Optimizing CO2 emission estimates in Paris through enhanced urban atmospheric monitoring

As part of the EU-funded PAUL project (ICOS Cities), the metropolitan area of Paris, in parallel with Munich and Zurich, has been instrumented with various observing systems to define the most-suitable approaches for CO₂ emissions monitoring. This effort is underpinned by an extensive urban atmospheric monitoring network, comprising nine towers equipped with high-accuracy and mid-cost sensors designed to capture  variations in atmospheric concentrations. Driven by 1-km meteorological fields (from WRF), the Stochastic Time-Inverted Lagrangian Transport (STILT) model has been employed for backward simulations of CO₂ enhancements based on state-of-the-art high-resolution inventories for 2023. Transport errors have been significantly reduced ( from about 4-5 m/s to  1~2 m/s) through the assimilation of three-dimensional wind profiles obtained from multiple Lidar data over Paris (Urbisphere project), using 3DVar data cycling assimilation. Fossil fuel emissions (TNO, AirParif) and biogenic emissions (using offline VPRM MODIS and Sentinel-2) serve as prior inventories in our inverse modeling framework. This framework employs a Bayesian inversion technique producing hourly fluxes with time-varied adaptive mesh grids (1 km in the downtown area, gradually aggregated to 100 km across the region), balancing computational efficiency with inversion accuracy near monitoring sites. However, direct comparisons revealed systematic discrepancies in the inversion results, particularly in the adjustments between anthropogenic and biogenic emissions. To address this, radiocarbon (¹⁴C) observations from two Parisian sites were incorporated as additional constraints, improving the partitioning of fossil fuel and biogenic contributions in the inversion.