2018年夏季高空数值预报产品在中国的适用性评价

Applicability Evaluation of Upper-Altitude Numerical Forecast Products in China in the Summer of 2018

摘要: 为研究高空数值预报产品在中国的适用性，本文收集了目前在我国气象服务中采用的主要数值预报产品(GRAPES、ECMWF)、气象台站大气探空数据，利用数理统计方法，分析了高空数值预报产品与实况场在2018年夏季的差异，进行不同数值预报产品在不同时效的预测能力综合评估，得出了以下结论：在温度预报上，模式对于北部的预报效果优于南部地区，其中ECMWF的平均偏差、均方根误差和相关系数在不同层次的12 h和24 h预报时效的对比中均小于GRAPES的结果，且在对流层低层的700 hPa和850 hPa，ECMWF结果的优势明显。在位势高度预报上，模式对于北部的预报效果优于南部地区，12 h预报时效的结果优于24 h预报时效的结果，在对流层低层的700 hPa和850 hPa，ECMWF结果的优势明显，且这种优势主要表现在我国西部地区。在风速预报上，GRAPES在平均误差、相关系数和均方根误差上明显优于ECMWF的结果，反映了GRAPES在风速预报中的优势。

Abstract: In order to study the applicability of high-altitude numerical forecast products in China, this paper collects the main numerical forecast products (GRAPES, ECMWF) and atmospheric sounding data of meteorological stations currently used in China’s meteorological services, and analyzes the numerical forecast products using mathematical statistics. Compared with the actual field in the summer of 2018, a comprehensive evaluation of the predictive capabilities of different numerical forecast products at different timelines was carried out, and the following conclusions were drawn: In terms of temperature forecasting, the model’s forecasting effect for the northern region was better than that of the southern region. The deviation, root mean square error, and correlation coefficient were all smaller than the results of GRAPES in the comparison of 12 h and 24 h forecast timeliness at different levels, and the ECMWF results had obvious advantages at 700 hPa and 850 hPa in the lower troposphere. In terms of geopotential height forecasting, the model’s forecasting effect for the north was better than that of the southern region. The results of 12 h forecast time were better than that of 24 h forecast time. At 700 hPa and 850 hPa in the lower troposphere, ECMWF results had obvious advantages, and this advantage was mainly shown in the western region. In wind speed forecasting, GRAPES was significantly better than ECMWF's results in average error, correlation coefficient and root mean square error, reflecting the advantages of GRAPES in wind speed forecasting.