华南西部欧洲细网格2m温度预报误差分析

Analysis on error of 2-meter-high temperature prediction from the ECMWF fine grid of West in South of China

利用2011年7月-2012年7月欧洲中期天气预报中心(ECMWF)模式细网格地面2m温度和广西区域自动站气温观测资料,对比分析了EC模式细网格2m温度24小时时效内在华南西部地区不同季节、不同天气系统影响下的预报性能。结果表明:(1)全年平均而言,低温预报误差整体较小,预报准确率达77.7%,高温预报误差变化较大,准确率只有32.8%,低温预报准确率比高温预报准确率高44.9%,低温预报具有较高的参考价值。(2)不同季节高温低温预报差异明显,在夏季(6月-8月)低温预报的准确率达80%,但最高温度的预报准确率只有10%左右;在冬季,最低温度准确率下降到65%左右,而最高温度准确率相反,上升至50%左右。(3)不同地理区域预报性能差别较大:最高温度预报1-3月桂西可信度较高,达60%,4-5月和11-12月四个月只有桂东部分地区的预报具有一定的参考价值。(4)从全年误差分布来看,高温预报在冬季是误差小的所占比重大,误差大的比重小,夏季的则相反,春秋的误差等级分布的较为均匀,每个等级所占比重相似。低温预报则分布的比较均匀,全年基本都是误差越小占比重越大,只是冬季误差小的比重相对较小。5)不同天气形势的温度预报性能亦不同:冬春季冷空气(锋面)影响过程和春季低温阴雨过程的高温预报有一定的参考价值;夏季区域性暴雨过程和副热带高压影响过程的高温预报参考价值较低,误差平均达31%和5.8%,可作为预报主观订正的幅度参考值,四种天气型的低温预报准确性都较高,达到70%以上。

Based on the 2-meter-high temperature data from the ECMWF fine grid and the temperature data of Guangxi automatic station from July 2011 to July 2012, under different seasons and synoptic systems, the forecasting performance of 2-meter-high temperature by the ECMWF fine grid within 24 h of west in south of China. Results show that： 1） annual average,the whole error of low-temperature forecast is small with 77% predicting accuracy, while high-temperature forecast has bigger errors with 32.8% accuracy, thus the low-temperature forecast has more reference value. 2） During different seasons, the differences between maximum-temperature and minimum-temperature forecast is obvious, accuracy of minimum-temperature forecast during the summer（ June to August） reaches 80%, while the prediction accuracy of the maximum-temperature is only about 10%; By contrast, the accuracy of minimum-temperature forecast is down to around 65%, and the maximum-temperature forecast accuracy rose to 50%. 3） The differences of forecast performance among different geographical regionsare bigger： the maximum-temperature forecast of western in Guangxi during Jan. is more reliable with60% accuracy. By comparison, forecast for eastern in Guangxi has certain reference value during April to May and Nov. to Dec.. 4） Referring to high-temperature forecast, in winter the proportion of small error is bigger and the proportion of big error is smaller, while in summer the distribution of errors is contrary,in addition, the error in the spring and autumn is well-distributed with the similar proportions of each grade. By contrast, the distribution of low-temperature forecast errors is rather balanced, smaller errors with bigger proportion, only in winter, there are smaller errors. 5） The forecast performance of different weather are different： the low-temperature forecast for the process of cold air affecting in winter and the process of cold and rainy weather in spring have some reference value; but the reference value of hightemperature forecast for the process of summer regional rainstorm and the process of subtropical high influencing are lower, with the error respectively up to 31% and 5.8%, which can be the reference value for the subjective correction forecast. The accuracy of low-temperature forecast for all four types of weather are high reaching 70%.