Morphometric methods for the analysis and classification of gastropods: a comparison using Littorina littorea

The study of morphology is a common means of biological grouping and classification. In recent years, morphometric studies have been dominated by quantitative geometric-morphometric methods of data extraction such as outline or landmark-based analysis. These methods are often used in conjunction with various classification methods such as linear discriminant analysis (LDA) and random forests (RF) in order to achieve inter-and intraspecific grouping based on environmental factors. Despite numerous studies incorporating these data-extraction and classification methods, comparisons of the effectiveness of these methods are largely lacking, especially for species which display low morphological variation. The aim of this study was to compare the effectiveness of two data-extraction methods, elliptic Fourier analysis (EFA) and generalized Procrustes analysis, and two classification methods, LDA and RF, using Littorina littorea as the study organism. The results show that the principal component scores derived from EFA, provided the optimal data input for classification while the greatest percentage of successfully classified individuals was achieved using LDA. However, based on this study RF is the recommended classification method as it is resistant to overfitting, makes no assumptions about the data, is well suited to morphometric data and produces similar rates of classification to LDA. The results are discussed in a biological context for L. littorea, based on the environmental factors of zonation and shore exposure.