Assessing different roles of baseflow and surface runoff for long-term streamflow forecasting in southeastern China

Accurate long-term streamflow forecast is essential to alleviate and solve the water security problems related to flood and drought disaster warnings. In this study, a new strategy for forecasting monthly streamflow is proposed and four scenarios are designed for the evaluation of different roles of baseflow and surface runoff on performances of long-term streamflow forecasting. The developed models are evaluated at multiple streamflow sites located in the Zhejiang Province of China. The results show that artificial intelligence (AI)-based models with two predictor variables (i.e. baseflow and surface runoff) performed better than that with a single predictor (streamflow) for all the months in a year, and the prediction accuracy of annual peak and monthly streamflow values is improved. Based on the comprehensive evaluations of all the models, the baseflow and surface runoff values are recommended as inputs to AI-based models for an improved prediction accuracy of streamflows.