ARRA: A kilometer-scale reanalysis over France with AROME

The ARRA (ARome ReAnalysis) project was launched at Météo-France in 2022, in order to replace the old existing reanalysis system SAFRAN by a system based on the numerical weather prediction (NWP) system AROME. The objective is to cover a period of 60 years at 1.3 km over France with the last version of the operational AROME system, and produce surface and upper air variables , which are of interest for many uses, especially in the energy sector. There is a strong interest in producing this reanalysis with a recent version of AROME in order to promote the joint use of the results of this reanalysis with real-time analysis or forecasting products , as is currently the case with SAFRAN. The production of this reanalysis is part of Météo-France's 2022-2026 Contract of Objectives and Performance (COP), which sets out the institution's strategic directions and objectives for the next five years to better meet the expectations of citizens, the French government and the business community, each of them facing the challenges of climate change at its own level. With this new reanalysis, small-scale phenomena, such as convection or orographic wind circulation, will be significantly improved compared to the existing SAFRAN system or the European Copernicus CERRA reanalysis, produced at 5.5 km resolution. Beyond the replacement of SAFRAN-based products and systems, this new reanalysis is also necessary for the post-processing of the next generation of kilometer-scale climate projections using AROME. It will also constitute a key dataset for the further use and deployment of machine learning (and other AI methods) for high- resolution weather and climate applications.

The ARRA reanalysis will have two components: the atmospheric part with a 3-hour cycle for the surface analysis (air temperature and humidity at 2 meters, snow and soil moisture) associated with the Increment Analysis Update (IAU) technique based on CERRA or UERRA reanalysis for the upper air and a specific product for the surface called ARRA-Land also at 1.3 km. ARRA-Land will be the SURFEX offline surface model driven by ARRA atmospheric forcing fields and by a 24-hour cumulative precipitation analysis based on MESCAN (Soci et al. 2016). The ARRA-Land product will contain hourly surface fields to produce estimates of surface and soil conditions (soil moisture, snow, etc.), with more advanced physics for the surface than in ARRA.

After a brief description of the project and system configuration, preliminary comparison with other reanalysis (SAFRAN, UERRA, CERRA) will be presented and discussed.