Integration of SWAT and remote sensing techniques to simulate soil moisture in data scarce micro-watersheds: A case of awramba micro-watershed in the upper Blue Nile Basin, Ethiopia

Abstract

Understanding soil moisture at a small scale is beneficial for predicting productivity and management of both rained and irrigated agriculture in mostly smallholder communities. This study aims to accurately represent micro-watershed scale soil moisture using the optimization capability of SWAT (SUFI2) model and soil information derived from Sentinel 2 A level 1 C satellite images with OPtical TRApezoid Model (OPTRAM) and MNDWI. The study was carried in the 700 ha Awramba watershed in the Upper Blue Nile, Ethiopia. Calibration and validation of SWAT were performed using in-situ stream flow data to enable the accurate simulation of water balance components such as soil moisture. The spectral water index was evaluated using MNDWI from the green band (560 nm) and short wave infrared band (2190 nm).The Results were evaluated based on the runoff response n and soil moisture fit to measured values. The runoff fit against the measured data using Nash Sutcliffe Efficiency (NSE) and R2 criteria is 0.7 is and 0.75, respectively. The simulated daily soil moisture against the in-situ constant soil moisture provided NSE = 0.51, R2 = 0.77, RMSE = 0.19 and PBIAS = −0.242. The simulation results indicate that validation of SWAT, OPTRA M and MNDWI models with in situ soil moisture data leads to acceptable accuracy with 0.0027 cm3 cm−3, 0.0022 cm3 cm−3 and 0.034 cm3 cm−3 standard errors, respectively. Furthermore, Sentinel 2A imagery is found to have a higher potential to simulate soil moisture compared to TDR data. The overall study indicates satellite-based soil moisture provides an encouraging pathway to setting up soil moisture-based prediction for smallholder agriculture in Ethiopia.