Enhancing top-down HFC-134a emission estimates through parameter space exploration

Bayesian inverse modeling is a widely used approach for estimating greenhouse gas (GHG) emissions from atmospheric measurements. However, this method is subject to various uncertainties, including errors in the transport model, inaccuracies in baseline mole fractions, and uncertainties associated with the parameters of the Bayesian inversion framework.

In this study, we investigated the impact of these uncertainties on the Bayesian inversion of a key hydrofluorocarbon contributing to climate change, HFC-134a. We first conducted a grid search to refine the nudging parameters for simulating three-dimensional initial HFC fields using the FLEXible PARTicle-Linear Chemistry Module (FLEXPART-LCM). Subsequently, we employed Latin Hypercube Sampling (LHS) to explore inversion uncertainties by sampling a broad parameter space within the Bayesian inverse modeling framework FLEXINVERT.

Through over 250 ensemble simulations for initial fields and 15,000 ensemble inversion runs, we identified the most influential parameters and optimized configurations for the inverse modeling of HFC-134a. These findings improve the reliability of HFC-134a emission estimates and provide insights into the role of inversion parameters, applicable to the inversion of other greenhouse gases.