Futuristic Streamflow Prediction Based on CMIP6 Scenarios Using Machine Learning Models

Accurate streamflow estimation is vital for effective water resources management, including flood mitigation, drought warning, and reservoir operation. This paper aims to evaluate four machine learning (ML) algorithms, namely, Long Short-Term Memory (LSTM), Regression Tree, AdaBoost, and Gradient Boosting algorithms, to predict the futuristic streamflow of the Swat River basin. Ten General Circulation Models (GCMs) of Coupled Model Intercomparison Project Phase 6 (CMIP6) under two Shared Socioeconomic Pathways (SSPs) 245 and 585 were used for futuristic streamflow assessment. The ML models were developed using maximum temperature, minimum temperature, and precipitation as the input variables while streamflow as the target variable. The performance of ML models was assessed via statistical performance indicators, namely the coefficient of determination (R 2 ), mean square error (MSE), mean absolute error (MAE), root mean square error (RMSE), Nash Sutcliffe Efficiency (NSE) and Percent BIAS (PBIAS). The AdaBoost exhibits exceptional performance (R 2 : 0.99 during training, 0.86 during testing). The futuristic streamflow projection shows an increase in mean annual streamflow between 2050 and 2080 s from 3.26 to 7.52% for SSP245 and 3.77–13.55% for SSP585. ML models, notably adaboost, provide a reliable method for projecting streamflow, will assist in hazard and water management in the area.