2019/07/05 by Unn Dahlén, Dahlen, Unn, Johan Linström +3
Computer Science · Environmental Science · #Applications (stat.AP) #Atmospheric and Environmental Gas Dynamics #Computation (stat.CO) #FOS: Computer and information sciences #Gaussian Processes and Bayesian Inference #Geochemistry and Geologic Mapping #Methodology (stat.ME)
paper · pdf · doi:10.48550/arxiv.1907.02706
openalex publication_date 2019/07/05 · openalex created_date 2022/07/28 · openalex updated_date 2026/07/28
Atmospheric inverse modelling is a method for reconstructing historical\nfluxes of green-house gas between land and atmosphere, using observed\natmospheric concentrations and an atmospheric tracer transport model. The small\nnumber of observed atmospheric concentrations in relation to the number of\nunknown flux components makes the inverse problem ill-conditioned, and\nassumptions on the fluxes are needed to constrain the solution. A common\npractise is to model the fluxes using latent Gaussian fields with a mean\nstructure based on estimated fluxes from combinations of process modelling\n(natural fluxes) and statistical bookkeeping (anthropogenic emissions). Here,\nwe reconstruct global CO flux fields by modelling fluxes using Gaussian Markov\nRandom Fields (GMRF), resulting in a flexible and computational beneficial\nmodel with a Mat 'ern-like spatial covariance, and a temporal covariance\ndefined through an auto-regressive model with seasonal dependence.\n In contrast to previous inversions, the flux is defined on a spatially\ncontinuous domain, and the traditionally discrete flux representation is\nreplaced by integrated fluxes at the resolution specified by the transport\nmodel. This formulation removes aggregation errors in the flux covariance, due\nto the traditional representation of area integrals by fluxes at discrete\npoints, and provides a model closer resembling real-life space-time continuous\nfluxes.\n