The viability of extended marine predators algorithm-based artificial neural networks for streamflow prediction

Precise streamflow prediction is necessary for better planning and managing available water and future water resources, especially for high altitude mountainous glacier melting affected basins in the climate change context. In the current study, a novel hybridized machine learning method, extended marine predators algorithm (EMPA)-based ANN (ANN-EMPA), is developed for streamflow estimation in the Upper Indus Basin, a key mountainous glacier melt affected basin of Pakistan. The prediction accuracy of the novel metaheuristic algorithm (EMPA) was also compared with several benchmark metaheuristic algorithms, including the marine predators algorithm (MPA), particle swarm optimization (PSO), genetic algorithm (GA), and grey wolf optimization (GWO). The results revealed that the newly developed hybridized ANN-EMPA outperformed the other hybrid ANN methods in streamflow prediction. ANN-EMPA improved the root mean square error, mean absolute error and Nash–Sutcliffe efficiency of ANN-PSO by 4.8, 4.1 and 0.5%, ANN-GA by 6.2, 5.6 and 0.6%, ANN-GWO by 3.7, 4.4 and 0.5%, and ANN-MPA by 3.2, 7.5 and 0.3%, respectively. Month number (MN) was also examined as input to the best models to assess its impact on the prediction precision. Obtained results showed that MN generally slightly improved the models’ accuracy. Results also showed that temperature-based inputs provided better prediction accuracy than only streamflow as inputs. Therefore, the ANN-EMPA model can be used for streamflow estimation from temperature data only when long-term streamflow data is unavailable.