Diversity patterns amongst herbivorous dinosaurs and plants during the Cretaceous: implications for hypotheses of dinosaur/angiosperm co-evolution

Authors


Richard J. Butler, Department of Palaeontology, The Natural History Museum, Cromwell Road, London SW7 5BD, UK.
Tel.: +44 20 7942 5582; fax: +44 20 7942 5546;
e-mail: r.butler@nhm.ac.uk

Abstract

Abstract Palaeobiologists frequently attempt to identify examples of co-evolutionary interactions over extended geological timescales. These hypotheses are often intuitively appealing, as co-evolution is so prevalent in extant ecosystems, and are easy to formulate; however, they are much more difficult to test than their modern analogues. Among the more intriguing deep time co-evolutionary scenarios are those that relate changes in Cretaceous dinosaur faunas to the primary radiation of flowering plants. Demonstration of temporal congruence between the diversifications of co-evolving groups is necessary to establish whether co-evolution could have occurred in such cases, but is insufficient to prove whether it actually did take place. Diversity patterns do, however, provide a means for falsifying such hypotheses. We have compiled a new database of Cretaceous dinosaur and plant distributions from information in the primary literature. This is used as the basis for plotting taxonomic diversity and occurrence curves for herbivorous dinosaurs (Sauropodomorpha, Stegosauria, Ankylosauria, Ornithopoda, Ceratopsia, Pachycephalosauria and herbivorous theropods) and major groups of plants (angiosperms, Bennettitales, cycads, cycadophytes, conifers, Filicales and Ginkgoales) that co-occur in dinosaur-bearing formations. Pairwise statistical comparisons were made between various floral and faunal groups to test for any significant similarities in the shapes of their diversity curves through time. We show that, with one possible exception, diversity patterns for major groups of herbivorous dinosaurs are not positively correlated with angiosperm diversity. In other words, at the level of major clades, there is no support for any diffuse co-evolutionary relationship between herbivorous dinosaurs and flowering plants. The diversification of Late Cretaceous pachycephalosaurs (excluding the problematic taxon Stenopelix) shows a positive correlation, but this might be spuriously related to poor sampling in the Turonian–Santonian interval. Stegosauria shows a significant negative correlation with flowering plants and a significant positive correlation with the nonflowering cycadophytes (cycads, Bennettitales). This interesting pattern is worthy of further investigation, and it reflects the decline of both stegosaurs and cycadophytes during the Early Cretaceous.

Introduction

Palaeobiologists have used evidence from the fossil record to propose a number of diffuse co-evolutionary scenarios (e.g. Janzen & Martin, 1981; Stebbins, 1981; Bakker, 1983; Stanley et al., 1983; Vermeij, 1983, 1987; Collinson & Hooker, 1991; Zavada & Mentis, 1992; Labandeira, 2002; Becerra, 2003; Gandolfo et al., 2004; Grimaldi & Engel, 2005; Rønsted et al., 2005; Moreau et al., 2006). Some of the best known of these hypotheses concern proposed interactions between nonavian herbivorous dinosaurs and early angiosperms (flowering plants) during the Cretaceous period (e.g. Bakker, 1978, 1986; Coe et al., 1987; Wing & Tiffney, 1987a,b; Weishampel & Norman, 1989; Tiffney, 1992, 2004; Coria & Salgado, 2005). The origin and radiation of angiosperms in the middle and Late Cretaceous represents one of the key events in the evolution of life, and played a profound role in shaping the extant biota (e.g. Crane et al., 1995; Soltis & Soltis, 2004; Friis et al., 2006; Frohlich & Chase, 2007; Crepet, 2008). The Cretaceous is also characterized by major faunal changes amongst the then dominant terrestrial herbivores: the nonavian dinosaurs (e.g. Sereno, 1997; Weishampel et al., 2004a). Considering the likely ecological impact of these, Bakker (1978, 1986) proposed a link with the origin and early diversification of angiosperms. He noted that Late Jurassic floras were composed largely of gymnosperms (including conifers, Bennettitales, ginkgos and cycads) and pteridophytes, whereas Late Jurassic herbivorous dinosaur faunas were dominated by sauropods and (to a lesser extent) stegosaurs, groups that Bakker considered as high browsers feeding largely upon tree canopies. By contrast, Early Cretaceous faunas were dominated by low-browsing ornithischian groups such as ornithopods and ankylosaurs, whereas high-browsing groups such as sauropods and stegosaurs were of much reduced diversity. Bakker (1978, 1986) proposed that this shift from a high-browsing fauna to a low-browsing fauna created highly disturbed environments that favoured the evolution of the weedy, r-selected (i.e. high fecundity and short generation times) early angiosperms. Continuing disturbance caused by low browsers favoured the subsequent angiosperm radiation, leading eventually to their ecological dominance. Other authors subsequently proposed that the middle to Late Cretaceous radiation of angiosperms might have driven the evolution and radiation of several herbivorous clades such as ornithopods, ceratopsians and titanosaurian sauropods (e.g. Coe et al., 1987; Wing & Tiffney, 1987a,b; Weishampel & Norman, 1989; Tiffney, 1992, 2004; Coria & Salgado, 2005).

Co-evolutionary hypotheses are easily formulated, but they are rather more difficult to test. Many of the hypotheses of co-evolutionary interactions between dinosaurs and angiosperms during the Cretaceous are couched in terms of putative co-radiations of major dinosaur clades and early flowering plants. Most attempts at testing dinosaur/angiosperm co-evolutionary hypotheses have assessed the evidence for co-radiations; the absence of evidence for co-radiations between Cretaceous angiosperms and herbivorous dinosaurs would falsify (on the basis of available data) many of the existing co-evolutionary hypotheses. Although co-radiations may indicate a co-evolutionary process, this is not necessarily the case; however, co-evolutionary hypotheses that do not involve co-radiation (speciation) are extremely difficult to test in the fossil record and are not considered here.

Weishampel & Jianu (2000) used species-level cladograms for Ornithischia and Sauropodomorpha to calculate absolute diversity patterns that included information on ‘ghost lineages’ (lineages that are missing from the fossil record but are inferred to have been present based upon information on the stratigraphic ranges and evolutionary relationships of fossil species; e.g. Norell & Novacek, 1992). Weishampel and Jianu suggested that few significant changes in dinosaur herbivore diversity corresponded to the onset of the angiosperm radiation (i.e. that there was limited evidence for co-radiations) and that co-evolutionary hypotheses could not be supported. Barrett & Willis (2001) carried out a detailed assessment of the fossil evidence, and concluded that spatiotemporal data did not support a causal role for dinosaurs in the origin of angiosperms, but that co-evolutionary interactions were plausible for the Late Cretaceous, although supported by little direct evidence at present. Despite the fact that neither study found significant evidence in support of co-radiations or co-evolutionary scenarios, these hypotheses continue to be invoked by more recent studies (e.g. Tiffney, 2004; Coria & Salgado, 2005).

This contribution takes a quantitative approach, building on previous qualitative tests of dinosaur–angiosperm interactions (e.g. Weishampel & Jianu, 2000; Barrett & Willis, 2001). We have compiled a new database of Cretaceous dinosaur and plant distribution to explore and quantify patterns of relative diversity. We provide statistical tests of temporal congruence between major faunal and floral turnover events and discuss whether the results of these tests support co-radiation events, and thus whether they support or undermine co-evolutionary hypotheses. As a result, this study represents one of the first attempts to provide a quantitative test of co-evolutionary hypotheses operating over extended geological timescales.

Methods

A relational database was built using Microsoft Access comprising information on the global distribution of herbivorous nonavian dinosaurs during the Cretaceous. This database was collated from an extensive review of the primary literature, beginning with references cited in Weishampel et al. (2004b) but also incorporating more recent references and a small amount of data from The Paleobiology Database (http://www.paleodb.org). All Cretaceous ornithischians and sauropodomorphs, and several theropod clades (Therizinosauroidea, Oviraptorosauria and Ornithomimosauria), are considered to be primarily herbivorous (cf. Weishampel & Norman, 1989; Weishampel et al., 2004a; Barrett, 2005). Taxonomic assignments generally follow Weishampel et al. (2004a,b), except where more recent information is available. For each locality we combined faunal information with data on palaeoecology, geological age, sedimentology and the inferred depositional environment. For each geological formation yielding herbivorous dinosaur material, we collected macrofloral information (localities, systematics, physiognomy and palaeoenvironments) from the primary literature (identified using standardized searches in GeoRef and Web of Science). At present, this database contains 6549 dinosaur and macroplant occurrences at 1952 localities, representing 407 dinosaur taxa and over 2300 macrofloral taxa.

Based upon this database, we generated plots of absolute and relative diversity through time for Cretaceous herbivorous dinosaurs and plants (Figs 1 and 2a,b), in terms of both numbers of valid genera and numbers of occurrences (an occurrence is defined as the presence of a taxon at a given locality) per Standard European Stage. The Cretaceous period (145.5–65.5 Ma) contains 12 Standard European Stages – the names of these stages and their dates are given in the caption to Fig. 1. Although generic plots provide information on relative taxonomic diversity, plots of occurrence data give some indication of the relative abundances of major clades (e.g. a clade may be relatively diverse in terms of named genera, but relatively rare in terms of occurrences). Counts of relative generic and abundance data can both be affected by taphonomy (biological and physical processes, including the processes of decay, fossilization, erosion and selective sampling, that occur between the death of an organism and its eventual discovery and study by palaeontologists, and which have the potential to bias our understanding of the history of life) and should be interpreted with caution.

Figure 1.

 Area plots showing temporal variation in the relative and absolute numbers of genera (a,c) and occurrences (b,d) of major groups of herbivorous nonavian dinosaurs included within our Cretaceous database (from Butler et al., in press: fig. 2). Geological stages of the Cretaceous are plotted along the x-axis, with the oldest stage (Berriasian) to the left of the figure, and the youngest (Maastrichtian) to the right. The y-axis shows either relative (a,c) or absolute (b,d) numbers of genera/occurrences. An occurrence is the presence of a particular taxon at a particular locality. Stage name abbreviations, in temporal order [with dates, in million years ago (Ma), in brackets] are as follows: BER, Berriasian (145.5–140.2 Ma); VAL, Valanginian (140.2–136.4 Ma); HAU, Hauterivian (136.4–130.0 Ma); BAR, Barremian (130.0–125.0 Ma); APT, Aptian (125.0–112.0 Ma); ALB, Albian (112.0–99.6 Ma); CEN, Cenomanian (99.6–93.5 Ma); TUR, Turonian (93.5–89.3 Ma); CON, Coniacian (89.3–85.8 Ma); SAN, Santonian (85.8–83.5 Ma); CAM, Campanian (83.5–70.6 Ma); MAA, Maastrichtian (70.6–65.5 Ma).

Figure 2.

 (a,b) Temporal variation in relative diversity of major plant groups in terms of genera (a) and occurrences (b) within our Cretaceous database (from Butler et al., in press: fig. 3). Geological stages of the Cretaceous are plotted along the x-axis, with the oldest stage (Berriasian) to the left of the figure, and the youngest (Maastrichtian) to the right. The y-axis shows either relative numbers of genera/occurrences. An occurrence is the presence of a particular taxon at a particular locality. The angiosperm radiation is most marked from the Aptian–Albian onwards. (c) Temporal variation in absolute taxic diversity of major plant groups, from the Lower Cretaceous to the present day, modified from Niklas et al. (1985: fig. 2). (d) Temporal variation in relative diversity of major plant groups during the Cretaceous, modified from Crane (1987: fig. 3a). Stage name abbreviations as in Fig. 1, with the exception of NEO, Neocomian (outdated term formerly used for the Berriasian–Hauterivian).

Where the dating for an occurrence is uncertain, we include it within the diversity estimates for all stages to which it is potentially assignable (e.g. occurrences from a formation dated as Barremian–Aptian are included in diversity counts for both the Barremian and Aptian). Relative diversity plots are subdivided by major dinosaur (Sauropodomorpha, Theropoda, Ceratopsia, Ornithopoda, Pachycephalosauria, Ankylosauria and Stegosauria) and plant (angiosperms, Coniferales, Cycadales, Bennettitales, Filicales, Ginkgoales, etc.) clades. Our database does not provide a comprehensive overview of global plant diversity through the Cretaceous, because we have only collected data from geographical areas and temporal periods that have yielded dinosaur material. However, our database is the first attempt to sample only those floras known to be directly contemporaneous, both spatially and temporally, with the sampled dinosaur record, thereby providing a unique perspective on changes in dinosaur and plant diversity through the Cretaceous. In addition, we compared our results with previously published analyses of Cretaceous plant diversity that were based upon independently derived databases (e.g. Niklas et al., 1985; Crane, 1987; Lidgard & Crane, 1990).

We chose to focus primarily on patterns of relative diversity rather than absolute diversity for several reasons. Most notably, recent work has demonstrated that much of the variation in marine invertebrate taxonomic diversity through time can potentially be explained by sampling biases, such as available rock volume and outcrop area (e.g. Peters & Foote, 2001; Crampton et al., 2003; Peters, 2005; Smith & McGowan, 2005, 2007). Recent studies have suggested that sampling biases also affect diversity patterns within the terrestrial record (e.g. Weishampel & Jianu, 2000; Upchurch & Barrett, 2005; Barrett et al., 2007; Lloyd et al., 2008). Of particular interest, the peaks in absolute dinosaur diversity that have often been identified in the middle and latest Cretaceous, and the diversity troughs identified in the earliest Cretaceous (‘‘Neocomian’’) and early Late Cretaceous, may be partially explicable as the result of sampling biases (Upchurch & Barrett, 2005; Barrett et al., 2007; Lloyd et al., 2008; P. M. Barrett, unpublished data). When ghost lineages are included in dinosaur diversity estimates many of the major diversity troughs identified on the basis of raw taxonomic counts can no longer be identified (see Weishampel & Jianu, 2000; Upchurch & Barrett, 2005; Lloyd et al., 2008).

By Contrast, relative abundance may be more robust to major geological sampling biases than absolute abundance. Although temporal changes in rates of sediment deposition/preservation will clearly affect absolute abundance patterns (i.e. more specimens/species will potentially become fossilized during periods of widespread deposition of sediments than during periods of erosion), it seems intuitively less likely that the relative abundances of different subclades with similar preservation potential (e.g. Ceratopsia and Ornithopoda) within a more inclusive clade (e.g. Dinosauria) would have been affected in a comparable manner. Nevertheless, it is possible to conceive of cases in which major geological sampling biases might effect relative diversity patterns: for instance, if certain dinosaur groups are more likely to be preserved in marine sediments as a result of coastal or even semi-aquatic habitat preferences (e.g. coastal habitat preferences have been proposed for nodosaurid ankylosaurids: Coombs & Deméré, 1996; Butler & Barrett, 2008), they could conceivably be over-represented in the fossil record during periods of global sea-level high stands. Such caveats must be borne in mind when examining relative diversity and patterns should be interpreted conservatively.

Nonparametric statistical tests were used to compare temporal changes in the relative diversity of angiosperms to those exhibited by the major dinosaur clades. Spearman’s rank correlation coefficient tests whether peaks in two data sets are correlated via a rank-ordered linear correlation, whereas Kendall’s tau coefficient assesses whether relative diversity patterns rise and fall in phase with one another. For either statistic, a significant positive result would indicate that as angiosperms became increasingly important components of global floras, the dinosaur clade in question would also have increased in relative importance (i.e. a co-radiation). A significant negative result would indicate the opposite, that the dinosaur clade in question became a less important component of global dinosaur faunas as angiosperms radiated. Significant negative results would, therefore, be inconsistent with a co-evolutionary relationship. By contrast, significant positive results would be consistent with each of the following alternative hypotheses: (i) that the dinosaur clade and angiosperms co-radiated in a co-evolutionary process; (ii) that the radiation of angiosperms drove the simultaneous radiation of the dinosaur clade, but that the radiation of the dinosaur clade did not affect angiosperm evolution; (iii) that the radiation of the dinosaur clade drove the simultaneous radiation of angiosperms, but that the radiation of angiosperms did not affect the evolution of the dinosaur clade; (iv) that the coincident radiations of both clades were driven by different processes or by a common external factor. Although diversity analyses alone cannot distinguish between these alternatives, they can at least identify clades for which co-evolution (the first of the four alternatives) can be considered a plausible hypothesis on the basis of available data – these clades can then be examined in greater detail using other approaches (e.g. GIS spatiotemporal analysis –Butler et al., 2008; direct examination of fossilized gut contents –Molnar & Clifford, 2001; Tweet et al., 2005).

Statistical analyses were carried out in past (Hammer et al., 2001) using a permutation test with 1000 random replicates and the results are presented in Tables 1 and 2. The same statistical tests were also used to compare temporal changes in the relative diversity of angiosperms with temporal changes in the relative diversity of other major plant groups.

Table 1.   Results of statistical comparisons (using Kendall’s tau coefficient and Spearman’s rank correlation coefficient) of temporal variation during the Cretaceous in the relative generic diversity (within our database) of angiosperms and (a) the relative generic diversity of major dinosaur groups; (b) the relative generic diversity of other major plant groups.
 Angiosperms – genera (Kendall’s)Angiosperms – genera (Spearman’s)
  1. ‘Cycadophytes’ includes cycads, Bennettitales and ‘cycadophyte’ material that cannot be assigned to either clade; *> 0.05.

(a) Dinosaur groups
 Sauropoda−0.0303 (= 0.8909; Ppermut = 0.951)−0.1608 (= 0.6175; Ppermut = 0.592)
 Stegosauria−0.7412 (= 0.0008; Ppermut = 0.002)*−0.851 (= 0.0004; Ppermut = 0.003)*
 Ankylosauria0.2901 (= 0.1892; Ppermut = 0.225)0.3923 (= 0.2072; Ppermut = 0.206)
 Ceratopsia−0.0909 (= 0.6808; Ppermut = 0.74)−0.0769 (= 0.8122; Ppermut = 0.823)
 Ornithopoda0.0763 (= 0.7297; Ppermut = 0.803)0.1261 (= 0.6962; Ppermut = 0.725)
 Pachycephalosauria0.3195 (= 0.1482; Ppermut = 0.219)0.3703 (= 0.2361; Ppermut = 0.237)
 Theropoda0.1706 (= 0.4401; Ppermut = 0.498)0.2958 (= 0.3506; Ppermut = 0.33)
(b) Plant groups
 Bennettitales−0.7576 (= 0.0006; Ppermut < 0.001)*−0.9091 (< 0.001; Ppermut < 0.001)*
 Cycadales−0.3877 (= 0.0793; Ppermut = 0.088)−0.5564 (= 0.0603; Ppermut = 0.065)
 ‘Cycadophytes’−0.7576 (< 0.001; Ppermut < 0.001)*−0.9021 (< 0.001; Ppermut < 0.001)*
 Coniferales−0.3939 (= 0.0746; Ppermut = 0.1)−0.5245 (= 0.08; Ppermut = 0.095)
 Filicales−0.8182 (< 0.001; Ppermut < 0.001)*−0.9440 (< 0.001; Ppermut < 0.001)*
 Ginkgoales−0.5455 (= 0.0136; Ppermut = 0.016)*−0.6643 (= 0.0185; Ppermut = 0.025)*
Table 2.   Results of statistical comparisons (using Kendall’s tau coefficient and Spearman’s rank correlation coefficient) of temporal variation during the Cretaceous in the relative number of occurrences (within our database) of angiosperms and (a) the relative numbers of occurrences of major dinosaur groups; (b) the relative numbers of occurrences of other major plant groups.
 Angiosperms – occurrences (Spearman’s)Angiosperms – occurrences (Kendall’s)
  1. ‘Cycadophytes’ includes cycads, Bennettitales and ‘cycadophyte’ material that cannot be assigned to either clade; *> 0.05.

(a) Dinosaur groups
 Sauropoda−0.1608 (= 0.6175; Ppermut = 0.592)−0.0303 (= 0.8909; Ppermut = 0.951)
 Stegosauria−0.8062 (= 0.0015; Ppermut = 0.002)*−0.6722 (= 0.0023; Ppermut = 0.002)*
 Ankylosauria0.4336 (= 0.1591; Ppermut = 0.185)0.303 (= 0.1702; Ppermut = 0.226)
 Ceratopsia0.5035 (= 0.0952; Ppermut = 0.114)0.3333 (= 0.1314; Ppermut = 0.159)
 Ornithopoda−0.1189 (= 0.7129; Ppermut = 0.703)−0.1515 (= 0.4929; Ppermut = 0.547)
 Pachycephalosauria0.5897 (= 0.0436; Ppermut = 0.062)0.4654 (= 0.0352; Ppermut = 0.065)
 Theropoda0.3287 (= 0.2969; Ppermut = 0.286)0.2121 (= 0.3371; Ppermut = 0.366)
(b) Plant groups
 Bennettitales−0.9510 (< 0.001; Ppermut < 0.001)*−0.8485 (< 0.001; Ppermut < 0.001)*
 Cycadales−0.662 (= 0.019; Ppermut = 0.019)*−0.4808 (= 0.0296; Ppermut = 0.036)*
 ‘Cycadophytes’−0.9021 (< 0.001; Ppermut = 0.001)*−0.7576 (= 0.0002; Ppermut < 0.001)*
 Coniferales−0.1259 (= 0.6967; Ppermut = 0.71)−0.0606 (= 0.7839; Ppermut = 0.853)
 Filicales−0.8652 (= 0.0003; Ppermut < 0.001)*−0.7176 (= 0.0012; Ppermut = 0.001)*
 Ginkgoales−0.6972 (= 0.012; Ppermut = 0.017)*−0.5428 (= 0.014; Ppermut = 0.013)*

In order to visualize the spatial and temporal distribution of angiosperm and dinosaur localities, we imported our database into a geographical information system, ArcGIS 9.1, using the procedure outlined by Rayfield et al. (2005). Localities were plotted onto palaeo-plate reconstructions, which are available for six different Cretaceous time slices (Scotese, 2001). Modern-day latitude–longitude coordinates for database localities were converted to palaeolatitude–longitude coordinates using PointTracker software (Scotese, 2004). We used a process of attribute selection within ArcGIS to divide our data set into 12 time slices, each corresponding to one of the Standard European Stages. The data points for each time slice were then plotted onto the most relevant (based upon the ages for stage boundaries given by Gradstein et al., 2004) palaeo-plate reconstruction: 140 Ma (Berriasian–Hauterivian), 120 Ma (Barremian–Aptian), 100 Ma (Albian–Cenomanian), 90 Ma (Turonian–Coniacian), 80 Ma (Santonian–Campanian) and 70 Ma (Maastrichtian). Attribute selection was then used to create new data layers demonstrating the spatial distribution of particular plant and dinosaur clades (e.g. sauropods and angiosperms) within a time slice. By overlaying these data layers, it is possible to compare spatial distributions with one another within a time slice and to assess changes in spatial distributions over time.

Results

Dinosaur diversity patterns

Sauropoda

Sauropod abundance is low in the Berriasian, as the clade comprises only 10% of genera and less than 25% of occurrences (Fig. 1a,b). However, sauropod abundance rises subsequently: during the Valanginian–Santonian, sauropods consistently account for about 30–40% of occurrences, whereas sauropod generic diversity increases to a peak of 43% of the total number of herbivorous genera during the Cenomanian. Relative sauropod generic diversity declines throughout the remainder of the Late Cretaceous (Turonian–Maastrichtian). By contrast, the relative number of sauropod occurrences declines only in the latest Cretaceous (Campanian–Maastrichtian). It is clear from our data set that there is a Campanian–Maastrichtian peak in sauropod abundance in terms of absolute numbers of genera and occurrences, but that the relative contribution of sauropods to global faunas declines during this time period.

Ankylosauria and Stegosauria

Ankylosaur abundance is relatively low during much of the earliest Cretaceous (particularly during the Valanginian and Hauterivian). The Barremian–Aptian saw an increase in ankylosaur occurrences and diversity (Fig. 1a,b). Although Berriasian–Hauterivian ankylosaurs are known only from Europe, the clade achieved a Laurasian distribution during the Barremian and occurred globally from the Aptian onwards. Following their Barremian radiation, the relative abundance of ankylosaurs is the greatest in the ‘middle’ to Late Cretaceous (Aptian–Santonian) but decreases during the latest Cretaceous (Campanian–Maastrichtian). However, as for sauropods (see above), there is a Campanian–Maastrichtian peak in ankylosaur abundance in terms of absolute numbers of genera/occurrences, but the relative contribution of ankylosaurs to global faunas declines. Stegosaurs form a significant component of herbivorous diversity, particularly in terms of generic diversity, during the earliest Cretaceous (Berriasian–Hauterivian). However, there is a marked decline in the relative number of stegosaur genera/occurrences from the Barremian onwards: stegosaurs are exceptionally rare components of Barremian–Albian herbivore faunas (accounting for only 1–3% of occurrences/genera) and no definite stegosaurs are known from post-Albian rocks (Galton & Upchurch, 2004).

Ornithopoda

Ornithopods played an important role in Cretaceous faunas (Fig. 1a,b), but there is little temporal variation in the relative abundance of the clade: with the exception of the Berriasian (during which time ornithopods account for nearly 52% of occurrences), ornithopods generally account for 30–43% of global occurrences and 25–33% of global genera throughout the Cretaceous. As already seen for the sauropods and ankylosaurs, there is a Campanian–Maastrichtian peak in ornithopod abundance in terms of absolute numbers of genera and occurrences, but the relative contribution of ornithopods to global faunas remains relatively constant during this time period.

Ceratopsia

Our database shows little variation in the relative importance of ceratopsians during most of the Cretaceous (Fig. 1a,b) – they comprise approximately 5–15% of occurrences and genera from the Berriasian until the end of the Santonian. However, the clade clearly underwent a major radiation during the Campanian and Maastrichtian: 30% of all Maastrichtian occurrences are of ceratopsians.

Pachycephalosauria

Undisputed pachycephalosaurs are limited to the Santonian–Maastrichtian of North American and eastern Asia (Maryańska et al., 2004) and form a minor component of faunas (Fig. 1a,b; 0.5–7% of occurrences; 3.5–11% of genera).

Theropoda

A radiation of potentially herbivorous theropods is first evident in the Barremian (Fig. 1a,b) and includes basal therizinosauroids, oviraptorosaurs and ornithomimosaurians. Thereafter, there is relatively little variation in the relative abundance of herbivorous theropods – they generally account for less than 7% of occurrences and 7–13% of genera, and are thus a minor component of ecosystems. All herbivorous theropod clades are confined to Laurasia.

Plant diversity patterns

Relative diversity plots for plant genera and occurrences show similar overall patterns (Fig. 2a,b). Plant floras in the earliest Cretaceous (Berriasian–Hauterivian; 130–145.5 Ma) are dominated by cycadophytes (cycads and Bennettitales), ferns (Filicales) and conifers (Coniferales), with significant components of Ginkgoales and minor components of other groups such as Czekanowskiales and Corystospermales. Interestingly, conifers comprise a much smaller proportion of floral diversity (20–35% of genera, 12–35% of occurrences during the Berriasian–Albian) than shown in the global diversity analyses of Niklas et al. (1985) (Fig. 2c) where they account for more than 50% of generic diversity. By contrast, the macrofloral diversity patterns obtained by Crane (1987) and Lidgard & Crane (1990) are very similar to the new compilation presented here. With the possible exception of lower conifer diversity, dinosaur-bearing formations of the Early Cretaceous do not stand out floristically from the general background pattern of floral diversity.

The Barremian (125–130 Ma) witnesses the appearance of the first widely accepted angiosperm macrofossils in our database, including assemblages from the Yixian Formation of China (e.g. Sun et al., 2001; Leng & Friis, 2003; Barremian–early Aptian) and the Potomac Group of Virginia (Doyle & Hickey, 1976; Barremian–Aptian), as well as possible angiosperms from the Lakota Formation of Wyoming (Fontaine in Ward, 1899; Barremian). Angiosperms are also present in a relatively small number of formations identified as Aptian in age – moreover, many of these have also been assigned broad age ranges and may be younger than Aptian (e.g. the Dakota and South Platte formations of the USA are both dated as late Aptian–early Cenomanian, the Haman Formation of Korea is dated as Aptian–Albian; Weishampel et al., 2004b). As in the earliest Cretaceous, it seems likely that angiosperms were rare floral components in Barremian–Aptian dinosaur ecosystems, an observation borne out by our occurrence data (Fig. 2b). Barremian–Aptian floras also show the onset of a marked decrease in occurrences and genera of cycadophytes. Some authors have ascribed this cycadophyte decline to competitive replacement by angiosperms (e.g. Crane, 1987; Watson & Cusack, 2005).

The Albian and Cenomanian exhibit a dramatic increase in angiosperm diversity, with angiosperms accounting for more than 50% of genera and more than 60% of occurrences. Between the Turonian and the Santonian dinosaur-bearing formations are characterized by a sharp decrease in angiosperm diversity and a major increase in the diversity of conifers. This pattern may be largely a taphonomic artefact (see Discussion, below). Finally, angiosperm abundance is high in the Campanian and Maastrichtian, as in other analyses of Cretaceous plant diversity (Niklas et al., 1985; Crane, 1987; Lidgard & Crane, 1990).

The differences between the relative diversity plots for plant genera and occurrences include: Bennettitales make a greater contribution in terms of numbers of occurrences (18–30%) than in terms of generic diversity (15–20%) during the earliest Cretaceous (Berriasian–Hauterivian); conifers make a greater contribution in terms of generic diversity (22–25%) than in terms of numbers of occurrences (11–18%) during the earliest Cretaceous (Berriasian–Hauterivian), but the reverse is true during the Turonian–Santonian (38–44% of occurrences; 51–62% of genera); angiosperms make a greater contribution in terms of generic diversity (43%) during the Aptian than in terms of numbers of occurrences (8%).

Statistical comparisons

In general, patterns of angiosperm relative diversity through the Cretaceous did not significantly correlate with relative diversity patterns for the dinosaur clades analysed (Tables 1 and 2). The only exception to this is Stegosauria: stegosaurs show a significant negative correlation with angiosperm diversity for both occurrence and genus data (Spearman’s, = 0.001–0.003; Kendall’s, = 0.001–0.002). Although pachycephalosaurs do not show a significant correlation with angiosperms when all data are included (Tables 1 and 2), a significant positive correlation is obtained when the controversial Berriasian taxon Stenopelix (Butler & Sullivan, 2008) is either excluded from the database or considered nonpachycephalosaurian (Spearman’s, = 0.001–0.044; Kendall’s, = 0.002–0.042). However, this result may be an artefact of sampling (see Discussion).

Significant negative correlations were identified between angiosperms and most other plant groups, with the exception of conifers (Tables 1 and 2). However, palaeoecological interpretation of these results is complicated by the fact that our data set comprises only a subset of the whole flora, and that the data set is further skewed during the Turonian–Santonian as a result of low sampling (see Discussion, below).

Discussion

Dinosaur diversity patterns

Sauropoda

The relative diversity patterns identified for sauropods differ somewhat from the absolute diversity patterns described by Upchurch & Barrett (2005) and from those also identified in our database (Fig. 1c,d). There is an apparent increase in absolute generic diversity between the Berriasian and the Albian, but it has been suggested that this is potentially explicable by an increase in the number of dinosaur-bearing formations, a proxy for temporal variation in the volume of dinosaur-bearing rock (see Upchurch & Barrett, 2005). If correct, sauropod diversity in real terms remained relatively constant from the Berriasian to the Albian. Absolute diversity patterns indicate that diversity declined in the Cenomanian–Coniacian, and that the Campanian–Maastrichtian saw a significant sauropod radiation composed exclusively of titanosaurian sauropods (Upchurch & Barrett, 2005).

Sauropods formed a significant component of herbivorous dinosaur diversity throughout the Cretaceous, although they were considerably less important than during the Late Jurassic. There is a strong regional signal in these data and it is clear that sauropods were much less important in North America than in Gondwana and other parts of Laurasia. Indeed, sauropods are unknown in the Turonian–Santonian of North America, and in the Campanian and Maastrichtian are represented by a single genus (Alamosaurus) that was restricted to the SW USA (e.g. Lehman, 2001).

Coria & Salgado (2005) proposed that a significant faunal turnover occurred among sauropods during the middle Cretaceous, with the extinction of diplodocoids and basal titanosauriforms followed by a Late Cretaceous radiation of titanosaurs: this event was linked to climate-driven floral changes. By contrast, Upchurch & Barrett (2005) suggested that a genuine decline in sauropod diversity occurred in the Cenomanian–Coniacian. Our data demonstrate a decline in relative generic diversity from the Cenomanian onwards, which is partly attributable to the radiation of other clades (primarily ceratopsians) during this interval, but there is little evidence for a decline in the relative number of sauropod occurrences at a global level prior to the Campanian (see above; Fig. 1b).

Ankylosauria and Stegosauria

Ankylosaurs were limited in diversity during the earliest Cretaceous (Berriasian–Hauterivian) and their record is confined to Europe at this time. However, ankylosaurs have been recorded from the Jurassic of North America and China (Vickaryous et al., 2004); so, this restricted distribution is likely to be an artefact. Nevertheless, the clade appears to undergo a radiation during the Barremian–Aptian: their relative diversity increases and they achieve a global distribution in this interval. This radiation is approximately coincident with declines in stegosaur occurrences and diversity. This coincidence is notable, given that: both ankylosaurs and stegosaurs are armoured, quadrupedal low browsers with similar body proportions and body sizes; each clade was predominantly Laurasian; and each possessed simple dentitions and jaw mechanisms, with expanded ribcages for bulk processing of plant matter (see Weishampel & Norman, 1989; Barrett, 2001; Galton & Upchurch, 2004; Vickaryous et al., 2004; although for evidence of complex jaw mechanics in ankylosaurs see Barrett, 2001; Rybczynski & Vickaryous, 2001). It is plausible that ankylosaurs replaced stegosaurs opportunistically, with ankylosaurs filling niches left vacant by stegosaur decline (Norman, 1985), but this hypothesis requires further detailed examination.

Ornithopoda

Ornithopods diversified significantly prior to the end of the Jurassic – for example, the Morrison Formation fauna of western North America (Late Jurassic: Kimmeridgian–Tithonian: 155.7–145.5 Ma) includes camptosaurids, dryosaurids and various basal ornithopods (Galton, 2007); however, they were relatively minor components of global faunas (Weishampel et al., 2004b). By contrast, our analysis demonstrates that ornithopods formed a major component of Cretaceous faunas from the Berriasian onwards. However, perhaps surprisingly, there is no discernable increase in the relative importance of ornithopods during the middle Cretaceous (see also Weishampel & Jianu, 2000; contraWeishampel & Norman, 1989) nor associated with the origin and diversification of hadrosaurs in the Late Cretaceous.

Ceratopsians

Ceratopsians increased in importance during the Early Cretaceous, but the clade was confined to eastern Asia at this time (e.g. Sereno, 2000). Ceratopsians form a minor component of global faunas throughout most of the Cretaceous. However, a major radiation of ceratopsians occurred during the Campanian–Maastrichtian (Fig. 1a,b), which was dominated by large-bodied ceratopsids. These taxa are known only from North America, although more basal ceratopsians are known from Asia and Europe.

Pachycephalosaurs

Although possible pachycephalosaurs have been reported from the Early and ‘middle’ Cretaceous (e.g. Galton, 1971; Sereno, 2000; Maryańska et al., 2004), all of the proposed records are problematic (Butler & Sullivan, 2008). Definite pachycephalosaurs appear in the fossil record during the latest Cretaceous (Campanian–Maastrichtian). As for Ceratopsia, the latest Cretaceous radiation of pachycephalosaurs seems to represent a genuine biological event.

Angiosperm diversity patterns

Critical review of the fossil record indicates that angiosperms were absent from the earliest Cretaceous (Berriasian–Valanginian) and older strata (e.g. Friis et al., 2006): the earliest angiosperm fossils are distinctive monoaperturate pollen from the Hauterivian (Hughes, 1994; Friis et al., 2006). Angiosperms were essentially absent, or exceptionally scarce, in dinosaur-inhabited ecosystems during the Berriasian–Hauterivian. Moreover, our analysis suggests that, although angiosperm diversity began to increase rapidly during the Barremian–Aptian (Fig. 2a), angiosperms remained scarce at this time (Fig. 2b).

The dramatic Aptian–Cenomanian increase in angiosperm diversity evident from our database (Fig. 2a,b) is similar to patterns detected by Crane (1987) and Lidgard & Crane (1990), but differs from that of Niklas et al. (1985), which show a slow rate of angiosperm diversification at this time. These differences may result from two biases in our data set: (1) our floral analysis is not global in extent, but is skewed to low-latitude temperate zones, where dinosaur fossils are best represented. The early angiosperm radiation exhibited a strong latitudinal signal, beginning in low palaeolatitudes (∼20°N to 20°S) and spreading to higher latitudes through the mid-late Cretaceous (Crane & Lidgard, 1989). Our sample from dinosaur-bearing formations therefore excludes many angiosperm-impoverished high latitude floras from the mid Cretaceous. (2) The Aptian–Cenomanian data are dominated by vast numbers of occurrences and genera from the Dakota Formation of the USA (e.g. Lesquereux, 1891) – angiosperms from this unit are represented by leaf impressions and their historical taxonomy is probably over-split (Wang & Dilcher, 2006). Regardless, the Aptian–Cenomanian represents the interval during which angiosperms first formed an important part of Cretaceous ecosystems in terms of both generic diversity and ecological abundance.

The abundance of angiosperms in dinosaur-bearing formations decreased in the Turonian–Santonian and was mirrored by an increase in conifer diversity. This pattern contrasts with the general background signal of increasing angiosperm diversification shown by macrofossils (Niklas et al., 1985; Crane, 1987) and pollen (Crane & Lidgard, 1989). Few formations of Turonian–Santonian age yield dinosaurs and only a handful of these have yielded published floras. The plant data for this interval are dominated by the marine Upper Yezo Group (late Turonian–Santonian) of Japan, which contains few angiosperms. Consequently, the scarcity of angiosperms in the Turonian–Santonian probably reflects a taphonomic signal. However, the analyses of Crane (1987) and Lidgard & Crane (1990) also show a minor decline in angiosperm diversity during this interval.

Dinosaur faunal turnover and angiosperm origins

The dinosaur faunal turnover hypothesis developed by Bakker (1978, 1986) posited a causal relationship between the replacement of high-browsing (tree canopy) sauropods (Jurassic) by low-browsing ornithischians (Cretaceous) and the origin and early diversification of angiosperms. Bakker proposed that this ecological change favoured small herbaceous plants with short life cycles, a characteristic of the earliest flowering plants. Here, we review the evidence for the proposed faunal change and its temporal fit to the well-documented angiosperm diversification.

Late Jurassic herbivorous dinosaur faunas are dominated by sauropods and stegosaurs, whereas ornithopods are only locally abundant (e.g. Dryosaurus in the Kimmeridgian Tendaguru Formation of Tanzania) and ankylosaurs are exceptionally rare (Weishampel et al., 2004b). Earliest Cretaceous (Berriasian–Hauterivian) faunas are poorly understood, but there does appear to be a major ecological shift, with ankylosaurs and ornithopods (in particular) playing a much more significant role, both in terms of numbers of occurrences and numbers of genera (Fig. 1a,b; Weishampel et al., 2004b). Sauropods and stegosaurs are of relatively low abundance in contrast to their acmes in the Late Jurassic; recent work has confirmed the existence of a major extinction event amongst sauropods at the Jurassic/Cretaceous boundary with an estimated 57%–89% decrease in diversity between the Tithonian and the Berriasian (Upchurch & Barrett, 2005). A similar stegosaur extinction may also have occurred at the Jurassic/Cretaceous boundary (Galton & Upchurch, 2004). The faunal turnover identified by Bakker (1978), from sauropod/stegosaur-dominated faunas to faunas in which ornithopods are much more important, therefore occurred relatively abruptly at the Jurassic/Cretaceous boundary. However, as discussed by Barrett & Willis (2001), this faunal event cannot be characterized as a change from a high-browsing fauna to a low-browsing fauna, nor is the turnover as extreme as claimed by Bakker (1978). Middle and Late Jurassic faunas already included a high proportion of low browsers, including stegosaurs (Sereno, 1997; Barrett & Willis, 2001; Galton & Upchurch, 2004; contraBakker, 1978), locally abundant small- to medium-sized ornithopods (e.g. camptosaurids, dryosaurids, Othnielosaurus; Norman, 2004) and rare ankylosaurs and ceratopsians (Weishampel et al., 2004b; Xu et al., 2006), as well as diplodocid and dicraeosaurid sauropods (e.g. Stevens & Parrish, 1999; Upchurch & Barrett, 2000; Rauhut et al., 2005; Sereno et al., 2007). Moreover, our database demonstrates that the main dinosaur group characterized as high browsers, the sauropods, continued to form a significant component of global faunas through the Cretaceous, even if they were less important than during the Jurassic. Although the faunal turnover identified by Bakker (1978) is real, its consequences in terms of dinosaur browse height are less marked, undermining the ecological underpinning of the hypothesis.

For the dinosaur faunal turnover hypothesis to hold, the faunal change should coincide with the angiosperm origin. Furthermore, we would anticipate a positive correlation between increasing diversity of low browsers (e.g. ornithopods and ceratopsians) and the early radiation of angiosperms (i.e. we would expect evidence for co-radiations of these groups). Our data demonstrate the contrary. There is a major hiatus between the faunal turnover identified by Bakker and the earliest fossil evidence of angiosperms. The earliest evidence for angiosperms in dinosaur-bearing formations (macrofossils in late Barremian–Aptian deposits) post-dates the general background signal (monaperturate pollen of Hauterivian age: Friis et al., 2006). Dinosaur faunal turnover at the Jurassic/Cretaceous boundary therefore precedes the earliest recorded angiosperm pollen by 10 Myr and the earliest records of angiosperm macrofossils by nearly 20 Myr. This stratigraphic incongruence is inconsistent with a causal relationship (see also Barrett & Willis, 2001). However, calibrated molecular phylogenies have proposed earlier dates for angiosperm diversification and conservative estimates have placed the crown group radiation in the Late Jurassic (e.g. Wikström et al., 2001; Bremer et al., 2004; Sanderson et al., 2004). There is reason to be cautious of calibrated molecular phylogenies (see, for example Graur & Martin, 2004; Nixon, 2008), but even if the origins of the crown group pre-dates the earliest fossil evidence, the complete absence of fossils argues against a prominent role for angiosperms in pre-Hauterivian ecosystems. With the exception of the decline of stegosaurs (see below), there are no clear changes in dinosaur diversity patterns during the Hauterivian–Barremian that could be linked to the origin or early diversification of angiosperms. Our analysis therefore does not support the faunal turnover hypothesis.

The middle and Late Cretaceous radiation of angiosperms

As discussed above, angiosperms start to become significant members of floras in dinosaur-inhabited ecosystems during the middle Cretaceous (Albian–Cenomanian) (Crane, 1987; Heimhofer et al., 2005). This is a major floristic event, but our analysis shows that few significant changes in dinosaur herbivore faunas correlate temporally with the angiosperm radiation (Fig. 1a,b). It is possible that a decline in sauropod diversity occurred during or after the Cenomanian, although this is contradicted by some evidence (relative abundance of occurrences; Fig. 1b). Coria & Salgado (2005) linked this diversity decline to a drop in the abundance of nonangiosperm plants, which have often been identified as potential sauropod feed (e.g. Hummel et al., 2008); however, further work is required to determine whether this apparent decline is genuine and to assess the dietary preferences of sauropods in greater detail.

We have identified a significant negative correlation between the relative diversity of stegosaurs and that of angiosperms. This reflects the fact that stegosaurs decline through the Early Cretaceous as angiosperms diversified, with their final extinction apparently occurring near the end of the Early Cretaceous. However, stegosaurs had already suffered a major extinction at the end of the Jurassic (see above), with another decline occurring by the Barremian. The latter event is approximately coincident with angiosperm origin, but pre-dates the onset of the major angiosperm radiation in the Albian–Cenomanian. Notably, however, this reduction does correspond to the extinction of numerous cycadophytes (Fig. 2), and stegosaur diversity shows a significant positive correlation with both Cycadales and Bennettitales in terms of both relative numbers of occurrences and genera (Table 3). Stegosaurs have been proposed as dispersers of cycad seeds (e.g. Mustoe, 2007) and the correlated decline of cycadophytes and stegosaurs is intriguing and could potentially support a co-evolutionary relationship. However, spatiotemporal comparisons of stegosaur and cycad distributions have thus far failed to identify statistically significant associations (R. J. Butler, P. M. Barrett, P. Kenrick & M. G. Penn, unpublished data).

Table 3.   Results of statistical comparisons (using Spearman’s rank correlation coefficient and Kendall’s tau coefficient) of temporal variation during the Cretaceous in the relative number of occurrences within our database of stegosaurs and of cycads, Bennettitales and ‘cycadophytes’.
 Stegosauria – genera (Kendall’s)Stegosauria – genera (Spearman’s)
Bennettitales0.6736 (= 0.002; Ppermut = 0.001)*0.8298 (= 0.0008; Ppermut = 0.001)*
Cycadales0.5917 (= 0.007; Ppermut = 0.016)*0.6929 (= 0.012; Ppermut = 0.015)*
‘Cycadophytes’0.6736 (= 0.002; Ppermut = 0.005)*0.8298 (= 0.0008; Ppermut = 0.003)*
 Stegosauria – occurrences (Kendall’s)Stegosauria – occurrences (Spearman’s)
  1. ‘Cycadophytes’ includes cycads, Bennettitales and ‘cycadophyte’ material that cannot be assigned to either clade; *> 0.05.

Bennettitales0.5822 (= 0.008; Ppermut = 0.006)*0.7705 (= 0.003; Ppermut = 0.003)*
Cycadales0.6046 (= 0.0062; Ppermut = 0.011)*0.7305 (= 0.007; Ppermut = 0.007)*
‘Cycadophytes’0.5469 (= 0.013; Ppermut = 0.021)*0.7630 (= 0.004; Ppermut = 0.006)*

Ankylosaurs radiated during the middle Cretaceous, but this event pre-dates the major angiosperm radiation; moreover, relative diversity patterns for ankylosaurs and angiosperms do not show a significant correlation. As discussed above, it is possible that the ankylosaur radiation was causally linked with stegosaur extinction, but this requires further investigation.

Other major herbivore groups (ceratopsians, pachycephalosaurs and ornithopods) show little evidence of significant changes in relative diversity during the Albian–Cenomanian. Ceratopsians are of relatively low diversity throughout the Early and middle Cretaceous, and only show a significant radiation in the latest Cretaceous (see below), whereas pachycephalosaurs are essentially limited to latest Cretaceous ecosystems. There is a high diversity of iguanodontian ornithopods in middle Cretaceous ecosystems (Norman, 2004); however, iguanodontian origins date back to the Jurassic, and there is no evidence that ornithopods increased in relative diversity in terms of either genera or occurrences during the middle Cretaceous (Fig. 1a,b). The herbivorous theropods did undergo several radiations (in each clade) during the middle Cretaceous, which occurred in Laurasian ecosystems from the Barremian onward. As with ankylosaurs, the beginning of this radiation in the Barremian appears to pre-date the major angiosperm radiation. In summary, co-evolutionary hypotheses that link middle Cretaceous dinosaur faunal changes to the angiosperm radiation (i.e. hypotheses that suggest co-radiation events) cannot be supported by the available data on taxonomic diversity (see also Lloyd et al., 2008).

Ceratopsians, pachycephalosaurs, hadrosaurs and titanosaurian sauropods underwent significant radiations during the latest Cretaceous (Santonian–Maastrichtian) and formed greater proportions of herbivore faunas than previously observed (although some of these are geographically restricted). However, only pachycephalosaurs (when Stenopelix is excluded) show significant positive correlations with angiosperm diversity (see above). This positive correlation may be artefactual, stemming from the poor sampling of the plant record during the Turonian–Santonian, which results in an apparent, but probably unrealistic, second major angiosperm ‘radiation’ during the Campanian–Maastrichtian. Moreover, the radiations of these dinosaur clades during the Late Cretaceous significantly post-date the onset of the major angiosperm radiation in the Albian–Cenomanian. It is possible that these radiations are linked to the continuing angiosperm radiation; however, this is difficult to test with the available fossil data (Barrett & Willis, 2001). Other approaches, such as GIS spatiotemporal analysis (Butler et al., 2008) may provide insights into dinosaur/angiosperm interactions during this time period. At present therefore dinosaur–angiosperm co-evolutionary hypotheses for the latest Cretaceous remain plausible, but require further investigation.

Acknowledgments

This research was funded by the award of a NERC Standard Grant (NE/C002865/1) to PMB, PK and MGP.

Ancillary