AGGLOMERATIVE HIERARCHICAL CLUSTERING FOR DELIMITING VENERIDAE SPECIES

Many species delimitation approaches use variation in geometric morphometric data rather than character traits used by taxonomists. We use agglomerative hierarchical clustering on different sets of character traits to delimit species and measure the information gain achieved by using additional characters. We illustrate this analysis on the bivalve family Veneridae, the most diverse Recent bivalve family (approximately 170 genera and 800 extant species). The baseline set of characters describes external shell sculpture, lunule, escutcheon and periostracum features and length to height ratio with 8 characters. Additional character sets tested include 23, 30 and 34 characters, respectively. We gathered feature data on 69 individuals. The specimens consist of 8 genera and a total of 14 species. We cluster the individuals using each character set and analyze the information gain and accuracy. We find this approach a useful indicator of species delimitation fidelity for the Veneridae family. Given the findings, practitioners and researchers alike can determine the extent to which gathering additional data will enhance the species identification task.