Assessing Lagrangian inverse modelling of urban anthropogenic CO2 fluxes using in situ aircraft and ground-based measurements in the Tokyo area

Abstract Background In order to use in situ measurements to constrain urban anthropogenic emissions of carbon dioxide (CO2), we use a Lagrangian methodology based on diffusive backward trajectory tracer reconstructions and Bayesian inversion. The observations of atmospheric CO2 were collected within the Tokyo Bay Area during the Comprehensive Observation Network for TRace gases by AIrLiner (CONTRAIL) flights, from the Tsukuba tall tower of the Meteorological Research Institute (MRI) of the Japan Meteorological Agency and at two surface sites (Dodaira and Kisai) from the World Data Center for Greenhouse Gases (WDCGG). Results We produce gridded estimates of the CO2 emissions and calculate the averages for different areas within the Kanto plain where Tokyo is located. Using these inversions as reference we investigate the impact of perturbing different elements in the inversion system. We modified the observations amount and location (surface only sparse vs. including aircraft CO2 observations), the background representation, the wind data used to drive the transport model, the prior emissions magnitude and time resolution and error parameters of the inverse model. Conclusions Optimized fluxes were consistent with other estimates for the unperturbed simulations. Inclusion of CONTRAIL measurements resulted in significant differences in the magnitude of the retrieved fluxes, 13% on average for the whole domain and of up to 21% for the spatiotemporal cells with the highest fluxes. Changes in the background yielded differences in the retrieved fluxes of up to 50% and more. Simulated biases in the modelled transport cause differences in the retrieved fluxes of up to 30% similar to those obtained using different meteorological winds to advect the Lagrangian trajectories. Perturbations to the prior inventory can impact the fluxes by ~ 10% or more depending on the assumptions on the error covariances. All of these factors can cause significant differences in the estimated flux, and highlight the challenges in estimating regional CO2 fluxes from atmospheric observations.