摘要: We developed statistical relationships between fractional cloud cover and acollection of variables drawn from forecast fields from a global numerical weather prediction model. These relationships were then applied to later forecas... 展开 We developed statistical relationships between fractional cloud cover and acollection of variables drawn from forecast fields from a global numerical weather prediction model. These relationships were then applied to later forecasts from the same weather prediction model to diagnose the cloud cover corresponding to the forecast states. The U.S. Air Force RTNEPH cloud analysis datasets in the Northern Hemisphere for January and July 1991 were used to represent cloud cover in separate winter and summer statistical relationship developments. Forecasts from the Phillips Laboratory Global Spectral Model at times corresponding to the cloud analyses were used to provide the weather model predictors. We used multiple linear regression and a hybrid regression estimation of event possibilities/multiple linear regression to develop the statistical relationships in the 5 percent cloud amount categories from 0 to 100 percent. We used multiple discriminant analysis with six cloud amount categories covering the range from 0 to 100 percent. Separate relationships were developed over 10-day periods for low, middle, and high cloud decks, and total cloud. The relationships developed were also distinct for different forecast durations. All relationships were applied to forecasts initialized on the day following the 10-day development period. Our results showed that multiple linear regression produced forecast diagnoses of cloud amount that were slightly better than the other methods in root-mean-square error. This method also modified the frequency distribution of cloud cover from that of the analysis. The hybrid method improved upon the frequency distribution of clear and overcast, but worsened the near-clear and near-overcast categories. Multiple discriminant analysis produced cloud cover diagnoses with the best combination of root-mean-square skill and preservation of the freque. 收起