Abstract
Fine-scale climate projections are essential for anticipating the effects of climate change, which is critical for both agricultural resilience and biodiversity conservation. The ROC-CHÂ project, which focuses on the impacts of climate change on chestnut (Castanea sativa Mill.), developed this topo-climatic study in the aim of accurately model these global changes and their local impacts. The incorporation of topography into the modelling process is curcial, as terrain characteristics can generate microclimates, which may lead to significant variations in plant spatial distribution. The modelling pipeline developed in this study was applied to a region in the Lozerian Cévennes (France) known for its numerous chestnut tree orchards. Combining empirical and machine learning approaches, this pipeline integrates topographical and meteorological inputs to produce 100-m resolution downscaled projections for the following climate variables: i) global solar radiation, and the associated weather variables of ii) minimum, maximum and mean temperatures, iii) potential evapotranspiration, and iv) precipitation. This modelling was carried out for both past (1990-2020) and future (2021-2100). For the historical period, observed weather station data were used, while for the future, the meteorological data used were based on simulations of the CNRM-CM5/ALADIN63 regional climate model applied to IPCC scenarios RCP 4.5 and RCP 8.5. The resulting topo-climatic predictions can be visualized through maps, offering a powerful tool for participatory research with chestnut producers. These maps serve as a basis for dialogue and decision-making aimed at adapting agricultural practices to future climate conditions.