Greenhouse gas emission mapping and quantification based on 3D transport modeling and Bayesian state estimation

New techniques for monitoring greenhouse gas (GHG) emissions are urgently needed in multiple spatial scales, to find ways to effectively mitigate emissions and for sustainable policy-making. This paper focuses on a novel facility-scale monitoring technique which uses sequential multi-beam open-path laser spectroscopy measurements and wind field data to infer spatio-temporal distributions of GHG emission sources. The problem of mapping and quantifying the GHG source is considered as a non-stationary, three-dimensional (3D) tomography problem in which the GHG source is reconstructed simultaneously with the temporally evolving 3D concentration of the GHG. The tomographic reconstruction problem is highly rank deficient due to the practically realizable limited-angle setup of sensors. However, the proposed approach for the reconstruction takes advantage of the non-stationarity of the inverse problem – we model the temporal evolution of the 3D concentration distribution and the contribution of the source to it by a convection-diffusion model in which the wind field data is incorporated. The reconstruction problem is formulated in the framework of Bayesian state estimation, using laser spectroscopy observations and the evolution model. Particularly, Kalman smoother estimates are computed. In this paper, the 3D tomographic reconstruction of GHG concentration and source is tested both with numerical simulation studies and with experimental field data. The results demonstrate the feasibility of the approach for mapping and quantifying various types of GHG sources and for quantifying their uncertainties in multiple wind conditions.