Specifying Monthly Mean Surface Temperatures in Canada and Alaska from the 500 mb Height Field

Abstract

Specification equations for monthly mean air temperature anomalies at 68 surface stations in Canada and Alaska are derived by applying a forward selection screening procedure to simultaneous monthly mean 500 mb height anomalies at 110 grid points in the Northern Hemisphere and the previous month's 1ocal temperature anomaly. For the annual average, these equations explain about 72% of the temperature variance (on dependent data) by means of only 4.3 variables, but with marked regional and seasonal differences. The average properties of the specification equations closely resemble those derived previously from the field of 700 mb heights. The 500 mb equation however, explain about 2% more of the temperature variance by means of 0.7 fewer terms in the equations. This small but consistent superiority is evident during each month of the year and for equations based on heights only, as well as those including previous local temperature. The superiority is produced by higher correlations of temperature with local heights and suggests that the 500 mb level is more equivalent barotropic than 700 mb. The specification equations are tested by means of several verification statistics computed for 5 years of independent data at 51 Canadian stations. The results show that equations based on concurrent 500 mb heights and previous local temperature perform slightly better than those based on heights only and much better than climatology or month-to-month persistence.