A Bayesian Framework to Quantify Uncertainty in Aerosol Optical Model Selection Applied to TROPOMI Measurements

• Main Authors: Kauppi, Anu, Kukkurainen, Antti, Lipponen, Antti, Laine, Marko, Arola, Antti, Lindqvist, Hannakaisa, Tamminen, Johanna
• Other Authors: Ilmatieteen laitos, Finnish Meteorological Institute, orcid:0009-0008-1000-0350, orcid:0000-0002-3371-7337, orcid:0000-0002-6902-9974, orcid:0000-0002-5914-6747, orcid:0000-0002-9220-0194, orcid:0000-0001-9202-906X, orcid:0000-0003-3095-0069
• Format: Article in Journal/Newspaper
• Language: English
• Published: MDPI 2024
• Subjects: aerosols; remote sensing; modelling (representation); atmosphere (earth); troposphere; tracking; uncertainty; Bayesian analysis; fine particles; probability calculation; aerosolit; kaukokartoitus; mallintaminen; ilmakehä; troposfääri; seuranta; epävarmuus; bayesialainen menetelmä; pienhiukkaset; todennäköisyyslaskenta; The Sentinel; Aerosol Robotic Network
• Online Access: http://hdl.handle.net/10138/577165

This article presents a method within a Bayesian framework for quantifying uncertainty in satellite aerosol remote sensing when retrieving aerosol optical depth (AOD). By using a Bayesian model averaging technique, we take into account uncertainty in aerosol optical model selection and also obtain a shared inference about AOD based on the best-fitting optical models. In particular, uncertainty caused by forward-model approximations has been taken into account in the AOD retrieval process to obtain a more realistic uncertainty estimate. We evaluated a model discrepancy, i.e., forward-model uncertainty, empirically by exploiting the residuals of model fits and using a Gaussian process to characterise the discrepancy. We illustrate the method with examples using observations from the TROPOspheric Monitoring Instrument (TROPOMI) on the Sentinel-5 Precursor satellite. We evaluated the results against ground-based remote sensing aerosol data from the Aerosol Robotic Network (AERONET).