基于ECMWF集合预报资料的乡镇温度预报误差订正方法

利用滑动平均法和递减平均法对2013—2014年江西省1 216个乡镇站点ECMWF集合预报2 m温度集合平均产品进行误差订正试验。结果表明:1)滑动平均法和递减平均法对江西地区乡镇温度预报为正的订正效果,订正后的预报准确率大于订正前,并且递减平均法的订正效果要略优于滑动平均法。2)误差订正方法对各时段温度TRMSE的订正能力都随预报时效的增加而减小,对高温预报准确率的提高明显大于低温,对山区预报准确率的提高大于平原,对有规律的预报误差的站点订正效果较好。3)随季节和站点变化的自适应递减平均法的预报结果较各季节和全年定常最优订正系数好,订正方法对秋季温度预报订正能力最强,春季最差。

Study on forecast error correction of township temperature based on ECMWF ensemble forecasting data

Error correction of ECMWF ensemble mean forecast of 2 m temperature during 2013 and 2014 was analyzed using moving average method and decaying averaging method in 1216 stations of township in Jiangxi province. The results showed that the decaying averaging method and moving average method had positive correction for the township temperature forecast, the forecasting accuracy was better after correction, and the forecasting accuracy of decaying averaging method was higher than that of moving mean method. The correction effect of correcting method decreased with forecast period of validity, the improvement of forecasting accuracy for high temperature was obviously higher than that for low temperature, and the improvement for mountainous area was better than for plain. The decaying averaging method had better performance on the site with regular forecast error. The forecast result of self adapting decaying averaging method changed with season and station, and was better than that of annual constant optimal correction. The correction effect for autumn was the best and the worst for spring.