Abstract
Multiple hydrological impact studies have assessed the hydrological consequences of climate change. Yet, especially in applied water management studies, often only one or a few GCMs are used and limited attention is given to the uncertainties present in future climate projections. Major investments for the implementation of adaptation measures are
... read more
based on the outcomes of these overconfident and possibly biased assessments. In this thesis an extensive analysis of the inherent uncertainties present in the modeling chain of hydrological impact studies is conducted. In addition, future global hydrological changes are derived and methods for the interpretation of the accompanying uncertainties are explored. Uncertainty analysis The first step of this study evaluates the influence of PCR-GLOBWB parameter and forcing data uncertainty on modeled discharge estimates. To this end PCR-GLOBWB was forced with three different forcing datasets and 250 different parameter combinations. The results show that, with simple Latin Hypercube sampling, parameter combinations can be obtained that outperform the default model parameterization. Yet, optimal parameter estimates highly depend on the forcing dataset used. This indicates that a hydrological model forced with often biased GCM datasets is preferably calibrated for each dataset independently. Otherwise, calibration attempts should be limited and parameter values could best be obtained from available global datasets, as has been done within this study. In a second step the global hydrological model PCR-GLOBWB is forced with meteorological datasets from an ensemble 12 GCMs. The resulting annual average discharge cycles are compared with discharge observations. This comparison reveals large variations between GCMs and large deviations of GCM derived discharge from discharge observations. The analysis is extended with the evaluation of the usability of GCM runoff for hydrological studies. The results indicate that especially for large catchments full hydrological modeling outperforms the use of GCM-generated runoff. Runoff routing and correct representation of subsurface hydrological processes, as present in hydrological models, introduce realistic flow velocities and consequently a correct reproduction of discharge extremes. Future hydrological impacts of climate change To asses future runoff changes PCR-GLOBWB is forced with datasets from 12 GCMs for the IPCC A1B scenario. The obtained change projections diverge widely. Yet, when discharge change is calculated relative to discharge amounts, obtained for the 20th century control experiment for each individual GCM, ensemble consistent and significant changes can be found. By 2100 runoff decreases are projected for southern Europe, southern Australia and the south and north of Africa and runoff increases are projected for sub-Arctic and Monsoon influenced regions. In a final step several performance based GCM weighting and selection methods have been investigated. These methods are used to derive weighted average future runoff changes from the ensemble of 12 GCMs and to quantify the accompanying weighted uncertainty ranges. The method which considered inter-model similarity, for both the current and future climate, performed best.
show less
