Modeling diurnal gross primary production in East Asia using Himawari-8/9 geostationary satellite data

作者：Yuhei YAMAMOTO, Kazuhito Ichii, Wei Yang, Yui Shikakura, Youngryel Ryu, Minseok Kang, Shohei Murayama, Su‐Jin Kim, Yuta Takao, Masahito Ueyama, Tomoko Akitsu, Hiroki Iwata, Ho‐Jin Lee, Jung‐Hwa Chun, Atsushi Higuchi, Takashi Hirano, Areum Kim, Hyun Seok Kim, Kenzo Kitamura, Yuji Kominami, Kazuho Matsumoto, Jun Suzuki, Kentaro Takagi, Yoshiyuki Takahashi, Satoru Takanashi, Hideaki Takenaka, Shingo Taniguchi, Yukio Yasuda · 发表于：Remote Sensing of Environment · 年份：2025 · DOI：10.1016/j.rse.2025.114866 · 被引用次数：4 · 研究领域：Plant Water Relations and Carbon Dynamics、Solar Radiation and Photovoltaics、Geophysics and Gravity Measurements

Gross primary production (GPP) is a key indicator of plant growth and ecosystem health, and accurately capturing its diurnal variation is crucial for understanding vegetation responses to extreme heat and drought. However, the applicability of satellite-based semi-empirical models to diurnal GPP estimation remains limited. This study refined diurnal GPP estimation in humid temperate climates by leveraging Himawari-8/9 geostationary satellite data to incorporate direct/diffuse radiation and the nonlinear GPP response to diurnal variations in absorbed photosynthetically active radiation (APAR). The eddy covariance-light use efficiency (EC-LUE) model was employed by adopting three approaches: the direct/diffuse (DD) setting to consider the direct/diffuse components of APAR, DD with nonlinear relationship (DD-NL) setting to additionally consider the nonlinear GPP-APAR relationship, and the baseline setting. The model was calibrated and validated using the eddy-covariance tower observations from 18 sites across Japan and South Korea. The DD-NL setting improved accuracy by correcting the baseline's overestimation of GPP under high APAR and underestimation under low APAR. Particularly for forest sites, the DD-NL setting reduced midday overestimations by 12–30 % on clear-sky days and morning/afternoon underestimations by 25–40 % on cloudy days. In the baseline setting, low-APAR biases progressively accumulated across daily to annual timescales, whereas the DD-NL setting reduced them ...