Calibration and validation of hillslope runoff and soil loss outputs from the Water Erosion Prediction Project model in Minnesota agricultural watersheds

There is growing interest in studying the impact of alternative agricultural management practices on runoff and soil loss under future climate change scenarios. In order to address this interest, it is important to demonstrate that runoff and soil loss can be accurately simulated under existing climates based on comparisons between modeled and experimental results. This study calibrates and validates the Water Erosion Prediction Project (WEPP) model to quantify the accuracy of predicting growing season runoff and soil erosion in agricultural hillslopes based on comparisons with experimental data from five Minnesota hydrologic unit code 12 watersheds. In order to accurately predict runoff and soil erosion in each watershed, the baseline effective hydraulic conductivity (K_be), interrill and rill erodibility (E_IR and E_R), and monthly precipitation standard deviations (P_stdev) were calibrated in WEPP using observed runoff and total suspended solids data from five Minnesota Discovery Farms field sites. Before calibration, Nash–Sutcliffe model efficiency (NSE) and percent bias (PBIAS) values for predicted versus measured monthly average total runoff (R_avg-T), runoff ratios (RR_T), and total soil loss were generally not in acceptable ranges. After calibration, the NSE values showed very good fits between measured and predicted monthly R_avg-T (0.64–0.98), RR_T (0.66–0.93), and soil loss (0.58–0.80). PBIAS values were also within acceptable ranges for R_avg-T and RR_T (±25%) and soil loss (±55%), except for RR_T at site BE1. NSE and PBIAS values during validation were within acceptable ranges, except for RR_T at site BE1. These findings suggest that the WEPP hillslopes calibrated in this study are sufficiently robust to accurately predict monthly runoff and soil erosion in Minnesota agricultural fields during the growing season.