Optimized Gross Primary Productivity Over the Croplands Within the BEPS Particle Filtering Data Assimilation System (BEPS_PF v1.0)

Abstract

Agricultural ecosystems play an important role in modulating the global carbon balance by taking up atmospheric carbon dioxide, while large differences and uncertainties exist in the estimated crop gross primary productivity (GPP) by terrestrial ecosystem models (TEMs). With the aim of reducing the parameter uncertainty in TEMs for crop GPP simulation, we developed a particle filtering data assimilation (DA) system based on the ecosystem model BEPS (Biosphere Exchange Process Simulator), that is, the BEPS_PF (v1.0). We investigated the feasibility of BEPS_PF on the multiple parameters optimization across typical crops (wheat, rice, soybean and corn) and on reducing the uncertainty of GPP over 32 cropland eddy covariance sites globally. With BEPS_PF DA, the average R² between GPP and observed data at the hourly scale has been efficiently improved by 0.36 and root mean square error reduced by 0.18 gC m⁻² hr⁻¹. The DA system has successfully corrected the GPP from the irrigated croplands which was severely underestimated by the model's prior parameters. We found that the maximum carboxylation rate at 25°C (V_cmax25) as well as the leaf nitrogen content (N_leaf) were co-varied with strong seasonal variations. The optimized V_cmax25 showed large differences among different crop types with ranges 27.07–62.95, 42.17–93.32, 31.89–105.81, and 38.34–89.29 μmol m⁻² s⁻¹ for corn, soybean, wheat, and rice respectively. We demonstrated that the BEPS_PF is an efficient tool for optimizing different processes in the ecosystems, and with the satellite data it can be extended to regional and global scales for more accurate estimation of carbon fluxes.