By Michael Le Page
Fossils of trilobites, one of the species diminished but not lost during the Devonian period
Kevin Schafer/Alamy Stock Photo
The best record yet of how biodiversity changed in the distant past has been created with the help of machine learning and a supercomputer. Among other things, it confirms that one of the five great mass extinctions didn’t really happen.
It was thought the oceans turned toxic around 375 million years ago, near the end of the Devonian period, wiping out many marine species including almost all trilobites. But the latest study shows no evidence of a sudden catastrophic change like the asteroid impact that wiped out the dinosaurs. Instead, there was a gradual decline over an immensely long time around 50 million years.
The late Devonian mass extinction isnt there, says Doug Erwin at the National Museum of Natural History in Washington DC. Theres a long decrease in diversity during the Devonian, as some people have suggested previously.
Fossils are used to date rocks. Because most species are only around for a few million years, if fossils of one species are present in rocks from different places, those rocks must be roughly the same age.
Roughly really does mean roughly, though. Previous studies of how biodiversity has changed over time have only been able to divide the past into huge chunks around ten million years long.
Now Shuzhong Shen at the Nanjing Institute of Geology and Palaeontology in China and his colleagues including Erwin have produced a dramatically improved record in which each chunk is just 26,000 years long. They did this by taking a statistical approach developed around a decade ago and using it to analyse 100,000 records of 11,000 marine species whose fossils have been found in China and Europe.
This approach is so computationally intense it would take dozens of years to do this on a normal computer. Instead, the team developed special machine-learning procedures and ran them on the Tianhe-2 supercomputer.
The record covers 300 million years overall, from the start of the Cambrian period 540 million years ago until just after the start of the Triassic period 240 million years ago.
The improved resolution is the equivalent of going from considering all people who lived in the same century to be contemporaries to considering only people who lived during the same six-month period to be such.
The mid-late Devonian diversity decrease is still very clear, but it is spread through the whole time and not concentrated in a single mass extinction, says palaeontologist Richard Bambach, now retired, who argued in a 2004 paper that there was no late Devonian mass extinction. That extends the conclusion I made.
The idea that there were five great mass extinctions when most plant and animal species on Earth died out was first proposed in a 1982 paper. Later studies have suggested there were anywhere between three and 20.
There is no formal definition of a mass extinction, so there is plenty of scope for debate. However, most biologists would agree that they involve a big increase in species extinctions over a relatively short time. At the end of the Permian period about 250 million years ago, for instance, most species died out in just 63,000 years, this new analysis shows.
In 2004, Bambach also suggested there was no mass extinction at the end of the Triassic, but better evidence for it has since emerged, both he and Erwin say.
The only issue is about the Devonian, so there would be four rather than five, says Erwin.
Journal reference: Science, DOI: 10.1126/science.aax4953
More on these topics:
By Michael Le Page