PERFORMANCE OF LINEAR MODELS IN PREDICTING CATION EXCHANGE CAPACITY OF CALCAREOUS SOILS

Abstract

Determination of soil cation exchange capacity (CEC) in lab is cumbersome, time consuming, and costly. Accordingly, this article attempted to formulate pedotransfer functions for predicting it using some soil physical and chemical properties e.g., sand (SA), silt (SI), clay (CL), organic matter (OM) and calcium carbonate (CC). This research included four steps: preparing soil database; selecting independent variables which are related to CEC value; formulating models using NCSS 12.0.2 software, and the last step is to achieve specific objective of the research which is the comparsion among models by a series of efficiency criteria: root mean square error (RMSE), Nash-Sutcliffe model efficiency coefficient (EF), average absolute percent error (AAPE), and percentage of improving model efficiency (PIME). The statistical results of the research indicated that CEC of calcareous soils could be predicted from models that have one variable (CL), two variables (CL and OM), and three variables (CL, OM, and CC) with slight decrease in the RMSE (2.95402, 2.81180, and 2.79268) respectively, and slight increase in the EF (0.887360, 0.898448, and 0.90023) respectively. While the reliable models to predict soil CEC are formulated from the fewer number of independent variables with having the lowest points of the standardized residual of CEC that greater than +2 cmol_c kg^-1).