Assimilating Eddy-covariance Flux into an Ecosystem Model: A Case Study in China's Forest Ecosystem

Process-based ecosystem model without the observed constraints may bias high and does not meet the demand for accurately estimating the regional carbon balance and assessing regional ecosystem responses to climate change. In this study, we collected and processed data from 6 eddy-covariance flux sites (5 plant types) and hereby optimized the parameters of a state-of-the-art ecosystem model in China's forest ecosystem. The prior simulations has overestimated the net ecosystem exchange (NEE) over all forest sites, with the mean biases ranging from 0.41 to 3.09 gC m-2 d-1. The optimization process has reduced these biases and enhanced the agreement of the simulated seasonal cycle of the latent heat (LE) and gross primary production (GPP), through alteration in parameters by reducing the vegetation (and soil) respiration and enhancing the plant photosynthetic capability at most forest sites. Substantially underestimated LE at 4 out of these sites has also been improved by 4% 20%. The effect of the optimized parameters has been further demonstrated by validation of the observed fluxes in the other years, implying the capability of the optimized model in systematically improving the model's performance in particular at the sites where errors of the prior simulations are marked. Simulations of separate scenarios of altered temperature and precipitation indicate that the modeled carbon cycling responses are linear to warming and asymmetric to alteration in precipitation despite the optimized parameters, whose impact is found complicated and nonlinear on the response of GPP to climate change while is quite simpler for that of the respiration processes.