基于降水遥相关型的夏季降水模式预测改进研究

The monthly-seasonal precipitation distribution in China has a spatial teleconnection pattern similar to that of the atmospheric circulation teleconnection in space. Based on the summer precipitation data in China for nearly 57 years, this paper studies the main spatial mode characteristics and their interdecadal variations of the spatial teleconnection pattern of summer precipitation in China, and evaluates and improves the prediction capabilities of the BCC_CSM model, the ECMWF_SYSTEM4 model and the NCEP_CFSV2 model for summer precipitation in China. The research results show that there are four significant spatial teleconnection patterns in the actual situation of summer precipitation in China, namely North China - the lower reaches of the Yangtze River, East China - central northern China, South China - the Yangtze River basin, and Southwest China - central Northeast China. Dynamic models can predict the large-scale precipitation distribution, but their prediction capabilities for the detailed characteristics of precipitation teleconnection between different regions, such as the rain belt, are relatively weak, and there are many false correlations. In order to improve the precipitation prediction skills of the model, a correction scheme is constructed with the precipitation teleconnection pattern in the actual situation as a constraint condition to correct the distribution of the precipitation teleconnection pattern in the model. The results show that the correction can effectively improve the prediction capabilities of the model for central Northeast China and the lower reaches of the Yangtze River. The results of a 4-year hindcast test show that the average anomaly consistency rate of the model prediction increases from 47% to 58%; the average root mean square error decreases from 153 mm to 120 mm; and the average Prediction Skill (PS) score of the trend anomaly comprehensive test increases from 64 to 73.

我国月-季降水分布在空间上存在类似于大气环流遥相关的空间遥相关型。本文基于中国近57年夏季降水资料研究了我国夏季降水空间遥相关型的主要空间模态特征及其年代际变化,评估并改进了BCC_CSM模式、ECMWF_SYSTEM4模式以及NCEP_CFSV2模式对中国夏季降水的预测能力。研究结果显示,中国夏季降水实况中存在华北-长江下游、华东-中国中北部、华南-长江流域、西南-东北中部等4类显著的空间遥相关型。动力模式可以预测大尺度的降水分布,而对于不同区域之间降水遥相关这种雨带细节特征的预测能力则较为薄弱,存在着较多虚假相关。为改善模式降水预测技巧,以实况中的降水遥相关型作为约束条件构建了修正方案,以此来修正模式中的降水遥相关型分布。结果显示,经过修正能够有效地改善模式对东北中部、长江下游的预测能力,4年的回报检验结果显示,模式预测的平均距平一致率从47%提高为58%;平均均方根误差从153 mm减小为120 mm;平均趋势异常综合检验(PS)评分从64提高为73。