GOSAT and in-situ based inversion of North American CO2 fluxes using FLEXPART

Temperate North America (TNA) consists of moisture-limited ecosystems in the west and mainly forests and croplands in the east, each exhibiting distinct responses of net ecosystem exchange (NEE) to climate variability. A dense in-situ network exists over TNA and is complemented with satellite XCO₂ measurements providing strong constraints on sub-continental flux variability, but NEE response to environmental drivers remains poorly understood (Byrne et al., 2020).  

In this work, we develop a regional inversion systems that can fully exploit the available observational density to better understand NEE response to climate anomalies on high spatial resolution.  

We combine in-situ observations with column-averaged CO₂ (XCO₂) from the Greenhouse Gases Observing Satellite (GOSAT) for a full year. For each observation we compute source-receptor relationships using the Lagrangian Particle Dispersion Model FLEXPART. These footprints serve as the forward model to link surface fluxes to atmospheric measurements. We optimize weekly total surface fluxes on a 2° × 2° grid across TNA and derive NEE by subtracting contributions from fossil fuel emissions, biomass burning, and ocean fluxes.  

We assess the sensitivity of inferred NEE to key methodological choices, including model data mismatch errors, assumed spatial and temporal error correlations and the representation of the diurnal cycle in biospheric exchange. For a preferred configuration, we discuss the spatial patterns of NEE in TNA. We also compare the resulting NEE fluxes to those from the global TM5-4DVar inversion and find good agreement in both spatial patterns and temporal variability, while our regional system provides enhanced spatial detail. We conclude by outlining next steps for improvements and discuss opportunities enabled by the high-resolution inversion for diagnosing TNA's carbon–climate processes.