Statistical Prediction Methods for North American Anticyclones

Abstract

Statistical prediction methods of forecasting the 24-hr movement and change of central pressure of North American winter anticyclones are developed utilizing multiple linear regression analysis. The three predictands obtained are the 24-hr change in central pressure and the eastward and northward displacements. This report is similar in some respects to one conducted by Veigas and Ostby relating to U. S. east coast cyclones. The dependent data consist of observations of 150 anticyclones. A moving coordinate system is employed: the predictor information is measured at certain predetermined grid points relative to the system center rather than at fixed geographical positions. Readily measured meteorological parameters are selected for the input data. These include point values of surface pressure, surface temperature, 500-mb height and their 24-hr changes; and the arithmetic mean temperature of the layer from the surface to the 500-mb level. Several different sets of regression equations are obtained for each predictand by slight modifications in the predictor-selection criteria. These regression equations are tested on a sample of 50 independent equations, and the per cent reductions of variance resulting from each method are compared.