Modeling the spatial distribution of carbon dioxide concentration by machine learning
DOI:
https://doi.org/10.5564/mjgg.v62i46.4252Keywords:
Machine learning, Carbon dioxide, Climate change, Greenhouse gas, Shapley Additive Explanations (SHAP)Abstract
The rapid increase in atmospheric carbon dioxide (CO₂) is a significant contributor to modern climate change and global warming. High-resolution data are essential for scientifically valid studies of CO₂ dynamics and changes. Although greenhouse gas observation satellites exist, they are still inadequate in terms of spatiotemporal resolution. We modeled to obtain a high-spatiotemporal resolution (16-day, 0.1x0.1) XCO₂ by machine learning algorithms (such as Random Forest (RF), eXtreme Gradient Boosting (XGB), Light Gradient Boosting Machine (LGBM), and Stacking (Stk)) using multisource data (such as weather parameters, vegetation index, land cover, elevation and population density) over Mongolia and China from 2010 to 2022. The results indicated that the XGB demonstrated good performance, with R² values exceeding 0.95 and 10-fold cross-validation R² (cv) values exceeding 0.55 during the study periods. Additionally, when calculating the Shapley Additive Explanations (SHAP) values based on the XGB model's results, the weather variables (53%) had the largest influence on the distribution of XCO₂. In comparison, population density alone had an 11% influence. As a result, the average XGB XCO₂ over the study period was 402.3±1.14 ppm, with the highest value observed in eastern China and the lowest value observed in Mongolia and the Qinghai-Tibet Plateau. This is related to the region's economic development and human activities. In terms of seasonal dynamics, the XGB XCO₂ values for winter, spring, summer, and autumn were 403.4±1.2 ppm, 403.3±1.3 ppm, 400.9±1.5 ppm, and 402.1±1.0 ppm, respectively. This shows that XCO₂ is highest in winter and lowest in summer. This seasonal dynamic is because plants absorb more CO₂ through photosynthesis in summer than in other seasons. The atmospheric ХCO₂ map produced in this study can serve as a baseline map for studying the carbon cycle in terrestrial ecosystems and climate change.
Нүүрсхүчлийн хийн агууламжийн орон зайн тархалтыг машин сургалтаар загварчлах нь
ХУРААНГУЙ: Агаар мандал дахь нүүрсхүчлийн хий (CO₂)-н агууламжийн хурдацтай өсөлт нь Дэлхийн дулаарал, орчин үеийн уур амьсгалын өөрчлөлтийн үндсэн шалтгаан болж байна. CO-н динамик, өөрчлөлтийн талаарх өндөр нарийвчлалтай өгөгдөлд тулгуурласан судалгаа хийх нь чухал юм. Хэдийгээр хүлэмжийн хийн ажиглалтын хиймэл дагуулууд байгаа боловч цаг хугацаа, орон зайн хувьд шаардлага хангахгүй хэвээр байна. Иймд энэхүү судалгаандаа, бид Монгол болон БНХАУ (Бүгд Найрамдах Хятад Ард Улс)-ын нутаг дэвсгэрийн хэмжээнд, 2010-аас 2022 он хүртэлх хугацааны Хүлэмжийн хийн ажиглалтын хиймэл дагуул (GOSAT)-ын хуурай агаарын босоо баганын дундаж ХCO₂ болон цаг уурын хэмжигдэхүүнүүд, ургамлын индекс, газрын бүрхэвч, гадаргын өндөршил, хүн амын нягтшил зэрэг 12 өгөгдлийг ашиглан машин сургалтын 4 өөр аргаар (Random Forest (RF), eXtreme Gradient Boosting (XGB), Light Gradient Boosting Machine (LGBM), Stacking (Stk)) XCO₂-г загварчлан харьцуулсан. Эдгээрээс XGB XCO₂ загвар хамгийн өндөр гүйцэтгэлтэй буюу детерминацийн коэффициент R²>0.95, солбин бататгах R²(cv)>0.55 байлаа. Түүнчлэн XGB загварын үр дүнд тулгуурлан Таамаглалыг тайлбарлах нэгдсэн тогтолцоо буюу Shapley Additive Explanations (SHAP) утгыг тооцоход XCO₂-н тархалтад цаг уурын хувьсагчид хамгийн их нөлөө (53%) үзүүлсэн бол хүн амын нягтшил дангаараа 11% нөлөө үзүүлсэн. Өндөр гүйцэтгэлтэй XGB загвараар судалгааны талбайг хамарсан өндөр нарийвчлалтай (16 өдрийн, 0.1°x0.1°) XGB XCO₂-н орон зайн тасралтгүй өгөгдлийг гарган авсан. Үр дүнд судалгааны хугацааны талбайн дундаж XGB XCO₂ нь 402.3±1.14 ppm, их утга нь БНХАУ-ын зүүн хэсгээр, бага утга нь Монгол болон Чинхай-Төвөдийн өндөрлөгт ажиглагдсан. Энэ нь бүс нутгийн эдийн засгийн хөгжил, хүний үйл ажиллагаатай холбоотой юм. Улирлын динамикийн хувьд XGB XCO₂ утга нь өвөл, хавар, зун, намрын улиралд харгалзан 403.4±1.2 ppm, 403.3±1.3 ppm, 400.9±1.5 ppm, 402.1±1.0 ppm байлаа. Эндээс үзвэл XCO₂ нь өвлийн улиралд хамгийн их утгатай, зун хамгийн бага утгатай байна. Улирлын динамик нь зуны улиралд ургамал фотосинтезийн процессоор СО-г бусад улиралтай харьцуулахад ихээр шингээдэгтэй холбоотой. Энэхүү судалгааны ажлын үр дүн нь биосферийн нүүрстөрөгчийн эргэлтийг загварчлах, уур амьсгалын өөрчлөлтийг судлахад суурь өгөгдөл болно.
Түлхүүр үгс: Машин сургалт, Нүүрсхүчлийн хий, Уур амьсгалын өөрчлөлт, Хүлэмжийн хий, Таамаглалыг тайлбарлах нэгдсэн тогтолцоо (SHAP)
Downloads
194
References
[1] P. Friedlingstein et al., “Global Carbon Budget,” Earth Syst. Sci. Data, vol. 14, pp. 4811–4900, 2022, https://doi.org/10.5194/essd-14-4811-2022.
[2] IPCC, “Global Carbon and Other Biogeochemical Cycles and Feedbacks,” in Climate Change 2021 – The Physical Science Basis, I. P. on C. C. (IPCC), Ed., Cambridge: Cambridge University Press, 2023, pp. 673–816. Available: https://doi.org/ 10.1017/9781009157896.007.
[3] Д. Дагвадорж, Агаар мандлын физик-химийн судалгаа. Улаанбаатар, 2023.
[4] M. Zhang and G. Liu, “Mapping contiguous XCO2 by machine learning and analyzing the spatio-temporal variation in China from 2003 to 2019,” Sci. Total Environ., vol. 858, p. 159588, 2023. Available: https://doi.org/ 10.1016/j.scitotenv.2022.159588.
[5] H. Yang, T. Li, J. Wu, and L. Zhang, “Inter-comparison and evaluation of global satellite XCO2 products,” Geo-spatial Inf. Sci., vol. 28, no. 1, pp. 131–144, Jan. 2025. Available: https://doi.org/ 10.1080/10095020.2023.2252017.
[6] R. Doughty et al., “Global GOSAT, OCO-2, and OCO-3 solar-induced chlorophyll fluorescence datasets,” Earth Syst. Sci. Data, vol. 14, no. 4, pp. 1513–1529, 2022. Available: https://doi.org/10.5194/essd-14-1513-2022.
[7] F. Kataoka, D. Crisp, T. E. Taylor, C. W. O’Dell, A. Kuze, K. Shiomi, H. Suto, C. Bruegge, F. M. Schwandner, R. Rosenberg, et al., "The Cross-Calibration of Spectral Radiances and Cross-Validation of CO2 Estimates from GOSAT and OCO-2," Remote Sens., vol. 9, p. 1158, 2017. doi: 10.3390/rs9111158.C. Rakauckas, A. Edelman, and R. Ferrari, “A Modular System for Increasing Predictiveness for Extreme Climate Predictions,” United States, 2021, Available: https://doi.org/10.2172/1769647.
[8] S. He, Y. Yuan, Z. Wang, L. Luo, Z. Zhang, H. Dong, and C. Zhang, "Machine Learning Model-Based Estimation of XCO2 with High Spatiotemporal Resolution in China," *Atmosphere*, vol. 14, p. 436, 2023, https://doi.org/10.3390/atmos14030436.
[9] C. He et al., “Deriving Full-Coverage and Fine-Scale XCO2 Across China Based on OCO-2 Satellite Retrievals and CarbonTracker Output,” Geophys. Res. Lett., vol. 49, no. 12, 2022 Available: https://doi.org/10.1029/2022GL098435.
[10] J. Li, K. Jia, X. Wei, M. Xia, Z. Chen, Y. Yao, X. Zhang, H. Jiang, B. Yuan, G. Tao, and L. Zhao, "High-spatiotemporal resolution mapping of spatiotemporally continuous atmospheric CO2 concentrations over the global continent," *Int. J. Appl. Earth Obs. Geoinf.*, vol. 108, p. 102743, Apr. 2022. Available: https://doi.org/ 10.1016/j.jag.2022.102743.
[11] А. Саруулзаяа ба Д. Сайнбаяр, “Нүүрсхүчлийн хий (CO₂) болон метан (CH₄) хийн агууламжийн орон зайн тархалт,” Монгол Орны Газарзүй-Геоэкологи Сэтгүүл - МОГГС, vol. 42, pp. 252–259, 2022.
[12] Д. Сайнбаяр, С. Эрдэнэсүх, ба А. Саруулзаяа, “Монгол орны нүүрсхүчлийн хийн хэлбэлзлийн судалгаа,” Geogr. Issues, vol. 22, no. 1, pp. 45–56, Mar. 2022. [Online]. Available: https://journal.num.edu.mn/GP/article/view/536.
[13] S. Adiya, S. Dalantai, T. Wu, X. Wu, J. Yamkhin, Y. Bao, E. Sumiya, G. Yadamsuren, D. Avirmed, and B. Dorjgotov, "Spatial and temporal change patterns of near-surface CO2 and CH4 concentrations in different permafrost regions on the Mongolian Plateau from 2010 to 2017," *Sci. Total Environ.*, vol. 800, p. 149433, Dec. 2021. Available: https://doi.org/ 10.1016/j.scitotenv.2021.149433.
[14] S. Yang, X. Wang, Z. Ge, G. Dong, M. Ma, and X. Han, “Global Per Capita CO2 Emission Trends,” 2023. Available: https://doi.org/ 10.3390/atmos14121797.
[15] F. P. Lopez, G. Zhou, G. Jing, K. Zhang, and Y. Tan, “XCO2 and XCH4 Reconstruction Using GOSAT Satellite Data Based on EOF-Algorithm,” 2022. Available: https://doi.org/ 10.3390/rs14112622.
[16] T. Yokota et al., “Global Concentrations of CO2 and CH4 Retrieved from GOSAT: First Preliminary Results,” SOLA, vol. 5, pp. 160–163, 2009. Available: https://doi.org/10.2151/sola.2009-041.
[17] T. A. Adjuik and S. C. Davis, “Machine Learning Approach to Simulate Soil CO2 Fluxes under Cropping Systems,” 2022. Available: https://doi.org/10.3390/agronomy12010197.
[18] X. Chen, Q. He, T. Ye, Y. Liang, and Y. Li, “Decoding spatiotemporal dynamics in atmospheric CO2 in Chinese cities: Insights from satellite remote sensing and geographically and temporally weighted regression analysis,” Sci. Total Environ., vol. 908, p. 167917, 2024. Available: https://doi.org/ 10.1016/j.scitotenv.2023.167917.
[19] J. Muñoz-Sabater, “ERA5-Land hourly data from 1950 to present,” 2019, Copernicus Climate Change Service (C3S) Climate Data Store (CDS). Available: https://doi.org/ 10.24381/cds.e2161bac.
[20] H. Boogaard, J. Schubert, A. De Wit, J. Lazebnik, R. Hutjes, and G. Van der Grijn, "Agrometeorological indicators from 1979 to present derived from reanalysis," Copernicus Climate Change Service (C3S) Climate Data Store (CDS), 2020, https://doi.org/10.24381/cds.6c68c9bb.
[21] H. Hersbach et al., "ERA5 hourly data on single levels from 1940 to present," Copernicus Climate Change Service (C3S) Climate Data Store (CDS), 2023, https://doi.org/10.24381/cds.adbb2d47.
[22] NOAA National Geophysical Data Center, “2-minute Gridded Global Relief Data (ETOPO2) v2. NOAA National Centers for Environmental Information”.
[23] A. J. Tatem, “WorldPop, open data for spatial demography,” Sci. Data, vol. 4, no. 1, p. 170004, 2017. Available: https://doi.org/10.1038/sdata.2017.4.
[24] T. Maganathan, S. Senthilkumar, and V. Balakrishnan, “Machine Learning and Data Analytics for Environmental Science: A Review, Prospects and Challenges,” IOP Conf. Ser. Mater. Sci. Eng., vol. 955, no. 1, 2020. Available: https://doi.org/10.1088/1757-899X/955/1/012107.
[25] P. Xu, “Review on Studies of Machine Learning Algorithms,” J. Phys. Conf. Ser., vol. 1187, no. 5, p. 52103, 2019. Available: https://doi.org/10.1088/1742-6596/1187/5/052103.
[26] B. Naveen and S. Manu, “Detecting DDoS Attacks Using Machine Learning Techniques and Contemporary Intrusion Detection Dataset,” Autom. Control Comput. Sci., vol. 53, no. 5, pp. 419–428, 2019. https://doi.org/Available: 10.3103/S0146411619050043.
[27] L. Breiman, "Random forests," *Mach. Learn.*, vol. 45, no. 1, p. 5, 2001.
[28] C. Zheng, A. Abd-Elrahman, V. Whitaker, and C. Dalid, “Prediction of Strawberry Dry Biomass from UAV Multispectral Imagery Using Multiple Machine Learning Methods,” 2022. Available: https://doi.org/10.3390/rs14184511.
[29] N. M. Shahani, X. Zheng, C. Liu, F. U. Hassan, and P. Li, “Developing an XGBoost Regression Model for Predicting Young’s Modulus of Intact Sedimentary Rocks for the Stability of Surface and Subsurface Structures,” Front. Earth Sci., vol. 9, no. October, pp. 1–13, 2021. Available: https://doi.org/10.3389/feart.2021.761990.
[30] G. Ke, Q. Meng, T. Finley, T. Wang, W. Chen, W. Ma, Q. Ye, and T.-Y. Liu, "LightGBM: A highly efficient gradient boosting decision tree," Microsoft Research and Peking University, 2017. [Online]. Available: https://proceedings.neurips.cc/paper_files/paper/2017/file/6449f44a102fde848669bdd9eb6b76fa-Paper.pdf.
[31] M. Kuhn and K. Johnson, Applied Predictive Modeling, vol. 26. 2013.
[32] S. M. Lundberg and S.-I. Lee, A Unified Approach to Interpreting Model Predictions. 2017.
[33] J. Zhao, G. Li, W. Cui, Q. Cao, and H. Zhang, “Important evidence of constant low CO 2 windows and impacts on the non-closure of the greenhouse effect,” Sci. Rep., vol. 9, no. 1, pp. 1–13, 2019, Available: https://doi.org/10.1038/s41598-019-41562-x.
[34] L. Wei, H. Liu, and L. Wu, “Spatial distribution of floating population in Beijing, Tianjin and Hebei Region and its correlations with synergistic development,” Math. Biosci. Eng., vol. 20, no. 3, pp. 5949–5965, 2023, Available: https://doi.org/10.3934/mbe.2023257.
[35] ҮТХН, “Парисын хэлэлцээрийг хэрэгжүүлэх үндэсний хэмжээнд тодорхойлсон хувь нэмрийн зорилт,” 2020, БОАЖЯ, Улаанбаатар. [Online]. Available: https://www.ccrcc.mn/report.
[36] Д. Сандэлгэр, Г. Болорчимэг, ба С. Эрдэнэсүх, “Мөрөн хот орчмын орчих мандлын дулааны судалгаа,” Монгол Орны Газарзүй-Геоэкологи Сэтгүүл - МОГГС, vol. Тусгай дуг, pp. 97–103, 2017.
[37] L. Shen, C. Zhao, Z. Ma, Z. Li, J. Li, and K. Wang, “Observed decrease of summer sea-land breeze in Shanghai from 1994 to 2014 and its association with urbanization,” Atmos. Res., vol. 227, pp. 198–209, 2019, Available: https://doi.org/10.1016/j.atmosres.2019.05.007.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2025 Sainbayar Dalantai, Erdenesukh Sumiya, Amarsaikhan Damdinsuren
This work is licensed under a Creative Commons Attribution 4.0 International License.
Copyright on any research article in the Mongolian Journal of Geography and Geoecology is retained by the author(s).
The authors grant the Mongolian Journal of Geography and Geoecology a license to publish the article and identify itself as the original publisher.
Articles in the Mongolian Journal of Geography and Geoecology are Open Access articles published under a Creative Commons Attribution 4.0 International License CC BY.
This license permits use, distribution and reproduction in any medium, provided the original work is properly cited.
