Objective The traditional chemical method for determining soil cation exchange capacity (CEC) is a time-consuming and laborious work, so many historical soil data were lack of CEC information. The study aims to develop a predictive model of soil CEC in Anhui Province based on easily accessible variables.

Method The soil CEC prediction model was established by using the information of soil organic matter content, particle composition and pH of 711 soil samples covering Anhui Province. A stepwise multiple linear regression method was used to establish the CEC prediction model, and its influence on the prediction accuracy of the model was examined by six groupings: soil horizon, soil type, parent material, land use, texture and calcareousness.

Result ① The accuracy of the prediction model built using ungrouped data from the province was low, with an adjusted R² of only 0.33. ② Grouping soil samples by soil type, land use and calcareousness improved the prediction accuracy of the model overall, with an adjusted R² ranged from 0.44 to 0.93. However, grouping soil samples by soil horizon, parent material and texture did not improve the model accuracy significantly or even decreased. ③ The important parameters for prediction of soil CEC in Anhui Province are soil clay content, followed by organic matter content and pH.

Conclusion The prediction accuracy of the CEC model based on the unsorted dataset for soils in Anhui Province is quite low. Grouping based on soil type, land use pattern and calcareousness can enhance the prediction accuracy of CEC. The most important variable for CEC prediction is clay content, followed by SOM and pH.