Vol. 10, Issue 4 (2021)

Author(s):
Daniel Prakash Kushwaha and Anil Kumar

Abstract:
Soil erosion and nutrient losses from hillslopes are the severe problems. North Western Himalayan Region (NWHR) of India always suffers from severe water erosion every year. This study was conducted in premises of Pantnagar, a town situated in Uttarakhand, NWHR. A field experiment was performed during monsoon season in 2018 under natural rainfall condition. Two soil amendments viz. biochar and anionic polyacrylamide (PAM) were used into the soil to reduce severe soil erosion and loss of major soil nutrients (N-P-K) from the plots. Twelve rainfall events were recorded which created runoff anyhow on the land surface and rainfall, runoff, sediment yield and major nutrient losses (N-P-K) were taken into account for the development of two models scenarios viz. sediment yield and nutrient loss. Two modeling techniques viz. multi-layer perceptron based artificial neural network (MLP-ANN) and multiple linear regression (MLR) were used. Every model was tested against quantitative performance evaluation criteria viz. Root mean square error (RMSE), Percent bias (PBIAS), Karl Pearson’s coefficient of correlation (CC) and Nash-Sutcliffe efficiency (NSE) along with line diagram and scatter plots. Results demonstrated that MLR is better than MLP-ANN to simulate nutrient loss modeling with CC value ranging from 0.955 to 0.962 during testing period, whereas MLP-ANN was found better than MLR in simulating sediment yield with CC value as 0.938 during testing period. Results also revealed that MLP-ANN was found better performer for simulating non-linear and inconsistent data of sediment yield, whereas MLR was found better performer in case of linear data.