Using of Specific Surface to Improve the Prediction of Soil CEC by Artificial Neural Networks

Cation exchange capacity (CEC) is one of the most important soil properties. Its direct measurement is difficult, costly and time-consuming. In spite of large number of researches done to predict CEC, its prediction improvement by adding new input variables, however, remains a challenging issue. To our knowledge no one has used the auxiliary variable of specific surface to predict CEC. In the present work, 1662 disturbed soil samples were collected from different parts of Guilan province. Soil properties including pH, sand, silt, clay, organic carbon, and CEC were measured. The entire particle size distribution (PSD) curve was extended from limited soil texture data. Using Skaggs et al moded. Then, total specific surface (TSS) and the product of the specific surface of clay fraction and its mass fraction (SS1) were calculated from the extended PSD curve to predict CEC by artificial neural networks. Strong nonlinear correlation was found between CEC, TSS and SS1. CEC predictions were improved by using TSS and SS1 in the PTFs. SS1 was the most important variable in the prediction of CEC. Partitioning the whole data into eight groups improved significantly the performance of the PTFs and increased the effect of TSS and SS1 in improving the CEC prediction. Using these PTFs is an easy and economical method and it would be a great step forward in improving the estimation of soil CEC.