Characterization of the river thermal regime in relation to its environment: a regional approach using in situ sensors in a temperate region (Wallonia, Belgium)

Due to climate change, the normal natural balance is being altered. In particular, the rise in temperature already observed is leading to numerous risks for humans and all other forms of life on Earth. Climate change also disrupts the conditions of the aquatic ecosystem such as water temperature (WT). WT is a major factor influencing aquatic ecosystems. WT influences many physicochemical factors in streams (e.g., oxygen availability, organic matter, toxicity and decomposition rates) that impact the metabolism, growth, survival, and swimming endurance of aquatic organisms (fish species, algae, insects, …). Climate change is also expected to lead to more frequent unusually high WT. These extreme thermal events defined as temperatures that lie outside the seasonal norms (abnormal), have repercussions on biodiversity (modification of productivity, spatial distribution and the extinction of species subordinated to their environment) and river managers whose actions currently in place may become ineffective in the coming years. The understanding of thermal processes in rivers is therefore crucial for coping with climate change. In this context, the objective of this thesis is to characterize the thermal regime of rivers in relation to their environment and to focus on thermal extreme events. All the research has been conducted on the hydrographic network of the southern region of Belgium (Wallonia) covered by a hundred stations recorded WT at intervals of 10 minutes. First, data is needed to characterize thermal regime. Specifically, continuous WT network is essential for effective research unfolding through time in a global change context. The networks of stations dedicated to WT measurement are scarce. However, many countries have large spatial networks of water quality (water quality monitoring) or water level (flow monitoring) measurement whose probes are equipped to measure WT. Therefore, we investigated the use of WT data from a regional water level monitoring network which measure WT as ancillary data (chapter 2). A Bland–Altman analysis with WT collected through a European monitoring network (Water Framework Directive) and WT data from our regional water level network (140 stations) was made to test the reliability of the water level network for continuous WT monitoring. We found that the water level stations were reliable tools in recording continuous WT in the streams of the study area. The temperature difference between the two WT monitoring networks was −0.57°C on average but showed a strong linear correlation. Our positive results allowed us to use WT from water level stations in order to perform both state‐of‐the‐art visualization of thermal regimes and spatio‐temporal queries for specific ecological monitoring at high frequencies. Second, based on WT data validated in chapter 2, we highlighted and studied « extreme WT ». Although many studies focus on maximum WT, « extreme WT » are more relevant to be identified and modeled in a climate change context. « Extreme » or WT above normal characterizes unusual weather events influenced by global change. Extremes can be damaging (and even lethal) for aquatic ecosystem. Indeed, for many aquatic species the different phases of the reproduction cycle are controlled by seasonal change (temperature, photoperiod, …). This coordination ensures that the offspring are produced under optimal environmental conditions for their survival. Hence, abnormal WT change may have negative impacts on species with seasonal spawning patterns. In addition, an increasing number of extreme events is observed as the climate continues to change. In chapter 3, we made a generalized additive modeling (GAM) to highlight extreme WT. GAM is a time series statistical method which allows to extract within-year variations (seasonality) and long-term changes (trend). WT which deviated significantly from commonly observed WT (or normal) were therefore be identified. Then, the WT temporal dynamics around extremes were modeled using specific differential equations in order to better understand the phenomenon. Thirdly, managers need to know how to mitigate rising WT due to climate change. We hypothesized that (i) extreme thermal events highlighted in chapter 3 are influenced by a limited set of environmental factors and (ii) the role played by those factors varies spatially. Therefore, in chapter 4, we determined which environmental variables affected the most WT. Environmental variables are composed of land cover, topographical (channel slope, elevation, shade) and hydromorphological (channel sinuosity, water level, watershed area, baseflow index) factors which are the main influential variables identified in the literature as affecting WT. The role of environmental parameters was investigated at different spatial scales (buffer, riparian and watershed scales) to identify the spatial scales over which those environmental variables influenced WT. To test these hypothesis, stepwise multiple linear regressions (MLR) were made between model variables (day thermal sensitivity, night thermal sensitivity, and non-convective thermal flux) calculated in chapter 3 and environmental variables. Environmental variables were evaluated for the six spatial scales during summer extreme events highlighted in chapter 3. Results showed that shade, baseflow index (a proxy of the influence of groundwater), water level and watershed area were the most significant variables influencing thermal sensitivity. Our results show also that a larger management scale is not more effective in reducing thermal sensitivity to extreme events. Finally, the main findings of the thesis were summarized and we discussed some limitations of our results and the main practical implications. Particularly, we discussed three questions: « With what data can we characterize the thermal regime? », « When and where is it relevant to characterize the thermal regime? » and « How managers can act to mitigate WT increases? ». Future research perspectives were also provided.