Multivariate analysis of Australian Eocene dispersed cuticle floras: influence of age, geography and taphonomy on biozonation

Abstract

Multivariate analyses of relative frequency data for cuticle parataxa were used to test for age-class separation of Australian Eocene cuticle samples. Seventy-two samples from 27 different localities were examined from which 705 cuticle types were identified. Clustering was carried out on the whole data set, using log-transformed frequency data excluding rare parataxa, Bray-Curtis dissimilarity between sample pairs, and UPGMA fusion. Ordination was by detrended correspondence analysis of the same data set. NEither classification nor ordination showed clear-cut correlation with age-classes. The inability of teh analysis to recognise the chronological is believed to be due vegetation heterogeneity, whose input into the depositional environment is influenced by a range of factors, including geography, climate, sedimentary facies changes and taphonomic biases. Stepwise multiple discriminant analysis carried out on a reduced data set (rare parataxa removed) was partially successful in eliminating the influence of vegetation heterogeneity. Fifty-one parataxa selected by preliminary X 2 teste were reduced to 22 principal indicators which were ranked according to their ability to distinguish age-classes. Combining these indicators into linear discriminant functions enabled separation of the Early Eocene, lower Middle Eocene and Late Eocene with 100% accuracy, and the upper Middle Eocene and middle Eocene with 95.7% and 83.3% accuracy, respectively. Increased accuracy of the biozonation is considered achievable by reducing further the influence of vegetation heterogeneity through the spatial and temporal expansion of the dispersed cuticle database and development of additional area-specific palynological dating schemes.