Summary of experience:
I have been working with data assimilation methods for more than sixteen years. My experience includes the implementation and use of data assimilation for various models (mostly geosciences models) and with various methods.
Between 2003 and 2012, I have been developing in Fortran the Valentina chemical data assimilation system. Valentina is a flexible and modular variational assimilation suite allowing global and regional chemical transport models to assimilate data with a 3D-Var, a 3D-FGAT or a 4D-Var method (see publications list). I further develop this activity within the framework of the series of the Monitoring Atmospheric Composition and Climate (MACC) European projects. These projects helped me to build a team of three people around Valentina, team I managed during several years.
Together we coupled Valentina with the global and the regional versions of the Meteo-France/CNRM Mocage chemistry transport model. The regional version of this coupled system is currently used operationally in the Copernicus Atmosphere Monitoring Service (CAMS) to produce regional air quality analyses.
Although I have mainly focused on variational assimilation methods for the development of Valentina, I am also very familiar with Kalman filter methods and ensemble methods. Part of my studies are based on hybrid methods and aimed to diagnose and formulate the covariance matrix of the background error, both for a global system assimilating satellite data (Pannekoucke and Massart, 2008, Massart et al., 2011) and a regional system assimilating in situ data (Jaumouille et al., 2011). My experience also covers some more methodological aspects of data assimilation like highlighting some issues of the 3D-FGAT method for transport models (Massart et al., 2010).
Apart from atmospheric chemistry, my work concerned several other components of the Earth system. I used to work with Nemo-Var, the assimilation system which is used for the ocean reanalysis ORAS5. I also participated to the development of an assimilation system for the French National Hydrometeorological and Flood Forecasting Center (SCHAPI) to further improve their operational flood forecast over France.
I joined ECMWF in 2012 to further develop the assimilation of the greenhouse gas satellite data as part of the MACC projects and then CAMS. In parallel, I explored and expanded hybrid assimilation methods similar to the ones used in the Research Department. I first adapted ECMWF Ensemble of Data Assimilations (EDA) to the CAMS greenhouse gases in collaboration with the Data Assimilation Methodology Team of the Earth System Assimilation Section (Massart and Bonavita, 2016). I also explored the use of ECMWF Ensemble Kalman Filter (EnKF) for the greenhouse gases in collaboration with the Numerical Methods team of the Earth System Modelling Section.
Since end of 2016 I moved to the data assimilation methodology team to work mainly on hydrib 4D-Var.