Spatiotemporal variability of methane concentrations driven by ruminant livestock in Mongolia: insights from satellite observations and machine learning

Methane (CH4) emissions from ruminant livestock (cattle, sheep, and goats) are a significant environmental concern because CH4 is a potent greenhouse gas. In Mongolia, livestock-driven CH4 emissions constitute 89.74% of total CH4 emissions, yet little is known about how the distribution of ruminant livestock is related to the spatiotemporal distribution of CH4 concentrations. In this study, we integrated quality-controlled, column-averaged methane concentration (XCH4) data from the GOSAT and GOSAT−2 L2 products with auxiliary data (natural and anthropogenic factors), and we compared the ability of four machine-learning algorithms to predict XCH4 from the auxiliary data. We selected the optimal algorithm, Gradient Boosting Regression (GBR), to generate monthly XCH4 data at a spatial resolution of 0.1° × 0.1° for Mongolia from 2010 to 2022. The GBR model achieved a coefficient of determination of 0.918, a root mean square error of 11.78 parts per billion (ppb), and a mean absolute percentage error of 0.47% on an independent validation dataset. We then analyzed the spatiotemporal distribution of XCH4 and its quantitative relationship with the numbers of ruminant livestock. Over the past 13 years, the average XCH4 in Mongolia increased by 9.16 ppb/yr. XCH4 reached an annual maximum in autumn and a minimum in winter. The spatiotemporal distribution of XCH4 was significantly influenced by the type and distribution of ruminant livestock. The distribution of cattle was most strongly correlated with XCH4 (r = 0.965), and cattle raising accounted for 69.33% of Mongolia’s total land area. In particular, rapid increases in the number of cattle in the mountainous regions of the aimags of Khangai and Khentii were associated with significant increases of XCH4 in those areas. The results indicated that effective management of the numbers and species of livestock is crucial for mitigating CH4 emissions. Understanding the spatiotemporal variability of CH4 concentrations can directly inform livestock management strategies, zoning policies, and national greenhouse gas reduction initiatives.