Objectified quantification of uncertainties in Bayesian atmospheric inversions
Classical Bayesian atmospheric inversions process atmospheric observations and prior emissions, the two being connected by an observation operator picturing mainly the atmospheric transport. These inversions rely on prescribed errors in the observations, the prior emissions and the observation oper...
Main Authors: | , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Copernicus Publications
2015-05-01
|
Series: | Geoscientific Model Development |
Online Access: | http://www.geosci-model-dev.net/8/1525/2015/gmd-8-1525-2015.pdf |
_version_ | 1811195227696791552 |
---|---|
author | A. Berchet I. Pison F. Chevallier P. Bousquet J.-L. Bonne J.-D. Paris |
author_facet | A. Berchet I. Pison F. Chevallier P. Bousquet J.-L. Bonne J.-D. Paris |
author_sort | A. Berchet |
collection | DOAJ |
description | Classical Bayesian atmospheric inversions process atmospheric
observations and prior emissions, the two being connected by an
observation operator picturing mainly the atmospheric transport.
These inversions rely on prescribed errors in the observations, the
prior emissions and the observation operator. When data pieces are
sparse, inversion results are very sensitive to the prescribed error
distributions, which are not accurately known. The classical
Bayesian framework experiences difficulties in quantifying the
impact of mis-specified error distributions on the optimized fluxes.
In order to cope with this issue, we rely on recent research results
to enhance the classical Bayesian inversion framework through a
marginalization on a large set of plausible errors that can be
prescribed in the system. The marginalization consists in computing
inversions for all possible error distributions weighted by the
probability of occurrence of the error distributions. The posterior
distribution of the fluxes calculated by the marginalization is not
explicitly describable. As a consequence, we carry out a Monte Carlo
sampling based on an approximation of the probability of occurrence
of the error distributions. This approximation is deduced from the
well-tested method of the maximum likelihood estimation. Thus, the
marginalized inversion relies on an automatic objectified diagnosis
of the error statistics, without any prior knowledge about the
matrices. It robustly accounts for the uncertainties on the error
distributions, contrary to what is classically done with frozen
expert-knowledge error statistics. Some expert knowledge is still
used in the method for the choice of an emission aggregation pattern
and of a sampling protocol in order to reduce the computation cost.
The relevance and the robustness of the method is tested on a case
study: the inversion of methane surface fluxes at the mesoscale
with virtual observations on a realistic network in Eurasia.
Observing system simulation experiments are carried out with
different transport patterns, flux distributions and total prior
amounts of emitted methane. The method proves to consistently
reproduce the known "truth" in most cases, with satisfactory
tolerance intervals. Additionally, the method explicitly provides
influence scores and posterior correlation matrices. An in-depth
interpretation of the inversion results is then possible. The more
objective quantification of the influence of the observations on the
fluxes proposed here allows us to evaluate the impact of the
observation network on the characterization of the surface fluxes.
The explicit correlations between emission aggregates reveal the
mis-separated regions, hence the typical temporal and spatial scales
the inversion can analyse. These scales are consistent with the
chosen aggregation patterns. |
first_indexed | 2024-04-12T00:39:52Z |
format | Article |
id | doaj.art-ad9d705ee9214af8b83a17395aee9264 |
institution | Directory Open Access Journal |
issn | 1991-959X 1991-9603 |
language | English |
last_indexed | 2024-04-12T00:39:52Z |
publishDate | 2015-05-01 |
publisher | Copernicus Publications |
record_format | Article |
series | Geoscientific Model Development |
spelling | doaj.art-ad9d705ee9214af8b83a17395aee92642022-12-22T03:55:03ZengCopernicus PublicationsGeoscientific Model Development1991-959X1991-96032015-05-01851525154610.5194/gmd-8-1525-2015Objectified quantification of uncertainties in Bayesian atmospheric inversionsA. Berchet0I. Pison1F. Chevallier2P. Bousquet3J.-L. Bonne4J.-D. Paris5Laboratoire des Sciences du Climat et de l'Environnement, CEA-CNRS-UVSQ, IPSL, Gif-sur-Yvette, FranceLaboratoire des Sciences du Climat et de l'Environnement, CEA-CNRS-UVSQ, IPSL, Gif-sur-Yvette, FranceLaboratoire des Sciences du Climat et de l'Environnement, CEA-CNRS-UVSQ, IPSL, Gif-sur-Yvette, FranceLaboratoire des Sciences du Climat et de l'Environnement, CEA-CNRS-UVSQ, IPSL, Gif-sur-Yvette, FranceLaboratoire des Sciences du Climat et de l'Environnement, CEA-CNRS-UVSQ, IPSL, Gif-sur-Yvette, FranceLaboratoire des Sciences du Climat et de l'Environnement, CEA-CNRS-UVSQ, IPSL, Gif-sur-Yvette, FranceClassical Bayesian atmospheric inversions process atmospheric observations and prior emissions, the two being connected by an observation operator picturing mainly the atmospheric transport. These inversions rely on prescribed errors in the observations, the prior emissions and the observation operator. When data pieces are sparse, inversion results are very sensitive to the prescribed error distributions, which are not accurately known. The classical Bayesian framework experiences difficulties in quantifying the impact of mis-specified error distributions on the optimized fluxes. In order to cope with this issue, we rely on recent research results to enhance the classical Bayesian inversion framework through a marginalization on a large set of plausible errors that can be prescribed in the system. The marginalization consists in computing inversions for all possible error distributions weighted by the probability of occurrence of the error distributions. The posterior distribution of the fluxes calculated by the marginalization is not explicitly describable. As a consequence, we carry out a Monte Carlo sampling based on an approximation of the probability of occurrence of the error distributions. This approximation is deduced from the well-tested method of the maximum likelihood estimation. Thus, the marginalized inversion relies on an automatic objectified diagnosis of the error statistics, without any prior knowledge about the matrices. It robustly accounts for the uncertainties on the error distributions, contrary to what is classically done with frozen expert-knowledge error statistics. Some expert knowledge is still used in the method for the choice of an emission aggregation pattern and of a sampling protocol in order to reduce the computation cost. The relevance and the robustness of the method is tested on a case study: the inversion of methane surface fluxes at the mesoscale with virtual observations on a realistic network in Eurasia. Observing system simulation experiments are carried out with different transport patterns, flux distributions and total prior amounts of emitted methane. The method proves to consistently reproduce the known "truth" in most cases, with satisfactory tolerance intervals. Additionally, the method explicitly provides influence scores and posterior correlation matrices. An in-depth interpretation of the inversion results is then possible. The more objective quantification of the influence of the observations on the fluxes proposed here allows us to evaluate the impact of the observation network on the characterization of the surface fluxes. The explicit correlations between emission aggregates reveal the mis-separated regions, hence the typical temporal and spatial scales the inversion can analyse. These scales are consistent with the chosen aggregation patterns.http://www.geosci-model-dev.net/8/1525/2015/gmd-8-1525-2015.pdf |
spellingShingle | A. Berchet I. Pison F. Chevallier P. Bousquet J.-L. Bonne J.-D. Paris Objectified quantification of uncertainties in Bayesian atmospheric inversions Geoscientific Model Development |
title | Objectified quantification of uncertainties in Bayesian atmospheric inversions |
title_full | Objectified quantification of uncertainties in Bayesian atmospheric inversions |
title_fullStr | Objectified quantification of uncertainties in Bayesian atmospheric inversions |
title_full_unstemmed | Objectified quantification of uncertainties in Bayesian atmospheric inversions |
title_short | Objectified quantification of uncertainties in Bayesian atmospheric inversions |
title_sort | objectified quantification of uncertainties in bayesian atmospheric inversions |
url | http://www.geosci-model-dev.net/8/1525/2015/gmd-8-1525-2015.pdf |
work_keys_str_mv | AT aberchet objectifiedquantificationofuncertaintiesinbayesianatmosphericinversions AT ipison objectifiedquantificationofuncertaintiesinbayesianatmosphericinversions AT fchevallier objectifiedquantificationofuncertaintiesinbayesianatmosphericinversions AT pbousquet objectifiedquantificationofuncertaintiesinbayesianatmosphericinversions AT jlbonne objectifiedquantificationofuncertaintiesinbayesianatmosphericinversions AT jdparis objectifiedquantificationofuncertaintiesinbayesianatmosphericinversions |