Multifaceted disparity approach reveals dinosaur herbivory flourished before the end-Cretaceous mass extinction

Page created by Perry Yates
 
CONTINUE READING
Multifaceted disparity approach reveals dinosaur herbivory flourished before the end-Cretaceous mass extinction
Paleobiology, 44(4), 2018, pp. 620–637
     DOI: 10.1017/pab.2018.26

     Multifaceted disparity approach reveals dinosaur herbivory
     flourished before the end-Cretaceous mass extinction

     Klara K. Nordén, Thomas L. Stubbs, Albert Prieto-Márquez, and Michael J. Benton

     Abstract.—Understanding temporal patterns in biodiversity is an enduring question in paleontology.
     Compared with studies of taxonomic diversity, long-term perspectives on ecological diversity are rare,
     particularly in terrestrial systems. Yet ecological diversity is critical for the maintenance of biodiversity,
     especially during times of major perturbations. Here, we explore the ecological diversity of Cretaceous
     herbivorous dinosaurs leading up to the K-Pg extinction, using dental and jaw morphological disparity
     as a proxy. We test the hypothesis that a decline in ecological diversity could have facilitated their rapid
     extinction 66 Ma. We apply three disparity metrics that together capture different aspects of morpho-
     space occupation and show how this approach is key to understanding patterns of morphological evo-
     lution. We find no evidence of declining disparity in herbivorous dinosaurs as a whole—suggesting that
     dinosaur ecological diversity remained high during the last 10 Myr of their existence. Clades show
     different disparity trends through the Cretaceous, but none except sauropods exhibits a long-term
     decline. Herbivorous dinosaurs show two disparity peaks characterized by different processes; in the
     Early Cretaceous by expansion in morphospace and in the Campanian by morphospace packing. These
     trends were only revealed by using a combination of disparity metrics, demonstrating how this approach
     can offer novel insights into macroevolutionary processes underlying patterns of disparity and eco-
     logical diversity.

     Klara K. Nordén, Thomas L. Stubbs, and Michael J. Benton. School of Earth Sciences, University of Bristol, 24
        Tyndall Avenue, Bristol BS8 1TQ, U.K. *Present address: Department of Ecology and Evolutionary
        Biology, Princeton University, Princeton, New Jersey 08544, U.S.A. E-mail: knorden@princeton.edu.
     Albert Prieto-Márquez. Field Museum of Natural History, 1400 S. Lake Shore Drive, Chicago, Illinois 60605,
        USA. Present address: Institut Català de Paleontologia Miquel Crusafont, Universitat Autònoma de
        Barcelona, Carrer de l’Escola Industrial 23, 08201 Sabadell, Spain.

     Accepted: 26 June 2018
     Data available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.pp07dm0

                               Introduction                                     taxonomic diversity did not equal losses in
        Studies of biodiversity in deep time are often                          ecological diversity (Edie et al. 2018), support-
     focused on taxonomic diversity (Jackson and                                ing the notion that these processes can be
     Johnson 2001). In contrast, how taxonomic                                  decoupled. This is important, because ecologi-
     diversity relates to a second aspect of biodi-                             cal diversity may increase the robustness and
     versity—ecological diversity—is little under-                              resilience of ecosystems, even as species diver-
     stood (Swenson 2011). Ecological diversity is                              sity declines. Resilience is defined as the ability
     broadly defined as the diversity of ecotypes, or                            of an ecosystem to recover after major pertur-
     niches, occupied (Magurran 1988). It is likely                             bations (Elmqvist et al. 2003; Hooper et al.
     that species diversity is positively correlated                            2005). Ecosystems that have high resilience can
     with ecological diversity—but it is not obvious                            adapt better and regain species diversity that
     that this relationship will be linear. High                                was lost during times of environmental stress,
     diversity in species can be accommodated                                   while low resilience facilitates extinction cas-
     either through increased packing within occu-                              cades (Walker 1995; Elmqvist et al. 2003; Folke
     pied ecological space (niche partitioning) or                              et al. 2004; Lindegren et al. 2016). Ecological
     through expansion into new niches (niche                                   diversity is therefore likely to be important for
     expansion [MacArthur 1965]). A recent study                                both the maintenance and recovery of
     exploring both taxonomic and ecological diver-                             biodiversity.
     sity of bivalves during extinction events in the                              These topics are of great interest to paleon-
     Phanerozoic found that great losses in                                     tologists, yet studies with a long-term

      © 2018 The Paleontological Society. All rights reserved. This is an Open Access article, distributed under the terms of the
      Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted
Downloaded from https://www.cambridge.org/core. IP address: 176.9.8.24, on 11 May 2020 at 04:50:56, subject to the Cambridge Core terms of use, available at
      reuse, distribution, and reproduction in any medium, provided the original work is properly cited.
https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2018.26
                                                                                                                                     0094-8373/18
Multifaceted disparity approach reveals dinosaur herbivory flourished before the end-Cretaceous mass extinction
DINOSAUR HERBIVORY FLOURISHED BEFORE EXTINCTION                                                            621

        perspective on ecological diversity have been                                 To gain a better understanding of how
        rare, especially in terrestrial systems. While                             ecological diversity evolves, particularly in
        many studies have explored morphological                                   the face of environmental change and how this
        disparity patterns in clades (Erwin 2007;                                  may relate to mass extinctions, there is a need
        Hughes et al. 2013), only a subset has focused                             for a long-term and detailed study of ecological
        on ecologically important traits (such as teeth                            diversity in terrestrial systems. Such
        and jaws). Many studies also limit the infer-                              approaches have previously been successful
        ences that can be made by calculating only a                               in uncovering the evolutionary dynamics of
        single metric, usually variance-based disparity.                           clades through time in marine biotas (Foote
        This is unfortunate, as using several metrics                              1999; Korn et al. 2013; Hopkins and Smith 2015;
        describing different aspects of morphological                              Marx and Fordyce 2015; Stubbs and Benton
        disparity is key to identifying patterns in                                2016; Smithwick and Stubbs 2018). To this
        ecological diversity (Foote 1999; Bush and                                 purpose, we explore the ecological diversity of
        Novack-Gottshall 2012; Novack-Gottshall                                    herbivorous dinosaurs during the last 80 Myr
        2016a, b). Novack-Gottshall (2016a) developed                              of their reign, ending with the K-Pg mass
        a conceptual framework describing the dispar-                              extinction 66 Ma. We use dental and jaw
        ity trends we might expect during different                                morphological disparity as a proxy for diver-
        processes of “ecospace” utilization, including                             sity in feeding habits and, by extension,
        contraction (loss of ecological diversity),                                ecotypes. The rationale for this is that morpho-
        redundancy, and niche partitioning. A key                                  logical variation in the jaws and dentition is
        message is that different modes of evolution,                              expected to have a strong ecomorphological
        such as loss of morphospace occupation versus                              signal, with major adaptive innovations and
        packing, cannot be distinguished using only                                modifications being intrinsically related to diet
        metrics describing variance or mean distance,                              and feeding mode (Norman and Weishampel
        for example, sum of variances (SoV) and mean                               1985; Barrett and Rayfield 2006; Cantalapiedra
        pairwise distances (Fig. 1; Ciampaglio et al.                              et al. 2013; Button et al. 2014; Gill et al. 2014;
        2001; Novack-Gottshall 2016a). Packing refers                              Stubbs and Benton 2016).
        to how densely taxa occupy a given region of a                                The K-Pg mass extinction event was one of
        morphospace. Close packing may be caused by                                the most severe in the history of life, leading to
        redundancy in functional morphology, where                                 the complete extinction of nonavian dinosaurs,
        the same morphology is shared among many                                   among many other groups. There is now
        taxa (Novack-Gottshall 2016a; Pigot et al.                                 abundant evidence that an asteroid impact
        2016). Yet these alternatives will lead to                                 coincided with the extinction (Alvarez et al.
        radically different interpretations (extinction                            1980; Schulte et al. 2010), and it is also
        on the one hand and packing on the other).                                 concurrent with a large igneous event, the
                                                                                   eruption of the Deccan Traps (Renne et al.
                                                                                   2015). The mass extinction was preceded by a
                                                                                   period of extreme climate changes (Friedrich
         A                        B                       C
                                                                                   et al. 2012)—possibly weakening ecosystems
                                                                                   before the final extinction. The extinction of the
                                                                                   dinosaurs is therefore an excellent test case of
                                                                                   whether lowered ecosystem resilience can
                                                                                   catalyze major extinction events, as has been
        FIGURE 1. Different disparity metrics capture different                    suggested for other mass extinctions (Roop-
        aspects of disparity. A, Original morphospace occupation                   narine et al. 2007; Dick and Maxwell 2015). The
        (shaded). A decrease in distance-based metrics (mean                       lead-up to the extinction of the dinosaurs is of
        pairwise disparity [MPD] or sum of variances [SoV])
        could be either due to an increase in packing in                           added interest, because it has been suggested
        morphospace (B) or a decrease in the volume occupied in                    that dinosaurs showed a broad-scale global
        morphospace (C). Measuring morphovolume in addition                        decline in speciation rates through the last 40
        to distance-based metrics allows us to separate these
        events, as morphovolume will not be affected by                            Myr of the Cretaceous (Sakamoto et al. 2016),
        morphospace packing.                                                       challenging the general consensus that there

Downloaded from https://www.cambridge.org/core. IP address: 176.9.8.24, on 11 May 2020 at 04:50:56, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2018.26
Multifaceted disparity approach reveals dinosaur herbivory flourished before the end-Cretaceous mass extinction
622                                                    KLARA K. NORDÉN ET AL.

     was no decline in diversity in their last 10 Myr                           understanding of how the disparity of all
     (Fastovsky et al. 2004; Wang and Dodson 2006;                              herbivorous dinosaurs was changing during
     Barrett et al. 2009; Lloyd 2012; Mitchell et al.                           the Cretaceous. Conflicting results for the same
     2012; Brusatte et al. 2015; Starrfelt and Liow                             clades (hadrosaurs and theropods) suggest
     2016). Herbivorous dinosaurs are of particular                             that different approaches may yield different
     interest, as they were hugely abundant and                                 results (e.g., different sets of taxa, different sets
     positioned at the bottom of the vertebrate food                            of morphological characters, different dispar-
     chain, and therefore changes in their diversity                            ity metrics). Further, it is reasonable to believe
     and ecology had the potential to create cascad-                            that disparity will vary depending on which
     ing effects through the ecosystem (Pringle et al.                          part of the organism one measures, and
     2007; Vila et al. 2016). Moreover, herbivores are                          whether geometric morphometrics or discrete
     expected to encompass more species with                                    characters are used (Mongiardino Koch et al.
     dietary specializations than carnivores (Ber-                              2017). Time binning and the temporal scope of
     nays and Graham 1988; Codron et al. 2016), so                              the analysis also matter, with some analyses
     herbivorous dinosaurs should have been parti-                              having investigated only the latest Cretaceous
     cularly sensitive to environmental perturba-                               (Brusatte et al. 2012), and others the whole of
     tions (Davies et al. 2004; Clavel et al. 2011).                            the Mesozoic, but with larger time bins
        Several studies have investigated dinosaur                              (MacLaren et al. 2017). Timescales matter:
     disparity leading up to the K-Pg extinction,                               peaks and troughs might be part of long-term
     three of which exclusively focused on ecomor-                              fluctuations, which would be missed if the
     phology (Larson et al. 2016; Strickson et al.                              temporal focus is too narrow, while using large
     2016; MacLaren et al. 2017). Larson et al. (2016)                          time bins may obscure short-term events and
     used a geometric morphometric approach,                                    trends.
     finding that the dental morphospace occupa-                                    This study extends, and complements, ear-
     tion of different maniraptoran clades remained                             lier disparity studies on dinosaurs by exploring
     stable throughout the Santonian–Maastrich-                                 this theme using a larger data set, multiple
     tian interval. Strickson et al. (2016) used                                disparity metrics, smaller time bins, including
     discrete dental characters of ornithopods                                  the great majority of herbivorous dinosaur
     and found that their disparity peaked in                                   taxa, and investigating global, regional, and
     the Coniacian–Santonian, with a trough in the                              clade-specific trends. By doing so, we uncover
     Campanian, followed by an increase in the                                  new detail and previously unaddressed
     Maastrichtian. MacLaren et al. (2017) explored                             aspects of dinosaur disparity dynamics pre-
     disparity using geometric morphometrics and                                ceding their extinction and demonstrate how a
     biomechanical characters of lower jaws from                                multifaceted approach to disparity is critical to
     herbivorous dinosaurs and showed a steady                                  understand the dynamics of morphological
     increase in disparity throughout the Cretac-                               evolution.
     eous, with no decline in the Maastrichtian.
     Based on discrete characters from the whole
     skeleton, Brusatte et al. (2012) detected declines                                          Materials and Methods
     in the disparity of ceratopsians and hadrosau-
     roids, but not among smaller ornithopods,                                    Taxon Sampling.—A list of all herbivorous
     theropods, and sauropodomorphs, from the                                   dinosaur taxa (excluding theropods) during
     Campanian to the Maastrichtian. This result                                the Cretaceous period was created using the
     both contrasts with (Strickson et al. 2016) and                            Paleobiology Database (http://paleodb.org) in
     agrees with (Larson et al. 2016) previously                                conjunction with the literature. All valid
     mentioned studies.                                                         species, according to Weishampel et al. (2004),
        These recent studies have been valuable                                 plus new species described after that paper’s
     in exploring both patterns of dinosaurian                                  publication (referenced in Supplementary
     adaptation and different data and methods,                                 Data) with any cranial and/or dental material
     but it is hard to make direct comparisons                                  were included. After the character matrix was
     between studies and build an overarching                                   constructed, taxa with >70% missing data were

Downloaded from https://www.cambridge.org/core. IP address: 176.9.8.24, on 11 May 2020 at 04:50:56, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2018.26
Multifaceted disparity approach reveals dinosaur herbivory flourished before the end-Cretaceous mass extinction
DINOSAUR HERBIVORY FLOURISHED BEFORE EXTINCTION                                                            623

        excluded. This cutoff was chosen to exclude as                             variation in all groups included and to add
        few taxa as possible (26), yet remove extreme                              characters describing jaw morphology. We
        outliers. This was checked visually by                                     used photographs and descriptions of
        inspecting morphospace plots for different                                 specimens in the literature for scoring of
        cutoff values. The final matrix used for                                    characters      (listed   and     referenced    in
        analysis consisted of 194 species, with 34%                                Supplementary Data 2).
        missing data overall (Supplementary Data 2).                                  Timescale and Time Bins.—The Cretaceous
           We excluded herbivorous theropods for                                   timescale used in this study is based on
        four reasons. (1) There is some debate about                               Gradstein et al. (2012) and the updates in the
        which theropods were herbivorous and which                                 2016 International Chronostratigraphic Chart.
        might have been omnivorous. (2) Character                                  Using this timescale, dinosaur-bearing
        choice: our characters are focused on tooth                                geological formations were given start and
        morphology, which would be a poor estimate                                 end dates based on Weishampel et al. (2004) or,
        of beak disparity. Adjusting the characters to                             if available, more recent publications
        accommodate beak morphologies would have                                   (Supplementary Data 2). The first and last
        drastically increased missing data in the data                             appearance dates of each taxon were then
        set, both from nonapplicable dental characters                             assigned based on these formation ages. We
        in beaked taxa and the addition of nonapplic-                              conducted a sensitivity test to evaluate the
        able beak-specific characters in toothed taxa.                              effect of different time-bin divisions (geological
        High percentages of missing data affect dis-                               substages, geological stages and equal-length
        parity analyses negatively; therefore, to main-                            time bins of 10, 5, and 3 Myr; Supplementary
        tain good quality we would need to cut a larger                            Figs. 1, 2). By evaluating the effects of these
        proportion of taxa, ultimately reducing the                                divisions, we constructed a modified series in
        sample size of the study. (3) Rarity: herbivor-                            which long time bins were divided into
        ous theropods constitute ~5% of herbivorous                                substages (Albian and Campanian) and two
        dinosaur species diversity in the Cretaceous                               short bins with particularly few taxa, Coniacian
        (Butler and Barrett 2008; Butler et al. 2009).                             and Turonian, were combined. This division
        (4) Minimal impact: all key herbivorous ther-                              allowed for time bins that were most equal in
        opod clades appear in the Early Cretaceous                                 sample size and duration (∼5 Myr), without
        (therizinosaurs, ornithomimids, and ovir-                                  either increasing error bars dramatically or
        aptorids) and continue through the Cretac-                                 losing temporal information in the data.
        eous, so if they were included, they would                                    Disparity Analyses.—All analyses were
        likely increase morphospace occupation, and                                conducted in R v. 3.2.2. (R Core Team 2015).
        increase expansion in the Early Cretaceous, but                            For constructing a distance matrix, we
        not change the main trends during the Cretac-                              calculated generalized Euclidean distances
        eous. Other methodological approaches would                                (GED) (Wills 1998, 2001) and maximum
        be more suitable for including beaked forms,                               observable distances (MORD) using the
        such as geometric morphometrics measuring                                  package Claddis (Lloyd 2016). GED has the
        the shape of the jaw (although theropods were                              benefit of always returning a complete matrix
        also excluded from the largest such study, by                              (required for ordination) by filling in a mean
        MacLaren et al. [2017] on the basis of lack of                             distance (based on calculable distances) into
        complete specimens).                                                       missing fields. MORD rescales the calculable
           Morphological Characters and Scoring.—                                  distances based on the information available,
        Eighty-eight      morphological     characters,                            by dividing each distance by the maximum
        encompassing dental and jaw morphology,                                    distance calculated from the observed
        were used in the analysis (Supplementary                                   characters (Lloyd 2016). Using both distance
        Data 1). Forty-two of the dental characters                                matrices, we then calculated preordination
        were adopted from Strickson et al. (2016), with                            (weighted mean pairwise disparity [WMPD])
        minor adjustments to the scoring of                                        and postordination (SoV, principal coordinates
        ornithopods. Another 46 were collected from                                analysis [PCO] convex-hull volume) disparity
        sources chosen to cover morphological                                      metrics.

Downloaded from https://www.cambridge.org/core. IP address: 176.9.8.24, on 11 May 2020 at 04:50:56, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2018.26
Multifaceted disparity approach reveals dinosaur herbivory flourished before the end-Cretaceous mass extinction
624                                                    KLARA K. NORDÉN ET AL.

        Mean pairwise disparity measures how                                    used as the cutoff point. A bootstrap method
     similar a given number of taxa are, that is, how                           (1000 replicates) was applied to estimate
     closely packed they are in morphospace. The                                90% confidence intervals. Following Kotrc
     WMPD is different, in that it gives more weight                            and Knoll (2015a), we also quantified the
     to pairwise distances based on more compar-                                strength of association between all discrete
     able characters (Close et al. 2015). To test for                           characters and the PCO axes using Cramér’s V
     effects of sample size and produce 90% con-                                coefficient (Supplementary Note 1, Supple-
     fidence intervals, we applied a bootstrap ana-                              mentary Fig. 4).
     lysis (1000 replicates with replacement).                                     We quantified the total amount of morpho-
     Confidence intervals were based on percentiles                              space occupied through time by measuring the
     of the bootstrap distribution.                                             PCO convex-hull volume. PCO volume based
        For postordination metrics (SoV, PCO                                    on a convex hull is sensitive to outliers (Kotrc
     volume) we applied a PCO to the GED and                                    and Knoll 2015b), which can create large peaks
     MORD distance matrices. A common problem                                   in volume from single data points. This can be
     when PCO is applied to a distance matrix                                   adjusted by changing the alpha shape para-
     based on discrete characters is the creation of                            meter, which allows large empty spaces (with
     negative eigenvalues. Negative eigenvalues in                              few data points) to be removed. The alpha
     a PCO can be corrected by adding a constant of                             shape parameter regulates the radius by which
     equal size to the largest negative eigenvalue to                           such spaces are “scooped out.” We calculated
     all distances (e.g., Cailliez correction; Cailliez                         volumes for three dimensions (PC 1–3) using
     1983). We explored how well “corrected” and                                different values of alpha. The range of values to
     “uncorrected” scores represent the original                                be tested was determined by visually inspect-
     distance matrix, and how this affected the SoV                             ing morphospace plots for different alpha
     statistic (Supplementary Note 2, Supplemen-                                values. We also calculated raw morphospace
     tary Figs. 3, 6). We note that the variation                               volumes. For volumes and alpha shapes, we
     explained by PC 1 and PC 2 is notably higher                               calculated 90% confidence intervals based on
     when no correction is used. Negative eigenva-                              percentiles of a bootstrap distribution (1000
     lues need not be a problem for the representa-                             replicates).
     tion of the first few PC axes if the greatest                                  To measure the amount of morphospace
     absolute negative value is smaller than the                                packing and expansion for each time bin, we
     greatest positive value (Legendre and                                      used the niche packing “flexible” metric
     Legendre 1998). This condition is met in our                               (NP flexible) created by Pigot et al. (2016). This
     data set. We discuss the fit of our ordination                              metric is calculated by estimating the number
     data to the original distances in the Supple-                              of species in a volume A1 that can be packed
     mentary Material. In summary, we found a                                   within another volume A2. A1 and A2 are
     high correlation (95%) between distances after                             sequential time bins, so, for example, to
     ordination axes and original distances when no                             calculate the percentage packing in the Cam-
     correction was applied. This suggests that                                 panian, A1 is the Campanian and A2 the
     results from SoV and WMPD should be                                        Santonian. A greedy algorithm sequentially
     comparable.                                                                removes species from A1, each time choosing
        The SoV statistic measures the spread of                                the data point that incurs maximum loss of
     taxa in morphospace. In this aspect, it is like                            morphovolume. Once the modified A1
     WMPD. SoV can be measured for any number                                   volume reaches the same volume as A2, the
     of axes, and how many are included varies                                  taxa remaining in A1 are considered to con-
     between studies. We included 30 axes based on                              tribute to packing, and the removed taxa are
     the R2 values of the correlation between dis-                              considered to expand trait space. In this way,
     tances in the PCO space with the original dis-                             percentages for expansion and packing can be
     tances in the GED or MORD matrices                                         calculated. The benefit of this algorithm is that
     (Supplementary Fig. 3). At the point where the                             it compares packing in the absolute volumes
     curve plateaus, additional axes do not add to                              of A1 and A2, irrespective of their positions in
     the variation explained, and this can hence be                             morphospace.

Downloaded from https://www.cambridge.org/core. IP address: 176.9.8.24, on 11 May 2020 at 04:50:56, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2018.26
DINOSAUR HERBIVORY FLOURISHED BEFORE EXTINCTION                                                            625

                                      Results                                      unchanged when using two approaches for
                                                                                   calculating morphological distances from the
           Distance-based Disparity Trends.—Distance-                              discrete character data, the GED (Fig. 2D,E)
        based metrics (WMPD and SoV) highlight                                     and MORD (Fig. 2A,B). There is a major
        three key trends in herbivorous dinosaur                                   discrepancy between the two distance
        disparity, revealing an Aptian peak, a                                     measures in the time from the Albian to the
        Campanian low, and a stable/increasing                                     Turonian–Coniacian, when the MORD metric
        trend in the Maastrichtian (Fig. 2A,B,D,E).                                records a sharp rise, but the GED metric
        As expected, both distance-based metrics,                                  suggests unchanged or decreasing disparity.
        within-bin WMPD (Fig. 2A,D) and the SoV                                    The Aptian peak is not pronounced using the
        (Fig. 2B,E), show almost identical trends. This                            MORD metric, but it does appear using smaller
        confirms that our ordination has not                                        time bins (Supplementary Figs. 1, 2). Apart
        distorted the original distances in any                                    from these key features, distance-based
        major way (we explore this more in depth in                                disparity remains relatively stable during the
        the Supplementary Material). They are also                                 Cretaceous.

                       A                                                                       D

                       B                                                                       E

                       C                                                                       F

        FIGURE 2. Disparity of herbivorous dinosaur teeth and jaws through the Cretaceous. A, Weighted mean pairwise
        disparity (WMPD), maximum observable distances (MORD)-based disparity; B, sum of variances (SoV), MORD-based
        disparity; C, morphovolume (principal coordinates analysis [PCO] convex-hull volume) based on MORD: upper graph
        raw volumes, lower graph alpha shapes; D, WMPD, generalized Euclidean distances (GED)-based disparity; E, SoV,
        GED-based disparity; F, morphovolume based on GED: upper graph raw volumes, lower graph alpha shapes. Shaded
        envelopes represent 90% confidence intervals based on 1000 bootstrap replicates. Stratigraphic abbreviations: Be,
        Berriasian; V, Valanginian; H, Hauterivian; B, Barremian; Ap, Aptian; Al, Albian; Ce, Cenomanian; T, Turonian; C,
        Coniacian; S, Santonian; Cm, Campanian; Ma, Maastrichtian.

Downloaded from https://www.cambridge.org/core. IP address: 176.9.8.24, on 11 May 2020 at 04:50:56, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2018.26
626                                                    KLARA K. NORDÉN ET AL.

     FIGURE 3. Jaw and dental morphospace occupation for groups of herbivorous dinosaurs. PC 1–PC 2 explains 32% of
     variance, and the correlation between pairwise distances from the first two PC axes and squared pairwise distances
     from the original matrices is 0.86 (Supplementary Note 1). A, Total morphospace occupation with convex hulls drawn
     around the major clades. B, Temporal trends in morphospace occupation through the Cretaceous. Both plots are based
     on the generalized Euclidean distances metric, and expanding taxa (shaded in B) are based on a two-dimensional
     packing analysis. Stratigraphic abbreviations: Ber, Berriasian; Val, Valanginian; Hau, Hauterivian; Bar, Barremian; Apt,
     Aptian; Alb, Albian; Cen, Cenomanian; Tur, Turonian; Con, Coniacian; San, Santonian; Cmp, Campanian; Maa,
     Maastrichtian.
        Silhouettes from PhyloPic.org: upper left image by Mathew Wedel, under Creative Commons Attribution 3.0
     unported license (https://creativecommons.org/licenses/by/3.0/); upper middle image by Raven Amos, under
     Creative Commons Attribution 3.0 unported license (https://creativecommons.org/licenses/by/3.0/); upper right
     image and lower image licensed under Public Domain Dedication 1.0 license (https://creativecommons.org/
     publicdomain/zero/1.0/).

        Morphospace Occupation.—Phenotypic similar-                             in morphospace (Supplementary Fig. 3A).
     ity between taxa can be visualized in a                                    Herbivorous dinosaur clades generally form
     morphospace based on the jaw and dental                                    clusters within the morphospace, with some
     character data (Fig. 3). The first two axes                                 groups occupying distinct regions, such as
     represent 32% of the total variation, high for a                           hadrosauroids and the highly divergent
     morphospace built on a discrete character data                             sauropods (Fig. 3). The separation between
     set (e.g., Brusatte et al. 2012; Halliday and                              clades in morphospace is significant between
     Goswami 2016), but low compared with a                                     all    groups     except   non-hadrosauroid
     landmark-based data set. However, a strong                                 ornithopods and neoceratopsians, using
     correlation (R = 0.86) between PC scores on the                            nonparametric multivariate analysis of
     first and second axes and original distances in                             variance    (NPMANOVA)        for   pairwise
     the distance matrix confirms that this is a good                            comparison (Supplementary Table 1). We
     representation of the relative placement of taxa                           note, however, that NPMANOVA can be

Downloaded from https://www.cambridge.org/core. IP address: 176.9.8.24, on 11 May 2020 at 04:50:56, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2018.26
DINOSAUR HERBIVORY FLOURISHED BEFORE EXTINCTION                                                            627

        misleading when dispersion differs greatly in                              necessity increase the density of taxa). If
        groups: a significant result can be confounded                              sample size does not increase, species can still
        by difference in spread, rather than solely                                expand, or lack expansion, but the latter does
        by difference in position (Anderson 2001).                                 not exemplify packing. Our morphospace
        In PC 1–PC 2 morphospace, ornithischians                                   packing analysis shows that the Aptian and
        with less derived teeth, such as early                                     Campanian peaks in morphovolume are
        ornithopods, ankylosaurs, psittacosaurids,                                 characterized       by     different   processes.
        and pachycephalosaurids, converge, forming                                 Morphospace packing was elevated in the
        a group. More derived ceratopsians and                                     Campanian, while the Aptian exhibits a
        ornithopods diverge and approach the area                                  combination of both packing and expansion
        occupied by hadrosauroids. Statistical tests,                              (Fig. 4). Packing is high in the Campanian, both
        based on Cramér’s V coefficient, confirm that                                in absolute numbers and as a percentage
        most characters contribute to the positioning of                           compared with expansion. Apart from the
        taxa in morphospace, and there is no bias for                              Aptian, expansion is high in the Valanginian,
        certain individual characters or categories of                             Barremian, and Santonian. The Hauterivian,
        characters (e.g., dentary and maxillary teeth or                           Albian, Cenomanian, Turonian–Coniacian,
        jaw-related features; Supplementary Fig. 4).                               and Maastrichtian all show drops in sample
           Morphovolume          Disparity     Trends.—                            size and a lack of morphospace expansion.
        Morphovolume (PCO convex-hull volume)                                         Regional Trends.—Our data set allows
        values during the Cretaceous show both                                     division between the most well sampled
        similarities and marked differences compared                               geographical regions, the Eurasian and North
        with distance-based metrics (Fig. 2C–F). The                               American continents (Fig. 5). The distance-
        volume of morphospace occupied is high in the                              based metrics for Eurasia lack the Aptian high
        Aptian, in agreement with distance-based                                   disparity but show the same low Campanian
        metrics, but it is also high in the Campanian,                             disparity. The North American trend shows
        when distance-based metrics give low values.                               both high disparity in the Aptian and low
        These patterns are unchanged using different                               disparity in the Campanian, in agreement
        distance metrics. Morphospace volumes are                                  with the global trend. In contrast to the
        sensitive to outliers, and we have accounted for                           global trend, however, there is an increasing
        this by changing the alpha shape parameter,                                trend from the Albian to the Turonian–
        which removes large areas of empty space, and                              Coniacian.      Morphovolumes         for    both
        by applying a bootstrap method. Alpha                                      continents retain the Aptian–Albian and
        volumes clearly reduce the variability in                                  Campanian peaks in disparity, mirroring the
        calculated volumes, suggesting this approach                               global trend (similar trends are recorded for
        is effective in removing the impact of outliers.                           MORD-based distances; see Supplementary
        Even with confidence intervals and for low                                  Fig. 8).
        alpha values, the peaks in the Aptian–Albian                                  Clade-Specific Trends.—Herbivorous dinosaur
        and Campanian remain (Fig. 2C–F), suggesting                               clades show distinct disparity dynamics
        they are not caused by outliers. Patterns are                              (Fig. 6). Both distance metrics generally show
        similar for both distance metrics, but the                                 identical patterns; therefore, only results based
        MORD metric shows high variability in the                                  on the GED distance metric are discussed here
        Turonian–Coniacian bin, probably reflecting                                 (see Supplementary Information for MORD
        the peak recorded by distance-based metrics.                               analyses). Neoceratopsians exhibit a decrease
        However, this increase does not persist at                                 in distance-based disparity in the Santonian
        decreasing alpha values. Similar patterns are                              and then increase throughout the rest of
        seen when six morphospace axes are included                                the Late Cretaceous, with morphovolumes
        (hypervolumes; Supplementary Fig. 5).                                      increasing in tandem in the Campanian, then
           Morphospace Packing and Expansion.—We                                   decreasing in the Maastrichtian (Fig. 6A).
        define morphospace packing as an increase in                                Hadrosauroids dramatically decrease in
        species within the same volume as recorded                                 distance-based disparity in the middle
        from the previous time bin (which will by                                  Campanian, and then recover previous levels

Downloaded from https://www.cambridge.org/core. IP address: 176.9.8.24, on 11 May 2020 at 04:50:56, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2018.26
628                                                    KLARA K. NORDÉN ET AL.

     FIGURE 4. Morphospace packing through time. Each value is calculated as the volume in the younger bin
     accommodated in previous bin. A, Proportion of unique taxa contributing to packing/expansion for two dimensions;
     B, number of unique taxa contributing to packing/expansion for two dimensions. “No packing/expansion” refers to
     bins where sample size decreases, therefore they do not constitute packing but a lack of expansion in morphospace. All
     volume estimates are based on the generalized Euclidean distances metric. Stratigraphic abbreviations as in Fig. 2.

                   A                                                                        D

                  B                                                                         E

                  C                                                                         F

     FIGURE 5. Regional disparity trends. Disparity trends for Eurasian taxa: A, Weighted mean pairwise disparity
     (WMPD), maximum observable distances (MORD) based; B, WMPD, generalized Euclidean distances (GED) based;
     C, principal coordinates analysis (PCO) volume, GED based. Disparity trends for North American taxa: D, WMPD,
     MORD based; E, WMPD, GED based; F, PCO volume, GED based. Shaded envelopes represent 90% confidence
     intervals based on 1000 bootstrap replicates. Stratigraphic abbreviations as in Fig. 2.

Downloaded from https://www.cambridge.org/core. IP address: 176.9.8.24, on 11 May 2020 at 04:50:56, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2018.26
DINOSAUR HERBIVORY FLOURISHED BEFORE EXTINCTION                                                            629

                      A                                                                        E

                      B                                                                        F

                      C                                                                        G

                      D

        FIGURE 6. Clade-specific disparity trends. Trends in principal coordinates analysis (PCO) volume (darker shading) and
        weighted mean pairwise disparity (WMPD) (lighter shading) through the Cretaceous for: A, neoceratopsians;
        B, hadrosauroids; C, non-hadrosauroid ornithopods; D, pachycephalosaurids; E, sauropods; F, ankylosaurids;
        G, psittacosaurids. All values based on the generalized Euclidean distances metric (see Supplementary Information for
        maximum observable distances–based disparity). Shaded envelopes represent 90% confidence intervals based on 1000
        bootstrap replicates. Stratigraphic abbreviations as in Fig. 2. Silhouettes from PhyloPic.org: A, By Raven Amos, under
        Creative Commons Attribution 3.0 unported license (https://creativecommons.org/licenses/by/3.0/); B and E, under
        Public Domain Dedication 1.0 license (https://creativecommons.org/publicdomain/zero/1.0/); C, by Jaime Headden,
        under Creative Commons Attribution 3.0 unported license (https://creativecommons.org/licenses/by/3.0/); D, by
        Emily Willoughby, under Creative Commons Attribution-ShareAlike 3.0 unported license (https://creativecommons.
        org/licenses/by-sa/3.0/); F, by Scott Hartman, under Creative Commons Attribution 3.0 unported license (https://
        creativecommons.org/licenses/by/3.0/); G, by Pete Buchholz, licensed under Creative Commons Attribution-
        ShareAlike 3.0 unported license (https://creativecommons.org/licenses/by-sa/3.0/).

        of disparity in the Maastrichtian (Fig. 6B).                               remain on a stable level of disparity through the
        Hadrosauroid morphovolumes, in contrast,                                   Aptian–Albian and Maastrichtian (Fig. 6C).
        increase through the Campanian, peaking in                                 High morphovolume is recorded in the Aptian.
        the late Campanian, with a decrease in the                                 Pachycephalosaurids show stable distance-
        Maastrichtian. Non-hadrosauroid ornithopods                                based disparity in the Campanian and
        increase in disparity in the Barremian, then                               Maastrichtian and increasing morphovolumes

Downloaded from https://www.cambridge.org/core. IP address: 176.9.8.24, on 11 May 2020 at 04:50:56, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2018.26
630                                                    KLARA K. NORDÉN ET AL.

     during the same interval (Fig. 6D). Sauropods                                 Morphospace saturation coupled with high
     exhibit high distance-based disparity values in                            morphovolume in the Campanian indicates
     the Aptian and middle to late Albian, which                                that this was not a time when dinosaurs were
     then plummet toward the Maastrichtian                                      declining in morphological disparity. On the
     (Fig. 6E). Similarly, sauropod morphovolumes                               contrary, dinosaurs were flourishing, with the
     peak in the Aptian, then decrease through the                              Late Cretaceous seeing the radiation of suc-
     rest of the Cretaceous. Ankylosaurs show the                               cessful groups such as hadrosaurids and neo-
     lowest levels of distance-based disparity in the                           ceratopsians. This exemplifies how low
     Santonian, compared with their relatively stable                           disparity values from distance-based metrics
     and higher disparity in the Early Cretaceous, but                          can be caused by increased packing and high
     show an increase in the Maastrichtian (Fig. 6F).                           morphodensity, rather than a loss of morpho-
     Ankylosaur morphovolume also remains stable                                volume (Novack-Gottshall 2016a). New spe-
     through the Cretaceous, with a slightly                                    cies filled in and saturated the morphospace
     increasing trend toward the late Campanian.                                already occupied, a process that was particu-
     Finally, psittacosaurids record the highest levels                         larly seen in hadrosauroids, which became
     of distance-based disparity in the Hauterivian,                            very diverse and abundant in the latest Cre-
     a decrease in the Aptian, but recovery in                                  taceous but did not show wide variation in jaw
     the Albian, albeit with slightly lower than                                and dental morphology (Strickson et al. 2016).
     Hauterivian values (Fig. 6G). In contrast,                                 Very low distance-based disparity coupled
     psittacosaurid morphovolume peaks in the                                   with high morphovolumes in the hadrosaur-
     Aptian.                                                                    oids confirms this (Fig. 6B).
                                                                                   The decreased morphodensity and mor-
                                                                                phovolume in the Maastrichtian represents an
                                Discussion                                      effect opposite to what happened in the Cam-
                                                                                panian: a thinning in morphospace. Whether
        Disparity Trends in the Cretaceous.—We find                              this is a true trend or an edge effect is unclear.
     three key results: (1) an Aptian peak in                                   Either way, our results do not indicate any
     disparity characterized by expansion and                                   long-term decline in dinosaur disparity pre-
     packing; (2) a second disparity peak in                                    ceding the end-Cretaceous extinction, in
     the Campanian characterized by intense                                     agreement with previous disparity studies
     morphospace packing; and (3) a thinning out                                (Brusatte et al. 2012; Larson et al. 2016; Strick-
     of morphospace in the Maastrichtian. These                                 son et al. 2016; MacLaren et al. 2017).
     results paint a picture of a largely resilient                                An exception can be seen in sauropods,
     dinosaurian ecosystem that was able to adapt                               which appear to have truly been in long-term
     and flourish in the face of climate-driven                                  decline. A combined decrease in distance-
     environmental changes.                                                     based disparity and volume suggests declin-
        Morphospace expansion in the Aptian sug-                                ing morphological diversity (as opposed to
     gests that this was a time of diversification of                            morphospace packing) (Fig. 6E), in agreement
     dental and jaw morphologies, particularly                                  with the known disappearance of broad-
     among sauropods and early hadrosauroids                                    crowned sauropods in the Late Cretaceous
     (Fig. 3). This includes Nigersaurus taqueti,                               (Chure et al. 2010). Sauropods, large herbivores
     which is the only sauropod known to have                                   with slow generation times, might have been
     evolved a specialized “dental battery” (Sereno                             more sensitive to extreme environmental
     and Wilson 2005), and early hadrosauroids                                  changes during the mid-Cretaceous (Gaston
     (Altirhinus kurzanovi, Jinzhousaurus yangi) with                           and Blackburn 1995; Purvis et al. 2000; Cardillo
     primitive versions of dental batteries. Dental                             et al. 2005).
     batteries would later become more advanced                                    In sum, our disparity results show a
     in hadrosaurids and were likely a key factor in                            dynamic pattern of waxing and waning of
     the success of this group in the Late Cretaceous                           clades—sauropods flourishing in the Aptian–
     (Norman and Weishampel 1985; Strickson                                     Albian then diminishing in the Late Cretac-
     et al. 2016).                                                              eous, while other clades rose to prominence

Downloaded from https://www.cambridge.org/core. IP address: 176.9.8.24, on 11 May 2020 at 04:50:56, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2018.26
DINOSAUR HERBIVORY FLOURISHED BEFORE EXTINCTION                                                            631

        (e.g., hadrosaurids, ceratopsians). Collectively,                             Differences to Previous Analyses.—Our
        herbivorous dinosaurian ecosystems show no                                 analyses reveal differences in temporal,
        signs of “weakening” preceding the Cretac-                                 regional, and clade-specific trends compared
        eous mass extinction.                                                      with earlier studies. MacLaren et al. (2017) did
           Contrasting Diversity and Disparity in                                  not identify a disparity peak in the Aptian, as
        Hadrosauroids.—Empirical studies suggest                                   in this study, but instead their distance-based
        that ecosystem species richness is positively                              disparity value rises steadily throughout the
        correlated with specialization, suggesting that                            Cretaceous. Neither did they recover a trough
        ecosystems with high species diversity mediate                             in the Campanian. Their study used 10 Myr
        the increase in species numbers mainly                                     time bins and a smaller sample, which might
        through niche partitioning rather than niche                               be responsible for the discrepancy. Large time
        expansion (Case 1981; Belmaker et al. 2012;                                bins can alter disparity curves substantially, as
        Pigot et al. 2016; Pellissier et al. 2018). Our                            is evident from adjusting time bins for our
        results show that a similar pattern might be                               own data, and this effect has also been
        true for one of the most species-rich groups of                            demonstrated previously (Deline and Ausich
        dinosaurs, the hadrosauroids. Once advanced                                2011). Applying 10 Myr bins, both distance-
        dental batteries had evolved in early                                      based metrics show a steadily decreasing
        hadrosauroids, these structures required little                            trend starting in the Early Cretaceous
        change to ensure huge success for the                                      (Supplementary Figs. 1, 2). Another reason
        clade. Our measures are currently limited to                               might be the difference in method and
        estimating morphospace packing, but an                                     morphology measured, as MacLaren et al.
        interesting further avenue of research is to test                          (2017) used geometric morphometrics and
        whether the taxa filling morphospace are                                    functional traits relating to lower jaw shape,
        doing so in an even manner or are more                                     while our study used discrete characters with a
        densely populating certain areas. Novack-                                  focus on both dental and jaw morphology.
        Gottshall (2016a) offers a framework to test                               Brusatte et al. (2012) found global decreases in
        these hypotheses, by measuring the degree of                               hadrosaur, neoceratopsian, pachycephalosaur,
        redundancy versus partitioning. Redundancy                                 and ankylosaur disparity from the Campanian
        refers to the occupancy of niches identical to                             to the Maastrichtian, and an increase in
        those previously occupied by existing species.                             sauropod disparity. This is the exact opposite
           Though suggestive, a high degree of redun-                              of the results reported here: sauropods show a
        dancy/partitioning in our morphospace does                                 decrease and all other groups record an
        not necessarily equal niche partitioning. “Niche                           increase in distance-based disparity from the
        space” has many more dimensions than dif-                                  Campanian to the Maastrichtian (Fig. 6). This is
        fering feeding modes. For example, differences                             intriguing and suggests that dental and jaw
        in body size incur differences in predation                                disparity, linked more tightly to feeding
        pressures (with higher predation pressures on                              ecology, might show a different pattern from
        small animals). In modern African savanna                                  the disparity of the entire skeleton. Further
        ecosystems, such differences can lead to dif-                              studies should explore the disparity of the
        fering habitat preference (open/vegetated)                                 entire skeleton, to clarify whether and how
        and, consequently, differing resource use                                  overall disparity dynamics differ from those of
        among herbivores (Riginos and Grace 2008). If                              dental and jaw morphology alone. Also in
        hadrosauroids are disproportionally spread                                 contrast to Brusatte et al. (2012), our main
        out along some other axis of niche space, it is                            trends are shared on the Eurasian and North
        possible that they might not have exhibited                                American continents (Fig. 5). This further
        greater niche partitioning than other groups. A                            strengthens the argument that herbivorous
        more detailed analysis should also include a                               dinosaurs worldwide were not experiencing
        spatial dimension, as hadrosaurs on different                              long-term declines (Brusatte et al. 2015) and
        continents could inhabit similar niches without                            suggests that the trends recovered herein are
        competing.                                                                 not restricted to regional patterns.

Downloaded from https://www.cambridge.org/core. IP address: 176.9.8.24, on 11 May 2020 at 04:50:56, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2018.26
632                                                    KLARA K. NORDÉN ET AL.

        The overall message of this study is different                          remove large peaks caused by outliers. We also
     from that of Sakamoto et al. (2016), who found                             use two dissimilarity metrics (MORD and
     declines in speciation rates, but the results need                         GED) to investigate how robust our results
     not be contradictory. Sakamoto et al. (2016) did                           are to different methodological approaches.
     find increasing speciation rates in neocera-                                   How effective are these methods for identi-
     topsians and hadrosaurids, indicating radia-                               fying spurious peaks? The Turonian–Con-
     tions of these groups, which agrees with the                               iacian peak using the MORD metric is
     morphospace packing results presented here.                                significantly higher than any other point.
     Furthermore, we present data suggesting                                    However, it is not recorded by the GED metric.
     sauropods were the only group declining in                                 This suggests that this peak in disparity is not a
     disparity, apparently coinciding with declining                            stable result, as it is not reproducible using two
     speciation rates. This is not to say that specia-                          alternative metrics. Similarly, the same Tur-
     tion rates and disparity must be coupled, but it                           onian–Coniacian peak does not persist at low
     is interesting to note that in the case of non-                            alpha levels in the volume estimates. The
     avian dinosaurs, both trends seem to co-occur,                             sample size in this bin is low, and upon
     with reduced speciation rates matched by low                               inspection of morphospace occupation
     morphological diversity in sauropods, and                                  through time, it appears that the MORD metric
     high speciation rates matched by high volume-                              estimates higher disparity between neocera-
     based disparity (but low distance-based dis-                               topsians, sauropods, and hadrosauroids than
     parity due to packing) in hadrosauroids.                                   does the GED metric. This suggests that the
     Therefore, the overall picture of dinosaur evo-                            Turonian–Coniacian peak is caused by a few
     lutionary dynamics preceding the extinction is                             very disparate taxa. Large areas of empty
     one of faunal turnover, with some clades                                   morphospace are often a sign of bias caused by
     prospering and others declining, rather than a                             missing data, and the ability of the MORD
     collective decline. In addition, the focus here on                         metric to handle missing data that are not
     morphospace volume and packing provides                                    randomly distributed (Lloyd 2016) should be
     different insights than a study of coupled spe-                            investigated. The implications of this are that
     ciation and lineage extinction rates.                                      we should be cautious to interpret the Tur-
        Potential Biases.—When using fossil data, the                           onian–Coniacian peak (and trough) as a real
     relative proportion of bias and true signal can                            phenomenon—it is probably a result of poor
     be difficult to estimate. Contrary to MacLaren                              sampling in this bin (Supplementary Fig. 7).
     et al. (2017), a decrease in disparity coinciding                          However, the Aptian and Campanian patterns,
     with an increase in sample size does not prove                             which we have focused on as our main result,
     that disparity measures are not biased by                                  persist with MORD and GED metrics, both for
     sample size. A decrease in distance-based                                  low alpha values and with the bootstrap con-
     disparity could be due to either reduced                                   fidence interval. Our bootstrap and alpha
     morphospace occupation or extensive packing                                shape analysis do reveal, however, that the
     (high morphodensity) (Fig. 1). Packing in                                  very high volume in the latest Campanian is
     morphospace is likely to be higher when a                                  unstable. We cannot be certain that the Cam-
     larger sample is included, and thus increased                              panian had a higher morphovolume than the
     sample size can also cause drops in disparity.                             Aptian—but high volumes in the Aptian and
     Here, we apply a bootstrap method to simulate                              the Campanian, as compared with other time
     the effect of random sampling, as done                                     bins, are well supported. We suggest our data
     elsewhere (e.g., Halliday and Goswami 2016;                                set is reliable enough to illuminate larger-scale
     Strickson et al. 2016; MacLaren et al. 2017). We                           patterns.
     also use alpha shape volumes, following Kotrc                                 The clade-specific analyses, particularly for
     and Knoll (2015a), to remove large areas of                                morphovolume, are influenced by low sample
     empty morphospace in volume estimates. The                                 size. Fine temporal trends can therefore not be
     volume analyses show that applying alpha                                   discerned, and peaks should be interpreted
     shapes strongly reduces the confidence                                      with caution. Higher sample sizes are achieved
     interval, suggesting it is a useful method to                              for neoceratopsians and hadrosauroids, and

Downloaded from https://www.cambridge.org/core. IP address: 176.9.8.24, on 11 May 2020 at 04:50:56, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2018.26
DINOSAUR HERBIVORY FLOURISHED BEFORE EXTINCTION                                                            633

        these are the only clades that can be interpreted                          Aptian–Albian and Campanian–Maastrichtian
        with certainty. Both clades are well sampled,                              are particularly well sampled (Starrfelt and
        and it is interesting to note that in the Campa-                           Liow 2016). It is therefore reasonable to con-
        nian bin, ceratopsians exhibit coupled increase                            sider the Albian–Aptian and Campanian–
        in volume and variance-based disparity                                     Maastrichtian disparity values as good esti-
        (expansion), while hadrosauroids show an                                   mates, while the other bins should be treated
        increase in volume but decrease in the variance                            with more caution, particularly for volume
        metric.                                                                    estimates. These two peaks are also the most
            Ultimately, disparity analyses can only                                interesting from the morphospace packing/
        record the disparity of the known fossil record.                           expansion perspective, with the combined
        Therefore, although we have shown that some                                expansion/packing in the Albian and intense
        trends are robust to random sampling, we                                   packing in the Campanian. Therefore, the
        cannot account for the fact that some geological                           pattern of morphospace packing in the Cam-
        intervals are relatively poorly sampled and                                panian is relatively well supported. Note,
        will underestimate the true disparity more                                 however, that the TRiPS method has been
        than other bins. PCO volumes will naturally be                             criticized (Close et al. 2018), because it can only
        closely linked to sample size when sample size                             truly standardize diversity data once sampling
        is low, as each species will add/remove some                               is sufficient in each time bin (i.e., has reached
        volume (as opposed to measuring variance).                                 the asymptote in the species discovery curve),
        For very large data sets, this effect is likely to be                      and when sampling is not sufficient, it over-
        less pronounced, because the effect of sample                              estimates binomial sampling probabilities.
        size tapers out (Ciampaglio et al. 2001). The                                 Different clades also show differing degrees
        extent and effect of sampling bias on the dino-                            of preservation: in particular, sauropods do not
        saur fossil record is a subject of much debate,                            often preserve skull material. Is the steady
        and a number of methods and models have                                    decline in sauropods just a sampling artifact?
        been developed to estimate bias (Barrett et al.                            This possibility cannot be excluded, and only
        2009; Brusatte et al. 2015; Sakamoto et al. 2016;                          increased sampling can resolve this. Never-
        Starrfelt and Liow 2016). Subsampling by                                   theless, the available fossil record of sauropods
        standardizing to very low sample sizes likely                              suggest broad-crowned forms were lost in the
        causes much of the real signal to be wiped out,                            Late Cretaceous, which saw the radiation of the
        resulting in a nearly flat diversity curve (Close                           narrow-crowned titanosaurs (Chure et al.
        et al. 2018). The phylogenetic diversity estimate                          2010). Therefore, as far as we can tell from the
        accounts for ghost lineages and, when applied                              known fossil record, there is a real loss in
        to the dinosaur fossil record, mostly follows                              sauropod morphological diversity in the Late
        that of the raw taxonomic count; therefore,                                Cretaceous, which fits with our disparity
        depending on position, it either confirms the                               curve—a coupled loss in volume and distance-
        major diversity trends or demonstrates that the                            based disparity. Hadrosaurs and ceratopsians,
        phylogenetic diversity estimate cannot remove                              on the other hand, are relatively well sampled
        biases effectively. The majority of recent                                 in the Campanian–Maastrichtian, and the
        diversity studies have been based on the resi-                             disparity trends recovered in this period for
        duals model (Barrett et al. 2009; Upchurch et al.                          these clades are better supported.
        2011; Lloyd 2012; Mannion et al. 2012; Brusatte                               Disparity Is Multifaceted.—“Disparity” is
        et al. 2015), which has since been shown to be                             sometimes discussed in terms of a uniform
        based on incorrect statistical assumptions and                             measure of morphological variety. The
        is therefore flawed (Sakamoto et al. 2017;                                  preferred metric in studies of the fossil record
        Dunhill et al. 2018). A different approach is to                           is distance-based metrics, as this measure is
        quantify the sampling rate of each bin to                                  least sensitive to sample-size biases. Such
        estimate true richness, the TRiPS method, as                               biases are particularly a concern for the
        applied by Starrfelt and Liow (2016).                                      vertebrate fossil record. But our data show
        Their results confirm diversity peaks in the                                explicitly that disparity has two aspects,
        Aptian and Campanian and suggest that the                                  variance and volume, and different metrics

Downloaded from https://www.cambridge.org/core. IP address: 176.9.8.24, on 11 May 2020 at 04:50:56, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2018.26
634                                                    KLARA K. NORDÉN ET AL.

     are needed to capture these aspects (Foote                                 pairwise distances or SoV, is the only feasible
     1999; Ciampaglio et al. 2001; Wills 2001;                                  option, particularly where sample size is very
     Kotrc and Knoll 2015a, b; Novack-Gottshall                                 small in some bins (as in our clade-specific
     2016a, b). Pairwise dissimilarity is a very                                analyses). A conservative approach is never to
     robust measure of disparity, but it has been                               apply volume-based metrics, as measures
     demonstrated that it “does not give an                                     might be so distorted by biases that no “real”
     adequate estimation of the amount of                                       signal will be visible, even on relatively good
     morphospace occupied” (Ciampaglio et al.                                   data sets. This is possible, but does not change
     2001: p. 711). Appreciating this can change                                the fact that distance-based metrics and
     the way we interpret disparity trends. The                                 volume-based metrics describe two different
     distance-based metric on its own suggests                                  aspects of disparity. If we can only measure
     “disparity” was low in the Campanian. But                                  one, our inferences of the evolutionary pro-
     was it low because the number of                                           cesses underlying this trend must equally be
     morphologies present was low, or because                                   limited.
     new clades were radiating into the same
     morphospace? This is an important
     distinction to make: the first scenario suggests                                                      Conclusion
     decline and extinction, the other the                                         Our results suggest that dinosaur ecological
     diversification of successful clades.                                       diversity, as far as it is reflected by dental and
        The contrast between disparity trends in                                jaw disparity, was not declining in the Late
     hadrosauroids and sauropods also demon-                                    Cretaceous, rejecting the hypothesis that dino-
     strates this. Both clades show approximately                               saur ecosystems were weakened preceding the
     the same decrease in distance-based disparity                              extinction. Instead, herbivorous dinosaurs
     from the Early to Late Cretaceous. Just con-                               were flourishing in the Campanian, with no
     sidering this metric, we could conclude they                               significant loss in disparity in the Maastrich-
     were decreasing equally in “disparity.” When                               tian. Clade-specific analyses reveal how differ-
     morphovolumes are measured, however,                                       ent clades rose and fell in disparity during the
     hadrosauroid disparity increases in volume in                              Cretaceous, notably a long-term decline in
     the Late Cretaceous, while sauropod disparity                              sauropod disparity and morphospace packing
     decreases. These distinct trends show that the                             in hadrosaurids, a group that was radiating in
     two clades show radically different patterns:                              the Campanian. These trends reveal complex
     hadrosauroids flourished, while sauropods                                   relationships among disparity, diversity, and
     were in decline.                                                           diversity dynamics. Sauropods increased in
        Our study explicitly underlines the impor-                              diversity but decreased in speciation rates and
     tance of considering several aspects of dis-                               disparity, and hadrosauroids, while diversify-
     parity. The measures used here are not                                     ing, did not expand proportionally in morpho-
     exhaustive but can be extended further to get a                            space, but rather retained a specific tooth/jaw
     more in-depth understanding of disparity pat-                              morphology. We show how disparity trends
     terns (e.g., measures of clustering/dispersion                             are multifaceted and cannot be described in
     in morphospace). Using modeling to identify                                simple terms of increases and decreases, or by
     particular modes of evolution in morphospace                               just a single metric. Thus, using a combi-
     is also a promising approach and would be an                               nation of metrics, each describing different
     interesting way to further our analysis                                    aspects of disparity, is key to understanding
     (Novack-Gottshall 2016a, b). We show that                                  the dynamics of morphological diversity
     measuring morphospace volume and packing,                                  through time.
     a relatively easy to implement extension of
     typical methods employed in disparity studies,
     can substantially increase the depth of the                                                    Acknowledgments
     analysis.                                                                    We thank M. J. Ryan (Cleveland Museum
        There might be cases in which a measure                                 of Natural History), C. Bullar (University
     robust to sample-size differences, such as mean                            of Bristol), J. Mallon (Canadian Museum of

Downloaded from https://www.cambridge.org/core. IP address: 176.9.8.24, on 11 May 2020 at 04:50:56, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2018.26
You can also read