An approach for the validation of climate model data as a basis for the interpretation of climate impact assessment in water management in Lower Saxony

摘要

Climate projections are used as input for impact models to simulate the effects of a changing climate for various applications. Assessing the reliability of these results is important because of their use for climate change adaptation. A first step to identify possible uncertainties is the validation of the climate model data. In this study, a simple approach is presented which uses statistical methods to estimate the accuracy of climate model data and provides a measure to assess the adequacy of the data for hydrological impact modelling. As an example the data of three regional climate models - including the statistical model WETTREG2006 and the dynamical models REMO and CLM, all driven by the global climate model ECHAM5/MPI-OM - are evaluated with respect to their ability to reproduce observed temperature and precipitation. The validation is carried out for interpolated areal means for the period from 1961-2000 based on daily values using different indices and efficiency criteria. The study area is the whole catchment of the rivers Aller and Leine (Lower Saxony, Germany), including nine subbasins of the catchment. The results show deviations for the dynamical climate models according to the observed temperatures as well as the drought indices. As regards WETTREG, however, all indices were in more or less good agreement. The hydrological model was able to adequately simulate discharge when driven by meteorological data and climate model data from the 20th century control run. However, the bandwidth of the goodness-of-fit of simulated historical high and low discharge was, however, larger than the bandwidth obtained for the corresponding extreme climate conditions. For a significance level of 95 % it could be evidenced by means of the Wilcoxon-Mann-Whitney test that the examined model chains are suitable for performing climate impact assessment. Albeit extreme conditions, especially low flows, should invariably be evaluated depending on the model chain in order to provide a more reliable basis for decisions and adaptation measures.