2017 年乌鲁木齐区域数值预报业务系统预报性能检验和评估

Performance Tests and Evaluations of DOGRAFS during 2017

摘要:

基于乌鲁木齐区域数值预报业务系统，运用MET检验工具，对2017年各季节DOGRAFSv1.0预报性能进行客观检验。结果表明：（1）2m温度日间预报温度整体偏低，夜间多数站点预报温度偏高；冬季预报温度偏高，其他三个季节温度预报整体偏低。10m风速冬季模拟性能最差，春季次之；所有季节风速预报均偏大。（2）夏季、秋季高空温度预报误差小，在3.0℃以内，冬季误差最大，温度预报整体呈冷偏差；不同季节高空位势高度随高度增加误差增大，误差约在6.5～12.0gpm，预报高度比实际高度偏低；不同季节高空U、V风随高度增加误差先增大后减小，均方根误差分别为2.4～6.2m/s和1.8～5.2m/s，U风预报整体比实况偏小，V风预报整体比实况偏大。（3）冬季大阈值降水漏报率较高，12.1mm阈值降水Bias评分仅为0.2，秋季大阈值降水空报率较高，12.1mm阈值降水Bias评分在2.0以上，夏季空、漏报率较低；在新疆地区，四个时段中14～20 BJT 、20～次日02 BJT空报站点数多于漏报，14～20 BJT空报率最高，02～08 BJT漏报率最高，08～14BJT晴雨预报以漏报为主；日间Ts评分高于夜间。

Abstract:

The forecast performance of DOGRAFS v1.0 in 2017 is objectively verified seasonally by MET. The results show that: (1) The daytime temperature is underestimated overall while the nighttime temperature at most of the sites is overestimated. The forecasted 2 m temperature is overvalued in winter and undervalued in the other three seasons as a whole. The predicted 10 m wind speed in 2017 is always higher than the corresponding observation with the maximum root mean square error in winter and followed by spring. (2) The modeled high-altitude temperature is generally colder than the observed value. It is worst simulated in winter and well predicted in summer and autumn with the RMSE less than 3 degrees centigrade. The predicted error of high-altitude geopotential height increases with the height increasing, and the RMSE is about 6.5 ~ 12.0 gpm. Furthermore, the forecasted geopotential height is lower than the actual height. The error of high-altitude U, V wind increase firstly then decrease with the height increasing in different seasons. The RMSE of high-altitude U, V wind are 2.4 ~ 6.2 m/s and 1.8 ~ 5.2 m/s respectively. The U component of wind speed is underestimated while the V component of wind speed is overestimated. (3) The missed event ratio of heavy rainfall in winter is high while the false alarm ratio of high threshold precipitation is high in autumn. In Xinjiang, the false alarm sites are more than the missed event sites from 14 to 20 BJT and 20 to 2 BJT. The false alarm ratio from 14 to 20 BJT is the highest and the missed event ratio from 02 to 08 BJT is the highest. The Ts score during daytime is higher than at night.