Abstract:

In the background of climate change, seasonal precipitation forecasting is of great importance for mid-long term allocation and comprehensive utilization of water resources. Given that each forecasting product has its unique advantages and disadvantages, we evaluated and compared the forecasting performance of deterministic and ensemble forecasts of eight models with 1-6 months forecast lead time (FLT) for the upstream of the Three Gorges Reservoir (TGR) to select the optimal model for each sub-basin. Results demonstrate that the optimal model may vary across different sub-basins and FLTs. In general, the ECCC3, ECMWF, CMCC models exhibit superior forecasting performance. Specifically, in the upstream of the TGR, the ECCC3 model performs best when the FLT is 1 month, while the ECMWF model excels for FLTs of 2-6 months. In the upstream of the Jinsha River and the Wudongde Project, the ECCC3 model performs the best. Conversely, in the downstream of the Jinsha River, Wujiang River, Xiangjiaba-Cuntan sub-basin, and Cuntan-Three Gorges sub-basin, the CMCC model outperforms other models. The ECMWF model performs the best for the Minjiang River while the MF model for the Tuojiang River. Based on the forecasting results of the optimal models in different sub-basins, we calculated the forecasting precipitation in the upstream of the TGR. Compared with the forecasting results of single best-performing models, the integration of forecasting result for each sub-basin reduces the root mean squared error by 9.33%-17.86%.

Key words: seasonal precipitation forecasting, deterministic forecasts, ensemble forecasts, evaluation of forecasting results, upstream of the Three Gorges Reservoir