基于误差权重的乡镇温度订正预报研究

Study on revised forecast method of station temperature based on error weight

使用2018-01-01-2022-05-22修水县共15个代表站点逐日日最高气温、日最低气温实况,以及江西省智能网格预报系统、CMA-GFS、CMA-SH9、ECMWF、JMA等模式高低温预报数据,在分析各模式预报误差的基础上,使用基于平均绝对误差权重的多模式集成订正方法对24 h和48 h高低温预报作出订正。结果表明:基于误差权重的订正方法适用于绝大多数站点,对最低气温的订正效果较好;有13个站点24 h最低气温的订正效果优于其他模式,其中溪口站预报准确率优于CMA-SH9模式的比例达5.9%;有10个站点48 h最低气温的订正效果优于其他模式,其中溪口站预报准确率优于CMA-SH9模式的比例达10.19%;最高气温订正效果优于最低气温,有14个站点24 h最高气温的订正效果优于其他模式,其中黄港站预报准确率优于江西省智能网格预报系统的比例达21.45%;所有站点48 h最高气温的订正效果均优于其他模式,其中余塅站预报准确率优于江西省智能网格预报系统的比例达23.05%。

Using the daily highest and lowest temperature data from 15 representative stations in Xiushui County from January 1,2018 to May 22,2022,as well as the high and low temperature forecast data from Jiangxi Province's intelligent grid forecasting system,CMA-GFS,CMA-SH9,ECMWF,JMA and other models,and analyzing the forecast errors of each model,a multi-mode integrated correction method based on the average absolute error weight was used to correct the 24-hour and 48-hour high and low temperature forecasts.The results indicate that the error weight based correction method is applicable to the vast majority of stations and has a good correction effect on the minimum temperature.The correction effect of the 24-hour minimum temperature at 13 station is better than other models,among which the accuracy of the forecast at Xikou station is 5.9%better than that of the CMA-SH9 model.The correction effect of the lowest temperature at 10 tations within 48 hours is better than other models,among which the accuracy of Xikou station's forecast is better than that of CMA-SH9 model by 10.19%.The correction effect of the highest temperature is better than that of the lowest temperature,and the correction effect of the 24-hour higest temperature at 14 stations is better than other models.Among them,the accuracy of Huanggang Station's forecast is better than that of Jiangxi Province's intelligent grid forecasting system,accounting for 21.45%;The correction effect of the highest temperature at all station within 48 hours is better than other models,among which the accuracy of the forecast at Yuduan Station is 23.05%higher than that of the Jiangxi Province Intelligent Grid Forecast System.