UNIVERSITY PARK, Pa. – Edward Spagnuolo has always been stuck in the past.
As a child, he dreamed of dinosaurs and hunting for fossils buried in the ground. He never outgrew the phase, and he eventually chased his dream to Penn State. Now a junior majoring in geobiology, Spagnuolo has found a new avenue for pursuing his passion – fossil leaves.
“I’ve always wanted to be a paleontologist, or someone who studies extinct organisms, ever since I was 3 years old,” Spagnuolo said. “I always thought that would be dinosaurs, but at Penn State I was exposed to paleobotany and I really fell in love with it.”
Spagnuolo works with Peter Wilf, professor of geosciences in his paleobotany laboratory, studying fossil leaves to better understand how the world looked millions of years ago and how ecosystems may respond to future environmental changes.
When he lost an internship opportunity last summer due to the COVID-19 pandemic, Spagnuolo found a silver lining in Wilf’s lab, starting a project to create and curate a database of rainforest vegetation data in Southeast Asia from Myanmar and Thailand to Indonesia and Papua New Guinea.
Researchers studying rainforests can provide powerful information on what species are abundant and ecologically important through defined study plots, but the data is not always published or easily accessible, Spagnuolo said. He combed through hundreds of papers and found 300 studies with published data hidden in regional and specialized journals in various languages.
“We are combining the data into a really powerful tool that can be used by anyone to study whatever they want in Southeast Asia, which has the highest levels of deforestation and extinction risk on the planet,” said Spagnuolo, who is minoring in biology, astrobiology, history, marine science, global and international studies and wildlife fishery science. “And for us, that means looking at paleo plant lineages.”
Wilf’s team uses data from rainforests in southeast Asia to better understand ancient fossils found half a world away in Patagonia, Argentina.
Scientists have found fossil evidence that some lineages of trees in southeast Asian rainforests are the descendants of those from Patagonia that migrated across Antarctica millions of years ago when the continents were still connected and when the Earth was much warmer than today.
Much of modern Patagonia is a dry steppe, but understanding the fossil history there could help predict how the Asian rainforests will fare with continuing human encroachment and changing climate conditions, Spagnuolo said.
“A lot of these are endangered species or they are what’s called a keystone species, which are critical to their ecosystem in the modern world,” he said. “Understanding their evolutionary history is really critical to their conservation.”
Spagnuolo’s extra time last summer also allowed him to continue working on a project on the other side of Wilf’s lab, which uses an artificial intelligence computer vision system to learn differences in leaves and categorize them into large evolutionary categories, like families.
The program produces heat maps that use small red squares on the sections of the leaf the computer found most important for classification. After analyzing the published heat maps, they noticed that some of the heat maps seem to emphasize leaf veins and other features long used by scientists to identify leaves. But other features deemed important by the computer are much more difficult to interpret. Spagnuolo and the team are studying the most important squares on thousands of these heat maps to learn from the algorithm.
“It’s really weird that we don’t know what’s driving some of these heat maps, but at the same time, it’s emphasizing that there’s so much to these leaves we just don’t understand and that we need further work to parse these out,” said Spagnuolo, who is a Schreyer Scholar, Millennium Scholar and member of the Presidential Leadership Academy and the EMS Academy of Global Engagement.
While the scientists used modern leaf samples in the study, they hope the work will allow for improved identification of fossil leaves, like the ones Wilf finds at digs in Patagonia and southeast Asia, and for a better understanding of flowering plant evolution.
“The future of understanding these leaf characteristics is merging what we know from botanical studies, paleobotanical investigations and leaf architecture with this computer vision side of things,” Spagnuolo said. “It’s just been really cool to train my eye on that.”