Specification of Monthly Precipitation over the United States from the Surrounding 700 mb Height Field

The relations between monthly precipitation in each of 60 climate divisions in the contiguous United States and the simultaneous field of monthly mean 700 mb height over North America and vicinity were investigated for both frequency and amount by applying a stepwise forward regression procedure to 30 years of data from January 1951 to December 1980. For winter, a good grid for specifying precipitation from 700 mb heights was found to consist of 79 points located every 5° of latitude from 20°–60°N and every 10° of longitude from 45°–150°W. Best results were obtained by expressing all data as ordinary anomalies and by pooling all cases for December, January and February. Fitting monthly totals by a gamma distribution and classifying into 21 percentiles gave as good results as those obtained by any other transformation of precipitation amount, with almost 42 percent of the precipitation variance, nationwide, explained by less than 3 heights. However, this was not as good as the 45 percent explained for precipitation frequency. For both frequency and amount, precipitation was specified best along the West Coast and worst just east of the Continental Divide. The assumption of a linear relation between monthly precipitation and the 700 mb height field was tested by constructing composite maps for dry and wet weather. Visual comparisons indicated that the linearity assumption is a good approximation for both frequency and amount. Likewise, relations between the large scale circulation and monthly precipitation are similar for frequency and amount. During the summer months of June, July and August, the percent of precipitation variance explained by the concurrent 700 mb height field was only 30% for amount and 37% for frequency on a nationwide basis. Poorer specification during summer than winter may be attributed to the frequent occurrence of convective and localized precipitation at that time. During both seasons, and for both frequency and amount, averaging the precipitation within coherent regions selected by factor analysis explained 12–13% more of its variance than yielded by individual divisions within each region. Thus, spatial smoothing minimizes small scale and random effects and significantly improves the relations between precipitation and the scale circulation.