The importance of baseline methods for the consistent estimation of regional and global emissions from inverse modeling

Inverse modeling provides a powerful statistical tool to verify national greenhouse gas (GHG) emission inventories by making use of in-situ atmospheric observations. Regional inversions are often based on atmospheric transport simulations with Lagrangian Particle Dispersion Models (LPDMs), where a large number of virtual particles are released from observation sites and traced backward for a limited amount of time to establish a relationship between atmospheric concentrations and emission sources within the simulated period. In order to account for all emissions prior to this simulation period, which also contribute to the corresponding observations, a baseline needs to be defined. This baseline definition is a crucial task, bearing a lot of uncertainties. Most studies investigating halocarbons use statistical methods to calculate the baseline by selecting low concentration observations at individual stations. In this study we show that statistical baseline methods have large systematic problems, that accumulate with increasing backward simulation periods and lead to a non-negligible underestimation of emissions that systematically increases with the length of the backward simulation time. As an alternative, we present a global distribution based (GDB) approach, where baseline concentrations are determined directly from global concentration fields at the termination points of the backward trajectories. These global fields are simulated with the FLEXible PARTicle dispersion chemical transport model (FLEXPART CTM) using a nudging routine to push modeled concentrations towards observed concentrations. We illustrate that this method is fully consistent with the length of the backward simulation, has the ability to account for meteorological variability, and leads to inversion results, that agree well with global emissions calculated with a simple box model.