Developing pedotransfer functions to investigate the relationship between soil cation exchange capacity and mean weight diameter

Pedotransfer functions are used increasingly to predict the cation exchange capacity (CEC), because, its measurement is difficult, costly and time-consuming. The 129 soil samples were collected from Hamedan and Guilan provinces. The soil particle size distribution, organic matter, CEC, pH and mean weight diameter (MWD) have been measured. In addition to conventional predictors of soil CEC, MWD have been used to predict CEC and their capability in the improvement of CEC predictions by regression and artificial neural networks (ANNs) have been evaluated. The results showed that entering MWD improved the accuracy of the models from 6 to 46 %. The indirect relationships between MWD and CEC were the reason for this improvement. According to root mean square error, ANN models performed better than regression models in the prediction of CEC in training step but, the different results were found in the testing step for most of the models. Therefore, MWD is the suitable predictor for the estimation of soil CEC by regression and ANNs methods.