Variability of River Flow Extremes in Flanders: Impact of Atmospheric and Catchment Drivers

Floods and droughts are severely impacting people's lives, worldwide. The magnitude and frequency of these extreme events have changed over time. Most likely, these extreme events will further change in the future, due to climate variability and climate change and other processes such as urbanization. Policy making and designs related to water management and engineering require a deep understanding of the magnitude and driving forces of such trends in river flow extremes. Attribution involves evaluation of relative contributions of multiple drivers to detected changes in environmental conditions. Explaining trends in hydrological extremes such as floods and droughts, by means of such attribution studies, are found to be challenging. This is mainly due to the low signal-to-noise ratio of the relevant hydro-meteorological variables and to the complexity of large hydrological systems, where different drivers often act in parallel. This dissertation aims to deliver a contribution to the abovementioned attribution science, with a focus on the Flanders and Brussels region of Belgium. We consider climate variability, climate change, and land cover changes as main drivers to explain changes in the river flow extremes. With this research, we have developed some novel methodologies and gained additional insights in the hydrology of Flanders. We firstly investigated the hydro-meteorological conditions in the study regions. It was found that long-term statistics of extreme precipitation do not show significant spatial differences across Flanders, based on a 12-year record of 1-minute rainfall at 43 rain gauges. Furthermore, well-known correlations between atmospheric conditions (in the form of weather types), and precipitation extremes and river peak flows were confirmed, based on a data record of 150 years for the Meuse river basin. This correlation was further successfully used in a novel method for flood frequency analyses, by conditioning the extreme value distributions on weather types. This novel method has been verified for the case study of the Meuse river, based on historical data going back to the 1500s. We have shown that, for 29 catchments over the past three decades, these weather types, representing climate variability, could explain up to 6% of changes in river peak flows. Future projections further show both low and peak flows would become more extreme in the future. These changes are stronger with increased radiative forcing. The magnitude of the changes, however, is highly uncertain. The most important sources of this uncertainty are, for low-flow projections, the global climate models (explaining 27 - 36% of the total variance) and the hydrological model structure (34 - 42%). For peak-flow projections, these are the global climate models (32 - 39%) and the statistical downscaling methods (21 - 26%). Also, interaction effects account for a significant part of the uncertainty. For selected case studies in Flanders, we have shown that river peak flows could increase by up to +75% by the end of the century and low-flow extremes could decrease by up to -50%. Also, the reliability of a selected drinking water supply facility, WPC Kluizen, could decrease by up to -20%, based on their current management practices. This, however, could be reversed by an additional water intake of 500 m3/h. Land cover changes also influence hydrological extremes: over the past three decades, these land cover changes could explain up to 8% of changes in river flows, where increased urban areas result in higher peak flows. To take future land cover changes into account for projections of hydrological extremes, we have developed an efficient approach by means of parameter updates in lumped conceptual models. We have shown that river peak-flow extremes could increase by +14% (Grote Nete catchment) and +8% (Zuunbeek catchment) for a Business-as-Usual scenario by 2040. For low-flow extremes, results showed less clarity. Investigating the impact of single drivers separately is not sufficient to make reliable projections. Indeed, we have shown that interaction effects could explain up to 32% of the variation in river peak flows in the past. This calls for an integrated model-based impact analysis, where our novel method to take land cover changes into account could be used. Other recommendations are to reduce uncertainty on climate change signals, by e.g. smart ensemble building, and improving the performance of hydrological models, mainly with respect to low-flow conditions.status: publishe