The method of factorial kriging based on the theory of coregionalization is developed by the combination of multi-statistics, geostatistics and GIS. Soil available heavy metals of 126 topsoil samples in Kunshan city, a typical region of Yangtze River Delta, were analyzed, and the spatial distribution pattern was investigated by the method of factorial kriging. Based on the analysis of multi-scale spatial structure characteristics of available heavy metals, we discussed the pollution source and cause of this spatial distribution by means of spatial scale-correlation analysis and spatial principal component analysis. Our results show that all the available heavy metals distribute normally or lognormally with great variability, and the contamination of available Cd is the biggest. The available heavy metals are categorized into three spatial scales, i.e. nugget, short-range (15 km) and long-range (40 km), respectively, and a linear model of coregionalization comprising these three spatial scales is fitted to the experimental auto-and cross-variograms of the soil available heavy metals. Significant relationship is found between Cd and Zn in the three scales. The spatial correlation of available heavy metals in short-range and long-rang are stronger than it in nugget, while the long-rang has more obvious negative correlation than the other two spatial scales. The results of spatial principal component analysis show the pollution sources are different in the three spatial scales. The kriging interpolation method was applied to work out the distribution maps of first and second principal component of available heavy metal, which indicate that available heavy metal concentrations in the soils are closely related to their industry activity, sewage irrigation and soil characteristics.