Predicting the evolution of intermittencies under climate change in France: exploitation of flow projections driven by CMIP5 climate models for Explore2 project

Funded by the French Ministry of Ecology, the French Biodiversity Agency (OFB) and project partners, Explore2 aims to update knowledge about the impact of climate change on hydrology in France, and to support stakeholders in adapting their water management strategies. A multi-scenario and multi-model approach is uniformly applied across the country to encompass a wide range of possible futures for the entire 21st century and to assess uncertainties at each step of the climate and hydrology modelling.

This study aims to extend the results of Explore2 towards the prediction of flow intermittence in headwaters streams, which is initially impeded by the coarse resolution of Explore2 simulations. A statistical approach is necessary to link Explore2 hydrological projections on main rivers to the daily probability of flow intermittence in headstreams (PFI). PFI observations on historical period are derived from data of the French Observatoire National des Etiages (ONDE), which carries monthly visual assessments since 2012, from May to September, at more than 3300 upstream river sites prone to drying  [1]. PFI is then considered as the proportion of ONDE sites observed under drying conditions on partitions of France (76 second-level hydroecoregions (HER2) with median size of 4690 km² paving France).

To predict PFI, logistic regressions are adapted from previous studies [2, 3] and are first calibrated in each HER2 using time series of daily discharge provided by the French hydrometric monitoring network, HYDRO [4]. A diagnosis analysis between 2012 and 2022 consistently demonstrates good performance, with a median Kling-Gupta Efficiency (KGE) around 0.83 across all HER2. Logistic regressions are then re-calibrated considering daily discharge time series simulated by five hydrological models (HMs) of Explore2 driven by SAFRAN meteorological reanalysis [5]. Performance varies according to the HM (KGE medians ranging from 0.60 to 0.82).

Finally, the logistic regressions are applied to simulate daily PFI values at each HER2 for the entire 21st century  with future discharge simulated by the five HMs driven by 17 climate projections under RCP8.5 scenario. Results suggest an increased probability of intermittence in most of the hydrological ensemble runs and under most scenarios. This presentation will focus on the spatial variability of PFI response to climate change projected at different time leads.

 

References

[1] Nowak and Durozoi. Guide de dimensionnement et de mise en œuvre du suivi national des étiages estivaux. ONEMA, 2012.

[2] Beaufort et al. Extrapolating regional probability of drying of headwater streams using discrete observations and gauging networks. Hydrology and Earth System Sciences, 2018. doi:10.5194/hess-22-3033-2018.

[3] Sauquet et al. Predicting flow intermittence in france under climate change. Hydrological Sciences Journal, 2021. doi:10.1080/02626667.2021.1963444Y.

[4] Leleu et al. La refonte du système d’information national pour la gestion et la mise à disposition des données hydrométriques. Houille Blanche, 2014. doi:10.1051/lhb/2014004.

[5] Durand et al. A meteorological estimation of relevant parameters for snow models. Annals of Glaciology, 1993. doi:10.3189/s0260305500011277.