六种数值模式对芜湖市地面气温和降水预报的对比检验分析

Contrastive Verification and Analysis of Surface Temperature and Precipitation Forecast Based Six Numerical Model in Wuhu City

对2008年7月至2009年6月JMA、T213、GRAPES、MM5、T639和GERMANY6种数值模式产品对芜湖市的地面气温和降水预报结果进行了对比检验分析，结果表明：各模式对最高气温的预报能力一般，其中JMA、GRAPES、T639相对较好；对最低气温的预报JMA表现突出，而GRAPES在冬季预报能力较好。各模式对于芜湖站降水的预报均无绝对优势，对备等级的降水预报效果各有千秋。对于≥0．1mm降水，JMA和GERMANY在24h时效内TS评分较高，且JMA漏报率很低，GERMANY空报率较低，T639则从48h时效起邢评分最高，且漏报率较低：对于≥10mm降水，T639的乃评分较高且漏报率较低，GERMANY和GRAPES在48h时效内评分也较高且空报率低，JMA在48h时效后空报率较高，成绩较差；JMA、T639和GERMANY对强降水预报能力相对较强，特别是T639对暴雨比较敏感，而各模式在72h之后对强降水的预报能力较差。另外，各模式对降水预报的邢评分均为夏季低、冬季高，空报率均为夏季高、冬季低。

Using six numerical prediction products （JMA, T213, GRAPES, MM5, T639 and GERMANY） from July 2008 to June 2009, ground temperature and precipitation forecast results are respectively verified and analyzed in Wuhu. Results show that firstly, all the six models is moderate on prediction capacity in maximum temperature, meanwhile, the result of JMA, GRAPES and T639 are better. In minimum temperature, JMA performs well all the year and GRAPES is better in winter. Secondly, all the six models are not dominant and have their own characteristics in different scale forecast results of precipitation. For rainfall ≥0.1 mm, the threat scores of 24 hours rainfall forecast made by JMA and GERMANY are high, and the missing forecast rates of JMA is low and the vacancy forecast rates of GERMANY is low. Besides, the threat scores of 48 hours rainfall forecast made by T639 is the highest and its missing forecast rates is low. For rainfall ≥ 10 mm, the threat scores of rainfall forecast made by T639 is high and its missing forecast rates is low. The threat scores of 48 hours rainfall forecast made by GERMANY and GRAPES is also high and its vacancy forecast rates is low. The vacancy forecast rates of JMA is high and its result is bad. JMA, T639 and GERMANY are good for heavy rain and especially T639 is sensitive to rainstorm. But all models are bad for 72 hours heavy rain forecast. Besides, the threat scores of rainfall forecast made by models are all low in summer and high in winter, meanwhile, their missing forecast rates are high in summer and low in winter.