Observational records and climate projections provide abundant evidence that water resources are vulnerable and have the potential to be strongly impacted by climate change, with wide-ranging consequences for human societies and ecosystems. Impact studies with hydrological models on the effects of climate change are important as they can indicate how the hydrological processes are likely to be affected and how strong they are going to be affected in the future. This is especially relevant for policy makers, who are charged with the responsibility of selecting appropriate adaptation measures. However, there are varieties of hydrological models with different levels of complexities and no model is perfect in characterising the real physical interactions. Therefore, this study will focus on the determination of the changes of the hydrological system to a number of meteorological factors, which are expected to change in the future for Belgium, by an ensemble of hydrological models. This will allow to assess the uncertainties related to hydrological model in the climate change impact assessment. For the intercomparison, three lumped conceptual models are considered, namely NAM, PDM and VHM, which have been frequently applied to describe the hydrological behaviour of Flemish rivers.
The three hydrological models were built and calibrated using identical datasets for the Grote Nete catchment, a medium scaled catchment in Belgium. Each model was calibrated in by a trial-and –error process in which the information from the available discharge series (sub-flows and routing parameters) was fully employed. This allows to have a more realistic representation of the hydrological behaviour. Calibration was executed for historical observation data for a 3-year period, validation for another verification period of 3-years. The simulations were evaluated using multiple criteria in order to create plausible models that could target the different aspects of the observed series. These criteria were grouped into different classes and include (1) classical model performance tests, (2) performance tests on the flow extremes and (3) the control of the peak flow changes in relation to rainfall changes. After calibration all three conceptual models appear to be fairly adequate for the Grote Nete catchment when comparing the different objective functions. The models could capture well the overall runoff processes and streamflow dynamics. High as well as low flows are simulated reasonably well. Sub-flows and their volumes match closedly to the observations. None of the calibrated models is superior with respect to all performance measures considered.
FORMULIER: F-WL-PP10-1 Versie 02 GELDIG VANAF: 17/04/2009 These three optimally calibrated models were forced with the climate scenarios for Belgium (Ntegeka et al., 2008) for two future time slices (2050s and 2100s). All three models estimated similar general runoff changes inder the climate change scenarios for both the 2050s and 2100s: the low flow or drought problems will become more important with expected decreases of the low flow minima, while the flood risk predictions are highly uncertain for the future. The low flow problems are expected to be more severe by the 2100s than the 2050s. When comparing the simulated changes between the models, it is found that the hydrological impacts of climate change depend on the type of hydrological model used, and that the magnitude of this impact differs between high flow and low flow periods and between the applied scenarios.
The study has provided additional insights in the climate change impacts on the hydrology in Flanders, but also on the uncertainty in the impact results produced by the hydrological models. As such it has provided important information for water managers in support of the decision making process. The study indicated that the role of hydrological model uncertainty is still considerable and that it would be wise to base water management decisions not only on one model. More research is required to investigate the validity of specific hydrological model schematizations for the specific conditions of catchment in Flanders, and depending on the application. This would allow model uncertainties to be quantified more rigorously; hence better decisions to be made on how to mitigate the impacts of climate change.