In a context where anticipating future trends and long-term variations in water resources is crucial, improving our knowledge about most types of aquifer responses to climate variability and change is necessary. Aquifers with variability dominated by seasonal (marked annual cycle) or low-frequency variations (interannual to decadal variations driven by large-scale climate dynamics) may encounter different sensitivities to climate change. We investigated this hypothesis by generating groundwater level projections using deep learning models for annual, inertial (low-frequency dominated) or mixed annual/low-frequency aquifer types in northern France from 16 CMIP6 climate model inputs in an ensemble approach. Generated projections were then analysed for trends and changes in variability. Generally, groundwater levels tended to decrease for all types and scenarios across the 2030-2100. The variability of projections showed slightly increasing variability for annual types for all scenarios but decreasing variability for mixed and inertial types. As the severity of the scenario increased, more mixed and inertial-type stations appeared to be affected by decreasing variability. Focusing on low-frequency confirmed this observation: while a significant amount of stations showed increasing variability for the less severe SSP 2-4.5 scenario, low-frequency variability eventually showed slight yet statistically significant decreasing trends as the severity of the scenario increased. For the most severe scenario, almost all stations were affected by decreasing low-frequency variability. Finally, groundwater levels seemed, in most instances, slightly higher in the future than in the historical period, without any significant differences between emission scenarios.

Watersheds are complex systems with multiple interactions between physical processes and human-induced socio-economic dynamics. Since the 2000s, numerous flooding and mudslide events have affected the territory in Normandy (France), leading to significant damages. Therefore, a public policy was adopted with the aim to reduce runoff and erosion, it includes: (i) the building of 4,000 hydraulic infrastructures (dams, fascines, hedges, etc.), (ii) the creation of turbidity water-treatment plants and, (iii) the conduction of animation and protection programs on soil and water resources. These investments are co-funded by several local authorities. This original research project aims evaluating the effectiveness of the above-mentioned public policy. Therefore, two complementary approaches are applied: (i) at the regional scale, the investments and damages between 2000 and 2017 were assessed and, (ii) for a pilot small scaled watershed (la Lézarde, 212 km²) a coupled modeling was conducted, taking hydro-sedimentary processes (flood envelopes, diffuse and concentrated erosion, karstic transfers) and associated socio-economic dynamics into account. Our results suggest that over the study period, at the regional scale 500 M\euro were invested to reduce erosion/runoff impacts and, 300 M\euro of damage were caused. Nevertheless, the effectiveness of the public policy since 2000s must be evaluated at the watershed scale using a Cost-Benefit Analysis (CBA) according to two main scenarios: S1 = pre-development (2000), and S2 = post-development (2017). The processes that govern the surface transfer are modeled for different design floods (Q10-50-100) coupling two semi-dynamic models (MikeSHE and Watersed), and the karstic transfer using a deep learning algorithm (Tensorflow). Additionally, three long-term scenarios (until 2050) are modeled taking into account the effects of climate change (RCP scenarios), the change in land use (-33% of grassland areas), and the modification of agricultural practices that limit runoff. These projections provide key elements for decision-makers to guide future public policies controlling runoff and erosion in this territory.