Estimating water runoff from downscaling climate change scenarios using soil and water assessment tool and machine learning: A case study of Lake Tana basin in Ethiopia

Abstract

Water is a fundamental natural resource necessary for life, and contributes to the development of the nation, including urbanization, shifting agricultural practices, and deforestation. These factors have both direct and indirect impacts on the watershed. This study presented machine learning for statistical downscaling as a means of hydrological modeling. A statistical downscaling model was created using a global circulation model from the community climate system model (version 4.0), and compared to different machine learning techniques, including linear regression, gaussian process, and support vector machine. The soil and water assessment tool (SWAT) was used to model climate-induced runoff and downscaling procedures. The output climate scenarios of the machine learning model were incorporated into SWAT to simulate water runoff in the study area of Lake Tana basin, Ethiopia. The simulation results of SWAT water runoff under deep learning climate conditions demonstrated the highest performance. The results could contribute to the hydrological analysis and improve the quality of statistical downscaling.