by Emma Dunne*1
Life on Earth is incredibly diverse. More than 1.7 million species have already been described and estimates suggest that there could be as many as 9 million in total. But exactly how this rich biodiversity has developed over the last 542 million years since the Cambrian remains the subject of debate amongst palaeontologists. Did biodiversity increase steadily from one geological period to the next, or did it wax and wane without any overall direction? These questions are crucial in a modern context: today, we are flooded with urgent reports on the state of biodiversity worldwide, with many scientists stating that we are in the middle of a biodiversity crisis driven by human impact, leading to what is being called the sixth mass extinction. To understand and appreciate what’s happening to Earth’s biodiversity today, we need to understand the dynamics that have led to the current levels of biodiversity.
Palaeontology, up until the middle of the last century, mostly involved descriptive tasks, such as classifying species or using fossils to date rock layers (biostratigraphy). But, as computers revolutionized many other aspects of the late twentieth century, they also transformed palaeontological research, allowing the emergence of a data-driven approach to investigating diversity through time. This new branch of palaeontology permitted in-depth investigations of the timing and extent of extinction events, the rate at which species appeared and how environmental conditions affected diversity, on both regional and global scales. Here, I highlight some key milestones in the history of palaeodiversity research, and describe recent advances in this area, including analytical approaches and databases.
The first palaeodiversity studies:
Technology may have permitted the field of modern palaeobiology to flourish, but some scientists were exploring biodiversity through geological time long before the digital era. In 1848–49, Heinrich Georg Bronn, a German palaeontologist, published a three-volume inventory of all known fossils that he had painstakingly collated by hand. From his tables, Bronn was able to identify a pattern of what he called dynamic succession: varying species appeared and disappeared from the fossil record at different intervals. To represent his data graphically, Bronn used what we now refer to as spindle diagrams (Fig. 1A). Later popularised by influential vertebrate palaeontologist Alfred Romer, these diagrams depict changes in the diversity of a taxonomic group over time by varying the thickness of the line. Despite Bronn’s monumental efforts in documenting the fossil record, this analytical approach to palaeontology failed to catch on.
Later in the nineteenth century, British geologist John Phillips, a friend and colleague of the famous geologist William Smith, used the British fossil record to produce the first estimate of diversity across the Phanerozoic. This ‘diversity curve’ (Fig. 1B), showing the gradual rise and fall of diversity across the eon, was constructed by counting the number of taxa (groups of related organisms) in each geological period, known as intervals. This approach might seem intuitive, but, as outlined later, it has several serious shortcomings. Yet it was more than a century before the matter of measuring diversity through time using fossil data was revisited, using a much more quantitative approach. This time, it finally received significant interest from the palaeontological community.
The emergence of quantitative palaeobiology:
During the 1970s, while many of his contemporaries were focused on excavating and describing fossils, Jack Sepkoski at the University of Rochester, New York, and later the University of Chicago, Illinois, was constructing a computer database of fossil marine invertebrate occurrence from the published literature. This simple digital database recorded the dates at which each group of marine invertebrates first appeared and disappeared from the fossil record. From this data, Sepkoski could plot the number of taxonomic groups in each interval of the Phanerozoic (the last 542 million years) and see how the patterns changed over time. The resulting ‘Sepkoski curve’ (Fig. 2), published in 1981, revealed a number of interesting patterns in the history of marine invertebrate diversity, including the Cambrian explosion (the sudden appearance of complex animals in the fossil record 542 million years ago), the Ordovician radiation (in which the animals of the Cambrian explosion were replaced by a new type of fauna, 485–444 million years ago), the Palaeozoic plateau (in which diversity remained constant from the end of the Ordovician period to the start of the Permian period (444–299 million years ago), and the Meso–Cenozoic diversification (an increase in diversity from the Triassic period, starting 251 million years ago, to the present day). Sepkoski’s work, like that of Bronn and Phillips, was first met with scepticism, particularly from evolutionary biologists — how could counting fossils be valuable for studies of evolution?
Nevertheless, Sepkoski’s most widely known contribution to the field of palaeodiversity came in 1982 in collaboration with David Raup, a palaeobiologist also based at the University of Chicago. Their study, which used Sepkoski’s database, identified five devastating extinction events in the marine fossil record, known as the ‘Big Five’ mass extinctions, including the event at the end of the Cretaceous period (66 million years ago) that wiped out the non-avian dinosaurs, and the most severe of all extinction events at the end of the Permian period (252 million years ago), when estimates show that over 90% of marine life perished (Fig. 3). This work set the scene for a plethora of studies on diversity and extinction in the history of life on Earth.
Following on from Sepkoski and Raup’s trailblazing research, other palaeontologists began to use this approach to investigate the palaeodiversity of other fossil organisms. In the late 1980s and early 1990s, studies began to emerge that focused on insects, plants and terrestrial and marine vertebrates. Sepkoski, this time in collaboration with Conrad Labandeira at the National Museum of Natural History in Washington DC, also examined insect palaeodiversity and concluded that the high diversity seen through time was the result of low extinction rates, and that the greatest radiation of insects predated the emergence of flowering plants, contrary to the assumption that the evolution of the two groups was tightly linked.
In the early 1990s, Michael Benton at the University of Bristol, UK, led the publication of The Fossil Record, a collection of the occurrences of fossil algae, fungi, protists, plants and animals, compiled by 90 experts on these groups. This dataset, and subsequent iterations of it, paved the way for quantitative palaeobiology to firmly enter the terrestrial realm; before then, most studies examining extinction and diversification through time had focused on invertebrates, particularly marine species. Benton’s work led to the idea that terrestrial diversity has increased steadily through time, punctuated only by the mass extinctions first documented by Raup and Sepkoski.
These early estimates of global palaeodiversity involved simply counting the number of taxonomic groups in each geological time interval. However, estimating true patterns of diversity through geological time is a tricky business, owing to the nature of the fossil record.
Biases in the fossil record:
The fossil record is notoriously incomplete and unevenly sampled. Not all organisms have the same potential of becoming a fossil, owing to many factors such as the environment they lived in or the structure of their bodies (remains made of organic minerals, such as shells or bones, are more likely to be preserved than soft tissues). However, even after they have formed, fossils are subject to further biases that act to remove them from the fossil record, through geological processes such as diagenesis (the formation and transformation of sedimentary rock) and erosion. Finally, fossils that make it through these relentless processes will need to be uncovered, correctly identified, and entered into the published literature before they are included in analyses of palaeodiversity. Researchers, when choosing areas to collect fossils from and what types of fossils to sample, introduce significant biases. It is clear from the available data that there are geological time intervals and regions of the world that are favoured more than others. For example, continents such as North America and Europe are particularly well sampled as they are generally more accessible, and there is no denying the popularity of dinosaurs and their relatives ever since the Bone Wars of the late 1800s.
Biases in the fossil record were acknowledged even as far back as Charles Darwin in On the Origin of Species (1859), when he noted that “our palaeontological collections are very imperfect”. Raup outlined several reasons why the decline seen in marine diversity during the Palaeozoic is not a real signal, but actually documents the ever-worsening quality of sampling as we move back in time. Recent studies of palaeodiversity have noted a correlation between the amount of sampling (rock area investigated, or a similar proxy for sampling effort) and the amount of diversity observed. Three explanations for this pattern have been put forward:
- Sampling is the primary driver of observed diversity — the more rock area sampled, the higher the diversity. This is sometimes referred to as the ‘bias hypothesis’.
- The common-cause hypothesis: both sampling and diversity are driven by some common factor, such as fluctuations in sea level or tectonic activity.
- The redundancy hypothesis: sampling and diversity are entirely or partially redundant with each other.
Mitigating the effects of biases:
It is difficult to say decisively which of the above three explanations is correct, and debate continues about how much sampling biases affect our ability to estimate genuine patterns of palaeodiversity from the fossil record. However, several analytical methods have been developed to estimate diversity patterns from an incomplete fossil record, many pioneered in only the past decade. These can be loosely arranged into three approaches: phylogenetic, residual and subsampling approaches. Phylogenetic approaches, more usually used in studies of modern or genetic diversity, attempt to fill in the gaps of a taxonomic group’s evolutionary history, known as ghost lineages, using evolutionary trees. The residual-diversity method is a modelling approach that attempts to remove the signal, or ‘effect’, of a chosen proxy for sampling, such as the amount of rock sampled. This method has been used widely in recent diversity analyses, including studies of dinosaur diversity, but it has come under scrutiny for the way in which the results from the model are generated.
The most widely used of the approaches are subsampling approaches, which, in the broadest sense, standardize unequally sized samples to allow us to compare diversity directly between assemblages (groups of fossils found together in the same time and place) that have been sampled with differing intensity. An early way to do this was to simply standardise samples by size (drawing down samples to equal numbers of specimens, individuals or localities) using a method known as rarefaction. However, methods such as rarefaction under-sample more-diverse assemblages and flatten diversity curves, which can lead to inaccurate estimates. In 2010, John Alroy at Macquarie University in Sydney, came up with a solution to this problem by developing a method that standardizes samples to equal levels of ‘completeness’ (how well a geological time interval or locality is sampled). This is known as shareholder quorum subsampling (SQS), and has become widely used by palaeobiologists examining the diversity of a variety of fossil organisms.
Since the development of SQS, other rigorous methods of subsampling have been proposed. One is TRiPS (true richness estimation using Poisson sampling), which generated a lot of attention by posing, and attempting to answer, the question of ‘how many dinosaur species were there?’ However, this method has also been subject to criticism due to the revalation that it can often track un-standardized diversity curves. Despite this, the authors of TRiPS have paved the way in data visualization by developing an online app that is plugged directly into the Paleobiology Database (discussed below), showing one way that palaeontology has firmly embraced the technological era.
Fossil occurrence databases have come a long way since the tables of Bronn and Phillips, and with the help of computers have developed even more rapidly since the 1990s. Today, the most widely known and frequently used database for palaeodiversity studies is the Paleobiology Database, which is this year celebrating its twentieth birthday. Since Alroy first set it up in 1998, the Paleobiology Database has grown immensely. It now contains more than 1.37 million occurrences of fossil plants, invertebrates and vertebrates, which have been collected from the published literature by over 400 palaeontologists across the world. The data collected include information on where a fossil was found, what age it is, what geological setting it was found in, and what taxonomic group it belongs to. The contents of the database are available to view online on an interactive map through the Navigator webpage (Fig. 5). The Neotoma Paleoecology Database, which focuses on the data belonging to selected vertebrate groups, goes one step further: alongside occurrence data, it incorporates data on fossil pollen, insects and charcoal, which allows studies not only of diversity, but also of how diversity correlates with environmental changes.
Current work and future directions:
With the increased availability of comprehensive databases and sophisticated quantitative approaches for estimating patterns of palaeodiversity in light of sampling biases, many of the hypotheses and common assumptions on the diversity of life through time are now being revisited.
One case that is being reassessed is the supposed steady increase in diversity on land from the Carboniferous period (starting 359 million years ago), when four-legged animals (tetrapods) first emerged on land, to the present day. Recent work, which uses almost 30,000 fossil occurrences from the Paleobiology Database and robust subsampling methods, found that during the Mesozoic era (252 million to 66 million years ago) tetrapod diversity remained relatively stable — that is, net diversification remained near-zero. During the Cenozoic era (the last 66 million years) the increase in diversity is thought to have been exponential, but this is also being called into question: is it a true signal, or is it an artefact of better sampling in more recent geological time intervals (the ‘pull of the recent’)?
Large databases and new fossil discoveries have allowed researchers to identify smaller-scale changes in diversity, even much further back in the fossil record. A recent paper by me and my colleagues focused on the first 100 million years of tetrapod diversity (Carboniferous to early Permian); again, using data from the Paleobiology Database and subsampling approaches, we were able to show how the collapse of the equatorial rainforests at the end of the Carboniferous (307 million years ago) caused early tetrapod diversity to decrease notably. We were also able to integrate the occurrence data with phylogenetic information, to further examine how this environmental change affected the global distribution of tetrapods. This study joins a growing body of work that is using fossil-occurrence data alongside phylogenetic and environmental information to further explore extinction and diversification through time, and the drivers behind these processes.
Even more exciting technological developments are on the horizon. A few years ago, researchers at the University of Wisconsin–Madison announced a project called PaleoDeepDive, a statistical machine-learning system that will automatically locate and extract fossil occurrence data from published scientific papers, thus removing the need for a researcher to individually examine a paper to extract data to input into the Paleobiology Database. Advances such as these have allowed the field of palaeontology to experience a sort of renaissance, akin to that which happened to genetics as genetic sequencing became faster and cheaper.
Palaeobiology, and quantitative palaeobiology in particular, remains a relatively young field. The speed at which new fossils are being discovered and new methods are being developed has led to an explosion of exciting and important research on the history of life on Earth. It is difficult to say how essential studies of palaeodiversity will be in aiding the current biodiversity crisis, or sixth mass-extinction event, but understanding the mechanisms behind and severity of past extinction and diversification events can certainly help us to manage our expectations.
Suggestions for further reading:
Benson, R. B. J., Butler, R. J., Alroy, J., Mannion, P. D., Carrano, M. T., & Lloyd, G. T. Near-Stasis in the Long-Term Diversification of Mesozoic Tetrapods. PLoS Biology 14(1): e1002359 (2016). (DOI: 10.1371/journal.pbio.1002359)
Benton, M. J. Diversification and extinction in the history of life. Science 268 52–58 (1983). (DOI: 10.1126/science.7701342)
Dunne, E. M., Close, R. A., Button, D. J., Brockelhurst, N., Cashmore, D. D., Lloyd, G. T., & Butler, R. J. Diversity change during the rise of tetrapods and the impact of the ‘Carboniferous Rainforest Collapse’. Proceedings of the Royal Scoiety: B. 285: 20172730 (2018). (DOI: 10.1098/rspb.2017.2730)
Raup, D. M. & Sepkoski J. J. Mass extinctions in the marine fossil record. Science 215: 1501–1502 (1982). (DOI: 10.1126/science.215.4539.1501)
Sepkoski, D. Rereading the Fossil Record: The Growth of Paleobiology as an Evolutionary Discipline. University of Chicago Press, Chicago, IL, USA. (2012) (ISBN: 978-0226272948)
Smith, A. B. & McGowan, A. J. (eds) Comparing the Geological and Fossil Records: Implications for Biodiversity Studies. Geological Society, London, Special Publications, 358, 1–7 (2011). (DOI: 10.1144/SP358.1)
1School of Geography, Earth & Environmental Sciences, University of Birmingham, UK.