MULTIVARIATE APPROACH TO GROUPING SOILS IN SMALL FIELDS : II. Soil Grouping Technique by Cluster Analysis

概要

論文の詳細を見る
Cluster analysis was applied in grouping soils based on their scores of the factors controlling soil variation in a cultivated field at the Experimental Farm of the International Institute of Tropical Agriculture in Ibadan, Nigeria. The procedure employed here consisted of two steps: 1) grouping of the samples into a given number of provisional clusters through non-hierarchical cluster analysis, and 2) analysis of the similarity among those clusters using their means with hierarchical cluster analysis and determination of the proper number of the final clusters. Two initial settings in the assignment of the number of the provisional clusters were tested in non-hierarchical cluster analysis. Both trials gave a consistent classification, which also agreed with that based only on hierarchical cluster analysis. This combination of non-hierarchical and hierarchical cluster analyses enabled us to prepare a soil map solely based on the factors controlling soil variation in the study area without introducing an a priori idea of soil classification. This method is not only an alternative to hierarchical clustering, but a more efficient tool in terms of capacity and computational speed to classify large numbers of samples which hierarchical cluster analysis is unable to handle.
社団法人日本土壌肥料学会の論文