Storm-based CSLE that incorporates the estimated runoff for soil loss prediction on the Chinese Loess Plateau

Abstract

The Loess Plateau region of China is one of the most erodible areas in the world. Several conservation practices have been considered to reduce soil and water losses in this region and a range of erosion models used to predict storm-based surface runoff and soil loss. In particular, the Chinese Soil Loss Equation (CSLE) has been widely used to assess soil losses for a range of different land-use types. However, the CSLE as originally developed was only based on estimating gross annual soil loss. The present study proposes the use of a rainfall–runoff erosivity factor to estimate storm-based soil loss using the CSLE. The factor was developed using data from three runoff-erosion plots in each of three different watersheds of the Loess Plateau over three different time periods (1956–1959; 1973–1980; 2010–2013). The modified CSLE was evaluated using data from 165 storm–runoff events from six plots in two watersheds. The performance of the modified CSLE was compared with that of the storm-based Revised Universal Soil Loss Equation (RUSLE). The findings during calibration (88 storms) and validation (77 storms) show the storm-based CSLE is highly accurate in terms of model efficiency as determined by the Nash–Sutcliffe efficiency (NSE) (calibration: 65.7%, validation: 75.1%) and root mean square error (RMSE) (calibration: 4.36 t ha–1, validation: 3.23 t ha–1). The modified CSLE also performs better than the storm-based RUSLE during both calibration (NSE = 58.3% and RMSE = 4.81 t ha–1) and validation (NSE = 48.3% and RMSE = 4.64 t ha–1). The storm-based CSLE was then used to predict soil loss in the three experimental plots of the third watershed using the parameters obtained from the previously monitored six plots. Unlike the first and second watersheds, surface runoff in the third watershed was estimated using the modified Soil Conservation Service curve number (SCS-CN) method and compared to measurements. The measured and estimated runoff for the storm-based CSLE has NSE values of 64.7 and 60.8%, respectively. High NSE values indicate that the proposed storm-based CSLE, which incorporates the modified SCS-CN, can accurately predict storm-based soil loss that is normally caused by sheet and rill erosion at the field scale on the Loess Plateau. This approach can be applied to other areas after calibrating for suitable model parameters.