EVALUATION OF REGIONAL PEDOTRANSFER FUNCTIONS BASED ON THE BP NEURAL NETWORKS

Abstract

The unsaturated soil hydraulic properties, including soil water retention curve and hydraulic conductivity, are the crucial input parameters for simulating soil water and solute transport through the unsaturated zone at regional scales, and are expensive to measure. These properties are frequently predicted with pedotransfer functions (PTFs) using the routinely measured soil properties. 110 soil samples at 22 soil profiles from Jiefangzha Irrigation Scheme in the Hetao Irrigation District of Inner Mongolia, China were collected for the analysis of soil properties i.e. soil bulk density, soil texture, particle size distribution, organic content, and soil water retention curve (SWRC). The Brooks-Corey (BC) model and van Genuchten (VG) model were used to fit the measured SWRC data for each soil sample by using the RETC software. Pedo-transfer functions (PTFS), which describes relationship between the basic soil properties and the parameters of the BC and VG models, were then established with the artificial neural networks (ANN) model. It is found that the ANN model has better effect on the clay loam, loamy clay, loam soil and silty clay to simulate BC model. However, it has better effect on the loam soil, loamy clay and sandy clay to simulate VG model. So, we can draw the conclusion that the ANN model can conveniently establish PTFS between soil basic feature parameters and SWRC model and has reasonable precision. This will be a good method to estimate soil water characteristic curve model and soil hydraulic parameter in the regional soil water and salt movement simulation and water resources evaluation.KeywordsArtificial Neural Network ModelClay LoamSoil Hydraulic PropertySoil Water Retention CurveBasic Soil PropertyThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.