Seasonal rainfall forecasting for the Yangtze River basin using statistical and dynamical models

Abstract

Summer monsoon rainfall forecasting in the Yangtze River basin is highly valuable for water resource management and for the control of floods and droughts. However, improving the accuracy of seasonal forecasting remains a challenge. In this study, a statistical model and four dynamical global circulation models (GCMs) are applied to conduct seasonal rainfall forecasts for the Yangtze River basin. The statistical forecasts are achieved by establishing a linear regression relationship between the sea surface temperature (SST) and rainfall. The dynamical forecasts are achieved by downscaling the rainfall predicted by the four GCMs at the monthly and seasonal scales. Historical data of monthly SST and GCM hindcasts from 1982 to 2010 are used to make the forecast. The results show that the SST‐based statistical model generally outperforms the GCM simulations, with higher forecasting accuracy that extends to longer lead times of up to 12 months. The SST statistical model achieves a correlation coefficient up to 0.75 and the lowest mean relative error of 6%. In contrast, the GCMs exhibit a sharply decreasing forecast accuracy with lead times longer than 1 month. Accordingly, the SST statistical model can provide reliable guidance for the seasonal rainfall forecasts in the Yangtze River basin, while the results of GCM simulations could serve as a reference for shorter lead times. Extensive scope exists for further improving the rainfall forecasting accuracy of GCM simulations.