MONTHLY PRECIPITATION TEMPERATURE RELATIONS AND TEMPERATURE PREDICTION OVER THE UNITED-STATES

The monthly mean precipitation-air temperature (MMP-MMAT) relation ove r the United States has been examined by analyzing the observed MMP an d MMAT during the period of 1931-87. The authors' main purpose is to e xamine the possibility of using MMP as a second predictor in addition to the MMAT itself in predicting the next month's MMAT and to shed lig ht on the physical relationship between MMP and MMAT. Both station and climate division data are used. It was found that the lagged MMP-MMAT correlation with MMP leading by a month is generally negative, with t he strongest negative correlation in summer and in the interior United States continent. Over large areas of the interior United States in s ummer, predictions of MMAT based on either antecedent MMP alone or on a combination of antecedent MMP and MMAT are better than a prediction scheme based on MMAT alone. On the whole, even in the interior United States though, including MMP as a second predictor does not improve th e skill of MMAT forecasts on either dependent or independent data dram atically because the first predictor (temperature persistence) has acc ounted for most of the MMP's predictive variance. For a verification p erformed separately for antecedent wet and dry months, much larger ski ll was found following wet than dry Julys for both one- and two-predic tor schemes. Upon further analysis, we attribute this to the differenc es in the climate between the dependent (1931-60) and independent (196 1-87) periods (the second being considerably colder in August) rather than to a true wetness dependence in the predictability. We found some evidence for the role of soil moisture in explaining negative MMP-MMA T and positive MMAT-MMAT lagged correlations both from observed data a nd from output of multiyear runs with the National Meteorological Cent er model. This suggests that we should use some direct measure of soil moisture to improve MMAT forecasts instead of using the MMP as a prox y.