Improved Consistency of Satellite XCO2 Retrievals Based on Machine Learning.

Quantifying atmospheric CO2 over long periods from space is crucial in understanding the carbon cycle's response to climate change. However, a single satellite offers limited spatiotemporal coverage, making comprehensive monitoring challenging. Moreover, biases among various satellite retrievals hinder their direct integration. This study proposed a machine learning framework for fusing the column‐averaged dry‐air mole fraction of CO2 (XCO2) retrievals from Greenhouse Gases Observing Satellite (GOSAT) and OCO‐2 satellites. The best model (R2 = 0.85) presented improved consistency of GOSAT retrievals by reducing 71.5% of the average monthly bias while using OCO‐2 retrievals as a benchmark, indicating the fusion data set's potential to enhance observation coverage. Incorporating the adjusted GOSAT XCO2 retrievals into the OCO‐2 data set added an average of 84.7 thousand observations annually, enhancing the yearly temporal coverage by 53.6% (from 14 to 21.5 days per grid). This method can be adapted to other satellites, maximizing satellite resources for a more robust carbon flux inversion. Plain Language Summary: CO2 sources and sinks are primarily regulated by anthropogenic emissions, photosynthesis and respiration on land and in the ocean, as well as by physical dissolution and carbonate chemistry with ocean circulation. The consistent long‐term quantification of atmospheric CO2 concentrations using satellite observations plays a pivotal role in understanding the response of global and regional carbon cycles to climate change. However, satellites have a revisit period, and factors like cloud and aerosol scattering impact the quality and quantity of their observations. A single satellite currently falls short of the demand to monitor global carbon sources and sinks, necessitating the integration of observations from various satellites to conduct carbon flux inversions. Different satellites come with distinct sampling patterns, instrument parameters, and retrieval algorithms, which leads to biases in their retrieval products. Our study, focusing on OCO‐2 and GOSAT, employs machine learning models to improve consistency between retrievals derived from these two satellites, thus generating a harmonized data set. The bias‐corrected GOSAT XCO2 retrievals exhibit high spatiotemporal consistency with OCO‐2 XCO2 retrievals, immensely enhancing the observational constraints for carbon flux inversions. This method holds promise for application to recently launched and future satellites, aiming to offer carbon flux inversions with amplified spatiotemporal observational constraints. Key Points: This study employs machine learning (ML) models to enhance XCO2 consistency between OCO‐2 and Greenhouse Gases Observing Satellite (GOSAT), reducing monthly inconsistency by 71.5%Integrating the OCO‐2 data set with GOSAT retrievals increased yearly observations by 56.2%Fusing satellite data through ML models can pave the way for improved carbon flux inversions in the future [ABSTRACT FROM AUTHOR]