Author:

Abstract:

Water scarcity has been increasing in recent years, even in temperate climates, and water scarcity will likely occur more often in the future. It is therefore important to gain better insights in the surface water availability at a regional scale, to estimate the impact of adaptation measures and to better understand the water system in regions where the water cycle is strongly influenced by human activities. Hydrological models are often used for these purposes. However, these are computational intensive and require a lot of background information for implementation, while their accuracy is limited, especially in areas with strong anthropogenic interferences. In some regions, such as Flanders, a relatively large network of observations is available. These open the option for non-parametric data-driven approaches. In this thesis, it is assessed whether non-parametric data-driven approaches are good alternatives or complementary tools to assess the water availability at a regional scale. To this purpose, two non-parametric data-driven approaches are used, a previously successfully applied standard drainage-area ratio method and an alternative interpolation method. It is found that non-parametric data-driven approaches can assess the water availability, represented by each component of the water balance, namely the river discharge, water uptakes and discharges by human activities, at a regional scale in a spatially detailed way. The methods take both the flow regulation effects and other anthropogenic influences as reflected in river flow observations into account. The approaches are computational efficient and do not require adjustments when river management and/or land use changes occur. After applying a validation, it is concluded that the non-parametric data-driven approaches are promising in their capacity to predict water availability at point locations. The overall performance in representing the relative changes of flows in space and time is similar to that of three state-of-the-art hydrological models, two independent local datasets and a local hydrological-hydraulic model. Due to the high spatial and temporal resolution of each component of the water balance, the alternative interpolation method can be used (1) to perform a historical analysis of hydrological drought, (2) to analyse the impact and potential for extreme conditions of a prohibition on water abstractions and reusing wastewater treatment plant effluent water on river flows, and the risks they entail and (3) to better understand the spatial variability in drought sensitivity of rivers. (4) Next to this, the two non-parametric data-driven approaches are implemented in a flexible framework that can be used to gain additional insights in hydrological similarity and assess the potential of combining hydrological models with the alternative interpolation method. The main findings are summarized hereafter. (1) The historical analysis for Flanders indicates that, as expected, there was more water scarcity in recent years. (2) At a regional scale, the impact of applying a prohibition on water uptakes regarding the number of unnavigable rivers that have a river discharge below the ecological minimal flow is limited. Only at the initial stages of a drought or for low flows with a small return period, the impact can be of relevance. Hereby, especially the prohibition on irrigation is of importance, since overall the water uptakes from industry on the unnavigable rivers is low. Reusing wastewater treatment plant effluent water should always be done with care. Application to alleviate the effects of extreme drought events seems limited due to the high dependence of some rivers to this external source of water. (3) The most important factor for the drought sensitivity of a river is the amount of water uptakes for irrigation relative to the available water in a river. A high percentage of irrigation results in a higher drought sensitivity. (4) Based on a first explorative analysis, it is observed that it is important to determine the preferred neighbour based on spatial proximity if one is interested in obtaining a good performance regarding the course of the hydrograph. In contrast, when one is interested in achieving a good performance regarding the water volume, then identifying the best donor catchment based on physical similarity is more important compared to using spatial proximity as a criterium. Moreover, it is concluded that there is much potential in combining hydrological models with non-parametric data-driven approaches. To assess the specific impact of combined sewer overflows and permanent untreated flows on the surface water availability in a data scarce area, the alternative interpolation method is not suitable. This thesis therefore proposes and successfully applies another method that uses conceptual models in combination with a hydrological-hydraulic model. In general, it is concluded that there is a lot of potential in using the alternative interpolation method, with or without hydrological models, for both management and research purposes to assess the surface water availability. The alternative interpolation method is as good as state-of-the art hydrological models, is computational efficient and delivers information on the river discharges, human water uptakes and discharges at a high spatial and temporal resolution.