[A-CG39] Biogeochemical cycles in Land Ecosystem

[ACG39-06] Data assimilation of solar-induced chlorophyll for better gross primary estimation in a rice paddy by a process-based VISIT-SIF Model

*LIANGXIAN FAN¹, TOMOMICHI KATO², TATSUYA MIYAUCHI², KANOKRAT BUAREAL², KEISUKE ONO³, TOMOKI MOROZUMI⁴, NAOHISA NAKASHIMA⁵ (1.Graduate School of Global Food Resources, 2.Research Faculty of Agriculture, Hokkaido University , 3.National Institute for Agro-Environmental Sciences, 4.National Institute for Environmental Studies, 5.Agro-environemental Science, Obihiro Univ. of Agriculture & Veterinary Medicine)

Keywords:SIF, GPP, Process-based model

Numerous studies have proved that solar-induced chlorophyll fluorescence (SIF) was a good proxy of gross primary production (GPP). SIF can be used to improve GPP simulations by optimizing critical model parameters through statistical Bayesian data assimilation techniques. A prerequisite is the availability of a functional link between GPP and SIF in terrestrial biosphere models. Here we modified the Vegetation Integrative Simulator for Trace gases (VISIT)model, a process-based terrestrial ecosystem model simulating carbon, nitrogen, and water cycles, to simulate the fluorescence quantum yield at near-infrared band at the canopy level (VISIT-SIF), allowing us to simulate land-atmosphere biogeochemical interactions via the SIF constrains.
We implemented the data assimilation of a ground measured SIF by VISIT-SIF at the eddy flux tower at a Mase rice paddy field (site code "MSE" in AsiaFlux network; 36.05°N, 140.03°E, 13 m a.s.l.) in Tsukuba, Ibaraki, Japan during 2018-2020. The site covers approximately 200 ha (2km × 1km along the Kokai River) and is divided into rectangular plots (100m × 50 m). The parameter optimization brought the better GPP estimation accuracy as the determinant coefficient (R²) in GPPs between simulation and observation was improved from 0.40 to 0.66 and the root-mean-square-error (RMSE) was reduced from 0.18 to 0.12 after the optimization at half-hourly scale, and R² was increased from 0.59 to 0.75 and RMSE was decreased from 0.10 to 0.05 after the optimization at daily scale. Our results also suggest a biome dependency of the SIF-GPP relationship that needs to be improved for some plant functional types.