Can molecular data validate morphometric identification of faecal pellets in Tanzanian forest antelope species?

Abstract

Forest-dwelling antelope species are often difficult to detect during surveys due to their cryptic behaviour and densely vegetated habitats. Dung counts have traditionally been used to infer forest antelope abundance but genetic identification has shown that visual identification of ungulate dung to species is often unreliable. This study attempted to use easily obtained morphometric data from faecal pellets to statistically assign antelope dung piles to species. We measured pellets from 238 dung piles collected from the Udzungwa Mountains, south-central Tanzania, a largely forested landscape with five forest-associated antelope species including the endangered Abbott’s duiker Cephalophus spadix. The species identity of sampled dung piles was determined by amplifying a c. 600 bp fragment of the mitochondrial control region and aligning DNA sequences with published references. We found no diagnostic differences in faecal pellet size between antelope species although there were significant differences in mean pellet length and width. We employed a single variable linear discriminant analysis to predict the species of dung piles based on pellet length. Despite significant differentiation between species we obtained an overall accuracy of 58.8 % that did not meet our specified probability threshold (P < 0.05). Abbott’s duiker dung piles were correctly assigned in the majority of cases (74 %). Overall, morphometric assignment of dung piles to species was not accurate enough to validate dung counts as a survey method for forest antelope although our results do not preclude the development of alternative field identification methods using additional non-molecular characters.