The rapid and accurate taxonomic identification of fossils is of great significance in paleontology, biostratigraphy, and other fields. However, taxonomic identification is often labor-intensive and tedious, and the requisition of extensive prior knowledge about a taxonomic group also requires long-term training. Moreover, identification results are often inconsistent across researchers and communities. Accordingly, in this study, we used deep learning to support taxonomic identification. We used web crawlers to collect the Fossil Image Dataset (FID) via the Internet, obtaining 415,339 images belonging to 50 fossil clades. Then we trained three powerful convolutional neural networks on a high-performance workstation. The Inception-ResNet-v2 architecture achieved an average accuracy of 0.90 in the test dataset when transfer learning was applied. The clades of microfossils and vertebrate fossils exhibited the highest identification accuracies of 0.95 and 0.90, respectively. In contrast, clades of sponges, bryozoans, and trace fossils with various morphologies or with few samples in the dataset exhibited a performance below 0.80. Visual explanation methods further highlighted the discrepancies among different fossil clades and suggested similarities between the identifications made by machine classifiers and taxonomists. Collecting large paleontological datasets from various sources, such as the literature, digitization of dark data, citizen-science data, and public data from the Internet may further enhance deep learning methods and their adoption. Such developments will also possibly lead to image-based systematic taxonomy to be replaced by machine-aided classification in the future. Pioneering studies can include microfossils and some invertebrate fossils. To contribute to this development, we deployed our model on a server for public access at www.ai-fossil.com.

The number of individuals of species varies, but estimating abundance, given incomplete and biased sampling in both contemporary and fossilized communities, is challenging. Here, we describe a new occupancy model in a hierarchical Bayesian framework with random effects, in which multispecies occupancy and detection are modeled as a means to estimate relative species abundance and relative population densities. The modeling framework is suited for temporal samples of fossil communities with repeated sampling including multiple species with similar preservation potential. We demonstrate our modeling framework using a fossil community of benthic organisms to estimate relative species abundance dynamics and changing relative population densities of focal species in nine (geological) time intervals over 2.3 Myr. We also explore potential explanatory factors (paleoenvironmental proxies) and temporal autocorrelation that could provide extra information on unsampled time intervals. The modeling framework is applicable across a wide range of questions on species-level dynamics in paleoecological community settings.

Conspicuous centers of biodiversity are frequently attributed to local conditions that promote speciation or resistance to extinction, but recent diversification studies indicate this mode of explanation might not be very general, so it may be fruitful to revisit the role of dispersal in concentrating biodiversity. Here we consider the processes underlying the marine diversity hotspot in the Indo-West Pacific among comatulid crinoids, suspension-feeding echinoderms conspicuous on modern tropical reefs. We used ancestral-range reconstruction on a phylogeny of extant crinoids, assembled a new occurrence database of fossil comatulids and interrogated it with probabilistic preservational models, and developed a morphological character matrix to estimate the relationships among living and fossil comatulids. Ancestral-range reconstruction on a phylogeny of extant comatulids recovers an origin outside the Indo-Pacific and elevated dispersal into it. A new occurrence database records the comatulid clade spreading out gradually from origin in the Early Jurassic of the West Tethys. Comatulids do not appear in their modern hotspot until the Oligocene, and taphonomic analyses show these results cannot be explained solely as a result of inadequate sampling in Asia and Oceania. Finally, phylogenetic analyses demonstrate that deeply nested crown-group comatulids had originated before the clade became well established in the East Tethys, implying many independent dispersals into the modern hotspot. These consilient results suggest a biodiversity hotspot that owes its existence to dispersals out of the ancient West Tethys rather than to elevated in situ diversification.

The ecogeographic rule known as Bergmann's rule suggests that there is a positive relationship between body size and latitude when comparing closely related taxa. The underlying mechanism or mechanisms to explain this pattern vary as widely as the taxa that seem to follow it, which has led to skepticism over whether Bergmann's rule should be considered a rule at all. Despite this, Bergmann's rule is widespread among modern birds, mammals, beetles, and some amphibians, but far fewer extinct taxa have been subjected to tests of Bergmann's rule. To examine whether Bergmann's rule is detected in extinct taxa, we compared body-size proxies in Lystrosaurus recovered from Early Triassic–aged strata in Antarctica, South Africa, India, and China. Our results reveal that average body size is largest at mid-northern paleolatitudes (∼45°N) instead of the highest southern paleolatitudes (∼70°S). Additionally, maximum body size is consistent across the Northern and Southern Hemispheres, indicating that Bergmann's rule did not apply for Lystrosaurus during the Early Triassic. To test potential sample size biases in our results, we used rarefaction and subsampling to show that only the Karoo Basin is well sampled and that large individuals are exceedingly rare, except in the Turpan-Junggar Basin of Xinjiang, China. Taken together, our results suggest that Lystrosaurus had the potential to reach large body sizes in each of the latitudinally widespread geologic basins studied here, but that local conditions may have allowed individuals at mid-northern paleolatitudes a greater chance of reaching a large size compared with southern congeners that suffered increased mortality when young or at a small size.

Dinosaur embryos cause a lot of excitement in the scientific literature and are often widely reported because of the general public's interest in dinosaur biology. Well-preserved, articulated oviraptorosaur embryos in eggs are usually interpreted as representing a stage of development close to hatching because of their large size and good level of skeletal ossification. Based on this evidence, a recent report suggested that the position of the one embryo's head was reminiscent of an avian-like hatching position. Here we explore how the developmental stage of well-preserved oviraptorosaur embryos can be estimated, rather than assumed. This will help in our understanding of their developmental biology and its evolutionary consequences. Using quantitative methods and comparison with modern crocodilian embryos, we show that all articulated oviraptorosaur embryos are small relative to the egg and most likely at a stage of development equivalent to around 50%–60% of the developmental period, that is, not even close to hatching. This conclusion is supported by the fact that many elements of the crocodilian skeleton are well ossified many weeks before hatching and the position of oviraptorosaur embryos' heads was also comparable to a crocodilian embryo many days before hatching. Misunderstandings about the stage of the developmental biology of these well-preserved oviraptorosaur embryos hampers our understanding of the true nature of their reproductive biology. We urge a more conservative approach to their interpretation. This is important, because misunderstandings in the minds of the public about dinosaur biology are hard to counter once poorly evidenced ideas have been reported around the world.

Understanding current and future biodiversity responses to changing climate is pivotal as anthropogenic climate change continues. This understanding is complicated by the multitude of available metrics to quantify dynamics and by biased sampling protocols. Here, we investigate the impact of sampling protocol strategies using a data-rich fossil record to calculate effective diversity using Hill numbers for the first time on Paleogene planktonic foraminifera. We sample 22,830 individual tests, in two different size classes, across a 7 Myr time slice of the middle Eocene featuring a major transient warming event, the middle Eocene climatic optimum (MECO; ∼40 Ma), at study sites in the midlatitude North Atlantic. Using generalized additive models, we investigate community responses to climatic fluctuations. After correcting for any effects of fossil fragmentation, we show a peak in generic diversity in the early and middle stages of the MECO as well as divergent trajectories between the typical size-selected community (>180 µm) and a broader assemblage, including smaller genera (>63 µm). Assemblages featuring smaller genera are more resilient to the climatic fluctuations of the MECO than those assemblages that feature only larger genera, maintaining their community structure at the reference Hill numbers for Shannon's and Simpson's indices. These results raise fundamental questions about how communities respond to climate excursions. In addition, our results emphasize the need to design studies with the aim of collecting the most inclusive data possible to allow detection of community changes and determine which species are likely to dominate future environments.

A dramatic shift from carbonate-rich to chert-rich marine strata occurred during the Permian and is frequently attributed to the increased activity of siliceous sponges and their biosiliceous sedimentation. The first-order ecologic consequences of this transition, if any, remain opaque. We analyze fossil occurrence data from the Phosphoria Basin (western North America) to test whether the presence of siliceous sponges, which are correlated with basin-wide chert strata, influenced the recruitment of benthic fauna. Using published lithologic descriptions, we categorized fossil collections by formation, facies, and lithology and used these data to code detrended correspondence analysis and nonmetric multidimensional scaling ordinations. We also analyzed the clustering of taxa into faunal units termed biofacies.

Results from these analyses indicate that fossil collections occurring in chert and carbonate are closely associated in faunal composition and community structure. These collections preferentially occur in the inner- to mid-ramp facies, in agreement with previous studies. Although largely similar in composition, collections of chert and carbonate lithology exhibit differences in the frequency and abundance of accessory brachiopod taxa (e.g., Composita and Hustedia), possibly a result of greater biosiliceous sedimentary input.

A length–frequency sample (n = 295) from a fossil population of the Ordovician trilobite Triarthrus eatoni Hall, 1838, assembled and analyzed by J. L. Cisne in 1973 is here reexamined using methods of length–frequency analysis commonly used in fishery science and marine biology. Theoretical considerations and the empirical data at hand suggest that the growth of T. eatoni was not “linear,” but asymptotic, as is the growth of most Recent marine invertebrates. The parameters of the von Bertalanffy growth function (L_∞ = 41 mm, K = 0.29 yr^–1) suggest that T. eatoni, which apparently lived in a challenging environment, grew somewhat more slowly than the extant marine isopod Ceratoserolis trilobitoides (Eights, 1833), used here as Recent analogue to T. eatoni. This trilobite probably lived up to 10 years, rather than the suggested 4 years, and its mortality rate was 15%–20% per year rather than 30%–40% per year. These represent the first estimates of trilobite absolute growth characteristics using methods known to accurately model growth in extant water-breathing ectotherms. These provide a baseline for trilobite growth that can be used to make inferences about growth in other species. The approach used here may also be applied to other trilobites for which suitable length–frequency data exist.

Ammonoid cephalopods were Earth's most abundant oceanic carnivores for hundreds of millions of years, yet their probable range of swimming capabilities is poorly constrained. We investigate potential hydrodynamic costs and advantages provided by different conch geometries using computational fluid dynamics simulations. Simulations of raw drag demonstrate expected increases with velocity and conch inflation, consistent with published experimental data. Analysis at different scales of water turbulence (via Reynolds number) reveals dynamic trade-offs between conch shape, size, and velocity. Among compressed shells, the cost of umbilical exposure makes little difference at small sizes (and/or low velocity) but is profound at large sizes (and/or high velocity). We estimate that small ammonoids could travel one to three diameters per second (i.e., a typical ammonoid with a 5-cm-diameter shell could travel 5–15 cm/s), but that large ammonoids faced greater discrepancies (a 10 cm serpenticone likely traveled <30 cm/s, while a 10 cm oxycone might achieve >40 cm/s). All of these velocities are proposed only for short bursts of jet propulsion, lasting only a few seconds, in the service of dodging a predator or conspecific rival. These analyses do not include phylogeny, taxonomy, second-order conch architecture (ribs, ornament, etc.), or hydrostatic consequences of internal anatomy (soft body, suture complexity). For specific paleoecological context, we consider how these results inform our reconstruction of Jurassic ammonite recovery from the end-Triassic mass extinction. Greater refinements will come with additional simulations that measure how added mass is influenced by individual shape-trait variations, ornament, and subtle body extensions during a single jet motion.

We examine temporal and spatial variation in morphology of the ammonoid cephalopod Discoscaphites iris using a large dataset from multiple localities in the Late Cretaceous (Maastrichtian) of the U.S. Gulf and Atlantic Coastal Plains, spanning a distance of 2000 km along the paleoshoreline. Our results suggest that the fossil record of D. iris is consistent with no within-species net accumulation of phyletic evolutionary change across morphological traits or the lifetime of this species. Correlations between some traits and paleoenvironmental conditions as well as changes in the coefficient of variation may support limited population-scale ecophenotypic plasticity; however, where stratigraphic data are available, no directional changes in morphology occur before the Cretaceous/Paleogene (K/Pg) boundary. This is consistent with models of “dynamic” evolutionary stasis. Combined with knowledge of life-history traits and paleoecology of scaphitid ammonoids, specifically a short planktonic phase after hatching followed by transition to a nektobenthic adult stage, these data suggest that scaphitids had significant potential for rapid morphological change in conjunction with limited dispersal capacity. It is therefore likely that evolutionary mode in the Scaphitidae (and potentially across the broader ammonoid clade) follows a model of cladogenesis wherein a dynamic morphological stasis is periodically interrupted by more substantial evolutionary change at speciation events. Finally, the lack of temporal changes in our data suggest that global environmental changes had a limited effect on the morphology of ammonoid faunas during the latest Cretaceous.

This study compares the δ¹⁵N values and the trophic position of two seabird species throughout the late Holocene in three regions in the southwestern Atlantic Ocean to assess the hypothesis that the decimation of megafauna led to changes in the trophic position of mesopredators. Modern and ancient mollusk shells were also analyzed to account for changes in the isotopic baseline through time. Results revealed that modern Magellanic penguins have higher δ¹⁵N values than their ancient conspecifics in the three regions, after controlling for changes in the isotopic baseline. This was also true for modern Imperial shags compared with ancient unidentified cormorants/shags from the two areas where ancient specimens were recovered (southern Patagonia and the Beagle Channel). Such temporal variability might be caused by three non–mutually exclusive processes: decreased availability of pelagic squat lobster resulting from decreasing primary productivity through the late Holocene, increased availability of small fishes resulting from the sequential depletion of other piscivores (South American fur seal and sea lion and Argentine hake) since the late eighteenth century, and modification of the migratory patterns of Magellanic penguins. Although disentangling the relative contribution of all those processes is impossible at this time, the results reported here demonstrate that the ecology of Magellanic penguins and Imperial shags has undergone major changes since the late Holocene.