Search

【Background】 Accurate prediction of soil water dynamics requires soil hydraulic parameters and the PedoTransfer Function (PTF) is an indirect method estimating soil hydraulic parameters based on easy-to-measure soil properties. 【Objective】 The purpose of this paper is to compare different machine learning methods as a pedotransfer function to estimate soil hydraulic parameters. 【Method】 We used the UNSODA soil hydraulic property database and compared three machine learning methods: artificial neural network (ANN), support vector machine (SVM), and K-nearest neighbor (KNN). The hydraulic parameters were described by the van Genuchten formula, and the relationship between its parameters with fractions of sand, silt and clay, as well as soil bulk density was analyzed using the three methods. The accuracy of each method was evaluated using measured water retention curves and saturated hydraulic conductivity from soils with different textures. 【Result】 The SVM model was most accurate and KNN the least to predict soil hydraulic parameters using these easy-to-measure soil properties. Evaluation of operating efficiency of all three methods revealed that the ANN model was least efficient and the KNN the most in model training. In contrast, the ANN model was most efficient while KNN the least in predicting soil hydraulic parameters. Comparison of the three models against the Rosetta model - a commonly used neural-network pedotransfer function with a single hidden layer - found that neural-network models with multiple hidden layers, as used in this paper, were more accurate. We also found that for all three models, increasing the number of input data improved their estimation accuracy. 【Conclusion】 Of the three models, SVM is most accurate for predicting soil hydraulic parameters using fractions of clay, sand and silt, and bulk density, followed by ANN, when the database was not large enough. With the size of the database increasing, the ANN model becomes increasingly more efficient. Since ANN can use the mini-batch method to train the model without increasing computational costs, our results suggest that selecting a suitable method to calculate soil hydraulic parameters should consider computational cost and estimation accuracy when the size of the database increases.

The soil water retention curve is the fundamental soil hydraulic property to characterize soil water movement and solute transport. Many efforts have been devoted in the past decades to developing models to describe soil water retention curves. However, most of them are empirical equations or assume that soil pore size distributions conform to a lognormal distribution. Yet, few effects have been undertaken to systematically propose and compare a series of possible alternative probability density functions to describe the sigmoid retention curves with parameters physically explainable. Here, we proposed a family of five soil water retention models based on sigmoid functions with parameters of clear physical implications coinciding with the statistical measures of soil pore size distribution. Compared with the widely used models (i.e., Brooks & Corey, 1964; Kosugi, 1996; van Genuchten, 1980), the proposed models have somewhat improved performances to characterize water retention data for a wide range of soil textures without introducing additional model parameters. Two of the proposed models are capable of characterizing the observed two local extrema in the moisture capacity curves. The associated unsaturated hydraulic conductivity models of the proposed soil water retention models are also derived, which show superior performance in characterizing the observed hydraulic conductivities compared with competing models, especially in macropore regimes. Additionally, we analyzed the parameter‐equivalent conversion between the proposed and the existing models, and a simple linear regression equation can be used to derive the parameters of the proposed models from the existing and other alternative different proposed models. Key Points: A family of soil water retention models based on sigmoid functions and related relative hydraulic conductivity functions are proposedParameters have statistical implications against pore size distributions and can be converted from existing or alternative new modelsCharacterization of hydraulic properties is improved using new models, especially for macropore properties, without additional parameters [ABSTRACT FROM AUTHOR]