Parametric Estimation of Water Retention Using Mgmdh Method and Principal Component Analysis

Mohammad Reza Neyshaburi,Mostafa Rastgou,Andrew S Gregory,Nader Nariman-Zadeh,Kourosh Mohammadi,Hossein Bayat

doi:10.17951/pjss.2016.49.1.29

Abstract

Performing a primary analysis, such as principal component analysis (PCA) may increase accuracy and reliability of developed pedotransfer functions (PTFs). This study focuses on the usefulness of the soil penetration resistance (PR) and principal components (PCs) as new inputs along with the others to develop the PTFs for estimating the soil water retention curve (SWRC) using a multi-objective group method of data handling (mGMDH). The Brooks and Corey (1964) SWRC model was used to give a description of the water retention curves and its parameters were determined from experimental SWRC data. To select eight PCs, PCA was applied to all measured or calculated variables. Penetration resistance, organic matter (OM), aggregates mean weight diameter (MWD), saturated hydraulic conductivity (Ks), macro porosity (Mp), micro porosity (Mip) and eight selected PCs were used as predictors to estimate the Brooks and Corey model parameters by mGMDH. Using PR or OM, Ks and MWD, improved the estimation of SWRC in some cases. Using the predicted PR can be useful in the estimation of SWRC. Using either the MP and Mip or the eight PCs significantly improved the PTFs accuracy and reliability. It would be very useful to apply PCA on the original variables as a primary analysis to develop parametric PTFs.

Highlights

Pedotransfer functions (PTFs) are functional relationships that transfer routinely, or cheaply measured properties into missing soil properties
The new feature of this article was using multi-objective group method of data handling for the point estimation of water content and comparing its results with those of the pedotransfer functions (PTFs) developed by artificial neural networks (ANN)
This study is a resumption of the research in which PTFs were developed by ANNs and a multi-objective group method of data handling to predict the Brooks and Corey water retention parameters [2] and all of the data were taken from Bayat, et al [2], Bayat, et al [4], Bayat, et al [5]

Summary

INTRODUCTION

Pedotransfer functions (PTFs) are functional relationships that transfer routinely, or cheaply measured properties (e.g. soil texture, structure and organic matter content) into missing soil properties C1M intercept of first domain of PTF1 – PTF4 Millan’s model c2M intercept of second domain PTF1 – PTF4 of Millan’s model

D Bi mass fractal dimension of R2

MATERIALS AND METHODS

Evaluation Criteria

RESULTS AND DISCUSSION

CONCLUSION