Scientists at the University of Michigan used artificial intelligence to uncover striking patterns in the evolution of passerines, a vast group of birds that includes most songbirds. Their analysis suggests that these birds did not evolve at a constant pace. Instead, major changes often arrived rapidly in parallel with changes in Earth’s climate.
Evolutionary theory has long proposed that life diversifies through periods of rapid change followed by slower stages. Fossils have provided clues that this pattern exists, and a new study identified it by analyzing skeletal measurements from specimens of modern birds. The researchers also found that some of the most important evolutionary explosions occurred during periods of major climate change.
sudden burst evolution
“This is very important for evolutionary theory, because it has a long history going back 100 years and predicts the emergence of new groups called evolutionary radiations, which are often associated with explosive bursts of diversification. Evolutionary theory predicts that adaptive radiations may account for a large portion of the diversity of life on Earth,” said Jake Barb, lead author of the study and a postdoctoral fellow in the UM School of Environmental Sustainability.
“This could be due to new ecological opportunities, or it could be because a population dispersed to a new continent and its rate of evolution accelerated dramatically. The idea is that over time, as evolution progresses, opportunities become fewer, so evolution slows down, and this happens in pulses over time. That’s what the theory predicts, and that’s what we seem to see in the data.”
The results were generated using artificial intelligence and large-scale statistical models. The survey results are natural ecology and evolution was supported primarily by Schmidt Sciences and the David & Lucile Packard Foundation.
AI measures thousands of bird skeletons
To reconstruct the evolutionary history of passerines, a team at the University of Michigan, including lead author Brian Weeks, studied more than 2,000 species and collected more than 170,000 individual skeletal measurements.
Data collection at this scale was made possible thanks to Skelevision, an AI tool developed by Weeks’ lab in collaboration with David Fouhey’s lab at New York University.
Skelevision photographs a specimen (in this case, a bird skeleton) in front of a grid that provides a consistent measurement scale. During their seven-year collaboration, Weeks and Fuhey created an AI model that can accurately measure 12 bones throughout a bird’s skeleton.
The researchers used the system to scan and measure more than 15,000 museum specimens. Most are from the UM Museum of Zoology’s collection. Each specimen can be scanned in about 45 seconds, allowing researchers to digitize entire museum collections much more efficiently than with other methods.
Reconstructing 45 million years of change
Berv also created a new statistical method called bifrost. This allowed the team to analyze complete skeletons of each species, rather than examining each bone individually. The researchers used this approach to estimate how passerine body shape has changed over about 45 million years of evolution.
“The whole organism is an integrated, complex form, and each part is interconnected with every other part of the body,” Barb said. “The question from a modeling perspective is, ‘What is the order of evolutionary changes that need to occur to explain the variation that we can see today?'”
The analysis reveals a period of particularly rapid evolution of body shape, about 35 million years ago. This explosion coincided with the Eocene-Oligocene transition, characterized by intense global cooling.
The statistical results also identified a series of evolutionary slowdowns about 15 million years ago, which coincided with another major geological event.
“Our discovery definitely changed my thinking about how the world works,” said Weeks, an associate professor of ecosystem science and management in UM’s School of Environment and Sustainability. “This pattern of rare large increases in evolutionary rate and many small decreases in evolutionary rate that we found is actually consistent with a pattern in which lineages are rapidly changing to explore new ecological space and take advantage of its opportunities.”
Climate and geography shape evolution
The researchers then tested the pattern by analyzing the global distribution of the birds in the dataset. They found that geography also helps predict the average rate of morphological evolution.
Bird assemblages at more extreme latitudes, where seasonal temperature changes are more dramatic, tend to contain species that evolve faster than those living closer to the equator. Because similar patterns have emerged over millions of years and across modern geographic regions, the findings suggest that environmental fluctuations may play an important role in driving changes in body shape.
“There appears to be an underappreciated link between latitudinal gradients and the rate of morphological evolution,” Weeks said. “We hope that our findings will inspire new integration of rates of morphological change into other areas of great interest, such as the very well-known latitudinal gradient in biodiversity.”
AI creates new value in museum collections
This study also highlights the scientific importance of museum collections. According to Weeks, artificial intelligence has made it possible to extract information from preserved specimens on a scale previously unimaginable.
“It is especially clear how important the museum investment is given the scale of this kind of analysis, which goes far beyond what could be done using specimens provided by private collectors,” he said. “It’s also fun to imagine how early collectors used the specimens they collected. They would be surprised to learn that photographs of these specimens were analyzed on a computer. This is just one example of how impossible it is to fully predict the future value of a specimen.”
Lessons for modern climate change
The researchers say the findings may also help scientists think about how species can respond to the rapid climate change currently occurring around the world.
“We’re at this moment in human history where we’re experiencing dramatic global climate change. We don’t know what’s going to happen in 10 years, let alone 10 million years,” Barb said. “To understand the long-term effects of human activity on the Earth, we need to study the relationship between events in Earth’s history and evolutionary changes.”
This research was also supported by the Michigan Institute for Social Data & AI, the Natural Sciences and Engineering Research Council of Canada, and the National Science Foundation.
