Estimation, variations, and impact factors of high-resolution global daily near-surface CO₂ during 2015–2021 based on OCO-2 and machine learning

Abstract

Carbon dioxide (CO₂) is a key component of the global carbon cycle and major driver of climate change. Although ground-based measurements and satellite remote sensing have been conducted globally, high-resolution, high-accuracy datasets of near-surface CO₂ concentrations remain scarce, limiting the understanding of their distribution and variation trends. To address this gap, we established a novel machine learning-based framework to generate global daily near-surface CO₂ concentrations at a spatial resolution of 0.5° × 0.625° during 2015–2021. By integrating satellite data, ground observations, meteorological and climatic variables, a vegetation index, and anthropogenic indicators, we developed a LightGBM prediction model that effectively captures spatiotemporal CO₂ patterns, achieving a correlation coefficient of 0.89 and a root-mean-square error of 3.5 ppm. The estimation results revealed that the global annual mean near-surface CO₂ concentrations during 2015–2021 were 408.71 ± 2.98 ppm, with an average growth rate of 2.66 ± 0.27 ppm/yr. High growth rates were observed in Southeast Asia, the South China Sea, West Africa, and the Amazon, with the greatest increase occurring in 2016. Compared with most other regions, South Asia, a high-carbon-emission region, exhibited greater and more variable near-surface CO₂ concentrations. Model interpretability analysis revealed that temperature strongly positively influenced the near-surface CO₂ in tropical regions, whereas evaporation and soil type positively affected these concentrations in arid regions. Soil improvements and large-scale afforestation were potential mitigation strategies to reduce the CO₂ levels in high-concentration areas. Daily high-resolution mapping enables the detection of short-term CO₂ increases resulted from large-scale wildfires, supporting regional emissions verification and near-term carbon emissions assessment.

Graphical abstract

Keywords

CO₂ / Near-surface / Variation patterns / Remote sensing inversion / OCO-2 / Machine learning

Highlight

	● LightGBM achieves high accuracy ( R 2 = 0.89), surpassing regional transport models.
	● Near-surface CO2 grows faster than XCO2 in the Amazon, reflecting surface dynamics.
	● Temperature, evaporation, and soil type affect CO2 in deserts and barren areas.
	● Wildfires elevate surface CO2, with increases of up to 3.45 ppm during large fires.

References

[1] Acharya S C , Babel M S , Madsen H , Sisomphon P , Shrestha S . (2020). Comparison of different quantile regression methods to estimate predictive hydrological uncertainty in the Upper Chao Phraya River Basin, Thailand. Journal of Flood Risk Management, 13(1): e12585

[2] Adiya S , Dalantai S , Wu T H , Wu X D , Yamkhin J , Bao Y H , Sumiya E , Yadamsuren G , Avirmed D , Dorjgotov B . (2021). Spatial and temporal change patterns of near-surface CO₂ and CH₄ concentrations in different permafrost regions on the Mongolian Plateau from 2010 to 2017. Science of the Total Environment, 800: 149433

[3] Amina M K , Chithra N R . (2023). Predictive uncertainty assessment in flood forecasting using quantile regression. H₂Open Journal, 6(3): 477–492

[4] Aragão L E O C , Anderson L O , Fonseca M G , Rosan T M , Vedovato L B , Wagner F H , Silva C V J , Silva Junior C H L , Arai E , Aguiar A P . et al. (2018). 21st Century drought-related fires counteract the decline of Amazon deforestation carbon emissions. Nature Communications, 9: 536

[5] Balamurugan V , Chen J , Wenzel A , Keutsch F N . (2023). Spatiotemporal modeling of air pollutant concentrations in Germany using machine learning. Atmospheric Chemistry and Physics, 23(17): 10267–10285

[6] Box E O . (2002). Vegetation analogs and differences in the Northern and Southern Hemispheres: a global comparison. Plant Ecology, 163(2): 139–154

[7] Brienen R J W , Phillips O L , Feldpausch T R , Gloor E , Baker T R , Lloyd J , Lopez-Gonzalez G , Monteagudo-Mendoza A , Malhi Y , Lewis S L . et al. (2015). Long-term decline of the Amazon carbon sink. Nature, 519(7543): 344–348

[8] Buchwitz M , Reuter M , Schneising O , Noël S , Gier B , Bovensmann H , Burrows J P , Boesch H , Anand J , Parker R J . et al. (2018). Computation and analysis of atmospheric carbon dioxide annual mean growth rates from satellite observations during 2003–2016. Atmospheric Chemistry and Physics, 18(23): 17355–17370

[9] Cao L Z , Chen X , Zhang C , Kurban A , Yuan X L , Pan T , de Maeyer P. . (2017). The temporal and spatial distributions of the near-surface CO₂ concentrations in Central Asia and analysis of their controlling factors. Atmosphere, 8(5): 85

[10] Chen J B , Hu R H , Chen L Y , Liao Z H , Che L L , Li T W . (2024). Multi-sensor integrated mapping of global XCO₂ from 2015 to 2021 with a local random forest model. ISPRS Journal of Photogrammetry and Remote Sensing, 208: 107–120

[11] Cheng W , Dan L , Deng X Z , Feng J M , Wang Y L , Peng J , Tian J , Qi W , Liu Z , Zheng X Q . et al. (2022). Global monthly gridded atmospheric carbon dioxide concentrations under the historical and future scenarios. Scientific Data, 9: 83

[12] Chi Y F , Ren Y , Xu C D , Zhan Y . (2024). The spatial distribution mechanism of PM_2.5 and NO₂ on the eastern coast of China. Environmental Pollution, 342: 123122

[13] Cleveland R B , Cleveland W S , McRae J E , Terpenning I . (1990). STL: a seasonal-trend decomposition procedure based on loess. Journal of Official Statistics, 6(1): 3–33

[14] Curran J C , Curran S A . (2016). Indications of positive feedback in climate change due to a reduction in Northern Hemisphere biomass uptake of atmospheric carbon dioxide.. Weather, 71(4): 88–91

[15] Danielson J J, Gesch D B (2011). Global Multi-Resolution Terrain Elevation Data 2010 (GMTED2010). Herndon: U.S. Geological Survey Open-File Report 2011-1073, 26p

[16] de Figueiredo Silva F , Fulginiti L E , Perrin R K , Braga M J . (2022). The increasing opportunity cost of sequestering CO₂ in the Brazilian Amazon forest. Empirical Economics, 62(2): 439–460

[17] Didan K (2015). MOD13C2 MODIS/Terra Vegetation Indices Monthly L3 Global 0.05Deg CMG V006 [Online]. https://doi.org/10.5067/MODIS/MOD13C2.006<b>(I searched online and couldn't find any reference information. Please supplement)</b>

[18] Donald P , Mayengo M , Lambura A G . (2024). Mathematical modeling of vehicle carbon dioxide emissions. Heliyon, 10(2): e23976

[19] Duan L F , Hu W , Deng D , Fang W K , Xiong M , Lu P L , Li Z L , Zhai C Z. . (2021). Impacts of reducing air pollutants and CO₂ emissions in urban road transport through 2035 in Chongqing, China. Environmental Science and Ecotechnology, 8: 100125

[20] Falahatkar S , Mousavi S M , Farajzadeh M . (2017). Spatial and temporal distribution of carbon dioxide gas using GOSAT data over IRAN. Environmental Monitoring and Assessment, 189(12): 627

[21] Gunasegaran T, Cheah Y N 2017. Evolutionary cross validation. In: Proceedings of the 2017 8th International Conference on Information Technology. Amman: IEEE, 89–95

[22] Gupta V, Mishra V K, Singhal P, Kumar A (2022). An overview of supervised machine learning algorithm. In: Proceedings of the 2022 11th International Conference on System Modeling & Advancement in Research Trends. Moradabad: IEEE, 87–92

[23] He C P , Ji M R , Grieneisen M L , Zhan Y . (2022a). A review of datasets and methods for deriving spatiotemporal distributions of atmospheric CO₂. Journal of Environmental Management, 322: 116101

[24] He C P , Ji M R , Li T , Liu X Y , Tang D , Zhang S F , Luo Y Z , Grieneisen M L , Zhou Z H , Zhan Y . (2022b). Deriving full-coverage and fine-scale XCO₂ across China based on OCO-2 satellite retrievals and CarbonTracker output. Geophysical Research Letters, 49(12): e2022GL098435

[25] Hersbach, Bell, Berrisford (2023). ERA5 Hourly Data on Single Levels from 1940 to Present. Climate Data Store. Reading: European Centre for Medium-Range Weather Forecasts (ECMWF)

[26] Hou Y F , Wang S X , Zhou Y , Yan F L , Zhu J F . (2013). Analysis of the carbon dioxide concentration in the lowest atmospheric layers and the factors affecting China based on satellite observations. International Journal of Remote Sensing, 34(6): 1981–1994

[27] Huang X F , Yang H , Qiao S W , Yao Y J , Cui L , Fan H W , Lv Q Z , Qiao Y N , Feng G F . (2025). Spatiotemporal estimation of near-surface CO₂ concentrations over the global continent based on hybrid machine learning model. Annals of GIS, 31(2): 191–206

[28] IPCC. (2023). Climate Change 2023: Synthesis Report. Contribution of Working Groups I, II and III to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Geneva: IPCC

[29] Jena M , Dehuri S . (2020). DecisionTree for classification and regression: a state-of-the art review. Informatica, 44(4): 405–420

[30] Jin J P , Zhang Q W , Wei C , Gu Q R , Huang Y J . (2024). Wildfire CO₂ emissions in the conterminous United States from 2015 to 2018 as estimated by the WRF-chem assimilation system from OCO-2 XCO₂ retrievals. Atmosphere, 15(2): 186

[31] Kaiser J W , Heil A , Andreae M O , Benedetti A , Chubarova N , Jones L , Morcrette J J , Razinger M , Schultz M G , Suttie M . et al. (2012). Biomass burning emissions estimated with a global fire assimilation system based on observed fire radiative power. Biogeosciences, 9(1): 527–554

[32] Karnosky D F . (2003). Impacts of elevated atmospheric CO₂ on forest trees and forest ecosystems: knowledge gaps. Environment International, 29(2−3): 161–169

[33] Ke G L, Meng Q, Finley T, Wang T F, Chen W, Ma W D, Ye Q W, Liu T Y (2017). LightGBM: a highly efficient gradient boosting decision tree. In: Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach: Curran Associates Inc., 3149–3157

[34] Kim M , Brunner D , Kuhlmann G . (2021). Importance of satellite observations for high-resolution mapping of near-surface NO₂ by machine learning. Remote Sensing of Environment, 264: 112573

[35] Kleidon A . (2021). What limits photosynthesis? Identifying the thermodynamic constraints of the terrestrial biosphere within the Earth system. Biochimica et Biophysica Acta (BBA) - Bioenergetics, 1862(1): 148303

[36] Kou X X , Zhang M G , Zhen P . (2013). Numerical simulation of CO₂ concentrations in East Asia with RAMS-CMAQ. Atmospheric and Oceanic Science Letters, 6(4): 179–184

[37] Lasslop G , Coppola A I , Voulgarakis A , Yue C , Veraverbeke S . (2019). Influence of fire on the carbon cycle and climate. Current Climate Change Reports, 5(2): 112–123

[38] Li H , Qiu J X , Zhang K X , Zheng B . (2025). Monitoring fossil fuel CO₂ emissions from co-emitted NO₂ observed from space: progress, challenges, and future perspectives. Frontiers of Environmental Science & Engineering, 19(1): 2

[39] Li J , Jia K , Wei X Q , Xia M , Chen Z L , Yao Y J , Zhang X T , Jiang H Y , Yuan B , Tao G F . et al. (2022). High-spatiotemporal resolution mapping of spatiotemporally continuous atmospheric CO₂ concentrations over the global continent. International Journal of Applied Earth Observation and Geoinformation, 108: 102743

[40] Li X C , Zhou Y Y , Zhao M , Zhao X . (2020). A harmonized global nighttime light dataset 1992–2018. Scientific Data, 7: 168

[41] Liu M , Zhu X Y , Pan C , Chen L , Zhang H , Jia W X , Xiang W N . (2016). Spatial variation of near-surface CO₂ concentration during spring in Shanghai. Atmospheric Pollution Research, 7(1): 31–39

[42] Lundberg S M, Lee S I (2017). A unified approach to interpreting model predictions. In: Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach: Curran Associates Inc., 4768–4777

[43] Mai B R , Deng X J , Zhang F , He H , Luan T , Li F , Liu X . (2020). Background characteristics of atmospheric CO₂ and the potential source regions in the Pearl River delta region of China. Advances in Atmospheric Sciences, 37(6): 557–568

[44] Meijer J R , Huijbregts M A J , Schotten K C G J , Schipper A M . (2018). Global patterns of current and future road infrastructure. Environmental Research Letters, 13(6): 064006

[45] Mercado L M , Bellouin N , Sitch S , Boucher O , Huntingford C , Wild M , Cox P M . (2009). Impact of changes in diffuse radiation on the global land carbon sink. Nature, 458(7241): 1014–1017

[46] Mustafa F , Bu L B , Wang Q , Shahzaman M , Bilal M , Aslam R W , Dong C Z. . (2022). Spatiotemporal investigation of near-surface CO₂ and its affecting factors over Asia. IEEE Transactions on Geoscience and Remote Sensing, 60: 4108316

[47] NASA/GSFC/GMAO Carbon Group (2021). OCO-2 GEOS Level 3 Daily, 0.5x0.625 Assimilated CO₂ V10r. Greenbelt: NASA Goddard Earth Sciences (GES) Data and Information Services Center (DISC）

[48] National Institute for Environmental Studies (2025). NIES GOSAT Level 4 Data Product Format Description. Tsukuba: National Institute for Environmental Studies (NIES)

[49] Oda T , Maksyutov S . (2011). A very high-resolution (1 km×1 km) global fossil fuel CO₂ emission inventory derived using a point source database and satellite observations of nighttime lights. Atmospheric Chemistry and Physics, 11(2): 543–556

[50] Pan G N , Xu Y , Huang B . (2021). Evaluating national and subnational CO₂ mitigation goals in China’s thirteenth five-year plan from satellite observations. Environment International, 156: 106771

[51] Park C , Jeong S , Kim C , Shin J , Joo J . (2023). Machine learning based estimation of urban on-road CO₂ concentration in Seoul. Environmental Research, 231(Pt 3): 116256

[52] Phillips O L , Brienen R J W , The RAINFOR Collaboration . (2017). Carbon uptake by mature Amazon forests has mitigated Amazon nations’ carbon emissions. Carbon Balance and Management, 12(1): 1

[53] Shapley L S (1952). A Value for N-Person Games. Santa Monica: RAND Corporation

[54] Sheng M Y , Lei L P , Zeng Z C , Rao W Q , Zhang S Q . (2021). Detecting the responses of CO₂ column abundances to anthropogenic emissions from satellite observations of GOSAT and OCO-2. Remote Sensing, 13(17): 3524

[55] Shi Y S , Matsunaga T , Saito M , Yamaguchi Y , Chen X H . (2015). Comparison of global inventories of CO₂ emissions from biomass burning during 2002–2011 derived from multiple satellite products. Environmental Pollution, 206: 479–487

[56] Siabi Z , Falahatkar S , Alavi S J . (2019). Spatial distribution of XCO₂ using OCO-2 data in growing seasons. Journal of Environmental Management, 244: 110–118

[57] Silvestre G D, dos Santos M R, de Carvalho A C P L F (2021). Seasonal-trend decomposition based on loess + machine learning: hybrid forecasting for monthly univariate time series. In: Proceedings of 2021 International Joint Conference on Neural Networks. Shenzhen, China: IEEE, 1–7

[58] Sims K, Reith A, Bright E, McKee J, Rose A (2022). LandScan Global, 2021. Oak Ridge: Oak Ridge National Laboratory (ORNL)

[59] Sun Q W , Zhao X S , Hua Y Y . (2024). Dynamic coupling between atmospheric CO₂ concentration and land surface temperature in major urban agglomerations in China: insights for sustainable urban development. Sustainability, 16(21): 9484

[60] Te T , Bagan H , Che M H , Hou X Y , Uudus B . (2023). Spatiotemporal variability of near-surface CO₂ and its affecting factors over Mongolia. Environmental Research, 236(Pt 1): 116796

[61] Tomohiro O, Shamil M (2015). ODIAC Fossil Fuel Emission Dataset. Center for Global Environmental Research, National Institute for Environmental Studies. Tsukuba: National Institute for Environmental Studies (NIES)

[62] Uddin M S , Smirnov O . (2022). Spatial distribution of the annual atmospheric carbon dioxide in the contiguous USA and their controlling factors. Environmental Modeling & Assessment, 27(1): 57–76

[63] Williams R G , Meijers A J S , Roussenov V M , Katavouta A , Ceppi P , Rosser J P , Salvi P . (2024). Asymmetries in the Southern Ocean contribution to global heat and carbon uptake. Nature Climate Change, 14(8): 823–831

[64] World Meteorological Organization (WMO) (2023). WMO Greenhouse Gas Bulletin NO. 19: The State of Greenhouse Gases in the Atmosphere Based on Global Observations Through. Geneva: World Meteorological Organization

[65] Xu W J , He Y Y , Cheng F Q . (2025). Mutual remediation of CO₂ emissions and industrial wastewater: feasibility, application, and ecological impact. Frontiers of Environmental Science & Engineering, 19(6): 84

[66] Yao T E , Lu S H , Wang Y Q , Li X H , Ye H X , Duan Y S , Fu Q Y , Li J . (2024). Revealing the drivers of surface ozone pollution by explainable machine learning and satellite observations in Hangzhou Bay, China. Journal of Cleaner Production, 440: 140938

[67] Zhang J B , De-oliveira-Ceciliato P , Takahashi Y , Schulze S , Dubeaux G , Hauser F , Azoulay-Shemer T , Tõldsepp K , Kollist H , Rappel W J . et al. (2018). Insights into the molecular mechanisms of CO₂-Mediated Regulation of Stomatal Movements. Current Biology, 28(23): R1356–R1363

[68] Zhang L F , Li T W , Wu J A , Yang H J . (2023). Global estimates of gap-free and fine-scale CO₂ concentrations during 2014–2020 from satellite and reanalysis data.. Environment International, 178: 108057

[69] Zhang L L , Zhang J Y . (2018). Analysis of the seasonal variation of CO₂ concentration in China based on GOSAT satellite data. Journal of the Indian Society of Remote Sensing, 46(10): 1667–1675

[70] Zhang L N , Yang C Y , Xiao Q Y , Geng G N , Cai J , Chen R J , Meng X , Kan H D . (2021). A satellite-based land use regression model of ambient NO₂ with high spatial resolution in a Chinese city. Remote Sensing, 13(3): 397

[71] Zhang M Q , Liu G J . (2023). Mapping contiguous XCO₂ by machine learning and analyzing the spatio-temporal variation in China from 2003 to 2019.. Science of the Total Environment, 858(Pt 2): 159588

[72] Zhang P F , Feng K S , Yan L , Guo Y Q , Gao B , Li J S . (2024). Overlooked CO₂ emissions induced by air pollution control devices in coal-fired power plants. Environmental Science and Ecotechnology, 17: 100295

[73] Zhao J, Cui W H, Sun Y H (2014). Spatial and temporal distribution characteristics of near-surface CO₂ concentration over China based on GOSAT data. In: Proceedings of SPIE 9265, Remote Sensing and Modeling of the Atmosphere, Oceans, and Interactions V. Beijing: SPIE, 92650N

[74] Zhou X Y , Yue X , Tian C G . (2024). Responses of ecosystem productivity to anthropogenic ozone and aerosols at the 2060. Earth’s Future, 12(1): e2023EF003781

Rights & permissions

Higher Education Press 2026