Abstract ID: 100
Assessment of S2S ensemble extreme precipitation forecasts over Europe
Lead Author: Pauline Rivoire
Institute of Earth Surface Dynamics, University of Lausanne, Switzerland
Keywords: precipitation, extreme, skill, assessment, binary loss index
Abstract: Heavy precipitation can lead to floods and landslides, hazards causing large damage and casualties. Some of these impacts can be mitigated if good forecasts and warnings are available.
We provide here a methodology to assess the skill of extreme precipitation forecasts on S2S time scales, for local as well as temporally and spatially aggregated extreme events. We apply this methodology to verify extreme precipitation events over Europe in the S2S forecast model from the ECMWF. The verification is conducted against ERA-5 reanalysis precipitation. The extreme events are defined as daily precipitation accumulations exceeding the seasonal 95th percentile. We use the classical Brier score. We additionally develop a skill metric based on a binary loss function to focus the assessment on extremes. The results are displayed with the last skilful day, that is, the last lead day for which the forecast is significantly more skilful than the climatology. We analyse daily local events, spatially aggregated events and counts of extremes in a 7-day window. Results consistently show a higher skill in winter compared to summer. The regions showing the highest skill are Norway, Portugal and the south of the Alps. The Mediterranean region has a relatively good skill in winter. The skill is increasing when aggregating the extremes spatially or temporally. The verification methodology can be adapted and applied to other variables, e.g. temperature extremes, river discharge, etc.
Co-authors:
Martius-Romppainen Olivia (Institute of Geography and Oeschger Centre for Climate Change Research and Mobiliar Lab for Natural Risks, University of Bern, Switzerland)
Naveau Philippe (Laboratoire des Sciences du Climat et de l’Environnement, ESTIMR, CNRS-CEA-UVSQ, Gif-sur-Yvette, France)
Tuel Alexandre (Institute of Geography and Oeschger Centre for Climate Change Research, University of Bern, Switzerland)
