RESULTS OF NUMERICAL FORECASTING WITH THE BAROTROPIC AND THERMOTROPIC ATMOSPHERIC MODELS

Abstract
Following a resume of the theory of thermotropic flow, a simple baroclinic model in which the direction of the thermal wind is assumed invariant with height, a discussion of the methods employed for the numerical integration of this model and the barotropic model is presented. On a finite difference grid of 414 points covering the United States and immediately surrounding regions, a series of sixty comparative 24-hour forecasts during January, 1955, at the 500 and 1000 mb levels was obtained by relaxation methods. The median correlation coefficients between the forecast and observed 24-hour height changes were 0.8J for both thermotropic and barotropic models at 500 mb and 0.69 for the thermotropic model at 1000 mb. In comparison with a method of pure interpolation, the numerical forecasts are shown to display a positive skill toward the center of the forecast region where the influence of the lateral boundaries is smallest. By normalizing the root mean square forecast errors to allow for the normal latitudinal variation, the Rocky Mountains are found to exert a marked influence on the forecasts at both 500 and 1000 mb over the south central United States. From a synoptic point of view, the numerical forecasts are found to compare favorably with conventional forecasts for the same period, although they appear to introduce a small but systematic tendency to move fully developed disturbances too slowly. This error is felt to stem from the truncation errors of the finite difference schemes employed. Recommendations for further research to reduce the several sources of error and to extend the physical basis of the model are made.

This publication has 0 references indexed in Scilit: