HEMISPHERIC NUMERICAL FORECASTING WITH THE BAROTROPIC MODEL, AND SOME REMARKS ON BOUNDARY-CONDITION ERROR

W. Lawrence Gates Air Force Cambridge Research Center

Search for other papers by W. Lawrence Gates in
Current site
Google Scholar
PubMed
Close
Restricted access

Abstract

After further experiments with a pilot case reported earlier, the results are summarized for a selected ten-day series of predictions obtained over the northern hemisphere by conventional numerical methods with the barotropic model. The overall accuracy of the forecasts is found to decay steadily with increasing forecast period up to 72 hours. In comparison with a series of barotropic forecasts prepared earlier for North America only, the hemispheric predictions are shown to be free of serious boundary-condition error in the middle-latitude regions of major synoptic activity. Outstanding among the errors remaining in the hemispheric integrations, however, are those due to the variation of the density of observational data (especially serious over the Pacific and Asiatic regions), those caused by excessive anticyclogenesis, those due to truncation error and the lack of smoothing, and those inherent in the model's neglect of baroclinic development. Research is in progress on these and other errors, in an effort to improve further the resolution of numerical prediction methods.

Abstract

After further experiments with a pilot case reported earlier, the results are summarized for a selected ten-day series of predictions obtained over the northern hemisphere by conventional numerical methods with the barotropic model. The overall accuracy of the forecasts is found to decay steadily with increasing forecast period up to 72 hours. In comparison with a series of barotropic forecasts prepared earlier for North America only, the hemispheric predictions are shown to be free of serious boundary-condition error in the middle-latitude regions of major synoptic activity. Outstanding among the errors remaining in the hemispheric integrations, however, are those due to the variation of the density of observational data (especially serious over the Pacific and Asiatic regions), those caused by excessive anticyclogenesis, those due to truncation error and the lack of smoothing, and those inherent in the model's neglect of baroclinic development. Research is in progress on these and other errors, in an effort to improve further the resolution of numerical prediction methods.

Save