news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Whether in the East or the West, "fortune-telling" sounds like a superstitious and mysterious thing, but when this word appears in the world of scientists, it means "inferring the evolution of life forms through inductive analysis of known information." Scientists can draw scientific conclusions by "fortune-telling" - for different animals, their different appearances may indeed affect their fate in extinction.

In June 2024, Song Haijun, a professor at the School of Earth Sciences of China University of Geosciences (Wuhan), led a team to conduct a study on "fortune telling" of fossils.

They used deep learning technology and automated methods, which is what we call AI, to study the evolution of biological morphology during the largest extinction event in history, the Permian-Triassic extinction event, and revealed how the different "appearances" of marine organisms determined their fate in this "devastating" mass extinction event.

Which was more likely to survive a mass extinction, the giant dinosaurs or the tiny Mesozoic mammals? This question may be easy, but would the conclusion be the same for other creatures, or other mass extinctions?

(Image source: Wikipedia)

Does survival or destruction have anything to do with appearance?

There have been five mass extinction events in Earth's history, the most famous of which is probably the asteroid impact at the end of the Cretaceous period, which was probably the cause of the extinction of all the tall and mighty dinosaurs at the time. However, the short mammals survived, and one of the species eventually evolved into us humans.

In fact, this fact shows that in the extinction event at the end of the Cretaceous period, extinction or not was strongly related to "appearance", that is, the shape of the animal - larger animals require more food and are more likely to die of starvation during the extinction event.

However, for the worst mass extinction event in history, the end-Permian extinction event about 252 million years ago, the correlation between animal morphology and extinction is less clear. This mass extinction, known as the "mother of all mass extinctions," wiped out 96% of marine life, including trilobites and horseshoe crabs.

The mass extinction at the end of the Permian period was the most severe mass extinction in history, and the famous trilobites disappeared in this extinction.

(Image source: Wikipedia)

The extinction event lasted a long time and occurred in two stages: a gradual period lasting about millions of years and a peak period lasting 1 million years. Some animals died in the gradual period, but more died in the peak period, such as the mass extinction of small crustaceans and ostracods (Ostracoda) and large, sessile filter-feeding brachiopods (Brachiopoda), which differed by 720,000 to 1.22 million years.

Since the causes and time of extinction of different species of animals are inconsistent, and there are so many extinct animals, almost all animals of all shapes and forms have become extinct, and only a few species have successfully survived the crisis, it is impossible to simply infer the correlation between shape and extinction. Therefore, in past studies, scientists have not had a definite answer to whether this extinction event selected the shape of animals.

How does AI “tell fortunes”?

In addition to the complexity of the extinction event itself, technological limitations also limit scientists' research on the end-Permian extinction.

In the past, studying the relationship between extinction and morphology required scientists to manually analyze the fossil morphology. They had to compare each fossil or fossil image, and classify the ancient creatures with the same morphology before and after the extinction event (such as pointed, thorny, smooth, with thin and flat shells, and wide and round shells) separately to observe whether the proportion of animals with the same morphology changed before and after the extinction event.

The results of this type of "traditional research" are greatly influenced by the research objects chosen by scientists and the research methods adopted.

For example, studies using traditional morphological description methods have shown that the morphological differences of ammonites (a distant ancient relative of the nautilus) barely decreased during the extinction event, indicating that the extinction event was not morphologically selective; on the contrary, the use of other research methods, such as comprehensive discrete feature analysis (based on the maximum and minimum values ​​of the morphological variation range, the sum of the data variance and the median of the data), showed that the morphological diversity of ammonites decreased significantly during the extinction event, supporting the conclusion that the extinction event was morphologically selective.

If we want to draw more accurate conclusions, we need to have a sufficient sample size and use a more accurate method to analyze. In this type of big data analysis, the emerging AI technology will undoubtedly have great potential.

To achieve this goal, Professor Song Haijun's team developed an analysis process called DeepMorph, which combines deep learning technology for extracting features from images with geometric morphometric methods to automatically analyze the outlines of fossil specimens, effectively capture the morphology of fossils, and simplify them into two-dimensional plane graphics, thereby clearly distinguishing various morphological types, and then repeating this process through multiple sampling.

To this end, Professor Song Haijun's team compiled a comprehensive database containing images of fossil specimens of six widely documented marine ancient creatures from the end-Permian mass extinction event, including the ancient biological relatives of the Nautilus, the ammonites, the bivalve, filter-feeding brachiopods, the "shrimp dumpling" ostracods with meat wrapped in two shells, the bivalves (clams) and gastropods (snails), and the vertebrate conodonts with sharp teeth.

This database includes 599 genera represented by 656 images before and after the extinction event, spanning the Changxing Period in the late Permian period to the Indian Period in the early Triassic period, from 254.14 million years ago to 250.7 million years ago, providing strong big data support for AI's deep learning.

a: The working principle of DeepMorph is to convert the type specimen images collected from publications into binary format through the U2-Net model segmentation, and then extract the fossil outline and morphological characteristics and include them in the database. b: Convert the morphology into multivariate normal distribution data; c: Use multivariate normal distribution data to simulate selective extinction and finally generate the extinction pattern map of Selective pattern.

(Image source: Reference 1)

Is the relationship between appearance and fate the same for different groups of animals?

DeepMorph's analysis of the data is similar to discrete feature analysis, using the sum of range (SOR, the entire range occupied by the data, determined by the most special morphology; for example, the smoothest shell is 0, the roughest is 10, and the range is 0-10), the sum of variance (SOV, the sum of the variance of each data and the mean, indicating the diversity of the data) and the position of the center of mass (POC, the median of the data) analysis as means to infer the selectivity of the extinction event for morphology.

The study found that the relationship between "appearance" and fate is not the same for different groups of animals. During the mass extinction, most of the extinctions were large animals with complex or strong shell patterns (such as spines, ribs, and tumors), while conodonts did not show signs of morphological selective extinction.

Before and after the extinction event, the ammonites that became extinct were mainly those with complex and highly decorative structures on their shells. This is reflected in the data as more extinctions on one side of the midpoint, which is called asymmetric selective extinction.

The Ceratitida and Prolecanitida, which had flat, smooth shells and less decorative features, survived the mass extinction and rapidly evolved into many new types. However, the morphology of the new types also generally maintained a smooth appearance, indicating that the appearance of ammonites was strongly correlated with whether they became extinct.

The morphological distribution ranges (total ranges) of various animals in the Changxing Stage of the Late Permian (orange), the transitional layer (grey), and the Indian Stage of the Early Triassic (blue). Ammonites (a), brachiopods (b), ostracods (c), bivalves (d), gastropods (e), and conodonts (f) show different extinction patterns.

(Image source: Reference 1)

The data of brachiopods have dropped significantly, and the richness of the genus level has dropped by 96.65%, indicating that most brachiopods have become extinct during this period. The reason why they were severely affected is mainly because their thick shells require a lot of calcium carbonate, and ocean acidification seriously hinders the formation of calcium shells, so species with complex, thick and decorative shells are almost all extinct.

Their survivors and newcomers were mostly descended from the simpler Spiriferid and Rhynchonellid clades, which had smaller bodies, simplified ornamentation, translucent shells, and reduced calcium usage, while the main extinct groups of ostracods were specialized groups with the most elongated and stubbiest shells.

These two groups show marginal selective extinction, where extinctions wipe out the most specialized groups like a gun that shoots out the first bird. In contrast to the more diverse forms of the Permian, the brachiopods and ostracods of the Triassic maintained roughly average forms, with the most common ones surviving.

The existing smallmouth mollusk, Terebratalia transversa, has a thin, translucent shell.

(Image source: Wikipedia)

The existing ostracods are like shrimps wrapped in two-valve shells, and their numerous shells are important fossils in the strata.

(Image source: Canada's Polar Life)

The extinction of the familiar groups, snails and bivalves - gastropods and bivalves - has no clear relationship to morphology.

Anyone who has kept or observed snails and clams should be impressed by their ability to survive turbidity, overheating, and lack of oxygen. Even without food, they can survive for a long time relying on their own reserves and the algae that grow on the walls of the aquarium. This is one of the reasons why they have survived major extinction events. All major morphological types have survived, and extinction has almost nothing to do with their morphology. It is just a matter of luck or misfortune.

Ambonychia ulrichi fossil from the Ordovician Fairview Formation in Warren County, Ohio, about 400 million years old, belongs to the Pterygia subclass and has similarities to modern scallops.

(Image source: sketchfab)

The gastropod (snail) fossils of the Paleozoic Era are very similar to today's snails.

(Image source: Reference 2)

The morphological space of another taxonomic group, the conodonts, was not significantly affected by the extinction event.

Unlike other evolutionary branches, the morphological diversity of conodonts decreased very little during the mass extinction event. On the contrary, after the first extinction pulse, their morphological space increased instead of decreasing, indicating that they still thrived during the extinction event, exploring a variety of new forms. Fish were similar, which may be related to the decrease in the number of their competitors (such as ammonites and nautiluses, which are also carnivorous).

The morphological changes of the extinct, survivors and newcomers of six branches during the Permian-Triassic mass extinction. Yellow is the newcomer, red is the extinct, and green is the survivor.

(Image source: Reference 1)

Four different selective extinction patterns, with red lines representing extinction events. a, horizontally selective extinction, such as ammonites; b, marginally selective extinction, including brachiopods and ostracods; c, non-selective extinction, including bivalves and gastropods; d, morphological extinction with negligible conodonts.

(Image source: Reference 1)

What is the significance of fortune-telling for fossils?

In history, the causes of the five mass extinction events were different, such as volcanic eruptions, climate change, planetary impacts, etc. Each extinction event had a different impact on the environment, and the organisms affected and extinct were also different.

For example, ammonites survived multiple mass extinctions by relying on their ability to tolerate low oxygen, but ultimately became extinct during the severe ocean acidification at the end of the Cretaceous period because they could not form calcium shells; conodonts were not greatly affected by the most severe mass extinction at the end of the Permian period, but failed to survive the less severe mass extinction at the end of the Triassic period.

A reconstruction of Ozarkodina, a jawless vertebrate that looked like a small fish. The tooth-like structures in their mouths became fossilized, called conodonts or conodonts. They survived the end-Permian extinction event but died out in a smaller extinction event at the end of the Triassic.

(Image source: drawn by the author)

In modern times, the impact of human activities on the Earth has caused many environmental problems, such as extreme high temperatures, acid rain, destruction of forests and habitats, biological invasions and environmental pollution, resulting in a new wave of extinction.

Since the advent of human civilization, 83% of wild animals have become extinct, and the rate of species extinction is estimated to be 100 times the average rate before the advent of humans. Under the influence of human beings on the environment, which species, which groups, and which ecosystems are more likely to become extinct?

Professor Song Haijun said that by analyzing the changes in morphological diversity in the fossil record, we can better predict and respond to the current threats to biodiversity. For example, taxa with a wide geographical distribution (such as birds) can survive occasional habitat destruction, but once the global environment changes at the same time, they cannot resist; while some taxa with strong survival ability but narrow distribution (such as cave fish and snails) may be able to resist environmental changes, but once the habitat is destroyed, they will also perish.

On January 9, 2019, the last known golden-topped agate snail, Achatinella apexfulva "George", died at the age of 14. These snails are only found in Hawaii. They were abundant in the past, but are endangered or have already become extinct due to invasive predators.

(Image source: Wikipedia)

By studying past extinct organisms, we can learn from history, reveal extinction mechanisms and predict the extinction risk of biological species, find groups with poor survival ability in the current environment, and protect them; in addition, the use of AI technology - DeepMorph automated method to analyze paleontological fossils can also serve as a starting point, providing more ideas and possibilities for future cross-disciplinary research on deep learning and geobiology.

references:

[1]Liu X, Song H, Chu D, et al. Heterogeneous selectivity and morphological evolution of marine clades during the Permian–Triassic mass extinction[J]. Nature Ecology & Evolution, 2024: 1-11.

[2] Frýda J, Nützel A, Wagner P J. Paleozoic gastropoda[J]. Phylogeny and Evolution of the Mollusca, 2008: 239-270.

[3]Ciampaglio, C. N. (2004). Measuring changes in articulate brachiopod morphology before and after the Permian mass extinction event: do developmental constraints limit morphological innovation? Evolution and Development, 6(4), 260–274.

[4]Villier, L. (2004). Morphological Disparity of Ammonoids and the Mark of Permian Mass Extinctions. Science, 306(5694), 264–266.

[5]Korn, D., Hopkins, M.J., and Walton, S.A., 2013, Extinction space—A method for the quantification and classification of changes in morphospace across extinction boundaries: Evolution , v. 67, p. 2795–2810,

[6]Peng, Y., Shi, G. R., Gao, Y., He, W., & Shen, S. (2007). How and why did the Lingulidae (Brachiopoda) not only survive the end-Permian mass extinction but also thrive in its aftermath? Palaeogeography, Palaeoclimatology, Palaeoecology, 252(1-2), 118–131.

Produced by: Science Popularization China

Author: Gu Ming Di Lian (popular science creator)

Producer: China Science Expo