中国地面气象要素格点融合业务产品检验

对2018年5 km分辨率中国地面气象要素2 m温度、10 m风速和24 h累积降水格点融合产品进行非独立和独立检验。非独立检验结果表明:(1)相比于站点观测,2 m温度格点融合产品整体偏暖,各月平均均方根误差在1℃左右,35℃以上高温和-20℃以下低温天气时均方根误差分别在1℃和2℃以上。(2)10 m风速格点融合产品可准确地描述0~2级风速,但对3级以上,特别是6级以上大风风速描述能力偏弱,主要表现为比实际偏小。(3)卫星-地面观测的二源融合和卫星-雷达-地面观测的三源融合降水格点产品在0~0.1 mm降水区间出现降水面积过大的现象;随着降水量级的增加,两种产品的均方根误差和平均偏差均随之增加,主要表现为降水融合产品的量级比观测偏小。相对而言,三源融合降水格点产品的准确性优于二源融合产品的。独立检验结果表明,三种要素的检验指标随时间或阈值的变化趋势与非独立检验基本一致,且更能表明格点融合产品与观测之间的偏差。主要是因为独立检验中使用到的观测均未参与格点融合产品的制作过程。综上所述,中国地面气象要素格点融合产品对一般天气描述较好,但在高低温、大风或强降水等极端天气时误差较大。

The Verification of Gridded Surface Meteorological Elements Merging Product in China

Independent and non-independent validations have been done to evaluate the merged 5 km resolution gridded products in China in 2018,including 2 m temperature,10 m wind and 24 h accumulated precipitation.The non-independent validation shows that:(1)compared with the in situ station data,the 2 m temperature product has a warm bias:the monthly-averaged Root-Mean-Square-Error(RMSE)is around 1 ℃,and for the hottest(≥35 ℃)and the coldest(≤-20 ℃)days,the RMSE is above 1 ℃ and 2 ℃ respectively.(2)The 10 m wind product can precisely re-produce the wind within grade 0~2,but it will underestimate the wind speed within grade 3 and after,especially over grade 6.(3)the precipitation product will over-estimate the rainfall area when the rainfall is between 0 to 0. 1 mm;and the RMSE and Mean-Error(ME)will increase along with the precipitation increasing,which is due to the under-estimation of the precipitation compared with the observation.Generally,the precipitation product merged from three sources(radar,satellite,and gauge)is better than that from two sources(satellite and gauge). The independent validation shows that the monthly RMSE and ME of the2 m temperature,10 m wind and 24 h accumulated precipitation data are very similar to the non-independent validation. And the independent validation can point out the bias,which are discussed in the non-independent validation,more clearly. This is mainly because the in situ station data used in the independent evaluation are not merged in the product. To summary,the merged gridded products have small biases in ordinary days,and larger bias in severe weather like very hot and cold days,windy or rainy days.