2025-07-04 HYDROLOGICAL SCIENCES JOURNAL 2025 70(卷), 9(期), (1517-1527页)
The Weihe River Basin, a typical watershed in the Loess Plateau, is used as the research object to examine the regulation of runoff by vegetation. Here, we employed two deep learning algorithms, FNN and LSTM, for runoff modelling and adopted two modelling schemes - one incorporating NDVI and the other excluding NDVI - to examine the role of vegetation in runoff simulation. Finally, the optimal algorithm and the most appropriate modelling scheme were applied to simulate runoff variations under different scenarios. The results showed that the overall trend of NDVI in the Weihe River Basin is increasing, and the overall trend of runoff is decreasing. Incorporating NDVI into the models will significantly improve the simulation accuracy. We also found that intra-annual vegetation cover variations impact runoff processes, with wet and normal year runoff highly sensitive to these changes. Our study highlights the role of vegetation in regulating runoff.