Landmark-based morphometrics with fewer landmarks: Some examples for medical entomology

Geometric morphometrics based on two-dimensional landmarks is a powerful tool for distinguishing morphologically similar or cryptic taxa, an important asset in the fight against medically and veterinary important arthropods. While it is commonly assumed that increasing the number of landmarks should improve discriminatory power by capturing more shape information, our findings challenge this assumption. In terms of shape discrimination (thus excluding size variation), we demonstrate that small subsets of landmarks can equal or even outperform full sets of landmarks. Fifteen examples of comparisons between closely related species were considered. These examples are drawn from published data covering six insect families: Culicidae, Glossinidae, Muscidae, Psychodidae, Reduviidae and Tabanidae. To assess the relevance of smaller subsets of landmarks, we compared the accuracy scores of unsupervised classification using full sets of landmarks (10–22 points) with those obtained using smaller subsets. To eliminate the potential influence of chance on reclassification scores, we validated our results by accounting for correct reclassifications due to chance alone. The strategy for selecting relevant landmark subsets employed two different approaches. The first relied on each landmark’s contribution to the total distance between shapes, thus establishing a hierarchy among them. The second, more comprehensive approach compared the reclassification scores of large random samples of landmarks, from the smallest subsets (3 landmarks) to the full set. From a public health perspective, the value of our approach lies in simplifying the tasks required for entomological surveillance: it could accelerate morphometric identification for large surveillance datasets, improve standardization among users, and reduce noise introduced by problematic landmarks. These gains are particularly relevant for distinguishing medically important but morphologically similar taxa, or when molecular tools are unavailable or too resource-intensive. The statistical procedures have been integrated into the XYOM online software, providing accessible tools for efficient landmark selection and improved morphometric analysis.Author summary: Landmark-based geometric morphometrics describes shape in direct relation to the number of landmarks used. It is commonly assumed that increasing the number of landmarks allows for more information about shape, and when discriminating between groups or taxa, this strategy is expected to improve classification accuracy. Our results challenge this assumption. We demonstrate that subsets of landmarks, as small as three or four, can equal or even outperform the species classification obtained by the full set of landmarks, and we propose two methods for identifying them. We discuss the possible causes of these counterintuitive results and the perspectives they could open up for morphometric studies.