Artificial Intelligence-Based Monthly Rainfall-Runoff Modeling in India’s Bardha Watershed

Recent rainfall-runoff modeling utilizing artificial intelligence (AI) models has shown high adaptability. The purpose of the current study was to simulate monthly runoff in the Bardha river basin using three different AI models. These models included the Long Short-Term Memory (LSTM) deep learning model, the Multilinear Regression Model (MLR), and the Support Vector Regression Model (SVR). The period from 2003 to 2007 was designated as the training phase, while 2008 to 2009 constituted the testing phase, spanning from 2003 to 2009. During the training period, the LSTM model achieved a coefficient of determination (R2) value of 0.90, a Nash–Sutcliffe simulation efficiency (NSE) value of 0.85, a root mean square error (RMSE) value of 15.5 m3/s, and an RMSE-observations standard deviation ratio (RSR) value of 0.36. Correspondingly, in the testing period, the LSTM model achieved an R2 value of 0.87, an NSE value of 0.79, an RMSE value of 7.65 m3/s, and an RSR value of 0.31. Among the AI models employed, the LSTM model stood out for its notably superior accuracy.