Statistical Models for Monthly Surface Temperature Predictions in Summer Season
夏季地面温度场月预报的统计模式

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.