中国分区大气加权平均温度模型的建立与应用研究

Study on Sub-Regional Weighted Mean Atmospheric Temperature Modeling and Its Application in China

基于中国气象数据网提供的30年平均地表气温与90个探空站点2015年间数据,采用含障碍的扩散插值法进行插值,按3.4 K温度间隔将插值后结果分成8级13区。通过实验分析发现探空数据层数和层顶高度对大气加权平均温度计算精度间存在近似指数函数关系,仅考虑单一因素时,层顶大于9 km或观测层数大于13层后,大气加权平均温度计算值与真值RMS约为1 K;其次,考虑实际样本层顶和层数分布情况,采用层高大于9 km、层数大于8的无错探空数据,分别建立了全国统一的以及分区域的大气加权平均温度回归模型,有效地削弱全国模型在不同地区适用性不一致的影响,拟合残差RMS平均降低了17.8%;最后,应用该分区模型反演了某区域的可降雨量序列,与同期的探空数据计算得到的可降雨量数据进行了比较与分析。

With the daily average surface temperature data from 1980 to 2010 provided by the China Meteorology Data Service Center and 64,131 observations in 2015 from 90 sounding stations,we firstly use the diffusion interpolation with barriers for interpolation and then divide the results into eight grades and thirteen districts at a temperature interval of 3.4 K.By experimental analysis,it is found that an approximately exponential function relationship between factors,i.e.,the number of sounding observation layers and the ceiling height,and the accuracy of weighted mean atmospheric temperature(T_(m)).Considering only one factor,the root-mean-square(RMS)between the calculated T_(m) using resampled and the measured one is about 1 K when the ceiling height is beyond 9 km,or the number of observation layers is beyond thirteen.Then,given the distribution of the actual ceiling height and the number of layers of samples,radiosonde observations with the ceiling height of over 9 km and the number of observation layers greater than eight are applied to build a national and a sub-regional regression model,respectively.In this way,the impact of inconsistent applicability of the national model in different regions is efficiently reduced,and the RMS of fitting residue declines by 17.8%on average.Finally,we apply the sub-regional model to retrieve the time series of precipitable water vapor(PWV)and compare them with contemporaneous PWV data observed by radiosondes.