用一维变分方法同化GMS-5反演湿度场及其在降水预报中的应用分析

在应用一维变分方法对GMS-5卫星资料反演的相对湿度场进行了同化分析试验的基础上,进一步分析了通过一维变分方法在MM5模式初始湿度场中融入GMS-5反演湿度场后,对降水预报的影响.分析试验结果表明,应用一维变分方法同化GMS-5反演相对湿度场后所得到的分析场,其均方根误差(rm se)在7个标准等压面上都比背景场要明显减小.均方根误差值在250 hPa下降幅度最大,达到7.1,这相当于12-h预报场rm se下降了48%;下降幅度最小的等压面高度是925 hPa,也达到了1.2,相当于12-h预报场rm se下降了15%.背景湿度场高湿区通常呈带状分布,普遍平滑了中尺度信息.运用一维变分方法融入了卫星信息后,分析场不仅明显改善了背景场湿区偏湿、干区偏干的现象,同时也很好地揭示出了三维湿度场的中尺度细微结构.个例分析表明,将一维变分同化后的GMS-5反演湿度场作为MM5模式的初始湿度场,与直接用NCEP资料的预报结果相比,MM5模式预报的降水强度、降水范围和降水强中心位置均有较明显的改善,更接近于实况观测.

Assimilation of GMS-5 Humidity Retrievals with a One-dimensional Variational Analysis Technique and its Application in the Forecast of Precipitation

Using one-dimensionally variational analysis technique,relative humidity(RH)fields retrieved from GMS-5 imagery are assimilated in this paper.Based on the analyzed RH fields resulted from the assimilation,the paper discusses the effect of initial moisture fields on the MM5 forecast of precipitation when GMS-5 retrievals are assimilated.It is found that there is a reduction of background field root-mean-squareerror(rmse) on each pressure level of seven levels(925 hPa,850 hPa,700 hPa,500 hPa,400 hPa,300 hPa,250 hPa).The maximum reduction of root-mean-square-error(rmse) is 7.1 at the 250 hPa level,which is a reduction of 48 % in the 12-h forecast root-mean-square-error(rmse).The minimum is 1.2 at the 925 hPa,which is an effective reduction of 15 % in the 12-h forecast rootrnean-square-error(rmse).The background field is so smooth that it covers some mesoscale information.When assimilating the GMS-5 retrievals by a one-dimensionally variational analysis technique,the analyzed field is not as smooth as the background field,which is desirable.The analyzed field avoids some dry zones from too dry and some humid ones from too moist,which appear in the background field.The analyzed field can give better information of moisture,which is one of MM5 initial condition for the forecast of precipitation. Two cases about the MM5 forecast of precipitation indicate that the GMS-5 satellite retrievals have a positive effect on MM5 forecast of precipitation.Compared with the forecast of precipitation when the NCEP data is used as the initial condition of MM5,the MM5 forecast quality of precipitation is better as the GMS-5 retrievals are assimilated by a one-dimensionally variational analysis scheme,and the distribution and precipitation of central rainfall is close to the actual.