It is of great practice significance to define baselines for soil heavy metals in the Yangtze River Delta Region to evaluation of status quo of the heavy metal accumulation in the soil. In the presence of anthropogenic heavy metal pollution, soil heavy metal data always significantly deviate from normal or lognormal distribution, which affects reliability of the results of classical statistics. In contrast, robust statistics is aimed at lowering the effect of outliers on statistical analysis. With the robust regression method regression relationships between Cu and Pb concentrations in topsoil and in subsoil were established for the study region of Zhangjiagang, which has thousands of years of agricultural history and has recently been developing rapidly in industry. Based on the topsoil Cu and Pb concentration baselines calculated through robust regression of their subsoil concentrations, the status of soil Cu and Pb accumulations were evaluated. Results show that the calculated geochemical baselines can reflect natural variation of soil Cu and Pb background values between different sampling sites. Compared with the robust statistical method, the evaluation of soil heavy metal accumulation in a region that has spatial variation of natural background values on the basis of the national standard for soil environmental quality or regional soil background values would generate bigger errors.