Artificial Intelligence Decodes Causes of Mass Extinction in the Permian

Photo: W. Foster We can even grasp how animals that became extinct millions of y
Photo: W. Foster We can even grasp how animals that became extinct millions of years ago lived.

Volcanic eruptions in Siberia caused massive climate change 252 million years ago. Approximately 75 percent of all land organisms and 90 percent of all ocean organisms perished. The paleontologist Dr. William Foster at the Center for Earth System Research and Sustainability (CEN) has now decoded the causes of this mass extinction in the oceans. To do so, he relied on a new form of machine learning. The findings have now been published in the journal Paleobiology.

A series of volcano eruptions in Siberia resulted in the largest mass extinction in history and massive greenhouse gas emissions. Temperatures rose 10 degrees over thousands of years. By analyzing the lives of extinct ocean organisms, Dr. William Foster and his team were able to trace back this mass extinction to 3 changes: oxygen loss in ocean water, increasing ocean temperatures, and most likely the acidification of the oceans.

These changes resemble current developments. -The findings from the Permian, however, cannot be transferred to the present one-to-one. Earth’s climate systems then and now are too different for that,- says Foster. -But we can show for the first time which characteristics were crucial to the extinction of certain organisms. This could provide clues about which animal groups are at risk in the future.-

The research team studied 1,283 ocean species whose fossilization could be precisely dated. To do this, they used a database with information about these organisms- living habits. They analyzed 12 aspects for each species. Was there a specific characteristic that made it more likely to survive the end of the Permian-or not?

With the help of machine learning, a type of artificial intelligence, all of these factors could be studied at the same time and brought together. -Using the previous applications of machine learning, we could not have said why an algorithm decides such and such,- according to Foster. However, a new method has now enabled Foster to do just this. -Some animals lived at greater depths. The machine revealed that the incipient loss of oxygen became risky. Animals that lived closer to the surface were more likely to suffer from the higher temperatures. Organisms that live in limited habits cannot really go anywhere when things get dicey.-

The findings thus show which characteristics of any given organism might prove deadly. There were 4 characteristics of particular significance: dispersal area in the water, the mineralization of shells, species diversity within the genus, and sensitivity to acidification. This means, conversely, that an animal genus with a large dispersal area and many species that could tolerate the decreasing pH of the water had the best chances of survival 252 million years ago. And this may also be true for the future.

Original publication: Foster WJ, Ayzel G, Münchmeyer J, Rettelbach T, Kitzmann N, Isson TT, Mutti M, Aberhan M (2022): Machine learning identifies ecological selectivity patterns across the end-Permian mass extinction; Paleobiology; DOI: 10.1017/ pab.2022.1