Abstract
In this study, three ensemble and decomposition methods (DMs), i.e., empirical mode decomposition (EMD), ensemble empirical mode decomposition (EEMD), and improved complete ensemble empirical mode decomposition with additive noise (ICEEMDAN) were coupled with artificial intelligence and machine learning based method AI-ML, i.e., multilayer perceptron (MLP), support vector regression (SVR) to develop six fundamental hybrid models to predict streamflow with one-month lead time. Developed models in this study were categorized into runoff models (RMs) and rainfall–runoff models (RRMs). Results indicated that (i) among standalone models (SMs), support vector regression (SVR) performs better than multilayer perceptron (MLP), (ii) decomposition methods (DMs) can improve the accuracy rate of the standalone models (SMs), and (iii) rainfall–runoff models (RRMs) have shown great accuracy throughout the investigation as compared to runoff models (RMs). To compare model performances, flow hydrographs (FHGs) were generated, and five performance evaluation criteria (PEC) were used to quantify the model precision. Two-step verification methods, i.e., extreme value analysis (EVA) and least value analysis (LVA) approaches, were proposed to verify the performances. Among all developed hybrid models (HMs), i.e., EMD-(MLP, SVR), EEMD-(MLP, SVR), and ICEEMDAN-(MLP, SVR), rainfall–runoff ICEEMDAN-(SVR) model was selected as an optimal model with MAE (59.56), RMSE (91.82), R (0.97) MAPE (8.75), and NSEC (0.97) for Mangla watershed, Pakistan.