基于MM5模式的精细化MOS温度预报

Fine MOS Temperature Forecast Based on MM5

利用2002年9月到2003年8月MM5模式每隔1 h的站点基本要素预报场和物理量诊断场资料,以及相应时段内宁夏25个测站的温度自记观测资料,同时采用多元线性和逐步回归2种MOS统计方法,预报宁夏25个测站48 h逐时温度。通过对2004年6月至2005年5月的预报效果检验,结果表明:MOS方法对宁夏48 h逐时温度有较强的预报能力,当天气形势变化较平稳时,MOS预报结果稳定,平均绝对误差控制在2℃之内;当有明显的冷空气活动时,误差变率起伏波动较大,预报结果不稳定,但与原MM5模式直接输出结果相比,MOS预报水平有明显提高,24 h极端温度TS评分个别月接近甚至超过预报员。

Based on hourly data of both basic elements forecasting fields and physical quantity diagnosing fields of MM5, and the temperature records of 25 stations in Ningxia from September 2002 to August 2003, multivariate and stepwise regressions were adopted to set up the MOS model of 48 hours hourly temperature forecast in 25 stations in Ningxia. The effect test conducted from June 2004 to May 2005 indicates that the MOS method has a good capability in forecasting 48 h hourly temperature in Ningxia. As weather changed smoothly, the forecast result of MOS was stable, and mean absolute errors were within 2 ℃ ; When the cold air invaded, the forecast result was unstable but it was much better than that of original product of MM5. Some of the TS grades of 24 hours extreme temperature forecast are close to or even better than that of forecasters.