[M-IS01] ENVIRONMENTAL, SOCIO-ECONOMIC, AND CLIMATIC CHANGES IN NORTHERN EURASIA

[MIS01-20] Large-scale partitioning of the territory of Russia in the context of carbon biogeochemical cycle using machine learning techniques

*Mikhail Krinitskiy^1,2,3、Tatiana Kharitonova³、Alexander Maksakov³、Vadim Rezvov^1,2 (1.Shirshov Institute of Oceanology, Russian Academy of Sciences、2.Moscow Institute of Physics and Technology、3.Lomonosov Moscow State University)

キーワード：Carbon biogeochemical cycle, Climate change monitoring, Natural ecosystems, Greenhouse gas fluxes, Machine learning, Clustering

Establishing methodologies for effective spatial monitoring stands as the foundational step for the implementation of initiatives aimed at curbing climatic impacts and adjusting to its shifts. The acquisition of data on the operation of natural ecosystems, especially in regions with significant emissions and capacity for atmospheric greenhouse gas sequestration, is essential for quantifying the anthropogenic influence on climate and its implications for economic sectors, highlighting its importance in the execution of climate policy. However, there is a critical need for adaptive methodologies capable of extrapolating localized observational data to perform spatial assessments, a process often hampered by the scale and variability of extensive territories. The goal of this study is to forge a strategy for the spatial classification of vast territories, focusing on the attributes of the carbon biogeochemical cycle, and synthesizing environmental monitoring data with an innovative machine learning strategy: clustering via the superpixel technique. The eco-regions delineated through this new schema were evaluated against existing zoning frameworks, encompassing administrative divisions as well as geological-geomorphological and bioclimatic-based methods. The proposed approach's merit lies in its integration of dynamic variables, offering flexibility in response to environmental shifts and accurately representing the natural segmentation of ecosystems. Additionally, the defined eco-regions are consistent with the boundaries of significant physiographic expanses – including the Eastern European Plain, Western Siberia, the Middle Siberian Plateau, the Urals, the Caucasus, and the mountainous regions of Southern and Northeastern Siberia, as well as the Far East – and the eco-regional mosaic closely mirrors environmental zonality. As such, the landscape-ecological analysis method put forth can be applied in future endeavors to demarcate territories based on the nature and intensity of carbon biogeochemical cycle processes, and proves valuable for the enhancement of data-driven models in the upscaling of point-based carbon cycle measurements.