State-wide California 2020 carbon dioxide budget estimated with OCO-2 and OCO-3 satellite data

Abstract. Satellite observations are instrumental in observing spatiotemporal variability in carbon dioxide (CO2) concentrations, which can be used to derive fluxes of this greenhouse gas. This study leverages NASA's Orbiting Carbon Observatory-2 and -3 (OCO-2 and OCO-3, respectively) CO2 observations with a Gaussian process (GP) machine learning inverse model, a Bayesian nonparametric approach well suited for integrating the unique spatiotemporal characteristics of these satellite observations, to estimate subregional CO2 fluxes. Utilizing the GEOS-Chem chemical transport model (CTM) to simulate column-average CO2 concentrations (XCO2) for 2020 in California – a period marked by the coronavirus disease (COVID-19) pandemic, drought conditions, and significant wildfire activity – we estimated the state-wide CO2 emission rates constrained by OCO-2/3. This study developed prior fossil fuel emissions to reflect reduced activities during the COVID-19 pandemic, while net ecosystem exchange (NEE) and fire emissions were derived based on satellite data. GEOS-Chem source-specific XCO2 concentrations for fossil fuels, NEE, fire, and oceanic sources were simulated coincident to OCO-2/3 XCO2 retrievals to estimate state-wide sector-specific and total CO2 emissions. GP inverse model results suggest that annual posterior median fossil fuel emissions were consistent with prior estimates (317.8 and 338.4 Tg CO2 yr−1, respectively; 95 % confidence level) and that posterior NEE fluxes had less carbon uptake compared to prior fluxes (−36.8 vs. −99.2 Tg CO2 yr−1, respectively; 95 % confidence level). Posterior fire CO2 emissions were estimated to be 68.0 Tg CO2 yr−1, which was much lower than a priori estimates (103.3 Tg CO2 yr−1). The total median annual CO2 emissions for the state of California in 2020 were estimated to be 349.6 Tg CO2 yr−1 (range of 272.8–428.6 Tg CO2 yr−1; 95 % confidence level), aligning closely with the prior total estimate of 342.5 Tg CO2 yr−1. This study, for the first time, demonstrates that OCO-2/3 XCO2 observations can be assimilated into inverse models to estimate state-wide source-specific CO2 fluxes on a seasonal and annual scale.