[A-CG37] 陸域生態系の物質循環

[ACG37-05] Intercomparison of Data-Driven Estimation of Soil Respiration in Japan

*山貫 緋称¹、市井 和仁^1,2、梁 乃申²、寺本 宗正³、高橋 善幸²、曾 継業²、高木 健太郎⁴、平野 高司⁴、石田 祐宣⁵、高木 正博⁶、楢本 正明⁷、中根 周歩⁸、近藤 俊明⁹、小嵐 淳¹⁰、安藤 麻里子¹⁰ (1.千葉大学環境リモートセンシング研究センター、2.国立環境研究所地球環境研究センター、3.鳥取大学乾燥地研究センター、4.北海道大学、5.弘前大学、6.宮崎大学、7.静岡大学、8.広島大学、9.国際農林水産業研究センター、10.日本原子力研究開発機構)

キーワード：炭素循環、土壌呼吸、リモートセンシング、機械学習、広域推定、二酸化炭素フラックス

Soil Respiration (SR), the sum of root respiration and heterotrophic respiration, is one of the most essential components of soil carbon cycles. However, large uncertainties remain in its temporal and spatial variations. So far, various efforts have been conducted to understand SRs. Many observation stations directly measure SR using chambers. Using these observation data and literature survey, several studies estimated spatial and temporal patterns of SR at global and regional scales based on semi-empirical equations and machine-learning methods. However, the database (e.g. Soil Respiration Database; SRDB) used in these large-scale studies contains inconsistently observed datasets. These inconsistencies may produce additional uncertainties in estimated fluxes. The largest SR observation network across Asia developed and maintained by NIES, Japan can be a good candidate to estimate spatio-temporal variations in SR across Asia, since these observations have been conducted with a consistent observation protocol and quality controls.

In this study, we updated our data-driven estimation of SR across Japan with observation data (eight sites across Japan), remote sensing data (MODIS land products), and random forest regression. Our estimation shows a reasonable performance with R²=0.87 for remote sensing only model and R²=0.91 for remote sensing and in-situ combined model. Based on the established model, we also produced upscaled estimations of SR across Japan with 1km spatial resolution from 2000 to 2020.

Intercomparison of our estimation with other available datasets was also conducted to understand advantages of our estimation. Our results show spatially more explicit variations compared with other global products. In addition, our advantage is to capture temporal variations (e.g. 8days). We also confirmed that previous estimations do not reproduce our observation network datasets, indicating consistent observation approach is important to upscale soil respiration.