In the descriptions presented in the previous chapter, the pebble lithology of each gravel member and the origins of the material have been discussed. However, if the lithostratigraphical scheme is valid it is necessary to demonstrate as far as possible that the lithological assemblages are sufficiently characteristic to identify the individual members. Such a comparison also provides insight into both local and regional catchment changes that may be present, together with other possible temporal patterns of lithological change.

In order to express and compare multivariate compositional data from the Lower Thames gravel members (see Appendix 2) as simply as possible, the analyses were compared numerically using the multivariate statistical technique, principal components analysis (PCA). Of the techniques available for the comparison of groups of samples consisting of an assemblage of variables, the most useful are those that ordinate all the samples together. Ordination procedures such as PCA seek to represent in a low-dimensional space the similarities between the individual samples. In this technique, samples of similar composition, irrespective of which group they are derived from, will, if the ordination is an accurate representation of the data, be positioned together in the ordination (Blackith & Reyment, 1971; Gordon & Birks, 1974). The geometrical disposition of samples will thus allow the detection both of clusters of similar lithological composition that may originate from the same lithostratigraphical unit and also of groups of dissimilar lithological composition that may originate from different units.