Improving biospheric models of net ecosystem exchange in the eastern U.S. to inform top-down anthropogenic CO2 emission estimates during the growing season in the DC/ Baltimore metropolitan area

Estimating urban greenhouse gas (GHG) emissions using atmospheric inverse models requires proper quantification of GHG mole fractions in incoming air masses to the region of interest. Urban inverse models are particularly complicated on the eastern coast of the U.S., given the large magnitude and variability of background CO₂ concentrations, influenced by upwind power plants and an active biosphere during the growing season. Biospheric fluxes in the local metropolitan area itself also complicate efforts to separately identify anthropogenic CO₂ emissions. This study forms part of a larger project aimed at estimating GHG emissions in the DC/ Baltimore metropolitan area using atmospheric observations from a recently-installed surface tower network in the region. To help improve anthropogenic CO₂ emission estimates during the growing season, this work focuses on improving biospheric models of net ecosystem exchange (NEE) in DC/ Baltimore and upwind areas. NEE estimates can then be used to either improve modeled background concentration fields around the DC/ Baltimore area, or as priors in a larger regional inversion, which can assign fluxes to both DC/ Baltimore and upwind locations. Towards this end, we run the Vegetation Photosynthesis and Respiration Model (VPRM) to estimate NEE for the eastern U.S. and southern Canada at 1 km² resolution using MODIS satellite data, WRF meteorological variables, and parameters optimized using flux tower data within our domain. Modifications are made in urban areas to reduce the respiration signal in a manner consistent with vegetation and impervious surface area coverage. NEE estimates from VPRM are also compared with estimates from two more complex process-based biospheric models: VEGAS, a prognostic dynamic vegetation model and CASA-GFED, a diagnostic model which, like VPRM, uses satellite data to capture seasonality in vegetation coverage. The three models are compared at diurnal, daily and seasonal timescales, and at different locations where flux tower-based estimates of NEE may be available. The quality of the three models is also assessed by transporting CO₂ surface fluxes forward to various regional towers and ACT-America aircraft campaign flight tracks using WRF-STILT footprints and comparing simulated to observed CO₂ concentration data at those locations.