Purpose: Soil management strategies for agricultural lands contaminated with potentially toxic trace elements, especially cadmium (Cd), are still inadequate and require a precise identification of soils that are not s afe for growing crops. Key soil variables need to be identified to connect soil safety with food safety by reliable models. Materials and methods: Soil variables that affect concentrations of metals in different portions of radish, Raphanus sativus L., were examined as part of a greenhouse experiment. 0.01 M di-sodium-di-hydroxy-ethylenediamine-tretra-acetic acid (Na2H2EDTA) solution was used in a strong rinsing experiment. Cd soil-radish relationships were derived by different modeling approaches and were used to develop local risk screening values for Cd in soil. Results and discussion: The current lab washing procedures readily remove surface Cd adherence but are not adequate to remove surface-deposited lead (Pb), thus overestimating bioaccumulation in plants by mean 111%. Shoot and root tissue Cd concentration in fresh weight basis do not present a significant difference and can be precisely predicted by regression models using different Cd pools and soil pH. Preferably a polynomial surface model can be used in developing local rick screening values that yield concentrations of Cd in radish at or below the Chinese food quality standard of 0.1 mg kg−1 (fresh weight). Conclusions: The bioaccumulation of Cd in radish depends on the Cd bioavailability in soil. But for Pb, surficial particle contamination masks the realistic bioaccumulation. We also demonstrated the usefulness of the polynomial surface model to develop local soil protection guidelines that are helpful to local farmers for proper soil management and avoidance of Cd exceedance in food.