|Title||Calibration of medium-range weather forecasts|
|Year of Publication||2014|
|Secondary Title||Technical Memorandum|
Statistical postprocessing techniques serve to improve the quality of numerical weather forecasts, as they seek to generate calibrated and sharp predictive distributions of future weather quantities. This document reviews the state of the art in statistical postprocessing, with focus on potential applications to the European Centre for Medium-Range Weather Forecasts (ECMWF)'s Integrated Forecasting System (IFS). At present, a recommended way to proceed is to apply well established, state of the art postprocessing techniques, such as nonhomogeneous regression or Bayesian model averaging, to each univariate weather quantity separately, with training data usefully augmented by reforecast datasets. Areas requiring further research are identified, in particular the suitable size and efficient use of reforecast datasets, and the generation and evaluation of probabilistic forecasts of combined events and spatio-temporal weather trajectories, thereby addressing spatial, temporal and cross-variable dependence structures.