An update of the LDEO fCO2-Residual method: algorithmic choices improve ocean carbon sink estimates

Le résumé fourni par la source

We evaluate the impact of various algorithmic design choices on reconstruction skill and estimated air-sea CO2 flux using the fCO2-Residual machine learning (ML) method (Bennington et al., 2022a) to reconstruct surface ocean fCO2. We reconstruct fCO2 globally over the period 1982-2023 by optimizing the hyperparameter selection process (ResidualOPT) and/or using ΔfCO2-Residual (subtracting fCO2atm from fCO2ocn) as a target variable (ΔResidualOPT). We compare these reconstructions to the original fCO2-Residual approach (ResidualORG) from Bennington et al. (2022a). We find general agreement between the three approaches for the majority of the analysis period, but there are significant differences towards the beginning and end. Compared to ResidualORG, ΔResidualOPT and ResidualOPT show stronger ocean carbon uptake in the 1980s, with a global mean difference over 1982-1987 of 0.6 and 0.4 PgC/yr, respectively. In 2023, the global mean air-sea CO2 flux for both approaches is reduced by 0.4 PgC/yr. The test error metrics show that the highest reconstruction skill is achieved for ΔResidualOPT. We present an additional analysis of reconstruction fidelity of the fCO2-Residual method by using a testbed of CMIP6 Earth System Models to reconstruct fCO2. The testbed analysis is in agreement with the observation-based test error metrics; both approaches improve reconstruction skill compared to ResidualORG, but ΔResidualOPT leads to the lowest reconstruction errors for the full analysis period (1982-2023). We conclude that, for the fCO2-Residual method, the impacts of algorithmic design choices (i.e., hyperparameters or target variable) on reconstruction skill and estimated air-sea CO2 fluxes are significant, and it is possible to improve reconstruction fidelity even without the availability of additional observations.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.