Numerical Retracing of Atmospheric Events

Abstract

FOR obvious practical reasons, all numerical integrations of the equations of atmospheric motion published so far have been made forward in time. To be useful, such numerical forecasts must be started from a position of uncertainty and extend up to, or beyond, the most clear-cut stage of the development in question. But the numerical approach can also be regarded as a method of experimenting with the atmosphere, and in that context the above procedure seems far from ideal. It would then be more logical to start from some well-defined point separating the initial and mature stages and to integrate forward and backward in time, possibly using different models for the two stages. Since the origins of atmospheric disturbances are often inconspicuous on the synoptic scale or obscured by lack of observations, the processes during the developing stage, and hence the results of the backward integration, would in general have the principal interest in this type of experimental work.