Based on spatial continuous datasets (NDVI) and local indicators of spatial association (LISA)——local Moran’s I (LMI), this paper deals with the process of landscape pattern fragmentation in a small watershed by detecting the change of local spatial autocorrelation of landscape features. As a spatial explicit indicator evaluating the landscape heterogeneity, LMI can detect the hot spots during the landscape fragmentation process. The relationship between the landscape pattern change and the land use / land cover change in the past 30 years is also studied, and the results show that the hot spots of landscape fragmentation are located in downstream areas where land use changes are very dramatic. The researches performed by the authors show that the land use change in the small watershed is the main driving force resulting in such changes of landscape pattern as landscape fragmentation. Studies show that the local indicators of spatial autocorrelation are helpful to the traditional analysis of landscape pattern.