Although computer power has increased to the point where operational weather forecasts can explicitly represent large convective storms on the model grid, these forecasts are still plagued by errors in storm timing, location, and precipitation amounts. These errors arise from two dominant sources: (1) uncertainties in the representation of important “subgrid” processes that are unresolved on the model grid, and (2) uncertainties in the initial state of atmosphere, which arise from errors in prior forecasts and incomplete (and uncertain) observations. Due to the non-linear and chaotic nature of the atmospheric fluid-dynamical system, the above errors grow rapidly and ultimately corrupt weather forecasts. My research group aims to improve convective-storm prediction by reducing the amplitude of these errors, accounting for model uncertainties through probabilistic forecasting approaches, and improving the basic physical understanding of convective processes
