Estimating Biases for Regional Methane Fluxes Using Spatially/Temporally Varying Random-Field Representations of Biases

Methane is a powerful greenhouse gas, and the development and improvement of emissions models rely on understanding the flux of methane released from anthropogenic sources relative to releases from other sources. Increasing production of shale oil and gas in the mid-latitudes and associated fugitive emissions are suspected to be a dominant contributor to the global methane increase. Landfills, sewage treatment, and other sources may be dominant in some parts of the U.S. Current approaches for measuring regional emissions yield highly uncertain estimates because of the sparsity of measurement sites and the presence of multiple simultaneous sources. Methods for effective assimilation of data from a variety of sources are critically needed to perform regional GHG attribution with existing measurements and to determine how to structure future measurement systems including satellites.

We present a hierarchical Bayesian framework to estimate corrections in methane fluxes based on atmospheric concentration measurements and a Lagrangian transport model (Weather Research and Forecasting - Stochastic Time-Inverted Lagrangian Transport). Spatial-temporal model corrections and parameter uncertainties are modeled as spectral random fields via Karhunen-Loeve expansions. We conduct the analyses at regional scales that are based on similar geographical and meteorological conditions. Sequential Monte Carlo algorithms are used to sample the high-dimensional posterior distributions and Bayesian model evidence is employed to select the appropriate model parsimony as informed by the data.