Key points are not available for this paper at this time.
The simultaneous relation between temperature and circulation during the winter season is studied by correlating anomalies of monthly mean surface temperature at 109 stations in the United States from 1948 through 1981 with monthly mean 700 mb height anomalies at a network of 133 grid points in North America and surrounding oceans. The data are screened by a stepwise forward selection procedure to yield multiple regression equations for specifying the monthly mean temperature anomaly at each city from the field of simultaneous 700 mb height plus the previous month's temperature anomaly. Several sets of multiple regression equations are derived on a dependent sample of 90 winter months by using two different grids of heights and applying various cut-off criteria, such as the F-ratio, additional explained variance and number of variables, to terminate the selection of additional predictors. Optimum termination points are determined by testing 19 alternate criteria on independent data from 1978 to 1981 in terms of five different verification statistics. It is concluded that use of 8 variables in the multiple regression equations is optimum in this lest. The 8-variable cut-off performed better than climatology, persistence, an older set of equations derived by screening 5-day mean data, or equations based on empirical orthogonal functions. By combining dependent and independent samples and by applying the best cut-off criteria plus synoptic experience, final specification equations are computed. These equations explain about three-fourths of the temperature variance by means of from 3 to 9 variables but with marked regional variation. They can readily give objective forecasts of the monthly mean surface temperature distribution over the United States by using any prognostic monthly mean 700 mb map as input. A short set of these equations, with about five terms per station, is currently being applied by the Climate Analysis Center of the National Weather Service on an operational basis.
William H. Klein (Fri,) studied this question.