About Us | Help Videos | Contact Us | Subscriptions
 

Abstract

 

This article in AJ

  1. Vol. 101 No. 4, p. 841-853
     

    * Corresponding author(s): j.triantafilis@unsw.edu.au
 View
 Download
 Alerts
 Permissions
 Share

doi:10.2134/agronj2008.0112

Digital Soil-Class Mapping from Proximal and Remotely Sensed Data at the Field Level

  1. John Triantafilis *,
  2. Belinda Kerridge and
  3. Sam M. Buchanan
  1. School of Biol., Earth and Environ. Sci., The Univ. of New South Wales, Sydney NSW 2052, Australia

Abstract

Effective agronomic management at the field level requires an understanding of the spatial distribution of soil because edaphological processes are a function of interrelationships among the physical and chemical properties. In precision agriculture, these interrelationships need to be considered to create soil management classes. In order to identify management classes, the age-old questions associated with soil classification are problematic: what properties need to be included and at what intensity; what repeatable methods can be employed and how many classes are present? To answer the first question with regards to farm-scale studies requires the collection of high-density soil data, which is often cost prohibitive. Therefore, ancillary information such as proximal sensing electromagnetic induction (EM) and remotely sensed data, with digital numbers (DN) in Red, Green, and Blue, are increasingly being used. To answer the second and third questions requires an exhaustive investigation of a quantitative method that is amenable to optimal selection of the number of classes in a data set. In this paper, we propose using ancillary data as a surrogate for soil properties to identify soil management classes by invoking the fuzzy k-means (FKM) algorithm. Using the fuzziness performance index (FPI) and normalized classification entropy (NCE), we identify k = 4 classes and a fuzziness exponent (φ) = 1.4 for further investigation. The classes form sensible soil management zones across a strongly sodic irrigated field. Fuzzy canonical analysis show the EM38 and EM31 contribute most to the discrimination of Vertosols and Dermosols based on texture and mineralogy, whilst Red and Green DN contribute to the discrimination of Vertosols based on larger organic matter content. Using ANOVA we conclude that the implementation of the FKM algorithm to classify proximal and remotely sensed ancillary data, produced soil management classes relevant to differential gypsum requirement (GR).

  Please view the pdf by using the Full Text (PDF) link under 'View' to the left.

Copyright © 2009. American Society of AgronomyCopyright © 2009 by the American Society of Agronomy