After randomizing the input information of age and distribution of forest species we simulated the long-term (2500 years) dynamic of forest landscape pattern of huzhong area in Great Xing'an Mountains using spatially explicit landscape model (LANDIS), Then we took the period when simulation results reached stable as the time range to estimate the historical range of variability (HRV). Furthermore, we analyzed the relationship between HRV and the forest landscape pattern of 1990 and 2000 using principal component analysis and kernel density estimation methods. The simulation results showed that the succession of almost all forest species arrived to their stable status at around 900th year of the simulation, and the period afterwards was utilized to estimate the HRV; as to the characteristics of forest, landscape pattern in 1990 fell out of the HRV in area of patches and the degree of landscape fragmentation, especially mature and over-mature forest deviated significantly from HRV in patch area and fragmentation; though forest harvest management was better planned after 1990, the forest landscape pattern in 2000 still deviated from the HRV.