INVESTIGATION OF SCS-CN METHODOLOGY ON EXPERIMENTAL PLOT AND CATCHMENT SCALES

Abstract

Runoff is one of the most important variables used in planning and design of hydraulic structures and assessing the water yield potential of a watershed. Runoff is a function of many variables such as rainfall duration and intensity, soil moisture, land use/land cover, soil infiltration capacity, watershed slope etc. There are a number of models available in literature considering different variables governing the surface runoff. Among them, the lumped conceptual models are quite useful for simple yet realistic analyses. The Natural Resources Conservation Service curve number (NRCS–CN) formerly known as the Soil Conservation Service curve number (SCS–CN) method is the most popular method to determine the storm event runoff from an ungauged small watershed for a given amount of rainfall. The SCS-CN method was developed in 1954. It is documented in Section 4 of the National Engineering Handbook (NEH-4) published by the Soil Conservation Service (now called the Natural Resources Conservation Service), United States. Department of Agriculture in 1956. The document has since been revised several times. The SCS-CN method is the result of exhaustive field investigations carried out during 1930s and 1940s. The method has since then witnessed myriad applications world over. It is one of the most popular methods for computing the surface runoff for a given rainfall event from small agricultural, forest, and urban watersheds. It is simple, easy to understand and apply, stable, and useful for ungauged watersheds. Due to its low input data requirements and simplicity, many erosion, hydrologic, and water-quality models have employed this method for determination of runoff. The primary reason for its wide applicability and acceptability lies in the fact that it accounts for most runoff producing watershed characteristics: soil type, land use/treatment, surface condition, and antecedent moisture condition. The only parameter of this methodology, i.e. the Curve Number (CN), is crucial for accurate runoff prediction. Based on exhaustive field investigations carried out in the United States, curve numbers were derived for different land uses, soil types, hydrologic condition, and management practices and these are reported in NEH-4. These numbers have seldom been verified for Indian watersheds. Evidently, most studies have concentrated on the application of the existing SCS-CN method utilizing CN derived from NEH-4 tables. No systematic effort appears to have been made for evaluating the SCS-CN methodology experimentally, particularly for Indian watersheds, which invokes the need of the study. The aim of present research was to enhance the understanding of SCS-CN methodology by investigating its different parameters employing naturally observed P-Q datasets. This study covers relative accuracy of different CNs VI determination methods and comparing them with NEH–4 tables CN values; evaluating the effect of initial abstraction coefficient and antecedent moisture on CN and runoff; evaluation of existing AMC–dependent CN formulae, which are otherwise developed using United States datasets. The AMC–dependent CN formulae incorporating initial abstraction coefficient effect is also tested for enhancing runoff estimation. The present study uses the rainfall (P)–runoff (Q) dataset of various climatic settings. Locally measured and published literature data have been used in the investigation of different parameters of SCS-CN methodology. For locally monitored data, the natural P–Q events were captured on 35 plots of 22m length and 5m width having different slope (5%, 3%, and 1%), land use (agricultural land use: Sugarcane, Maize, Black gram, Fallow land, Lentil, and Chana), and hydrologic soil group (HSG) during August 2012–April 2015 (or three crop growing seasons in study area) for the experimentation work carried out at Roorkee, India. The experimental field (Lat.: 29° 50′ 09″ N and Long.: 77° 55′ 21″ E) is situated at the right bank of Solani River, a tributary of Ganga River, the largest river basin in India. Precipitation was recorded with the help of Tipping Bucket rain gauge and a non-recording rain gauge installed within the experimental site. The surface runoff generated during rain storms was collected in separate chambers equipped with multi-slot divisor (5-slot) (1m × 1m × 1m) constructed at the downstream end of each plot and the variation in depth of water stored with respect to time was monitored regularly, but manually. Infiltration tests were conducted for each plot using the double ring infiltrometer. Soil water measurements were taken by time domain reflectometry (TDR) probe of the ‘Fieldscout TDR-300’. Besides, the published literature P-Q data were collected for 36 plots/watersheds having different size, land use, slope and soil consisting heterogeneous climatic conditions. The rainfall (P)runoff (Q) behaviour pattern was analysed using naturally observed PQ data from experimental study plots located at Roorkee site and it was found that nonlinear variation of runoff coefficient (Rc) with P is similar to the variation of Q with P, but the correlation between Rc and P is much lower than that between Q and P. As expected, the mean runoff coefficient (Rcm) was higher for the plots having HSGs C followed by B and A. The concept of runoff initiation threshold (I) also called rainfall threshold for runoff generation confirms the runoff generation phenomenon of generating low runoff from lighter soils as the values of I was highest for HSGs A followed by B and C. These finding indicates that HSG (or indirectly soils infiltration capacity, fc) seems to play a major role in controlling runoff in the plots. The KruskalWallis (KW) test analysis performed to analyse the effect of land use, soil VII type, and plot slope on Q (or Rc) show that, Q is more significantly influenced by soil type rather than land uses or slopes as fc is the main explanatory variable for runoff (or CN) production in the study plots. In present study experimental plots, CN is inversely related to fc, which supports the applicability of NEH4 tables CNs declining with fc (or HSG). Further to check the dependency of observed CN on in-situ antecedent moisture content, CN (or, potential maximum retention, S) values showed a higher degree of dependence on the physically observed 1-day antecedent soil moisture (θo1) than other duration antecedent soil moisture values. The performance of eight different CN estimation methods, viz. storm event mean and median, rank-order mean and median, log-normal frequency, S-probability (SP), geometric mean and least square fit, was evaluated using P–Q data measured on small agricultural plots located in India. The KruskalWallis test multiple comparison analysis show that there was no single method which has produced significantly higher (or lower) CNs than other. The least square fit method was observed to estimate significantly lower CN than other methods except log-normal frequency method. Based on the overall score and ranking system calculated from different goodness of fit indices, the method performance in runoff estimation was as follows: S-probability > geometric mean > storm event mean > rank-order median > rank-order mean > least square fit > storm event median > log-normal frequency. The comparison of observed P-Q data based CNs with tabulated CNs show that, on the whole, the CN estimates from NEH-4 tables do not match those derived from observed P–Q dataset. As a result, the runoff prediction using former CNs was poor for the data of experimental plots of Roorkee site. However, match was little better for higher CN values, consistent with general notion that the existing SCS-CN method performs better for high P–Q (or CN) events. The reason for tabulated CNs to have performed most poorly is that these are the generalized values derived from the watersheds of United States, consistent with the results of other studies. The plot-data optimization yielded initial abstraction coefficient (λ) values ranging from 0 to 0.659 for ordered dataset and 0 to 0.208 for natural dataset (with 0 as the most frequent value for both datasets). Mean and median λ values were, respectively, 0.030 & 0 for natural P– Q dataset and 0.108 & 0 for ordered P–Q dataset, quite different from standard λ =0.2, but consistent with the results of other studies carried out elsewhere. Notably, the existence of Ia-S relationship for different plots was also investigated; and in contrast to the existing notion, Ia when plotted against S exhibited no correlation for both natural and ordered datasets, consistent with the findings of Jiang (2001). Runoff estimation was very sensitive to λ and it improved consistently as  changed from 0.2 to 0.03. Compared to traditionally assumed λ=0.2, a refined VIII λ=0.03 is recommend for the use in regions of similar to study site. Further, a relationship between CN0.20 (λ = 0.20) and CN0.03 (λ = 0.03), useful for CN conversion for field application is established. It is well established phenomenon that accurate estimation of the surface runoff is one of the most important bases for planning and management of water resource systems and environmental quality assessment of water and soil. Therefore, in popular SCS–CN method, correct estimation of AMC–dependent CN values is always necessary. Since CNs varies with climatic condition of watersheds, there is need of using AMC-dependent CN-formulae developed utilizing data of watersheds having heterogeneous climatic conditions. The formulae developed from heterogeneous and large data sets will tend to have wider applicability. The present work evaluated the five existing (Arnold et al. 1990; Chow et al. 1988; Hawkins et al. 1985; Mishra et al. 2008b; Sobhani 1975) and three proposed (MC6, MC7, MC8) CN-AMC formulae. For developing the proposed formulae, CNs were derived for datasets from a large number of naturally observed P–Q events for an agricultural field located at Roorkee, Uttarakhand, India and available published data around the globe using standard initial abstraction ratio (λ) values as 0.20 and 0.030. The analysis shows that the existing Hawkins et al. (1985) formulae performed the best for conversion of CN2 into CN1 and CN3, when tested on NEH–4 AMC defining Tabular CNs considered as targeted values. It might be because the existing formulae were derived from the same datasets used as targeted values (i.e. NEH–4 AMC defining tables). However, all the three proposed MC6, MC7, and MC8 were best of the existing formulae in their application to field data. MC8 incorporating the effect of λ = 0.030 performed the best of all, and MC7 and MC6 better than the other existing formulae. Among the existing formulae, Mishra et al. (2008b) was superior followed by Hawkins et al. (1985). A comparison of the results derived from the eight different methods concluded that the MC8 formula that incorporates the effect of λ into standard SCS–CN method showed a superior performance in runoff simulation than the others. Since the proposed formulae performed the best in field application, these are recommended for field use to improve the accuracy of SCS–CN model. Keywords: Agricultural field; Curve number; Antecedent moisture condition; Runoff; NEH-4 Table; SCS-CN; NRCS-CN; Initial abstraction coefficient; Infiltration capacity.