| description abstract | Multiple regression equations for predicting 5-day mean temperatures in the United States were originally computed from 5-day mean values of both 700-mb. height and surface temperature, but they gave better results operationally when applied to properly weighted 46-hour forecasts of height and temperature. Since re-derivation from daily instead of mean data yielded poorer results, it appears that use of prognostic daily values as input in equations computed from mean data produces the best mean forecast under current operating conditions. In an effort to obtain daily temperature forecasts for several days in advance, 5-day mean objective temperature predictions were tested a s forecasts of daily mean temperature on each of the individual days comprising the forecast period. Although perfect mean forecasts would have been most accurate for the middle day of the period, the objective prognoses attained maximum accuracy a day or two earlier. Comparison is made with chance, persistence, climatology, and daily predictions prepared at local forecast offices. The objective forecasts were superior to these controls on each day of the 5-day period, with maximum difference on day 3. Additional tests of the skill of the objective predictions as 2- and 3-day forecasts are described, and it is concluded that the objective method can be of assistance in the routine preparation of 72-hour forecasts. | |
