Machine Learning-Based Observation Operators to Assimilate Microwave and SIF Satellite Observations into the ECMWF Integrated Forecast System

The CO2MVS Research on Supplementary Observations (CORSO) project aims at reducing the uncertainties in the predicted biogenic carbon fluxes by leveraging new types of satellite observations, machine learning and data assimilation. The objective of this work is to implement the assimilation of active microwave and solar-induced chlorophyll fluorescence (SIF) satellite observations in the Integrated Forecast System (IFS) developed at the European Centre for Medium-Range Weather Forecasts (ECMWF) to jointly analyze soil moisture and leaf area index (LAI). The assimilation of observation requires the development of an observation operator to predict the model-simulated counterpart of the observation from the model fields. In this paper, we present the development of machine learning-based observation operators for the normalized backscatter at 40° from the Advanced Scatterometer (ASCAT) instrument and SIF derived from the TROPOspheric Monitoring Instrument (TROPOMI). Results show that deep learning methods provide accurate predictions of the satellite signals from the IFS model fields. The next step will be the implementation of these operators in the IFS and the evaluation of the impacts of analyzing both soil moisture and vegetation variables on the estimation of carbon and water fluxes along with near-surface meteorological variables.