Relationship between Mechanical Properties and Unsaturated Hydraulic Conductivity Curves of Soils

Soil unsaturated hydraulic conductivity is one of the most important soil physical properties in recognizing, investigating and modeling the transport of water, solutes and pollutants in the soil. The objective of this study was to estimate the soil unsaturated hydraulic conductivity using easy to measure soil physical, mechanical and chemical properties by regression and artificial neural networks (ANNs) methods. In this study, 148 soil samples were taken from five provinces of Mazandaran, Kermanshah, West and East Azarbaijan and Hamedan. Pedotransfer functions were developed using soil physical, chemical and mechanical properties to estimate unsaturated hydraulic conductivity parameters of the van Genuchten- Mualem model (n and α) in 8 steps. The parameters estimated in each step, were used to simulate unsaturated hydraulic conductivity curve in the range of 0 – 1500 kPa. The accuracy of the estimated curves in each step was evaluated using curve by curve comparison with the fitted (measured) unsaturated hydraulic condoctivity curve. The ANNs performed better than the regression method, because, the AIC criterion values were obtained between -4101 and -1169 for the ANNs and between -1379 and -382 for the regression. Among the ANNs developed pedotransfer functions, the step 8 which utilized the tensile strength as an estimator along with basic soil properties, performed better than the other models in estimating the hydraulic conductivity. The results showed that the variables with little variability did not improve the estimates of hydraulic conductivity but the parameters with high variability such as tensile strength improved estimates of hydraulic conductivity.