Computationally efficient methods for large-scale atmospheric inverse modeling

Abstract. Atmospheric inverse modeling describes the process of estimating greenhouse gas fluxes or air pollution emissions at Earth's surface using observations these gases collected in atmosphere. The launch new satellites, expansion observation networks, and a desire for more detailed maps have yielded numerous computational statistical challenges standard frameworks that were often originally designed with much smaller data sets mind. In this article, we discuss computationally efficient methods large-scale atmospheric focus on addressing some main practical challenges. We develop generalized hybrid projection methods, which are iterative solving problems, specifically case fluxes. These algorithms confer several advantages. They efficient, part because they converge quickly, exploit matrix–vector multiplications, do not require inversion any matrices. also robust can accurately reconstruct fluxes, automatic since regularization covariance matrix parameters stopping criteria be determined as algorithm, flexible paired many different types models. demonstrate benefits study from NASA's Orbiting Carbon Observatory 2 (OCO-2) satellite. then address challenging problem model when mean is known priori; so by reformulating problem, thereby extending applicability to include hierarchical priors. further show exploiting mathematical relations provided method, efficiently calculate an approximate posterior variance, providing uncertainty information.