{"title":"Estimating Saturated Hydraulic Conductivity from Soil Physical Properties using Neural Networks Model","authors":"B. Ghanbarian-Alavijeh, A.M. Liaghat, S. Sohrabi","country":null,"institution":"","volume":38,"journal":"International Journal of Environmental and Ecological Engineering","pagesStart":58,"pagesEnd":64,"ISSN":"1307-6892","URL":"https:\/\/publications.waset.org\/pdf\/13216","abstract":"Saturated hydraulic conductivity is one of the soil\r\nhydraulic properties which is widely used in environmental studies\r\nespecially subsurface ground water. Since, its direct measurement is\r\ntime consuming and therefore costly, indirect methods such as\r\npedotransfer functions have been developed based on multiple linear\r\nregression equations and neural networks model in order to estimate\r\nsaturated hydraulic conductivity from readily available soil\r\nproperties e.g. sand, silt, and clay contents, bulk density, and organic\r\nmatter. The objective of this study was to develop neural networks\r\n(NNs) model to estimate saturated hydraulic conductivity from\r\navailable parameters such as sand and clay contents, bulk density,\r\nvan Genuchten retention model parameters (i.e. r\r\n\u03b8 , \u03b1 , and n) as well\r\nas effective porosity. We used two methods to calculate effective\r\nporosity: : (1) eff s FC \u03c6 =\u03b8 -\u03b8 , and (2) inf \u03c6 =\u03b8 -\u03b8 eff s , in which s\r\n\u03b8 is\r\nsaturated water content, FC \u03b8 is water content retained at -33 kPa\r\nmatric potential, and inf \u03b8 is water content at the inflection point.\r\nTotal of 311 soil samples from the UNSODA database was divided\r\ninto three groups as 187 for the training, 62 for the validation (to\r\navoid over training), and 62 for the test of NNs model. A commercial\r\nneural network toolbox of MATLAB software with a multi-layer\r\nperceptron model and back propagation algorithm were used for the\r\ntraining procedure. The statistical parameters such as correlation\r\ncoefficient (R2), and mean square error (MSE) were also used to\r\nevaluate the developed NNs model. The best number of neurons in\r\nthe middle layer of NNs model for methods (1) and (2) were\r\ncalculated 44 and 6, respectively. The R2 and MSE values of the test\r\nphase were determined for method (1), 0.94 and 0.0016, and for\r\nmethod (2), 0.98 and 0.00065, respectively, which shows that method\r\n(2) estimates saturated hydraulic conductivity better than method (1).","references":null,"publisher":"World Academy of Science, Engineering and Technology","index":"Open Science Index 38, 2010"}