AI-Based Rainfall-Runoff Modelling for Sustainable Water Management in Potteruvagu Watershed, India

Planning and managing water resources requires accurate prediction of streamflow data since it influences decisions. However, selecting the right model to simulate a watershed is challenging, and researchers typically rely on field testing to identify the most suitable model for their needs. In recent years, artificial intelligence (AI) has gained interest in rainfall-runoff modeling due to its promising adaptability in this area. The research project aimed to predict monthly streamflow in the Potteruvagu river basin, utilizing three distinct AI models, namely, the random forest regression model (RF), artificial neural network (ANN), and k-nearest neighbour regression model (KNN). The study used the calibration period or phase of 1998–2004 and the validation period of 2005–2006, covering the time span of 1998–2006. In contrast to the other AI models utilized in the study, the RF model demonstrated significantly higher accuracy results. During the training or calibration period, the RF model has a Nash-Sutcliffe efficiency (NSE) value of 0.92 and a coefficient of determination (R2) value of 0.94. During the testing or validation period, the RF model likewise produced an R2 of 0.74 and an NSE value of 0.67. However, the outcomes of the alternative models were satisfactory. The outcomes suggest that the RF model could be a useful tool for applications for sustainable management of water resources.