Inter-comparison of statistical downscaling methods for projection of extreme precipitation in Europe
- 1DTU Environment, Lyngby, Denmark
- 2Helmholtz-Zentrum Potsdam, Deutsches GeoForschungsZentrum, Potsdam, Germany
- 3Norwegian Water Resources and Energy Directorate (NVE), Oslo, Norway
- 4DHI, Hørsholm, Denmark
- 5Hydraulics Laboratory, KU Leuven, Heverlee, Belgium
- 6Department of Hydrology and Hydraulic Engineering, Vrije Universiteit Brussel, Brussels, Belgium
- 7Faculty of Environmental Sciences, Czech University of Life Sciences Prague, Prague, Czech Republic
- 8Institute of Earth and Environmental Science, University of Potsdam, Potsdam, Germany
- 9T. G. Masaryk Water Research Institute, Prague, Czech Republic
- 10Lithuanian Energy Institute, Kaunas, Lithuania
- 11Department of Civil Engineering, University of Thessaly, Volos, Greece
- 12Department of Hydrology and Hydrodynamics, Institute of Geophysics, Polish Academy of Sciences, Warsaw, Poland
- 13Middle East Technical University, Civil Engineering Department, Ankara, Turkey
Abstract. Information on extreme precipitation for future climate is needed to assess the changes in the frequency and intensity of flooding. The primary source of information in climate change impact studies is climate model projections. However, due to the coarse resolution and biases of these models, they cannot be directly used in hydrological models. Hence, statistical downscaling is necessary to address climate change impacts at the catchment scale.
This study compares eight statistical downscaling methods (SDMs) often used in climate change impact studies. Four methods are based on change factors (CFs), three are bias correction (BC) methods, and one is a perfect prognosis method. The eight methods are used to downscale precipitation output from 15 regional climate models (RCMs) from the ENSEMBLES project for 11 catchments in Europe. The overall results point to an increase in extreme precipitation in most catchments in both winter and summer. For individual catchments, the downscaled time series tend to agree on the direction of the change but differ in the magnitude. Differences between the SDMs vary between the catchments and depend on the season analysed. Similarly, general conclusions cannot be drawn regarding the differences between CFs and BC methods. The performance of the BC methods during the control period also depends on the catchment, but in most cases they represent an improvement compared to RCM outputs. Analysis of the variance in the ensemble of RCMs and SDMs indicates that at least 30% and up to approximately half of the total variance is derived from the SDMs. This study illustrates the large variability in the expected changes in extreme precipitation and highlights the need for considering an ensemble of both SDMs and climate models. Recommendations are provided for the selection of the most suitable SDMs to include in the analysis.