Measurement Scale Effect on Prediction of Soil Water Retention Curve and Saturated Hydraulic Conductivity

摘要

Soil water retention curve (SWRC) and saturated hydraulic conductivity (SHC)are key hydraulic properties for unsaturated zone hydrology and groundwater. Inparticular, SWRC provides useful information on entry pore-size distribution,and SHC is required for flow and transport modeling in the hydrologic cycle.Not only the SWRC and SHC measurements are time-consuming, but also scaledependent. This means as soil column volume increases, variability of the SWRCand SHC decreases. Although prediction of the SWRC and SHC from availableparameters, such as textural data, organic matter, and bulk density have beenunder investigation for decades, up to now no research has focused on theeffect of measurement scale on the soil hydraulic properties pedotransferfunctions development. In the literature, several data mining approaches havebeen applied, such as multiple linear regression, artificial neural networks,group method of data handling. However, in this study we develop pedotransferfunctions using a novel approach called contrast pattern aided regression(CPXR) and compare it with the multiple linear regression method. For thispurpose, two databases including 210 and 213 soil samples are collected todevelop and evaluate pedotransfer functions for the SWRC and SHC, respectively,from the UNSODA database. The 10-fold cross-validation method is applied toevaluate the accuracy and reliability of the proposed regression-based models.Our results show that including measurement scale parameters, such as sampleinternal diameter and length could substantially improve the accuracy of theSWRC and SHC pedotransfer functions developed using the CPXR method, while thisis not the case when MLR is used. Moreover, the CPXR method yields remarkablymore accurate soil water retention curve and saturated hydraulic conductivitypredictions than the MLR approach.