Climate Hazards Enhanced Reanalyses & Reasoning with AI

CHERRI aims at developing new and innovative processing and methods for future Sentinels and other satellites for reanalysis. This will be achieved by exploring a multi-technology approach, by combining existing methods with machine-learning.Firstly, existing physics-based data assimilation systems will be enhanced using new land data assimilation methodologies and new observations such as visible satellite observations. Secondly, hybrid machine-learning and physics-based data assimilation systems will be developed to emulate the existing systems, producing analyses in a much more efficient manner and combining observations with data driven forecasts. Thirdly, end-to-end data driven methods will be devised to produce reanalyses exclusively from observations.Each of the three technologies will use the latest available rescued historical satellite and in-situ data to improve the quality of reanalyses prior to 1979. In addition, methodologies to better exploit recently launched satellites, such as the Copernicus Sentinels, will be explored. Novel approaches to inter-calibration and the correction of systematic errors will be applied to the observations before they are collected into a single easy-to-use database which will be available within the project and beyond via the Anemoi framework.The three technologies and use of latest observations will be thoroughly tested in reanalysis demonstrators covering different time periods to analyse high impact case studies as well as multi-decadal runs to assess the reliability of climate trends. These demonstrators will be evaluated using novel diagnostics against independent observations of key ECVs across all of the Earth-system components. CHERRI will provide recommendations to C3S on the potential configurations for the next generation of reanalyses.