Undergraduate Mathematical Sciences Seminar
Thursday, 18 May, 2006, 12:30--1:50pm
Probabilistic Weather Forecasting with Spatial Dependence
Veronica Berrocal
UW Statistics Graduate Student
Probabilistic weather forecasts are obtained by running numerical weather prediction models with varying initial conditions and/or model parameters, resulting in ensembles of deterministic forecasts. However, forecast ensembles are often underdispersive, and therefore uncalibrated. Several postprocessing method have been developed to calibrate ensemble forecasts, most of which corrects the weather forecasts location by location without accounting for the weather field's spatial dependence. We introduce a statistical postprocessing technique, called Spatial Bayesian model averaging (Spatial BMA), to calibrate forecast ensembles of whole weather fields. Spatial BMA provides statistical ensembles of weather field forecasts that take the spatial structure of observed fields into account and honor the information contained in the original ensemble. The technique was applied to 48-h forecasts of surface temperature over the North American Pacific Northwest using the University of Washington mesoscale ensemble, with good results.