OBJECTIVE PREDICTION OF FIVE-DAY MEAN TEMPERATURES DURING WINTER

Abstract

A statistical screening procedure is used to derive linear multiple-regression equations which express 5-day mean surface temperature as a function of 5-day mean 700-mb heights centered 2 days earlier. Application of these equations to heights obtained from barotropic prognoses would have produced temperature predictions of positive skill during the test winter of 1957–58. The forecasts can be considerably improved by including as a predictor the local value of 5-day mean surface temperature for a period 4 days earlier than the forecast period. When this term was combined with the barotropically-estimated heights, objective temperature predictions comparable in accuracy to conventional forecasts were made by multiple-regression equations. Further work is in progress to obtain additional improvement by screening the entire field of surface temperature.