There is an inherent limit to the predictability of atmospheric numerical models. The principal reason for this limit is generally considered to be the nonlinear interactions between different components of the wave spectrum; uncertainties and errors in the wave spectrum resolvable by the model and errors introduced by the neglect of unresolved waves grow with integration in time, spreading throughout the spectrum, and eventually destroy the forecast. Meteorologists now have some understanding of predictability of large scale models used for weather forecast; the spatial resolution is roughly hundreds of kilometers. Numerical models for convective- and meso- scale meteorological events have grown in sophistication in recent years, the question of predictability can now be investigated. The spatial resolution of convective-scale models is roughly 1-2 km, and 20-100 km for mesoscale models. Under this award, Professor Droegemeier will investigate the predictability of a convective-scale model, which has been used to study the dynamics of thunderstorm outflow and microbursts and demonstrated some success in simulating these convective-scale events. Besides examining the sensitivity of the model prediction to different aspects of the initial conditions, Professor Droegemeier will look for ways to initialize the model with direct Doppler radar observations; he will also continue his work on adapting the Piecewise Parabolic Method to meteorological problems.
