Neural Network Analysis for Hierarchical Prediction of Soil Hydraulic Properties

Abstract

The solution of many field‐scale flow and transport problems requires estimates of unsaturated soil hydraulic properties. The objective of this study was to calibrate neural network models for prediction of water retention parameters and saturated hydraulic conductivity, Ks, from basic soil properties. Twelve neural network models were developed to predict water retention parameters using a data set of 1209 samples containing sand, silt, and clay contents, bulk density, porosity, gravel content, and soil horizon as well as water retention data. A subset of 620 samples was used to develop 19 neural network models to predict Ks. Prediction of water retention parameters and Ks generally improved if more input data were used. In a more detailed investigation, four models with the following levels of input data were selected: (i) soil textural class, (ii) sand, silt, and clay contents, (iii) sand, silt, and clay contents and bulk density, and (iv) the previous variables and water content at a pressure head of 33 kPa. For water retention, the root mean square residuals decreased from 0.107 for the first to 0.060 m3 m‐3 for the fourth model while the root mean square residual Ks decreased from 0.627 to 0.451 log(cm d‐1). The neural network models performed better on our data set than four published pedotransfer functions for water retention (by ≈0.01–0.05 m3 m‐3) and better than six published functions for Ks (by ≈0.1–0.9 order of magnitude). Use of the developed hierarchical neural network models is attractive because of improved accuracy and because it permits a considerable degree of flexibility toward available input data.

Hydraulic properties and embolism in small-diameter roots of five temperate broad-leaved tree species with contrasting drought tolerance

“…The TDR sensors at lower depths were used for monitoring the depletion of soil water in the profile. Soil matrix potential (Ψ S ) was estimated with the program Rosetta Version 1.2 (Schaap et al 1998) from the soil water content data. This program implements pedotransfer functions to predict water retention parameters after van Genuchten (1980) based on soil bulk density and textural distribution.…”

Determining Soil – Water Content by Data Driven Modeling When Relatively Small Data Sets Are Available

“…This approach has been described e.g. in [4] or [7]. Briefly summarized, a neural network consists of input, hidden and output layers, which contains processing elements.…”

“…Besides the standard regression methods, artificial neural networks (ANNs) have become the tool of choice in developing PTFs, e.g., [1], [5], [7], [10], etc.). Authors of above works confirm that they received better results from ANN-based pedotransfer functions than from standard linear regression-based PTFs.…”

Markov chain Monte Carlo simulation using the DREAM software package: Theory, concepts, and MATLAB implementation

“…An explicit prior distribution is used for the soil hydraulic parameters to make sure that their posterior estimates remain in close vicinity of their respective values derived from surrogate soil data using the Rosetta toolbox of hierarchical pedo-transfer functions (Schaap et al, 1998(Schaap et al, , 2001). The initial state of each chain is sampled from the prior distribution, and boundary handling is applied to enforce the parameters to stay within the hypercube specified by min and max.…”