[A-CG35] グローバル炭素循環の観測と解析

[ACG35-06] Data-driven estimation of soil CH₄ absorption in Japan

*山貫 緋称¹、市井 和仁¹、山本 雄平¹、寺本 宗正²、孫 力飛³、小嵐 淳⁴、安藤 麻里子⁴、永野 博彦⁵、平野 高司⁶、高木 健太郎⁶、石田 祐宣⁷、高木 正博⁸、近藤 俊明⁹、高橋 善幸³、梁 乃申³ (1.千葉大学環境リモートセンシング研究センター、2.鳥取大学乾燥地研究センター、3.国立環境研究所、4.日本原子力研究開発機構、5.新潟大学、6.北海道大学、7.弘前大学、8.宮崎大学、9.国際農林水産業研究センター)

キーワード：メタン吸収、機械学習、リモートセンシング、広域推定

Accurate prediction of atmospheric greenhouse gases (GHGs) concentrations is important for understanding climate change such as global warming. Forest soils are considered a sink for CH₄, which has 28 times the greenhouse effect of CO₂. Still, a lack of observational data makes it unclear whether the absorption capability will be maintained with long-term warming. Therefore, accurately estimating the CH₄ sink of forest soils is crucial in predicting future climate change. So far, a field observation network for continuous automatic measurement of soil CH₄ absorption is being developed in Asia, and it has become clear that soil CH₄ absorption capability and global warming response vary greatly from region to region. Spatiotemporal variations in soil CH₄ absorption are considered to be influenced not only by climate but also by the physical and chemical properties of soils. In this study, we estimated soil CH₄ absorption in Japan by applying a machine learning method to data from the largest soil CH₄ absorption observation network in Asia, which has been developed and conducted by National Institute for Environmental Studies (NIES) using the same observation methods, soil properties, organic carbon properties, and microbial properties obtained by Japan Atomic Energy Agency (JAEA) and other organizations, and satellite observation data.