Machine learning identified three main drivers of the great Permian extinction

(ORDO NEWS) — The Earth experienced the most severe extinction in its history 252 million years ago. In a new article, paleontologists share how they used machine learning to identify features that were critical to the survival of organisms during that difficult time.

During its long history, the Earth has experienced several major episodes of mass extinction of species. The most massive was the great Permian extinction, which occurred on the border of the Paleozoic and Mesozoic eras, about 252 million years ago.

Then about 75% of the species of terrestrial creatures and 90% of the inhabitants of the ocean disappeared. The most likely cause is considered to be the eruptions of huge volcanoes in Siberia, which emitted a lot of greenhouse gases into the atmosphere. As a result, the Earth’s climate began to change rapidly: the average temperature increased by as much as ten degrees.

The first author of a new paper in Paleobiology , William J. Foster , believes that similar climate trends are observed on Earth today: “Of course, our results for the Permian period cannot be directly applied to modern climate changes. The two climate systems are too different.”

The researchers examined over 25 thousand samples belonging to 1283 different genera of living organisms. Among them were bivalves and gastropods, crustaceans, sponges and algae, which were found in southern China. Notably, all of these species had a mineralized skeleton or shell. Paleontologists have also used data that describe the lifestyle of these creatures.

Each of the genera was characterized by 12 parameters, which made it possible to identify among them the most significant for survival. Scientists used machine learning to analyze all these features simultaneously. So it was possible to establish which of them most strongly distinguish the two lists of species – surviving and extinct.

What makes this work valuable is the use of specific machine learning techniques. As a rule, such an analysis of the data leads to a result that is difficult to interpret. In this case, machine learning is called a “black box”.

The flow of work is as follows: there is data at the input, they fall into the notorious “black box”, in which something mysterious happens, and, finally, machine learning produces a result. What exactly guided the algorithms, we do not know in this case.

However, the team led by Dr. Foster took a different approach based on game theory. In this case, the “box” of machine learning turns out to be not so “black” and we can understand which features of organisms influenced the decision more.

Foster describes his results as follows: “Some animals lived in deep layers of water. In this case, machine learning indicates that the risk was due to an increasing lack of oxygen.

The animals that lived near the surface, in turn, faced an increase in water temperature. In addition, if only some habitats are suitable for you and they become unsuitable, you will not be able to move somewhere else.

Thus, the researchers established the main vulnerabilities of the inhabitants of the Permian period, which caused their death.

The new result also confirms that falling oxygen levels, rising ocean temperatures, and acidification were indeed the main drivers of this extinction. It should be borne in mind that they can play an important role in the event of new mass extinctions.

Online:

Contact us: [email protected]

Our Standards, Terms of Use: Standard Terms And Conditions.