Abstract: Ecological process model provides an effective way for studying water, carbon and nitrogen cycles of terrestrial ecosystem. The combination of the model and data assimilation could build a link between observation and simulation, which improves the effectiveness of model and observation methods to reveal the real state of land surface, but the accuracy of simulation and assimilation is directly affected by model parameters. Parameters are constants in traditional opinion; however, some current studies have focused on the timespace heterogeneity of parameters. In this paper, we used Harvard Forest Environmental Monitoring Site (EMS) data to simulate water and carbon flux in Harvard Forest area. Firstly, the sensitivity of parameters was analyzed. Additionally, in order to acquire monthly optimal values of parameters, the values of parameters were changed repeatedly in reasonable range, simulated annealing algorithm was used and objective function was built. Also, the time heterogeneity of sensitive parameter was analyzed through coefficient of variation. The results showed that the ecological model parameters were not constants, and the property of parameters changed with time. In addition, the heterogeneity of parameters was different. In this paper, the sensitive parameters in the Biome-BGC model were divided into corresponding levels based on their time heterogeneity. The results of this study could promote indepth understanding of ecological model parameters and provide an idea for identification and optimization of these parameters, which is helpful to improve the accuracy and effectiveness of model and data assimilation.