夏季地面温度场月预报的统计模式

Statistical Models for Monthly Surface Temperature Predictions in Summer Season

用１５年北半球夏季地面温度的资料分析了月平均场与该月前后一周左右的逐日平均场在时间和波数上的相关。发现越新、越靠后的日平均场与该月的平均场相关越强。用第０天外推的未来３０天平均场与实况的相关比用前３０天平均场作外推的相关要显著得多。由此得出结论认为各种预报方法的效果检验应与前者定义的惯性预报作比较。文中提出了几个用于作月预报的统计模式，分别用不同时间或不同波数作预报因子。其结果均使预报准确率有所提高。既用不同时间又用不同波数作预报因子的模式更为理想。文中最后讨论了一个源于动力学考虑的统计模式。近１４００个月预报例子的结果与实况的平均相关系数高达０．７５，显著优于惯性预报。

The correlations between monthly mean and daily mean fields which are one week before or after that month have been analysed with 15-year record of the 1000 hPa temperature for the Northern Hemisphere.It is found that the newer the daily data are,the stronger the correlations. Persistent skills for 30-day averaged forecasts are much greater when using instantaneous daily data rather than ahead30-day averages.It suggests that model skill measures should be compared to the persistence defined by the former.Several statistical models are developed in this paper.Skills of all these models are better than those of persistence for monthly forecasts.A dynamically oriented statistical model is the best one.The mean correlation coefficient between monthly predictions with this model and the observations of about 1400cases achieved 0.75 is higher than the persistence,0.58.