Abstract
This paper describes an operational system for specifying monthly precipitation amounts in the contiguous United States from the concurrent 700-mb monthly mean height field over North America and adjacent oceans. Multiple regression equations are derived for each month of the year at 60 climate divisions by applying a quasi-objective, forward selection procedure to 30 yr of data for 1951–80. The resulting specification equations explain an average of 37% of the precipitation variance, but values range from 70% along the Pacific Coast in January to 10% in southern New England in July. When applied to prognostic 700-mb charts for 1987 and 1988, the equations have shown more skill than persistence but less skill than official monthly outlooks.
Four attempts to improve the specifications are discussed. Best results were obtained by screening the mean precipitation amounts within 10–12 coherent regions, selected by factor analysis, instead of 60 smaller climate divisions. This procedure raised the explained precipitation variance during each month of the year, with a mean increase of almost 11%. The average increases of explained variance produced by the other attempts were about 4% for precipitation frequency instead of amounts, 2% for seasonal instead of monthly means, and 1% for previous precipitation and variables derived from 700-mb heights as additional predictors. Consequently, average precipitation amounts within large coherent regions are now being specified routinely and used as additional guidance at the Climate Analysis Center.
