杏吧原创

Waves of animals died at an ancient Spanish lake and now we know why

Fossil forensics and artificial intelligence have shed light on how multiple groups of large mammals died by a lake in what is now Spain 9 million years ago
fossil skeleton
Fossilised skeleton of Decennatherium rex, an ancient ancestor of giraffes
M. Mart铆n-Perea et al, Palaeogeogr. Palaeoclimatol. Palaeoecol. (2021)

A bunch of large, now-extinct mammals 鈥 including European ancestors of giraffes, primitive horses and sabre-toothed deer 鈥 died 9 million years ago at a watering hole in Spain. Now, artificial intelligence and painstaking fossil analyses have helped solve the mystery of what happened to them.

鈥淚 like to jokingly say we are like crime scene investigators: When, how and why did it happen?鈥 says at the National Museum of Natural Sciences in Madrid. 鈥淭he only difference is our 鈥榗rime scenes鈥櫬燼re millions of years old.鈥

Palaeontologists discovered the fossil site, dubbed Batallones-10, in 2007 near Madrid. It took 14 dig sessions to unearth all of the bones buried between 1 and 5.5 metres deep in an area around a quarter the size of a tennis court.

So far, the digs have revealed the presence of 15 large mammal species comprising 68 individual animals. Most were plant-eaters. One of the most abundant and well-preserved species was an early European ancestor of giraffes called Decennatherium rex. It was first described after a (see image above).

Other animals included now-extinct hipparionine horses, two of which were pregnant, a mastodon, and deer that were relatives of modern-day musk deer, . Rhinos, giant tortoises, a monitor lizard, sabre-toothed cats, frogs, birds and small mammals were also dug up.

Researchers originally suspected the animals died over an extended period from natural causes at a watering hole. Over time, it gradually filled up with mud and dead animals, fossilising their remains.

However, in 2020 Mart铆n-Perea and his colleagues used machine learning to seek patterns in the fossils鈥 precise location coordinates recorded during digs. three distinct fossil-forming layers too subtle for humans to discern.

Mart铆n-Perea and his colleagues have now used the same artificial intelligence technique to assign all 7968 fossil items to a particular layer. By studying how the fossils formed 鈥 a field called taphonomy 鈥 and comparing the different levels, a different picture started to emerge.

鈥淭he taphonomic evidence seemed to come together perfectly, as if putting together a jigsaw puzzle,鈥 says Mart铆n-Perea. There were resemblances in the way the bones accumulated in each layer, hinting that three similar but separate and relatively sudden events caused the larger animals鈥 demise.

What鈥檚 more, the fossils had no marks caused by carnivores or trampling and they kept the shape of near-complete skeletons due to the animals鈥 tissues drying out quickly. Also the bones weren鈥檛 weathered, suggesting they were buried rapidly.

The researchers now think that animals lingered close to the watering hole during three or more episodes of drought leading them to overgraze the surrounding area. But the pond never actually dried up 鈥 the team found no geological evidence of mud-cracks usually seen on dry lake beds 鈥 so the animals didn鈥檛 die of thirst. Instead, weaker, younger and pregnant animals probably died of starvation or became trapped in the mud of the shrinking pool, much like what can happen nowadays in the African savannah.

The team thinks the lack of vegetation and returning rainfall after each drought caused flash floods that quickly filled the carcass-containing cavity with mud, explaining how the bones were so well preserved.

鈥淭his is an impressively thorough study and the hypothesis is well supported by the evidence,鈥 says , a pioneer of taphonomy at the Smithsonian Institution in Washington DC. 鈥淭he use of AI represents an important advance in analysis of excavation data.鈥

Palaeogeography, Palaeoclimatology, Palaeoecology

Topics: Artificial intelligence / fossils