Analysis on the station-based and grid- based integration for dynamic-statistic combined predictions

Summer precipitation prediction in China is important to society and economic development, but still a challenging issue in current meteorological studies, due to the uncertainty of the climate system. This paper developed a weighted integration method founded on dynamic-statistic prediction methods. The advantages and disadvantages of station-based and grid-based integration on summer precipitation prediction, along with the underlying reasons, are respectively analyzed by calculating the anomaly correlation coefficients (ACC), prediction score (PS), and root-mean-square errors (RMSE). The main manifestations indicated that 1) The weighted integration method can provide better skill of summer rainfall prediction in China than the single dynamic-statistic combined prediction method; 2) For the station-based integration, the 10-year ACC mean of summer precipitation prediction is 0.098–0.106, passing the 90% significance level, which is also obviously higher than that of the grid-based integration. The station-based integration has a higher symbol consistency ratio (SCR) than the grid-based integration, and the probability distribution of anomaly percentiles of station-based integration is closer to the observation, which causes the corresponding PS score to be 69.2–70.0 and higher than the grid-based integration. The independent sample validation of 2020 and 2021 further confirmed that station-based integration had a higher ACC than grid-based integration. It is indicated that station-based integration may have better performance in improving the accuracy of the summer precipitation prediction in China, which needs to be deeply considered in the scientific study and real prediction issues.