Prediction of soil cation exchange capacity using stepwise regression based on texture and organic matter content
DOI:
https://doi.org/10.25252/SE/2025/253911Keywords:
Cation exchange capacity (CEC), gradient slope, soil properties, total organic carbon, sand, silt, clay, statistical modelingAbstract
Cation exchange capacity (CEC) is an essential index for soil fertility and nutrient adsorption, but the determination of CEC by direct method still requires a great deal of time and resources. We present a simple and practical stepwise regression model for estimating soil CEC based on easily measured cultural parameters: sand, silt, clay, and total organic carbon (TOC). A total of 114 soil samples data set was subjected to analysis by including quadratic and interaction terms, which also implies the non-linear behaviour of soil properties. In addition, a high predictive accuracy for the resulting model (R² = 97.6%, RMSE = 1.83, MAPE = 6.3%) was verified by residual analysis as well as sensitivity test and comparison with decision tree model. More importantly, the decision tree focused on the sand content as an important threshold variable and the regression model could also quantify how those effects interacted. Distinct from most studies that used complex black-box models regime (ANN, ANFIS, and SVR), the present study has shown that stepwise regression gives a good compromise between simplicity and interpretability as well as accuracy of prediction and is particularly useful in practical site-specific conditions commonly characterizing resource-poor environments. TOC showed highest relative (%) importance (35–40%) for predicting CEC followed by clay-particulate matter with 30%, then siltite (20%) and limitedly sandite at 10%. These findings demonstrate the model’s stability and the strong effect of TOC, clay on soil CEC. These results emphasize that classical statistical techniques have a role in environmental modeling, especially when resources or accessibility to the field are limited.
Downloads
Published
Issue
Section
License
Copyright (c) 2025 Soil Science Society of Pakistan

This work is licensed under a Creative Commons Attribution 4.0 International License.