An atmospheric inversion over the city of Cape Town: sensitivity analyses

• Published in: Atmospheric chemistry and physics 2019-06, Vol.19 (11), p.7789-7816
• Main Authors: Nickless, Alecia, Rayner, Peter J, Scholes, Robert J, Engelbrecht, Francois, Erni, Birgit
• Format: Article
• Language: English
• Subjects: Air pollution research; Anthropogenic factors; Atmospheric research; Backup software; Biosphere; Building components; Carbon dioxide; Carbon dioxide concentration; Carbon dioxide measurements; Components; Correlation; Covariance matrix; Distribution; Error detection; Estimates; Fluctuations; Fluxes; Fossil fuels; In situ measurement; Information sources; Inversion; Inversions; Methods; Reduction; Sensitivity; Sensitivity analysis; Spatial distribution; Telephone cables; Temperature inversions (Meteorology); Uncertainty; Uptake
• ISSN: 1680-7324; 1680-7316; 1680-7324
• DOI: 10.5194/acp-19-7789-2019

An atmospheric inversion was performed for the city of Cape Town for the period of March 2012 to June 2013, making use of in situ measurements of CO.sub.2 concentrations at temporary measurement sites located to the north-east and south-west of Cape Town. This paper presents results of sensitivity analyses that tested assumptions regarding the prior information and the uncertainty covariance matrices associated with the prior fluxes and with the observations. Alternative prior products were considered in the form of a carbon assessment analysis to provide biogenic fluxes and the ODIAC (Open-source Data Inventory for Anthropogenic CO.sub.2 product) fossil fuel product. These were used in place of the reference inversion's biogenic fluxes from CABLE (Community Atmosphere Biosphere Land Exchange model) and fossil fuel emissions from a bespoke inventory analysis carried out specifically for the Cape Town inversion. Our results confirmed that the inversion solution was strongly dependent on the prior information, but by using independent alternative prior products to run multiple inversions, we were able to infer limits for the true domain flux. Where the reference inversion had aggregated prior flux estimates that were made more positive by the inversion - suggesting that CABLE was overestimating the amount of CO.sub.2 biogenic uptake - the carbon assessment prior fluxes were made more negative by the inversion. As the posterior estimates tended towards the same point, we could infer that the best estimate was located somewhere between these two posterior fluxes.