Revealing centurial changes in butterfly wing shape via the automated analysis of museum specimens

Abstract Understanding how environmental change is shaping organismal morphology of key functional guilds is important, since morphological traits determine the delivery of ecosystem services we depend upon. Adult butterflies, as pollinators, are commercially and culturally important, but the longitudinal data across appropriate time frames to understand trait response trends are generally limited. Here we leverage digitised museum butterfly specimens to help fill this data gap by employing automated methods (machine learning and Elliptical Fourier Analysis) to achieve high-throughput measurements of traits across large collections. We analyse how wing shape (the outline of the whole butterfly) varies in over 178,000 specimens of 65 butterfly species collected across the UK and Ireland over two centuries. We identify two major biologically attributable axes of butterfly wing shape: aspect ratio and asymmetry. Analysing a subset of these specimens (108,511 from 34 species), we find directional trends in wing shape. Over a period of 105 years (1901-2006), wings generally became narrower and longer in over half (18) of the studied species. This response may be affecting butterfly behaviour by favouring gliding flight. Ultimately, we highlight that historical processes including temperature change have historically shaped butterfly wing shape. We demonstrate that studying whole wing outline using automated methods allows us to extract the high-throughput, long-term data required to understand the responses of ecologically important taxa at scales that we were previously unable to achieve.