Seamless reconstruction and spatiotemporal analysis of satellite-based XCO2 incorporating temporal characteristics: A case study in China during 2015–2020

Abstract

Carbon dioxide (CO2) is a crucial greenhouse gas, and its concentration and spatiotemporal characteristics are among the principal sources of uncertainty in global warming assessments. Satellite remote sensing is a widely adopted, high-accuracy approach for monitoring atmospheric CO2. However, limited swath width and cloud cover significantly reduce satellite observation coverage. This study addresses the temporal changes in CO2 concentration and utilizes a machine learning-based fusion of multiple data sources to generate daily, full-coverage, 0.05° spatial resolution column-averaged CO2 concentration data for China from 2015 to 2020. Ten-fold cross-validation yielded a determination coefficient R2 of 0.97, root mean square error of 0.92 ppm, and mean absolute error of 0.59 ppm. Compared to other datasets, this study’s dataset exhibits superior accuracy and spatiotemporal detail. Using the produced CO2 concentration data in this study, we conducted a spatiotemporal analysis of CO2 concentrations in China. The results indicate that, in general, the Western region exhibits a higher growth rate in CO2 concentration than the Eastern and Central regions, with areas of lower CO2 concentration experiencing higher growth rates while regions with higher CO2 concentration have lower growth rates. Moreover, the highest increase in CO2 concentration occurred in 2016, with a substantial decrease in CO2 concentration growth observed in 2018. Notably, the reduction in CO2 concentration in the Qinghai-Tibet Plateau region during the summer is considerably smaller than in other regions, possibly due to atmospheric transport from the Indian Peninsula.