Evaluation of Intelligent Models Due to Estimating Saturated Hydraulic Conductivity in Loamy Soils

Saturated hydraulic conductivity is one of the main parameters in agricultural and environmental studies that is essential for the estimation and management of water and solute transport in soil. In this research, 1200 series of data related to loamy soils were used to prediction and modeling saturated hydraulic conductivity using soil physical properties such as bulk density, available water capacity, organic carbon content and percent of clay, silt and sand content. From studied data series, 900 series were allocated for use in training models and 300 series for testing models in three designed scenarios. The performance of SVM, MLP and M5 models was evaluated in estimating saturated hydraulic conductivity of loamy soils. The performance of the models was compared using the statistical indices such as coefficient of determination (R2), root mean square error (RMSE) and mean bias error (MBE). The results showed that all three models used have good ability in saturated hydraulic conductivity modeling, but the M5 model with a high coefficient of determination over 0.95 in all three scenarios and a lower RMSE than other models was selected as the superior model. The results showed that intelligent data mining models make it possible to estimate unknown values based on ready patterns in a database. Therefore, the applied models can be used to predict solute transport in soil and soil physical parameters.