The Application of Quantitative Methods to Vegetation Survey: I. Association-Analysis and Principal Component Ordination of Rain Forest

Abstract

Classification is likely to be more satisfactory at high levels of variation in composition of vegetation and ordination more satisfactory at lower levels, and thus some combination of the 2 approaches will be more informative than either alone. Enumerations of 110 stands of forest, each of area 0.12 ha, on Kolombangara, British Solomon Islands Protectorate, were used as test data. Classification was by association-analysis, ordination by principal component analysis of Orloci''s weighted similarity coefficient. The 1st division of the association-analysis separated several western groups of stands, together with one northern group (A) (+ Teysmanniodendron) from the remaining northern groups([long dash]Teysmanniodendron). Ordination of the association-analysis groups showed that group A is misplaced with the western stands, and ordination of the + Teysmanniodendron stands revealed several cases of misclassification. After the resulting transfer of stands from the western to the northern series, each was again subjected to association-analysis. This resulted in 96 stands being distributed among 6 classes ("forest types"), the remaining 14 falling into 8 small classes. Ordination within classes was used to rectify misclassifcations. The external criterion of geographical location was used to group 15 stands, rejected from the larger classes, into a 7th forest type. Ordination within forest types revealed correlations between composition and environment not otherwise apparent. Examination of the distribution on the ordination of the more abundant large tree species, and of 4 species of Calophyllum, showed correlations with environment. The procedure used is potentially a valuable tool in the interpretation of enumeration data from rain forest.