Abstract: Spatial distribution of outliers and magnitude of point source effects of heavy metals in soil caused by point source contamination will affect accuracy of the environment risk assessment of and making of soil remediation schemes, and further environmental management of the contaminated site. A typical heavy metal contaminated site was selected for the study and Cd concentration in the topsoil was set as object. A robust variogram model was introduced to screen out a suitable variogram model. Outliers of Cd concentration in the topsoil were identified and processed and point source effects were effectively isolated with the aid of the winsorization method. Results show that the spatial variation of Cd concentration in the topsoil is great with obvious outliers. The Cressie-Hawkins robust variogram is more suitable for estimation of semi-variogram of Cd content in the topsoil. The use of the robust variogram model in couple with the winsorization method can effectively identify outliers and separate point source effects. The outliers are mainly distributed at the locations with significant human activity. As the quasi-point process often leads to sharp variation, the isolated point source effects can explain most of the Cd concentration in the topsoil of the sites of outliers.

Key words: heavy metal, outliers, point source effect, contaminated site, robust variogram