“`html

Scholars from U-M, NYU, and other institutions joined efforts to examine myriad museum specimens, affirming that an age-old ecological principle pertains to avian wings

ANN ARBOR—For ages, researchers have noted that fauna in soothing climates exhibit longer appendages, a phenomenon referred to as Allen’s Rule. Historically associated with the necessity to control body temperature, the exact process leading to this phenomenon has remained inadequately comprehended.

A novel computer vision system has now validated that this principle is relevant to bird wings as well. This discovery reformulates our comprehension of avian wing evolution to encompass the necessities of temperature management alongside flight mechanics. This insight could assist scientists in deciphering how birds might adjust to climate shifts.

“Gathering skeletal measurements on a grand scale enables us to address significant inquiries regarding how species evolve and engage with their surroundings,” stated Brian Weeks, the principal author of the new investigation and assistant professor at the University of Michigan’s School for Environment and Sustainability.

Published in the journal Global Ecology and Biogeography, the research signifies the culmination of a six-year partnership between ecologists and computer scientists from the University of Michigan and New York University.

The team’s “Skelevision” system leverages artificial intelligence to autonomously recognize and gauge bird bones from images.

“We employ a deep neural network to identify individual bones in specimen photographs, classify their type, and produce an accurate digital outline of each,” explained David Fouhey, a senior author of the study and assistant professor at NYU Tandon School of Engineering and NYU’s Courant Institute.

“Together with the jointly designed hardware, we can simplify 3D measurements to a 2D task, where contemporary computer vision systems excel.”

Prior to this technology, scholars tended to examine skeletal characteristics in relatively limited sample sizes. The tedious process required manually handling delicate bones and measuring each component with calipers, leading to a bias towards more easily measured external features.

The research team created a comprehensive system for analyzing bird skeletons—including both the imaging hardware featuring a high-resolution camera positioned above a setup for bird bones and the advanced AI software that evaluates these images to identify and measure individual bones.

This combined hardware-software strategy reduces the time for specimen handling from approximately 15 minutes to about one minute each. The methodology was detailed in a 2023 report in Methods in Ecology and Evolution, demonstrating Skelevision’s accuracy across 12,450 bird specimens.

This efficiency enabled the researchers to analyze wing-bone measurements from 1,520 species of passerine birds across 80 families from all continents except Antarctica. The specimens predominantly originated from the University of Michigan Museum of Zoology, and the dataset has since been enhanced with specimens from Chicago’s Field Museum of Natural History.

“Wing bones serve a distinct function in thermal regulation,” Weeks noted. “As birds take flight, the muscles supported by these bones are essential for dissipating the substantial heat produced by flight. This implies that the pattern we are observing—longer wing bones in warmer areas—is primarily motivated by the necessity for effective cooling rather than warmth conservation.

“Even features as vital as wings, which we have traditionally analyzed solely for flight mechanics, are being influenced by thermal regulation requirements. This carries significant implications for how birds could react to climate change.”

The technology is being further advanced with an enhanced 3D scanning system to evaluate additional attributes such as volume and surface area. The researchers have also made their dataset and open-source code publicly available.

Alongside Weeks and Fouhey, the authors of the study include Christina Harvey from the University of California Davis; Joseph Tobias from Imperial College London; Catherine Sheard from the University of Bristol; and Zhizhuo Zhou from the University of Michigan and Carnegie Mellon University.

Research funding was provided by the David and Lucile Packard Foundation.

“`