Probabilistic Wind Vector Forecasting using Ensembles and Bayesian Model Averaging

Probabilistic forecasts of wind vectors are becoming critical as interest grows in wind energy. The current weather forecasting paradigm is deterministic, with uncertainty assessed through ensembles, collections of deterministic forecasts. Ensemble forecasts are often uncalibrated. Bayesian model averaging (BMA) is a statistical ensemble postprocessing method that creates calibrated predictive probability density functions. It represents the predictive PDF as a mixture of component PDFs based on individual forecasts.

In this talk we extend BMA to bivariate distributions, enabling us to provide probabilistic forecasts of wind vectors. BMA is applied to 48-hour ahead forecasts of wind vectors over the Pacific Northwest, and is shown to provide better-calibrated probabilistic forecasts than the raw ensemble which are also sharper than probabilistic forecasts derived from climatology.