Artificial Intelligence Based Runoff and Sediment Yield Prediction under Climate Change

Abstract

To minimize the impact of food shortages and to ensure the safety of hydraulic structures and newlineinfrastructure, it is essential to make accurate long-term as well as daily rainfall runoff and newlinesediment yield predictions. Artificial intelligence-based machine learning and data science newlinemodels have been showing great success in a wide range of commercial fields, including the newlinecommercial industry, but they are generally struggling in many scientific fields, including newlinehydrology, despite showing great success in many commercial areas. This study investigates newlinethe capability of Artificial intelligence based multilayer perceptron (MLP), support vector newlinemachine (SVM), genetic algorithm-based hybrid multilayer perceptron (MLP-GA), genetic newlinealgorithm-based hybrid support vector machine (SVM-GA) models in daily rainfall runoff as newlinewell as sediment yield modeling at Ghatora, Jondhara and Sigma sites of Seonath River basin newlinein the Chhattisgarh. Their results are compared with a observed runoff and sediment yield newlinedata, rainfall and runoff data gathered from Ghatora, Jondhara and Sigma sites of Seonath newlineRiver basin in the Chhattisgarh, were divided into two parts, and were model developed and newlinevalidated considering each part by swapping training (70% data) and testing datasets (30% newlinedata). A novel approach Gamma Test (GT), is used to selection of the ideal input variables in newlinea data set. Artificial intelligence based all four model were compared with respect to seven newlinestatistics, Nash Sutcliffe model Efficiency coefficient (NSE), Willmott Index of Agreement newline(d), mean absolute error (MAE), mean bias error (MBE), root mean square error (RMSE), newlinePearson correlation coefficient (PCC) and coefficient of determination (R2 newline). According to the newlineresults of the comparison, the newly developed hybridized SVM-GA method performed newlinebetter than MLP, SVM and MLP-GA in predicting runoff and sediment yield than the other newlinemethods. An impact assessment under the climate change scenario developed by the newlineIntergovernmental Panel on Climate Cha