云南夏季降水异常的影响因子及物理统计预测方法

云南夏季降水年际变化较大,影响因子众多,夏季降水的预测较为困难。使用1965—2017年云南省122个气象观测站的逐日降水资料和NCEP大气环流资料,采用年际增量的方法来预测云南夏季降水。文中基于云南夏季降水年际增量变化规律和影响夏季降水的环流形势及物理过程,选取了6个具有物理意义的预测因子,包括:前期2月南太平洋海温异常、前期2月东亚北部海平面气压异常、前期4月北美500hPa位势高度异常、前期5月太平洋北部海平面气压异常、前期1月印度半岛北部500hPa位势高度异常及前期2月澳洲以南地区200hPa高度场偶极子异常,来建立云南夏季降水预测模型。并对预测模型进行逐年交叉检验和1998—2017年逐年独立样本检验。交叉检验中夏季降水年际增量预测值和观测值的相关系数为0.85,相对均方根误差为8.0%。回报检验中夏季降水年际增量的相对均方根误差为9.1%,63.0%的异常年份预测值能够准确地预报出夏季降水异常。该预测模型有较好的预测能力。

Influencing Factors and Physical Statistical Prediction Methods of Summer Rainfall Anomaly in Yunnan

Because of the obvious interannual variation of summer precipitation in Yunnan and various influencing factors, it is difficult to predict summer precipitation. The daily precipitation observation data from 122 meteorological stations in Yunnan Province from 1965 to 2017 and NCEP atmospheric circulation data and the year to year increment method are used to predict summer precipitation in Yunnan. In order to provide a theoretical basis for the prediction of summer precipitation in Yunnan, it is indispensable to analyze the varying regularities and physical processes affecting the year to year increments of summer rainfall and atmospheric circulation. The prediction model is established based on the method of multiple linear regression analysis. Six predictors that have explicitly physical meaning are selected: the anomaly of the SST (Sea Surface Temperature) in the South Pacific in February, SLP (Sea Level Pressure) in Northeast Asia in February, 500 hPa geopotential height in May over the North America in April, SLP in the northern Pacific in May, 500 hPa geopotential height in the northern India in January, and 200 hPa geopotential height in South Australia in February. Using the above six predictors, the prediction model of summer rainfall is established. In addition, not only the crossing test verification is conducted on the prediction model is with the independent samples from 1965 to 2017, but also the prediction test verification is conducted from 1998 to 2017. In the crossing test verification, the correlation coefficients between predicted and observed interannual increments of summer rainfall is 0.85, and the root mean square relative error is 8.0%. In the prediction test verification, the root mean square relative error of is 9.1%. The prediction model makes good predictions, about 63.0% of the summer rainfall anomaly. The prediction model shows satisfactory forecasting ability.