[A-CG39] グローバル炭素循環の観測と解析

[ACG39-P06] Estimation of half-hourly gross primary productivity in rice paddy fields using five years of ground-based solar-induced chlorophyll fluorescence observations

*Gyeong-Min Kim¹、Hyun-Dong Moon^1,2、Bo-Kyeong Kim^1,2、Subin Choi¹、Hayeon WON^1,2、Hyun-Hwan Yang¹、Jae-Hyun Ryu³、Jong-Sung Ha⁴、Seung-Taek Jung⁴、Jaeil Cho^1,2 (1.Dept Applied Plant Sci, Chonnam Natl Univ, Republic of Korea、2.BK21 FOUR Ctr IT-Bio Converg Sys Agric, Chonnam Natl Univ, Republic of Korea、3.Natl Inst Agric Sci, Rural Dev Admin, Republic of Korea 、4.Satellite Appl Div, Korea Aerospace Res Inst, Republic of Korea)

キーワード：Solar-induced chlorophyll fluorescence, Gross primary productivity, Near surface remote sensing, Rice paddy, Random forest

Gross primary productivity (GPP) of cropland refers to the rate of organic matter production during photosynthesis, which is related to the growth and yield of crops. The eddy covariance technique is the only method to observe GPP in situ at the canopy scale without damage to the vegetation. However, the major assumptions for the eddy covariance, such as horizontal and uniform terrain and a homogeneous surface, impose restrictions on observation sites. Chlorophyll fluorescence is red to far-red light emitted by plants when illuminated by photosynthetically active radiation, representing energy loss during the conversion of light energy absorbed by plants into biochemical energy in photosynthesis. It is expected that the solar-induced chlorophyll fluorescence (SIF) can offer a promising approach for estimating GPP, as the intensity of fluorescence is related to the fraction of energy used for carbon fixation and photosynthetic activity. Continuous measurements should be performed in wide-ranging meteorological conditions to understand the response of SIF and GPP and to enhance the accuracy of GPP estimates using SIF. Most studies used remote sensing data from satellites, whose spatio-temporal scale is large, while only a few studies carried out multiyear observations at the same location. In this study, a hyperspectral field sensor was installed with eddy covariance instruments on a 4-meter tower in a rice paddy over a period of five years to establish an SIF and GPP dataset under various meteorological conditions. The observed SIF and GPP showed a non-linear regression. It seems that SIF alone had limitations in estimating GPP because SIF represents only the light reaction of photosynthesis, whereas GPP is the results of both light reaction and CO₂ fixation. When comparing the estimated GPP using SIF and other factors (such as meteorological parameter and vegetation indices) through random forest model with the observed GPP, the correlation coefficient was about 0.9. It is necessary to compare the constructed random forest model's performance with observation sites in other regions.

Acknowledgement
This work was supported by Korea Environment Industry and Technology Institute (KEITI) through Project for developing an observation-based GHG emissions geospatial information map, funded by Korea Ministry of Environment (MOE) (Project No. RS-2023-00232066, 2023-2027).