 605e615. Soulis, K.X., Ntoulas, N., Nektarios, P.A., Kargas, G., 2017. Runoff reduction from extensive green roofs having different substrate depth and plant cover. Ecol. Eng. 102, 80e89. Steenhuis, T.S., Winchell, M., Rossing, J., Zollweg, J.A., Walter, M.F., 1995. Scs runoff equation revisited for variable-source runoff areas. J. Irrigation Drain. Eng. Asce 121 (3), 234e238. Stewart, R.D., 2018. A dynamic multidomain green-ampt infiltration model. Water Resour. Res. 54 (9), 6844e6859. Stewart, D., Canfield, E., Hawkins, R., 2012. Curve number determination methods and uncertainty in hydrologic soil groups from semiarid watershed data. J. Hydrol. Eng. 17 (11), 1180e1187. Thorndahl, S., Willems, P., 2008. Probabilistic modelling of overflow, surcharge and flooding in urban drainage using the first-order reliability method and parameterization of local rain series. Water Res. 42 (1e2), 455e466. Tramblay, Y., Bouvier, C., Martin, C., Didon-Lescot, J.F., Todorovik, D., Domergue, J.M., 2010. Assessment of initial soil moisture conditions for event-based rainfallrunoff modelling. J. Hydrol. 387 (3e4), 176e187. Triadis, D., Broadbridge, P., 2012. The Green-Ampt limit with reference to infiltration coefficients. Water Resour. Res. 48. W07515. Tyagi, J.V., Mishra, S.K., Singh, R., Singh, V.P., 2008. SCS-CN based time-distributed sediment yield model. J. Hydrol. 352 (3e4), 388e403. USDA, Soil Conservation Service (SCS), 1954. National Engineering Handbook. Section 4. Hydrology. U.S Development of agriculture, Washington, D.C. USDA, Soil Conservation Service (SCS), 1972. National Engineering Handbook. Section 4. Hydrology. U.S Development of agriculture, Washington, D.C. Wang, D.B., 2018. A new probability density function for spatial distribution of soil water storage capacity leads to the SCS curve number method. Hydrol. Earth Syst. Sci. 22 (12), 6567e6578. Wang, S., Wang, H., 2018. Extending the Rational Method for assessing and developing sustainable urban drainage systems. Water Res. 144, 112e125. Wilson, L.E., Ramirez-Avila, J.J., Hawkins, R.H., 2017. Runoff Curve Number Estimation for Agricultural Systems in the Southern Region of USA. World Environmental and Water Resources Congress 2017: Watershed Management, Irrigation and Drainage, and Water Resources Planning and Management, pp. 353e366. Yang, Y.Y., Toor, G.S., 2017. Sources and mechanisms of nitrate and orthophosphate transport in urban stormwater runoff from residential catchments. Water Res. 112, 176e184. Yeh, T.C., Liao, C.S., Chen, T.C., Shih, Y.T., Huang, J.C., Zehetner, F., Hein, T., 2018. Differences in N loading affect DOM dynamics during typhoon events in a forested mountainous catchment. Sci. Total Environ. 633, 81e92. Yen, H., Wang, X.Y., Fontane, D.G., Harmel, R.D., Arabi, M., 2014. A framework for propagation of uncertainty contributed by parameterization, input data, model structure, and calibration/validation data in watershed modeling. Environ. Model. Software 54, 211e221. Yen, H., Hoque, Y.M., Wang, X.Y., Harmel, R.D., 2016. Applications of explicitly incorporated/post-processing measurement uncertainty in watershed modeling. J. Am. Water Resour. Assoc. 52 (2), 523e540. Zeng, Z.Y., Tang, G.Q., Hong, Y., Zeng, C., Yang, Y., 2017. Development of an NRCS curve number global dataset using the latest geospatial remote sensing data for worldwide hydrologic applications. Remote Sense. Lett. 8 (6), 528e536. Zhang, W., Villarini, G., Vecchi, G.A., Smith, J.A., 2018. Urbanization exacerbated the rainfall and flooding caused by hurricane Harvey in Houston. Nature 563, 384e388. Zhang, D.J., Lin, Q.Y., Chen, X.W., Chai, T., 2019. Improved curve number estimation in SWAT by reflecting the effect of rainfall intensity on runoff generation. Water 11, 163. Zoure, C., Queloz, P., Koita, M., Niang, D., Fowe, T., Yonaba, R., Consuegra, D., Yacouba, H., Karambiri, H., 2019. Modelling the water balance on farming practices at plot scale: case study of Tougou watershed in Northern Burkina Faso. Catena 173, 59e70.