Uncertainty in continuous ΔCO-based ΔffCO₂ estimates derived from ¹⁴C flask and bottom-up ΔCO ∕ ΔffCO₂ ratios

Abstract

Measuring the 14C / C depletion in atmospheric CO2 compared with a clean-air reference is the most direct way to estimate the recently added CO2 contribution from fossil fuel (ff) combustion (ΔffCO2) in ambient air. However, as 14CO2 measurements cannot be conducted continuously nor remotely, there are only very sparse 14C-based ΔffCO2 estimates available. Continuously measured tracers, like carbon monoxide (CO), that are co-emitted with ffCO2 can be used as proxies for ΔffCO2, provided that the ΔCO / ΔffCO2 ratios can be determined correctly (here, ΔCO refers to the CO excess compared with a clean-air reference). In the present study, we use almost 350 14CO2 measurements from flask samples collected between 2019 and 2020 at the urban site Heidelberg, Germany, and corresponding analyses from more than 50 afternoon flasks collected between September 2020 and March 2021 at the rural ICOS site Observatoire pérenne de l'environnement (OPE), France, to calculate average 14C-based ΔCO / ΔffCO2 ratios for those sites. For this, we constructed a clean-air reference from the 14CO2 and CO measurements of Mace Head, Ireland. By dividing the hourly ΔCO excess observations by the averaged flask ratio, we calculate continuous proxy-based ΔffCO2 records. The mean bias between the proxy-based ΔffCO2 and the direct 14C-based ΔffCO2 estimates from the flasks is – with 0.31 ± 3.94 ppm for the urban site Heidelberg and −0.06 ± 1.49 ppm for the rural site OPE – only ca. 3 % at both sites. The root-mean-square deviation (RMSD) between proxy-based ΔffCO2 and 14C-based ΔffCO2 is about 4 ppm for Heidelberg and 1.5 ppm for OPE. While this uncertainty can be explained by observational uncertainties alone at OPE, about half of the uncertainty is caused by the neglected variability in the ΔCO / ΔffCO2 ratios at Heidelberg. We further show that modeled ratios based on a bottom-up European emission inventory would lead to substantial biases in the ΔCO-based ΔffCO2 estimates for both Heidelberg and OPE. This highlights the need for an ongoing observational calibration and/or validation of inventory-based ratios if they are to be applied for large-scale ΔCO-based ΔffCO2 estimates, e.g., from satellites.