This is a proposal to carry out Phase II of a project re- sponding to Program Solicitation No. NSF 89-30 of the SBIR Program. In Phase I, a simplified model that calculates the complex of phy- sical processes that locally modulate large scale meteorological fields (forecasted, e.g., by operational numerical weather predict- ion models at the National Meteorological Center), was developed and tested for feasibility in local forecasting applications. This differs from conventional procedures, which use statistics derived from past weather records to link local weather with large scale model variables. These have the disadvantages that they are biased toward specififying average conditions, and that they cannot be applied at sites with short periods of historical weather data. It was demonstrated in Phase I that the parameters of the model can be objectively adjusted to minimize expected forecast errors at a site. In Phase II it is proposed that the complete model be deve- loped, together with its complementary set of algorithms to search for sets of model parameter values that will minimize some measure of the difference between model-generated and observed values. The result can then be tested with real data in a variety of situa- tions. If successful, this phase of the work should lead to a weather forecasting system that can be tailored to the needs of local users (utilities, airports, agriculture, construction proj- ects, etc.) and implemented on a local workstation-size computer.
