Application of Artificial Neural Networks in Prediction of Saturated Hydraulic Conductivity Using Soil Physical Parameters

Soil hydraulic properties such as saturated and unsaturated hydraulic conductivity play an important role in environmental research. Since direct measurement of these soil hydraulic properties is time-consuming and costly, indirect methods such as pedotransfer functions and artificial neural networks (ANN) were developed based on readily available parameters. In this study, the use of ANN to predict saturated hydraulic conductivity using a measured soil moisture curve and bulk density was investigated. Measured bulk density and soil moisture curves were used to estimate saturated hydraulic conductivity from calculated fractal dimensions, air entry values, bulk density and effective porosity using ANN. In the training and testing steps of ANN, 114 and 28 measured soil samples were used, respectively. R2 and MSE were 0.76 and 0.0028, respectively, for the ANN method with four inputs. A comparison of ANN with Rawls et al. (1993, 1998) models showed that the neural network more accurately predicts saturated hydraulic conductivity.