Numerical Taxonomy of Soils from Nine Orders by Cluster and Centroid-Component Analyses1
- J. E. Cipra,
- O. W. Bidwell and
- F. J. Rohlf2
Multivariate statistical procedures of numerical taxonomy were used to investigate the patterns of similarities among 59 soils. Twenty-one morphological and laboratory characteristics of modal soil profiles from nine orders of the new classification system were used. Characters were selected to avoid high intercorrelations. Raw character values were transformed to give each character a mean of zero and variance of unity. Centroidcomponent analysis with projections of soils onto centroidcharacter axes was employed. This facilitated expression of similarities among soils in three dimensions in a manner which did not assume that clusters of soils existed. Correlation and distance matrices also were computed to obtain two estimates of the similarity of each individual to every other. Similarities among soils as indicated by these two matrices were summarized by cluster analysis. Results were displayed in phenograms yielding a system of nested clusters of soils.
The three methods of estimating and summarizing similarity relationships gave comparable results. The numerical classifications agreed, in general, with the new classification system with respect to Mollisols, Alfisols, Inceptisols, Entisols, and Spodosols. Vertisols showed little affinity for one another, Aridisols clustered with Mollisols and Alfisols, and the single Oxisol clustered with Ultisols. Ultisols showed no special affinity for other Ultisols. Numerical taxonomic methods are capable of summarizing large amounts of new data efficiently, making them a valuable soil classification tool.Please view the pdf by using the Full Text (PDF) link under 'View' to the left.
Copyright © .