Probabilistic Weather Forecasting

Operational probabilistic weather forecasts at leads times of days ahead depend on ensembles of numerical weather prediction (NWP) model runs. However, as ensemble forecasts tend to be biased and underdispersed, some form of statistical postprocessing is required, with Bayesian model averaging and nonhomogeneous regression being state of the art approaches for doing this. Challenges and opportunities for future work include postprocessing efforts for probabilistic forecasts of multivariate quantities, including the case of spatial, temporal and spatio-temporal weather trajectories