Based on data collected from literatures on cropland contamination of China, relationship between cadmium contents in soil and in crops was explored and human health risk of cadmium through soil-crop-human pathway analyzed. The data used in this paper partly come from published data in the last 3 decades in China and unpublished data in our own laboratory. The data for model parameterization were distinguished between field studies and salt-added pot experiments. Uptake factors and single-variable regression models were developed based on soil Cd content and used to predict cadmium content in edible parts of rice, wheat and vegetables, while multiple regression models based on soil Cd content and soil pH were to predict Cd content in those parts. Results show that the regression models are much better than the median of the uptake factors in predicting Cd content in those parts. Generally, the upper 95% prediction intervals of regression models provided a better conservative prediction than the 90 percentile uptake factors did. Soil pH as a variable generally contributed significantly to model fit. The main factors which restricted model development and model predictive utility were uneven distribution and inadequacy of data. Finally, soil environmental benchmarks of Cd content in agricultural land were derived by applying multiple-regression model based on Cd content limits for crops as human food.