Thermal sensitivity of rivers, an indicator for ecological refuges and hydrological model diagnostic

Climate change constraints impose to identify streams and rivers that are, either resilient to atmospheric warming and can serve as ecological refuges, or highly sensitive and require restoration measures, e.g. tree planting, restoration of groundwater connections. In order to characterise the streams’ resilience or sensitivity, predict their evolution under climate change, and analyse scenarios of riparian restorations, we used data-driven analysis and physical-based modelling using thermal sensitivity (TS), as a key indicator linking climate forcing, hydrological processes and ecological relevance.

We first aggregated and harmonised a large database of French streams temperatures, i.e. 5515 (summer records) and 3143 (annual records) monitoring sites recorded between 2008 and 2024. We developed linear regressions for each site to predict weekly water temperature from weekly air temperature. The slope of the regression was qualified as the sensitivity of stream temperature (TS) of a given site to change in air temperature. Based on this large dataset, we identified several predictors of the TS, including surface area upstream sampling stations, baseflow index, riparian vegetation, summer air temperature. We found a significant positive correlation between TS and summer water temperatures, i.e. cold-streams are less sensitive to air temperature, and could therefore be important refuge areas for native species conservation.

We then used TS indicator to evaluate how a physically based thermal model (T-NET) coupled with a hydrological model (J2000) on a meso-scale contrasted lithology catchment could simulate daily water temperatures and discharges.  TS enables the attribution of water temperature model’s biases to specific hydrological processes, rather than merely quantifying overall model performance. In particular, low-TS streams, typically groundwater-dominated headwaters, highlight structural limitations in the representation of groundwater-driven heat fluxes.

Finally, we used the coupled model in sub-catchments with high-TS to study the influence of several riparian vegetation restoration scenarios on the stream temperatures. We could rank streams and rivers according to the temperature attenuation gain for specific riparian scenarios.

Our results suggest that TS indicator could be a relevant metric for biological applications, such as stream species distribution modelling.