基于LPSC算法的逐小时格点气温客观预报及检验评估

Objective forecast of hourly grid temperature based on the LPSC algorithm and its evaluation

基于中国气象局陆面数据同化系统(Land surface Data Assimilation System of China Meteorological Administration,CLDAS)逐小时气温实况融合数据,检验评估了ECMWF、CMA-MESO-3km不同尺度模式对甘肃省逐小时气温的预报性能,并利用低频滑动平均订正算法(LPSC)对模式的系统性误差进行订正;同时对SCMOC和订正后两种模式的逐小时气温预报效果进行了统计对比。结果表明:(1)ECMWF、CMA-MESO-3km模式对甘肃省逐小时气温的预报具有相对稳定的系统性误差,夜间预报准确率明显低于白天,主要表现为夜间预报显著偏高,白天为小的负偏差。(2)LPSC算法能够有效改善ECMWF和CMA-MESO-3km对甘肃省逐小时气温预报的系统性误差,订正效果显著。订正后ECMWF、CMA-MESO-3km的预报准确率分别较模式本身提高了20.24%、20.25%,平均误差减小至±0.3℃之内;空间分布亦表明,订正后全省平均误差均明显降低至±2℃之内。(3)同类产品对比检验表明:订正后ECMWF、CMA-MESO-3km两种逐小时气温预报产品的预报效果整体上均优于SCMOC,预报准确率分别较SCMOC高20.65%、13.55%,平均绝对误差在各个时次也明显低于SCMOC。技巧评分的空间分布表明,订正后ECMWF在全省大部分地方均为正技巧,其中酒泉南部山区可达80%以上;而订正后CMA-MESO-3km的预报效果各个季节分布存在差异,主要体现在陇中和陇东南地区,冬春季以弱的正技巧为主,夏秋季基本为负技巧。另外,业务应用结果表明,对于转折性天气过程,使用该方法需要特别注意。

In this study,based on the hourly temperature fusion data collected by the Land Surface Data Assimilation System(CLDAS)of the China Meteorological Administration,the respective performances of the ECMWF and CMA-MESO-3km models for hourly temperature forecast in Gansu Province were tested and evaluated,and the low-frequency moving average correction algorithm(LPSC)was used to correct the models’systematic errors.In addition,the hourly temperature forecast effects of the two revised models and SCMOC were statistically compared.The results show the following:(1)The ECMWF and CMA-MESO-3km models have relatively stable systematic errors in the hourly temperature forecast of Gansu Province,while the forecast accuracy at night is significantly lower than that in the daytime,which is mainly characterized by significantly higher forecast at night and weak negative deviation in the daytime.(2)The systematic error of the hourly temperature forecast of the ECMWF and CMA-MESO-3km models in Gansu Province can be effectively improved with the LPSC algorithm,and the correction effect is significant.After revision,the forecast accuracies of ECMWF and CMA-MESO-3km are respectively improved by 20.24 and 20.25%,and the mean error is reduced to within±0.3℃.The spatial distribution also showed that the mean error provincewide was reduced to within±2℃.(3)The comparison test of similar products show that the forecast effects of hourly temperature of the revised ECMWF and CMA-MESO-3km were better than that of SCMOC on the province as a whole.The forecast accuracies were respectively 20.65 and 13.55%higher than that of SCMOC,and the mean absolute error was significantly lower than that of SCMOC at each time.The spatial distribution of the skill score shows that the revised ECMWF is positive in most places throughout the province,more than 80%of which is in the southern mountainous area of Jiuquan.However,there are some differences in the forecast effect of the revised CMA-MESO-3km,which are mainly reflected in central and southeastern of Gansu Province,with weak positive technique predominating in winter and spring,and basically negative technique in summer and autumn.The results of the special application show that special attention should be paid to using this method for transforming weather processes.