Integrated forecasting of monthly runoff considering the combined effects of teleconnection factors

This is the authors' abstract. We don't add key points for this paper.

The Danjiangkou Reservoir, China Teleconnection factors significantly influence long-term runoff forecasting. Existing methods primarily focus on the overall impact of these factors on runoff, often neglecting the consideration of extreme values to some degree. Additionally, the interaction among different teleconnection factors can influence runoff forecasting results, which can be defined as the combined effects of these teleconnection factors. This study examines the combined effects of teleconnection factors on the differences in global and extreme runoff forecasting performance and proposes a parallel forecasting strategy to further improve long-term runoff forecasting accuracy. The research results indicate that the proposed model in this study demonstrates exceptional performance, with the Nash-Sutcliffe Efficiency (NSE) improving by 4.1 % and the Root Mean Square Error (RMSE) decreasing by 6.1 % compared to the model that does not consider the combined effects of teleconnection factors. Notably, the forecasts exhibit significant enhancements in predicting extreme runoff values. Furthermore, the combined effects of the Solar Flux Index, the North American Polar Vortex Intensity Index, and the Western Pacific Subtropical High Intensity Index may be a crucial cause of extreme runoff in the Danjiangkou Reservoir. The findings of this study can provide valuable insights for long-term predictions. ● The 5-month-ahead Western Pacific Subtropical High Ridge Position Index has the greatest contribution to Danjiangkou monthly runoff, per XGBoost-SHAP. ● The Solar Flux Index may reduce the contributions of the North American Polar Vortex and Western Pacific Subtropical High Intensity Indices to runoff. ● The combined forecasting strategy in this study effectively improves monthly runoff forecasting accuracy.