Assessing climate change and human impacts on runoff and hydrological droughts in the Yellow River Basin using a machine learning-enhanced hydrological modeling approach.

Abstract

Analyzing the impacts of climate change (CC) and human activities (HA) on hydrological events is important for water resource management. This study quantifies the impacts of CC and HA on runoff and hydrological drought characteristics (HDC) in the Yellow River Basin (YRB) of China. Trends and abrupt change points in runoff at 16 hydrological stations were detected. The Soil and Water Assessment Tool (SWAT), Random Forest (RF), and five bias-correction models, including SWAT_RF_bias3 which integrated SWAT outputs with RF using maximal precipitation and temperature inputs, were evaluated for their efficacy in monthly runoff simulation. The "simulated-observed" method was employed to assess the contributions of CC and HA to runoff and HDC variations. Results indicated a general decrease in runoff across the stations during 1961-2016. SWAT_RF_bias3 emerged as the superior model, highlighting the importance of high precipitation in the headwater region and near the main channel of the midstream for accurate runoff simulation. HA was found to contribute significantly more (68-95 %) to runoff reductions than CC. Additionally, CC predominantly influenced the frequency decrease in severe and extreme hydrological droughts, while HA was the main driver behind the increased magnitude and duration of extreme droughts. These findings underscore the complex interplay between CC and HA in water resource management and the effectiveness of bias-correction models in enhancing hydrological simulations in the YRB.