Abstract: Effectively revealing landscape pattern and its spatial heterogeneity are basic for researches of landscape function and dynamic. Based on the land use classification data, this paper explored the spatial pattern, spatial heterogeneity and its scale-dependence for effective mesh size using ArcGIS 9.3, FRAGSTATS 3.3 and GS+7.0, with a case study of Wujiang River Basin in Guizhou Province, China. The results show that the relation between spatial heterogeneity of landscape fragmentation and extent exhibits significant power law with a characteristic scale of 4590 m. The regions with lower values of effective mesh size are the northeast mountain-hill area, the west of central hill-basin area and the north of west plateau-mountain area, which possess the similar characteristic of dissected topography; the regions with higher values of effective mesh size are the east of central hill-basin area and the north of west plateau-mountain area with comparatively subdued topography. The structural components such as topography and climate play the main role in the spatial heterogeneity; the data series is complex due to the influences of topography, climate, and human activities, and exhibits significant spatial auto-correlation with range of 33 km. The fractal dimensions are bigger, and the auto-correlation scales are smaller in the directions of northeast-southwest and east-west than the north-south and southeast-northwest directions. Isotropic heterogeneity and anisotropic heterogeneity are in accordance with the physical geographical characteristics. The results can provide a basis for biodiversity conservation, land use management and regional landscape planning and construction.