Regional climate models (RCM) provide high resolution meteorological data for different climate change scenarios. These series however cannot be used directly for the assessment of possible hydrological impacts of climate change due to the bias in its variables which is present on all spatial temporal scales. The purpose of this contribution is to develop a methodology to correct these biases. The first step is to compare the spatial structure of RCM generated rainfall for the control period and the observed precipitation both conditioned on circulation patterns (CPs). As a next step the spatial structure is corrected using a copula based approach and a subsequent correction of the dependence structure. Subsequently CP dependent quantile-quantile (Q-Q) transforms between observed and RCM-estimated rainfall are defined. The Q-Q transforms are derived from a 20-year calibration period and validated during a 10-year period of observations. This Q-Q transformation is than used to estimate the rainfall patterns and amounts from RCM predictions of rainfall fields for a future period. The methodology is applied to daily rainfall fields on a 25 km grid over the Rhine basin in Germany. Distributions and extremes are compared on different spatial and temporal scales for the control period. The raw RCM predictions for climate scenarios are finally compared with the bias corrected results.
Biography: Andras Bardossy is professor for hydrology and geohydrology at the University of Stuttgart. He studied Mathematics in Budapest and holds a doctoral degree both in mathematics (1981) and civil engineering (1993). His main interest is the stochastic modelling of precipitation in space and time and the space time interpolation of environmental variables. He published more then 100 papers in differenrt scientific journals.