Combined Terrestrial Evapotranspiration Index prediction using a hybrid artificial intelligence paradigm integrated with relief algorithm-based feature selection

Drought is a common environmental disaster strongly influenced by the potential production of agricultural products, lack of water resources, and yields destructive effects on the economy. In this study, the prediction of a novel monthly Combined Terrestrial Evapotranspiration Index (CTEI) was considered as a measure of all three types of drought (meteorology, hydrology, and agriculture) in the Ganga river basin (GRB) over the period of 14 years. For this purpose, a combination of hydro-meteorological and satellite-based data, including 11 input variables, was implemented. A new Artificial Neural Network (ANN) integrated with a Whale Optimization Algorithm (WOA-ANN) and a Relief algorithm-based Feature Selection (FS) method was applied to simulate the monthly CTEI index and find the optimum input combinations. Besides, the standalone ANN and Least Square Support Vector Regression (LSSVR) models were examined to validate the WOA-ANN performance. The results indicated that WOA-ANN with (R = 0.9391), (RMSE = 0.241), (WI = 0.968), and (U-95% = 0.669) had a high ability to predict CTEI and reduced the RMSE in the ANN and LSSVR models by 27% and 30%, respectively. WOA-ANN had the best predictive performance, followed by LSSVR and ANN models, respectively, concerning various graphical validations and diagnostic analyses. Besides, the outcome of this research, despite fewer parameters (seven parameters), considerably outperformed the study of (Elbeltagi et al., 2021) with (R = 0.9055 and RMSE = 0.33) accuracy employing 11 inputs.