Capturing hydrological variability : Spatio-temporal aspects of the Rhine basin

Variability is present on multiple levels in the hydrological cycle, both spatially and temporally. Spatial variability can be caused by variations in catchment characteristics and by changes in dominant hydrological processes. Temporal variability occurs as a result of typical variations in diurnal and seasonal cycles, but also on longer time scales resulting from a change in climate. Spatial and temporal variability across a basin ultimately affects the hydrological response. This thesis describes how variability in different components of the hydrological cycle (precipitation, runoff generating processes, and evaporation) influence the hydrological response in the Rhine basin. The studies cover three important steps in hydrological research: understanding past events through analysis of observations (Chapter 2 and 3), conceptualising hydrology through testing and developing models (Chapter 4 and 5), and understanding the hydrological response under changing conditions (Chapter 6 and 7). The results in this thesis improve how hydrological variability resulting from variability within and across different components of the hydrological cycle is captured.The response of five major components of the hydrological cycle to drought conditions is investigated in Chapter 2. With a special focus on the Dutch province of Gelderland, the responses of precipitation, soil moisture, vegetation, groundwater levels, and surface water levels are quantified. Droughts are defined based on a frequency-of-occurrence approach, and the actual drought frequency is tested against this definition. It is showed that each variable is more frequently in drought conditions (27-38% of the time) than expected based on the drought definition (20% of the time). This is a direct result of spatial variability within each variable. At least one variable is in drought conditions for 73% of the time, and 3--4 variables need to be simultaneously in drought conditions to have a match between actual drought frequency and the definition. This chapter shows how the drought definition does not reflect actual drought occurrence in a spatial heterogeneous region.The interaction between soil moisture content and vegetation productivity during the summer drought of 2018 is described in Chapter 3. This chapter compares in situ soil moisture observations at several depths with satellite derived vegetation productivity indices. It was shown that the anomaly in vegetation productivity lagged between two and three weeks behind the anomaly in soil moisture content. Furthermore, it was shown that the critical soil moisture content (indicating the transition point between energy-limited and water-limited conditions) increased linearly with integration depth. This chapter highlights the complex interaction between vegetation productivity and soil moisture content.While it seems obvious that high spatial resolutions are required to optimally represent the spatial variability, this is not always possible. Limitations in data availability and/or computational power restrict models to be ran on high spatial resolutions. Chapter 4 compares simulations at low resolution with simulations at high resolution, using the SPHY model. While both the high- and low-resolution models correctly simulated the discharge, local differences were present. The high-resolution model showed that both extremely positive and extremely negative anomalies can occur simultaneously in the same basin during seasonal extremes. This was not captured by the coarse resolution model. A newly proposed metric, the density-weighted distance, showed that the coarse resolution model missed on average two standardised anomalies when compared with the results from the high-resolution model. This difference was larger in the basins with glaciers than in basins without glaciers. In conclusion, it was shown that results from a coarse resolution model cannot easily be translated to the results from a high-resolution model.The dS2 model is introduced in Chapter 5. This distributed rainfall-runoff model is based on the simple dynamical system approach and is developed with an emphasis on computational efficiency. This allows simulations to be performed at high spatial and high temporal resolutions, without high computational costs. Snow and flow routing modules were added to ensure that the model can be applied to mesoscale basins. The dS2 model is tested to a basin in the Swiss Alps, a subbasin of the Rhine, and showed that it can simulate the hydrological response with high accuracy. This model is used for the simulations in the next two chapters.The role of increasing temperatures on the discharge is described in Chapter 6. This chapter shows how temperature-driven changes in evaporation and snow processes influence the hydrological response. Simulations of two decades (1980s and 2010s) were performed to quantify the contribution of precipitation and temperature-driven changes in evaporation and snow processes on the discharge. This showed that temperature-driven changes in evaporation and snow processes were just as important as the precipitation-driven changes. Stepwise temperature increases showed that increases in evaporation reduced discharge throughout the year, and that snow processes caused a different seasonal pattern. Higher temperatures reduced the amount of snowfall and caused the snowpacks to melt earlier in the year. Additionally, melting of the glaciers was increased over the entire year. Overall, the discharge was reduced throughout the year, and the change in snow seasonality caused a large discharge reduction during late spring and early summer. This chapter showed the importance of these temperature-driven changes on discharge simulations, as they can both counteract and amplify changes in discharge.Chapter 7 describes how CO2-induced changes in vegetation and temperature affect evaporation and discharge in the Rhine basin. The relative importance of CO2-induced changes in stomatal resistance, leaf area index, and temperature are quantified. Increases in evaporation caused by temperature reduced the discharge with 17% (with respect to the average discharge over 2018). Higher stomatal resistance led to reductions in evaporation, effectively increasing the discharge with 4%. The increase in leaf area index had only a very limited effect on the discharge. Overall, discharge is expected to decrease with 14% as a result of these CO2-induced changes. These results are confirmed in a sensitivity study. While changes in temperature are typically accounted for in hydrological studies, the changes in vegetation (and especially the increase in stomatal resistance) are not. This chapter highlights that the response of vegetation needs to be accounted for, otherwise it is likely that evaporation is overestimated, and discharge is underestimated.By combining data-analysis studies with simulation studies, this thesis showed how spatial and temporal variability influence the hydrological response. The complex interactions between different components of the hydrological cycle, as well as the contrasting responses to changing climate conditions, need to be represented correctly to provide reliable discharge estimates. The results from this thesis give insight on how to capture the hydrological variability, in order to improve how hydrology can support and protect society