Abstract: Data assimilation provides an effective way to integrate the model simulation and remote sensing observation, through the integration of remote sensing data in the run of the model, adjusting the model trajectory to reduce model error and improve simulation accuracy. This paper uses the ensemble Kalman filter (EnKF) assimilated MODIS LAI into the BiomeBGC model in growing season to simulate the water and carbon fluxes in a broadleaved Korean pine forest in Changbai Mountains. At the same time, the simulated snow sublimation and the parameters of the calculation method of soil temperature are improved, which can effectively reduce the error of the ecological respiration in winter. The result shows that as compared with the original model simulated without data assimilation, the improved Biome-BGC model with the assimilation of the MODIS LAI makes the correlation coefficient between the simulated values and the observed values of the gross ecosystem primary productivity (GPP) increased by 0.06, and reduced the centered rootmeansquare error (RMSE) by 0.48 g C·m^-2·d^-1, ecosystem respiration (RE) correlation coefficient increased by 0.02, centered rootmeansquare error decreased by 0.20 g C·m^-2·d^-1; the correlation coefficient of net ecosystem exchange of carbon (NEE) increased by 0.35, centered rootmeansquare error decreased by 0.50 g C·m^-2·d^-1. Meanwhile, data assimilation has no significant effect on the simulation precision of evapotranspiration (ET), but the improved model improves the correlation coefficient of ET. The data assimilation based on EnKF algorithm improves the accuracy of the carbon flux simulation in the broadleaved Korean pine forest in Changbai Mountains, and has an important significance on more accurate estimation of carbon flux at regional scale.

Key words: non-growing season, greenhouse gas, temperate Korean pine plantation, main controlling factor, light-felling