Abstract: Quantitative and spatially explicit forest change and fragmentation information is critical for addressing many pressing issues,including global climate change, carbon budgets, and sustainability. In this study, multitemporal TM images in 1984, 1994 and 2005 were used to classify 105 spectral classes respectively by implementing the unsupervised ISODATA algorithm. and the spectral classifications were conversed into the actual land cover classes involving forest, nonforest and missing value based on the local reference data including China’s land use for 1980’s and 1995, as well as the Globcover data for 2004—2006 and forest inventory data for 2005 in this analysis. Built on the aggregated patterns of forest versus nonforest, a forest fragmentation model was employed to extract the forest fragmentation status and trends over this collective forest area. Analysis results showed that forest area increased in the period from1984 to 1994 but it gradually reduced from 1995 to 2005. An accelerated forest fragmentation was observed in the whole study period, which was evidenced by a gradual decreasing share of interior forest accompanied by an increasing trend for perforated forest, patch forest and transitional forest, while with a relative stable status for edge forest and undetermined forest. The findings has implied that the unreasonable anthropogenic disturbances should be prohibited to impose on the existing forests over this region and an informative regulation plan sponsored by the local administrative agencies should be implemented when pursuing a philosophy of sustainable forest management to mitigate even eliminate the potential threats to forest ecosystems and biodiversity.