Developing climate projections and services in data-scarce regions: the case of French tropical overseas territories

Over the past couple of decades, thanks to the sustained development of Global Climate Models (GCMs) combined with dedicated downscaling strategies such as regional climate modelling or statistical downscaling, climate projections and associated services are now increasingly available across many regions, particularly in nations of the Global North like France. However, whereas this is the case for continental France, the country includes numerous overseas territories, most of them being small islands in the tropical Atlantic, Indian and Pacific Oceans, where this information was only partially available until recently, if at all.

Here we present the recent development of ensembles of climate projections for most French tropical overseas territories (French West Indies and Guiana, Reunion Island, Mayotte, New Caledonia and French Polynesia), complemented with services in the form of climate information provided for different regional warming levels. The ensembles consist in the blending of data from global climate models (CMIP6), regional climate models (e.g. CORDEX), high-resolution global models and convection-permitting models where available. The models are evaluated against gridded and local observations with a focus on important regional climate processes, and selected accordingly for each domain. Fine-scale reference products for daily surface temperature and precipitation are developed for each territory. They combine long-term weather station observations with high-resolution data from either evaluation simulations driven by the ERA5 reanalysis or numerical weather prediction models. These products are then used to bias-correct and statistically downscale model fields, thereby providing kilometer-scale ensembles of transient climate simulations for each territory over both the historical and future periods, which are made freely available on national climate data portals (DRIAS, Climadiag Commune).

In addition, local temperature observations are used to constrain warming projections from CMIP6 for each territory, in order to compute regional warming levels corresponding to global warming levels +1.5°C, +2°C and +3°C. Following the national reference warming trajectory for adaptation to climate change (TRACC), a framework that has been previously applied over continental France to guide adaptation policies, climate indices are computed for these regional warming levels from the aforementioned climate projections and also made freely available. In addition to generic indices (e.g. number of hot days/nights, of heavy precipitation days etc.), tailored indices for the agriculture, water resource, energy, public health and disaster management (wildfires, coastal hazards) sectors are being developed using local impact data from various stakeholders.

Future extensions include regional climate model emulators previously developed for European domains. They show encouraging results for tropical islands and are expected to make key contributions to the characterization and understanding of climate projection uncertainties in these data-scarce regions.