How has our knowledge of dinosaur diversity through geologic time changed through research history?

Ornithischians

Raw ‘global’ ornithischian diversity (Fig. 5A) is constant and stable throughout publication history. The apparent magnitude of longer-term trends is obscured by the relative over-sampling of the Campanian and Maastrichtian, which are almost an order of magnitude higher than any other Jurassic or Cretaceous stage interval. Indeed, the Campanian shows no sign of slowing down in increasing diversity, and is the highest and most rapidly increasing of any time interval. In spite of this, the overall trends in raw diversity remain, with steadily increasing Middle to Late Jurassic diversity, a small earliest Cretaceous decline followed by a ‘middle’ Cretaceous peak in the Aptian, a shallow decline into the early Late Cretaceous, and an increase in the Campanian.

Raw diversity in Europe shows increasing diversity across the J/K transition before an earliest Cretaceous decline (Valanginian–Hauterivian), constant ‘middle’ Cretaceous diversity, and an increase from the Campanian to Maastrichtian (Fig. 5B). Raw African ornithischian diversity is too inconsistent to analyse any changes through geological time or publication time (Fig. 5C). Raw Asian diversity is fairly constant through the Cretaceous, until an apparent major Campanian peak and Maastrichtian decline (Fig. 5D). In North America, empirical diversity is flat and low throughout the Late Jurassic and most of the Cretaceous (Fig. 5E). There is a Campanian peak, and order of magnitude higher than any prior interval, which is rapidly increasing through publication time. Diversity decreases from this into the Maastrichtian, in which diversity has remained relatively stable through publication time. Sampling in South America is also relatively poor, with apparent diversity remaining low and flat where a signal is obtained (Fig. 5F).

Subsampled ‘global’ ornithischian diversity shows a distinctly different pattern from the raw curve, both in terms of overall trends, and in terms of the magnitude of the effect of publication history (Fig. 6A). The Jurassic is generally too poorly sampled to reveal a constant signal, but there is evidence of a decline through the J/K transition, which remains constant through publication time. This is followed by a middle-Cretaceous increase, in which ornithischian diversity is at its second highest level throughout their history. The magnitude of this Albian radiation has rapidly increased over publication time, the result being that originally what appeared to be increasing subsampled diversity over the Early/Late Cretaceous transition now shows a major decline from the Albian to Coniacian. Santonian subsampled diversity remains unknown, but when we see a signal emerge in the Campanian, diversity is higher than the Albian, reaching its highest level before declining by more than half into the Maastrichtian. This overall structure, besides the Albian, remains consistent throughout publication time with no major perturbations to the apparent ‘global’ curve.

Subsampled European diversity reveals increasing diversity across the Tithonian/Berriasian transition, followed by overall gradually decreasing diversity throughout the remainder of the Early Cretaceous (Fig. 6B). In Africa, the signal is too poor to reveal anything besides a Kimmeridgian/Tithonian subsampled diversity drop (Fig. 6C), and in Asia, there is evidence of a decline in subsampled diversity across the Albian/Cenomanian transition (Fig. 6D). In North America, subsampled diversity reveals a decline across the Early–Late Cretaceous transition, and a major decline from the Campanian to Maastrichtian, a pattern that remains stable through publication history (Fig. 6E). In South America, the subsampled signal is too poor to comment on ornithischian diversity (Fig. 6F).

If we look at how coverage has changed through publication history (based on Good’s u), we should expect that subsampled diversity patterns are reflective of this pattern. At a global level, coverage in the Cretaceous is much better than the Jurassic (Fig. 7A). Much of this, however, is based on patchy regional records. In Europe, we find that coverage increases across the J/K interval (Fig. 7B), and is the only place where a consistently reliable record here can be obtained. In Africa, coverage is generally poor, besides in the latest Jurassic (Fig. 7C). In Asia, coverage is poor up until the late Early Cretaceous (Fig. 7D). In North America, coverage is good in the latest Jurassic and ‘middle’ to Late Cretaceous, but non-existent in Early to Middle Jurassic and earliest Cretaceous (Fig. 7E). Coverage is generally poor for the entire South American ornithischian record (Fig. 7F), explaining why obtaining a subsampled diversity signal here is difficult.

Theropods

The overall shape of the raw ‘global’ theropod diversity curve remains consistent through publication history for the Jurassic (Fig. 8A), similar to ornithischians, where we see steadily increasing Middle–Late Jurassic diversity. ‘Middle’ Cretaceous raw diversity fluctuated, followed by a major Campanian–Maastrichtian rise. The lowest apparent diversity is in the Coniacian, reaching earliest Cretaceous levels. Notable variations due to publication history are in the Barremian–Cenomanian, where diversity increases in magnitude through time, gradually exceeding that for Late Jurassic diversity. Raw European diversity is fairly constant through publication history (Fig. 8B), with a Middle Jurassic diversity peak in the Bathonian, followed by a Callovian–Oxfordian trough, a second larger Kimmeridgian peak, and then constant decline from the Tithonian to the Valanginian. Barremian diversity is increases through publication time, and is as high as Kimmeridgian levels. Aptian and Albian diversity is relatively low through publication history. Campanian and Maastrichtian diversity levels are slowly increasing through publication history. As with ornithischians, African theropods are generally too poorly sampled at the stage level to recognise any consistent empirical patterns (Fig. 8C). There is a Cenomanian raw diversity spike, but how this compares with much of the rest of the Cretaceous is obscured by patchy sampling. In Asia, raw Late Jurassic diversity is generally lower than for the Cretaceous (Fig. 8D). The Cretaceous sees three peaks in apparent diversity during the Aptian, Turonian, and Campanian–Maastrichtian, with the latter being considerably higher than any previous one, and growing rapidly through publication history. In North America, raw diversity levels are dwarfed by the intensive sampling of latest Cretaceous theropods, with major gaps in the Middle–Late Jurassic and earliest Cretaceous records (Fig. 8E). Campanian and Maastrichtian raw diversity is constantly increasing at a faster rate than any other time interval, and consistently reveals a slight apparent diversity decline into the end-Cretaceous. Raw South American diversity estimates are changing rapidly through publication history, with almost every interval in which dinosaurs are available to be sampled doubling or tripling since 1991 (Fig. 8F). Of note is a recently emerging Late Jurassic theropod fossil record in South America, which at the present reveals an apparent low diversity.

When subsampling is applied, in the Late Jurassic we see a switch from steadily increasing subsampled diversity to a major Oxfordian peak and subsequent decline in diversity through the J/K transition decline, a pattern that is consistently recovered through publication time (Fig. 9A). Subsampled diversity is at its highest level during the Aptian than at any other stage during theropod history, and has doubled in the last 20 years of publication history. Campanian and Maastrichtian diversity are as high as the Cenomanian, a pattern that remains consistent through publication time. We see the ‘global’ J/K transition decline reflected in Europe (Fig. 9B), and a strong Barremian peak, which is not captured on a ‘global’ scale. Latest Triassic subsampled diversity is higher than at any other point in the Jurassic in Europe. Maastrichtian subsampled diversity remains high, reaching the same level as that for the Kimmeridgian. In Africa, as with ornithischians the signal is very patchy after subsampling is applied (Fig. 9C), but captures an Albian–Cenomanian diversity increase, which remains constant throughout publication history, and flat diversity in the latest Cretaceous. The subsampled theropod diversity signal is also patchy in Asia, but does reveal a very high latest Cretaceous diversity level, which is not otherwise seen throughout theropod evolutionary history (Fig. 9D). In North America, the subsampled record is as patchy as that for ornithischians, but remains stable through publication history (Fig. 9E). Here, we see slightly increasing subsampled diversity in the latest Jurassic, a large decline from the Aptian to Albian, and a major diversification from the Santonian to Campanian. In South America, a subsampled diversity signal is almost entirely absent, although we do see a reduction in almost half from the Norian to Rhaetian, which remains stable through publication history (Fig. 9F).

Theropod coverage levels are quite patchy at the ‘global’ level, remaining constant in the Late Triassic, fluctuating in the Middle Jurassic to earliest Cretaceous, but remaining fairly stable in the ‘middle’ and latest Cretaceous through publication history (Fig. 10A). On a regional level, this apparent ‘global’ signal across the J/K transition is again emphasised in Europe, but in the Valanginian and Albian, coverage is getting notably worse through publication history (Fig. 10B). Coverage in Africa (Fig. 10C) and Asia (Fig. 10D) is very patchy, and does not appear to have changed in the last 20 years overall, besides the origin of moderate coverage levels in the Oxfordian and Aptian of Asia. In North America, coverage levels are moderately high in the latest Jurassic, Aptian, and Albian, and latest Cretaceous, only improving in the latest Jurassic through publication history (Fig. 10E). In South America, coverage is generally poor throughout the Jurassic and Cretaceous, but appears to be declining in the Norian and Rhaetian theropod records (Fig. 10F).

Sauropodomorphs

Sauropodomorph empirical diversity emphasises some more changes in raw patterns through publication time, particularly in the ‘middle’ and Late Cretaceous (Fig. 11A). Late Jurassic patterns are fairly consistent, with a rising Kimmeridgian and Tithonian raw diversity emphasising an apparent major decline across the J/K interval. In Europe, sauropods show a consistent and major decline in raw diversity from the Kimmeridgian to the Berriasian (Fig. 11B). Much of the rest of the Cretaceous is too poorly sampled, but raw sauropod diversity never attains Kimmeridgian levels in Europe for the rest of their evolutionary history. Sauropodomorph dinosaurs are generally better sampled than theropods and ornithischians in Africa, showing an apparent decline through the Triassic/Jurassic transition, a latest Jurassic raw diversity peak, and low levels through the ‘middle’ to Late Cretaceous transition (Fig. 11C). In Asia, raw taxonomic diversity is generally low compared to the Maastrichtian, in which diversity is relatively high and still rapidly increasing through publication history (Fig. 11D). The North American sauropod record is very patchy, with the latest Jurassic showing a shift from rapidly increasing raw diversity from the Oxfordian to a slight drop from the Kimmeridgian to Tithonian (Fig. 11E). The South American Jurassic sauropod record is patchy, but raw diversity is increasing throughout the ‘middle’ to Late Cretaceous through publication history (Fig. 11F).

At a ‘global’ level, Jurassic sauropodomorph subsampled diversity remains consistent through publication history (Fig. 12A). Here, we see steadily increasing diversity levels through the Middle and Late Jurassic, before a decline through the J/K transition, which might have been initiated before the J/K boundary itself. The greatest change in subsampled diversity is in the Albian, which has almost doubled in the last 20 years, with implication for the ‘mid-Cretaceous sauropod hiatus’ (Mannion & Upchurch, 2011). Subsampling reduces the European diversity signal due to poor sampling of sauropods, although there is evidence for the sauropod decline beginning prior to the J/K transition (Fig. 12B). In Africa, when subsampling is applied, the few intervals in which a signal emerges reveal a fairly constant level of diversity through the Jurassic and Cretaceous, and through publication time, with the notable exception being an increase in subsampled diversity in the latest Jurassic (Fig. 12C). In Asia, the signal is also fairly poor after subsampling is applied (Fig. 12D). Here, we see an increase in subsampled diversity across the Triassic/Jurassic transition, and the highest diversity level is in the Maastrichtian, where subsampled estimates have increased by more than double in the last 20 years. In North America, the subsampled signal is highly degraded, although of note is a near doubling of Albian diversity levels in the last 20 years (Fig. 12E). In South America, the signal is very inconsistent, but improving through publication history, with a patchy Late Cretaceous signal beginning to emerge (Fig. 12F). Full subsampling results are provided in Supplemental Informations 7 and 8.

Sauropodomorph coverage varies greatly at the ‘global’ level, with high levels in the Triassic-Jurassic transition, the Middle and Late Jurassic (with the exception of the Callovian), and the Maastrichtian (Fig. 13A). As with theropods and ornithischians, however, this is a composite of a very patchy regional record. In Europe, coverage is high during the latest Triassic, Middle Jurassic, and Late Cretaceous, and this does not seem to have varied with publication time (Fig. 13B). In Africa, moderate levels of coverage also have not changed substantially since 1991 (Fig. 13C). In Asia, coverage is generally high in the Late Jurassic, but the Cretaceous record is incredibly poor with just two data points (Aptian and Maastrichtian; Fig. 13D). In North America, the latest Jurassic has high coverage levels, which are increasing through publication history in the Kimmeridgian, and moderately high coverage in the Aptian and latest Cretaceous (Fig. 13E). In South America, coverage is very patchy and inconsistent, with the only noteworthy change through publication history being an increase for the Rhaetian interval (Fig. 13F).

Ornithischians

The J/K interval decline in subsampled diversity remains constant and recognisable throughout publication history, with this stability suggesting a real biological signal and not a publication artefact (Tennant, Mannion & Upchurch, 2016b). However, more focussed sampling needs to occur on J/K interval deposits to reveal the true global signal, as much of this pattern is based on fossils exclusively from historically well-sampled European localities (Tennant et al., 2016) (Figs. 6, 10 and 12). Ornithischian subsampled diversity decreases steadily through the Early Cretaceous in Europe, with a possible radiation in the Campanian to Maastrichtian, perhaps explained by an increase of recent occurrences of latest Cretaceous dinosaurian findings mainly in Spain, Portugal, France, and Romania (Riera et al., 2009; Csiki et al., 2010). However, many of these latest Cretaceous European dinosaur faunas are not particularly well-resolved stratigraphically compared to the well-studied North-American sections, which makes the timing of any regional extinction here and comparison with North America and Asia difficult at the present. Advanced ornithischian faunas, including ceratopsians and hadrosaurids, appear to have diversified extremely rapidly in the latest Cretaceous, but this is classically explained by the oversampling of North American Late Campanian localities, like Dinosaur Park Formation and its approximate temporal equivalents. Although a small rise in subsampled diversity is recovered from the Campanian to the Maastrichtian in Europe, this is considerably less marked than the decline in North America, where subsampling reveals that ornithischian diversity was actually declining from the Campanian to Maastrichtian (Brusatte et al., 2015).

Ornithischian subsampled diversity in Asia has been increasing steadily through publication time in the ‘middle’ Cretaceous, filling in the gap from equivalent latitude European deposits at this time. This is plausibly due to the radiation of Parksosauridae and Ankylopollexia clades, two of the most dominant Late Cretaceous dinosaurian taxa around this time. Together with the North American record, this manifests as a great global decline across the Early–Late Cretaceous interval, a pattern that was not recognised until more recent years due to the discovery of more Konzentrat-Lagerstätten in Mongolia and China around this time, such as the Jehol Biota (Lambert et al., 2001; Godefroit et al., 2008; Upchurch et al., 2011). A perceived Late Cretaceous subsampled diversity increase for Asian taxa, particularly hadrosauroids, could be due to a renaissance in the discovery of Cretaceous Asian dinosaurs over the past two decades (Lloyd et al., 2008; Barrett, McGowan & Page, 2009; Zhou & Wang, 2010; Upchurch et al., 2011; Mo et al., 2016). Despite the increasing availability of Early Cretaceous dinosaur-bearing formations (DBFs) in Africa in the last 20 years (e.g. Tunisia, Niger; Taquet & Russell, 1999; Anderson et al. (2007)), sampling here is still too limited to reveal any consistent patterns in ornithischian subsampled diversity (Mannion et al., 2011; Upchurch et al., 2011; Tennant, Mannion & Upchurch, 2016b) (Figs. 6 and 7).

This regional distinction could be due to the tie between ecomorphological function and biological diversity, as Asian hadrosauroids increased in morphological disparity during the latest Cretaceous, whereas in North America large-bodied bulk-feeding ornithischians decreased in their disparity (Vavrek & Larsson, 2010; Campione & Evans, 2011; Brusatte et al., 2012; Mitchell, Roopnarine & Angielczyk, 2012). In North America, several abiotic factors, including extreme fluctuations of the Western Interior Sea, and the Laramide orogeny and proposed biogeographic provincialism, may have affected the evolution of North America dinosaurs in distinct ways from species on other continents (Gates, Prieto-Márquez & Zanno, 2012; Arbour, Zanno & Gates, 2016), meaning that the North American record is unlikely to be representative of global diversity pattern (Sampson et al., 2010; Brusatte et al., 2012).

Theropods

As already shown elsewhere (Barrett, McGowan & Page, 2009; Brusatte et al., 2012), ‘global’ theropod diversity trends are overall very similar to that of Ornithischia, with subsampled diversity increases during the Late Jurassic (Oxfordian and Tithonian peaks punctuated by a Kimmeridgian decline), late Early Cretaceous (Aptian), early Late Cretaceous (Cenomanian) and latest Cretaceous. Moderately high Middle and Late Jurassic diversity subsampled levels represent the radiation of major avetheropodan clades, and a wealth of new discoveries in recent years, particularly from Asia (Upchurch et al., 2011; Xu et al., 2011; Carrano, Benson & Sampson, 2012; Benson et al., 2014; Tennant, Mannion & Upchurch, 2016b).

European subsampled theropod diversity is more constant than in other regions, with a Bajocian peak followed by a Bathonian–Oxfordian trough, and a Kimmeridgian peak followed by a Tithonian to Valanginian drop. This can, at least in part, be explained by an abundance of well-sampled Late Jurassic formations from across Western Europe (Upchurch et al., 2011; Benson et al., 2013; Tennant et al., 2016). Barremian diversity is increasing rapidly through publication history, and is now as high as calculated for the Kimmeridgian. As with the Late Jurassic, at least part of this signal represents the influence of a Lagerstätten effect (e.g. Las Hoyas, Spain) (Buscalioni et al., 2008; Upchurch et al., 2011; Sánchez-Hernández & Benton, 2012), highlighting that single, well-sampled formations can have a profound historical effect on our understanding of regional diversity patterns, even when subsampling methods are applied. The European Aptian–Albian record is increasing slower through time compared to the Campanian–Maastrichtian. However, this might possibly change in the future, as the ichnological record in southern Europe is quite abundant for the Aptian–Albian interval, and suggests a currently unrecognised dinosaurian diversity present there (Dalla Vecchia, 2002; Meyer & Thuring, 2003).

The North American theropod record is dwarfed by an oversampling of latest Cretaceous dinosaur-bearing formations (e.g. Dinosaur Provincial Park, Hell Creek Formation). An increasingly even representation of latitudinally diverse localities from the Cenomanian–Campanian of Utah, Colorado, New Mexico, and Mexico (e.g. Wahweap Formation), may increase the magnitude of the small subsampled diversity drop through the Maastrichtian. Subsampling highlights a latest Jurassic peak in diversity (due to the abundance of remains from the well-sampled Morrison Formation; Foster (2003)), although Jurassic subsampled diversity never attains that of the Cretaceous highs during the Aptian and Campanian. In contrast to Brusatte et al. (2015), who found no evidence for a progressive Campanian-Maastrichtian decline in North American theropod faunas using similar SQS analyses (implemented in R; see (Tennant, Mannion & Upchurch, 2016a, 2016b) and (Alroy, 2010a, 2010c) for comparative discussions), we find a very slight decline that remains constant through publication history, that likely relates to our usage of a slightly different subsampling approach. Aptian subsampled diversity is relatively high due to the more heavily sampled localities from Montana to Texas (Kirkland et al., 1997; Cifelli et al., 1999; Kirkland & Madsen, 2007).

In Africa, there is a Cenomanian radiation (Fig. 9C) mainly due to the multitaxic theropod dominated Kem Kem beds and other Albian–Cenomanian (‘middle’ Cretaceous) equivalents in Northern Africa, but this signal might have been altered by time averaging effects constraining a more temporally diluted diversity in a single unit (Mannion & Barrett, 2013; Evers et al., 2015; Chiarenza & Cau, 2016). Asian subsampled diversity peaks in the Aptian, Campanian, and Maastrichtian might be explained by a Lagerstätten ‘bonanza’ effect, especially considering the high quality preservation deposits discovered and heavily sampled in the last 20 years (e.g. Liaoning) (Lloyd et al., 2008; Zhou & Wang, 2010; Godefroit et al., 2013; O’Connor & Zhou, 2015; Tennant et al., 2016), although coverage remains only moderate (around 0.5) in each of these intervals (Fig. 10). Similarly to the pattern in Africa and Asia, South American theropod subsampled diversity stands out compared to other North America and Europe, remaining relatively signal deficient. Despite an increasing rate of discovery of new taxa, which often alter our knowledge of dinosaur phylogeny and biogeography from the ‘middle’ Cretaceous of Patagonia and Brazil (Novas et al., 2005a, 2013; Novas & Pol, 2005; Canale et al., 2009), coverage remains poor at the stage level, emphasising the need for greater stratigraphic resolution of the theropod-bearing formations here.

Sauropodomorphs

Subsampled diversity patterns of sauropodomorphs share some characteristics of those of theropods and ornithischians, despite having a different fossil record due to taphonomic differences (i.e. larger, more robust skeletons being preferentially preserved in different environmental settings) (Mannion & Upchurch, 2010, 2011; Dean, Mannion & Butler, 2016). This is compounded by a difficulty in assigning a large number of taxa to specific stage bins, which unfortunately excludes many of them from our analyses (Supplemental Information 1). Differences in diversity patterns between sauropodomorphs and ornithischians have classically been interpreted as being due to exclusive competition between the two main herbivorous dinosaurian subtaxa (Butler et al., 2009), with an explosive radiation in ornithischians during the Early Cretaceous resulting from the apparent decline in diversity of sauropodomorphs. In fact, the J/K transition represents a major extinction ‘event’ for sauropodomorphs, reflecting the decline of non-neosauropods, diplodocoids and basal macronarians (Mannion et al., 2013; Tennant, Mannion & Upchurch, 2016b). Sauropodomorph faunas have a low subsampled diversity in the earliest Cretaceous, coupled with a generally poor fossil record (Mannion & Upchurch, 2010), but at a time when we otherwise see rapid increases in theropod and ornithischian diversity and a prolonged phase of faunal turnover (Upchurch & Mannion, 2012; Tennant, Mannion & Upchurch, 2016b). Sauropodomorph subsampled diversity levels fluctuate from the ‘middle’ Cretaceous until the final latest Cretaceous radiation, with a possible small decline in the Maastrichtian. This finding is somewhat contrary to that of Sakamoto, Benton & Venditti (2016) who found that their decline was initiated in the Early Cretaceous, and that the diversification of titanosaurs was at an insufficient rate to compensate for the overall loss of sauropodomorph lineages throughout the rest of the Cretaceous. This discrepancy could be due to the differences in datasets used, and that several recently named titanosaurs taxa have yet to be included in published phylogenies, or the distinction between estimation diversity levels against diversification rates. However, we find that sauropodomorphs are at their most diverse during the Albian (Fig. 12). Sauropodomorphs appear to be overrepresented with respect to what we might expect for almost the entire duration of the Jurassic, whereas the opposite is true for the Cretaceous (Mannion et al., 2011; Upchurch et al., 2011; Tennant, Mannion & Upchurch, 2016b). The general patterns of ‘global’ subsampled diversity shows a steady increase from Middle to the end of Jurassic with a decline through J/K transition (Upchurch & Mannion, 2012; Tennant, Mannion & Upchurch, 2016b). The relatively high Late Cretaceous subsampled diversity levels can at least be partially explained by the constant discovery of new titanosaurian taxa, especially from Gondwanan continents (Vieira et al., 2014; de Jesus Faria et al., 2015; Bandeira et al., 2016; Poropat et al., 2016), and only recently a more appreciated diversity of diplodocoids (e.g. dicraeosaurids, rebbachisaurids) from relatively poorly sampled regions such as Africa (Mannion & Barrett, 2013; Wilson & Allain, 2015; Ibrahim et al., 2016).

Large-bodied sauropodomorph diversity in the Tithonian is certainly influenced by the intense sampling history of the North American Morrison Formation, where there is an unusually high diversity and cranial disparity of megaherbivores within a relatively resource-poor environment (Button, Rayfield & Barrett, 2014). Here, high diversity remains in spite of our accounting for large collection biases associated with Konzentrat-Lagerstätten (Alroy, 2010a, 2010c), implying that sauropodomorphs reached their zenith in diversity during the Late Jurassic. Sauropodomorphs appear to be better sampled than theropods and ornithischians in Africa (Fig. 13C), although their records remain largely too inconsistent and patchy record to reveal any major patterns. Asian subsampled diversity is constantly low until the Maastrichtian, where it increases moderately due to a series of recent discoveries from Pakistan and China (Malkani, 2010; Junchang et al., 2013). However, the Asian Cretaceous sauropodomorph record is otherwise very poorly sampled, especially compared to ornithischians and theropods. This phenomenon could be explained by a taphonomic size bias discriminating against the preservation of larger-bodied animals in pre-Late Cretaceous Konservat-Lagerstätten, while they are more present although more rare in the dense bone assemblages from the latest Cretaceous of Mongolia, China, and India (Kidwell, 2001).

There is a notable subsampled diversity decline in European sauropodomorphs through the J/K transition, as with other dinosaurian groups (Upchurch & Mannion, 2012; Tennant, Mannion & Upchurch, 2016b). This is distinct from results obtained with other methods (e.g. TRiPS, True Richness estimated using a Poisson Sampling) which do not find any evidence for such a decline (Starrfelt & Liow, 2016). Subsampling also reveals that sauropodomorph diversity in the latest Cretaceous of Europe was relatively flat. The sauropodomorph record in South America is poor and mostly confined to the Late Cretaceous, with diversity levels rising and resolution improving through publication time as coverage increases and as new taxa get identified from emerging Patagonian and Brazilian deposits (Novas et al., 2005b; Novas, 2009).

Here, it is worth noting the distinction between global and regional sauropodomorph records. On a global level, our results provide strong evidence for a substantial sauropod subsampled diversity decline from the Campanian to Maastrichtian. However, this decline is not represented in any of the regional sauropodomorph diversity signals. Instead, the ‘global’ signal in the Maastrichtian is comprised of a medley of regional records, which are only continuous with the Campanian record in Europe and North America. Therefore, the ‘global extinction’ of sauropods in the latest Cretaceous is actually due to regionally heterogeneous sampling signals that are summed into a misleading ‘global’ curve. A similar case can be made for the apparently ‘global’ radiation in the Albian, which is primarily a reflection of a well-sampled North American Albian sauropodomorph record (Fig. 12). Thus, when looking at diversity signals, interpretation of global patterns without considering structural changes on a regional level is not recommended.