Journal of Watershed Management Research

Mendeley  

Zotero  

Farzadmehr M, Dastourani M, Khashei Siuki A, Jalali Moakhar V. (2020). Estimating the Saturated Hydraulic Conductivity of Soil Using Gene Expression Programming Method and Comparing It with the Pedotransfer Functions. J Watershed Manage Res. 11(22), 155-164. doi:10.52547/jwmr.11.22.155
URL: http://jwmr.sanru.ac.ir/article-1-1045-en.html

Saturated hydraulic conductivity of soil is an important physical property of soil that affects water movement in soil, Since the measurement of saturated hydraulic conductivity by direct methods in the field or in the laboratory is hard, time-consuming and costly, the indirect methods are being used.The aim of this study is to estimate the saturated hydraulic conductivity from other soil properties by using the Gene Expression Programming (GEP) method and some well-known pedotransfer functions and the Rosetta model, and then to compare their performances. A dataset including 151 soil samples obtained from a site in Bojnord province was used in this study. The soil properties used were sand, silt and clay percentage, organic carbon percentage, TNV, EC, saturated moisture, pH, bulk density and particle density. Modeling process using the GEP was done by using all of these properties as input parameters. The GEP model used only four properties, sand and silt percentages, bulk density and particle density, in its developed function to estimate saturated hydraulic conductivity. This model with RMSE=2.84 (cm/d) and R²=0.91 showed the best performance in comparison to the other pedotransfer functions. After the GEP, the Jabro (1992) pedotransfer function with RMSE = 4.74 (cm/d) and R² = 0.82 was the best model in comparison to the rest of the pedotransfer functions and the Rosetta model. Saxton et al (1986), had the least accurate k_s estimation among all methods. Since different data sets had been used to develop each of the pedotransfer functions and also because of high spatial variability of k_s, there was a large difference between RMSE, MAE and MBE errors of the used methods. For the dataset of this study, the GEP model showed the best performance in estimating k_s among other methods and its advantages were choosing model structure and important parameters to estimate k_s.