Title: Spatially consistent postprocessing of probabilistic cloud cover forecasts
Authors: Stephan Hemri - MeteoSwiss (Switzerland) [presenting]
Christoph Spirig - MeteoSwiss (Switzerland)
Jonas Bhend - MeteoSwiss (Switzerland)
Jan Rajczak - MeteoSwiss (Switzerland)
Lionel Moret - MeteoSwiss (Switzerland)
Mark Liniger - MeteoSwiss (Switzerland)
Abstract: Even though numerical weather prediction models (NWPs) are run at increasingly high resolutions, raw ensemble forecasts still tend to be biased and underdispersed. Hence, statistical postprocessing is expected to improve forecast skill and to provide a more reliable estimation of forecast uncertainty. At MeteoSwiss a project on statistical postprocesssing of ensemble forecasts for spatial fields of different variables has recently been launched, integrating the available NWP ensemble forecasts and observations. Predictions of cloud cover are impaired by Switzerland's complex terrain and the high frequency of low stratus over the Swiss Plateau, which is poorly represented in the raw model predictions. First tests based on ensemble model output statistics and analog based approaches showed an increase in univariate forecast skill. While the application of empirical copula based approaches like ensemble copula coupling or the Schaake shuffle turned out to provide promising baseline scenarios, the methods to generate physically realistic forecast scenarios still need to be improved. A consistent representation of the spatial structure of cloud cover and realistic forecast scenarios are desired for graphical forecast products serving the general public. Possible methods and first results will be discussed to overcome this challenge.