An international team of scientists from the United States, Denmark, the United Kingdom and Spain has employed state-of-the-art machine learning techniques to provide the most complex picture yet of the reverberating consequences of land mammal declines on food webs over the past 130,000 years.
“While about six percent of land mammals have gone extinct in that time, we estimate that more than 50 percent of mammal food web links have disappeared,” said study lead author Evan Fricke, who conducted the study during a faculty fellowship at Rice University, and is now a research scientist at the Massachusetts Institute of Technology (MIT). “And the mammals most likely to decline, both in the past and now, are key for mammal food web complexity.”
Complex food webs – containing all the links between predator and prey in a specific geographical area – are crucial for regulating populations in ways that allow more species to coexist, thus supporting biodiversity and stability. However, declines in animal populations has a strong impact on food webs, undermining ecosystem resilience. Although land mammal declines have been thoroughly studied in the context of the current biodiversity crisis, it is less clear how those losses have degraded the world’s food webs.
By using data on current predator-prey interactions, the scientists trained a machine learning algorithm to recognize how the traits of species influenced that one species would prey upon another. “This approach can tell us who eats whom today with 90 percent accuracy,” said study senior author Lydia Beaudrot, an assistant professor of Biosciences at Rice. “That is better than previous approaches have been able to do, and it enabled us to model predator-prey interactions for extinct species.”
This method offers a comprehensive global view into the food webs that linked ice age mammals, as well as on what food webs would look today if creatures such as saber-tooth cats, marsupial lions, giant ground sloths, or wooly rhinos still coexisted with surviving mammals. “Although fossils can tell us where and when certain species lived, this modeling gives us a richer picture of how those species interacted with each other,” Professor Beaudrot added.
By charting food webs changes over a long period of time, the analysis showed that food webs worldwide are collapsing due to declines in animal populations. “The modeling showed that land mammal food webs have degraded much more than would be expected if random species had gone extinct,” Dr. Fricke said. “Rather than resilience under extinction pressure, these results show a slow-motion food web collapse caused by selective loss of species with central food web roles.”
However, since many declines stem from contractions in the geographical ranges of various species, restoring those species to their historic ranges has a significant potential to reverse the declines.
“When an animal disappears from an ecosystem, its loss reverberates across the web of connections that link all species in that ecosystem. Our work presents new tools for measuring what’s been lost, what more we stand to lose if endangered species go extinct and the ecological complexity we can restore through species recovery,” Dr. Fricke concluded.
The study is published in the journal Science.