Ecomorphospace occupation of large herbivorous dinosaurs from Late Jurassic through to Late Cretaceous time in North America

Brief comment on terminology and ecological theory

Many palaeontological studies investigate ecomorphospace occupation of an assemblage to approximate niche relationships (Van Valkenburgh, 1994; Stroik, 2014). This assumes that the relationship between form and function, imperfect as it may be (Lauder, 1995), is strong enough to reflect the ecological niche of the organisms under study. Within this context, competitive potential (i.e., that groups could have competed) is established when niches of two or more groups overlap (Stroik, 2014 and references therein). In the present study, the term niche is used to refer to the dietary ecology of a group or groups and ecomorphology to refer to aspects of the morphology related to dietary ecology. To refer to the broader concepts associated with dietary and non-dietary niches and dietary and non-dietary ecomorphology, I will use the phrases “general niche” and “general ecomorphology”.

Within the context of general niche theory, in order for competitive exclusion to be possible several conditions must be met. This includes overlap not only in dietary requirements but also in time and space as well as resource limitation (Stroik, 2014 and references therein). For the purposes of the present study, I have assumed that these and some other important conditions (see ‘Assumptions’) necessary for competition to occur are met as Lucas & Hunt (1989) had originally dismissed the competitive exclusion hypothesis on the grounds of morphological dissimilarity and by extension lack of dietary overlap. It should also be made clear that even if overlap is observed between clades in ecomorphospace, that this does not mean competition was likely to have occurred and been responsible for the ‘sauropod decline’ and North American stegosaur extinction. Rather, this overlap would be indicative that the overlapping clades could have potentially competed on the basis of ecomorphological similarity but the validity of the competitive exclusion hypothesis would need further testing to assess if overlap along other non-dietary niche dimensions was also present.

Assumptions

To begin to assess if morphological disparity alone was enough to have precluded competition between megaherbivore clades, especially sauropods and iguanodontians (sensu Lucas & Hunt, 1989), several simplifying assumptions are made but should be tested in future assessments of the competitive replacement hypothesis as a whole. The first involves resource availability. Although multiple studies on Mesozoic North American megaherbivore communities have established that resources were likely limiting and/or that niche partitioning was present (e.g.,  Foster, 2003; Mallon & Anderson, 2013; Mallon, 2019), resource limitation should be established for the entirety of the Late Jurassic through to the end of the Cretaceous.

Overlap in time and space is also assumed between taxa in the coarse temporal bins used (discussed below). The taxa included here are sampled from various localities across Alberta, Saskatchewan, Montana, North Dakota, Wyoming, Texas, Utah, Kansas and New Mexico that span from the Upper Jurassic Morrison Formation to the Upper Cretaceous Hell Creek Formation encompassing around ∼90 million years in duration (from 156 to 66 Ma) (Geological Society of America, 2018; Maidment & Muxworthy, 2019) (Table 1). Given very few specimens from the clades of interest from the Early Cretaceous of North America have been recovered and published on in the literature, and the extinction of North American stegosaurs prior to the Early Cretaceous, it is only possible to conduct a preliminary investigation into ecomorphospace at this time. For simplicity, I have split the ecomorphological dataset (see next section for details) into three main blocks of time: Late Jurassic (156–147 Ma; Kimmeridgian-Tithonian; Maidment & Muxworthy, 2019), Early Cretaceous (145.0–100.5 Ma; Barriasian-Albian; Geological Society of America, 2018), and Late Cretaceous (100.5–66 Ma; Cenomanian-Maastrichtian; Geological Society of America, 2018). Although some time bins are easier to split into smaller intervals while maintaining relatively larger (n > 5) sample sizes for each clade (e.g., Campanian Dinosaur Park Formation), other time intervals such as the Early Cretaceous, have sample sizes for some clades that are too small (n < 3) to reasonably separate into smaller time slices. Further, stegosaurs and most sauropods disappear after the Early Cretaceous and would not be represented in subsequent time periods preventing any comparison in ecomorphology between potential replacement clades (e.g., iguanodontians). There is also the added difficulty of finding suitable material for the groups considered that have been collected from the same formation, or even more ideally, the same locality.

Any investigation into potential competitive interactions among herbivores must also consider potential differences in feeding heights, as such differences have been shown to facilitate resource partitioning (Lamprey, 1963; Bell, 1971; Leuthold, 1978; McNaughton & Georgiadis, 1986; Coe et al., 1987; Foster, 2003; Mallon et al., 2013; Barrett, 2014). Stegosaurs, ankylosaurs and ceratopsids have been previously suggested to have fed on vegetation at or below 1 m in a quadrupedal posture (but see Mallison, 2010 for alternative feeding poses in stegosaurs), whereas hadrosaurids (and presumably other similar-sized iguanodontians) are reconstructed as having fed up to 4 m and Diplodocus, Apatosaurus, Camarasaurus and Brachiosaurus are variably reconstructed with feeding heights below 2.9 or 3.4 m, below 3.4 or 5.3 m, between 3 and 5.7 m, and between 5 and 9.4 m, respectively (Foster, 2003; Mallon et al., 2013; Barrett, 2014).

Ecomorphospace reconstruction and analysis

general procedure and data collection

Twelve cranial measurements known to correlate with feeding behaviour, feeding height, plant mechanical properties and other aspects of feeding ecology in modern animals (e.g., Janis, 1990; Janis, 1995; Spencer, 1995; Mendoza, Janis & Palmqvist, 2002; Mitchell et al., 2018) and have been previously used in other dinosaur ecology studies (e.g., Whitlock, 2011; Mallon & Anderson, 2013; Mallon & Anderson, 2015; MacLaren et al., 2017; Osi et al., 2017; Mallon, 2019; Wyenberg-Henzler, 2020; Wyenberg-Henzler, Patterson & Mallon, 2021; Wyenberg-Henzler, Patterson & Mallon, 2022) (see Fig. 1 and Table S1 for details) were used assess niche overlap. These measurements were collected from the literature and/or by travelling to museums to measure in person (AMNH, CMN, GPDM, MOR, ROM, TMP, UALVP, USNM) (Table S1).

Snout shape index was taken following the modified methodology of Dompierre & Churcher (1996) used in Wyenberg-Henzler (2020) and (Wyenberg-Henzler, Patterson & Mallon, 2021; Wyenberg-Henzler, Patterson & Mallon, 2022). This snout shape proxy uses photos of the snout taken in dorsal or ventral view to distinguish between wide, square shaped snouts and narrow, pointed snouts. Outlines taken of the snout are overlaid by a right-angle triangle with two sides bordering the anterior-most extension and widest part of the snout with a hypotenuse at an angle of 26° from the vertical (Fig. 1). The ratio of the area of the triangle anterior the outline of the snout and the area of the triangle itself yields the snout shape index which reflects the relative broadness of the snout—a feature known to reflect feeding selectivity in modern ungulates (e.g., Spencer, 1995; Dompierre & Churcher, 1996).

To linearize relationships between variables, numerical data were log-transformed prior to subsequent analyses. Due to the incomplete nature of most specimens and the nature of the analyses, only specimens for which ≥50% of these measurements could be taken were included, subset by family, and subjected to Bayesian principal components analysis (PCA) to produce a dataset without missing values (Hammer & Harper, 2006; Brown, Arbour & Jackson, 2012) (Code S1, File S1). Bayesian PCA is complex method wherein missing values are iteratively calculated from known values using PCA expectation maximization combined with a Bayesian model (Brown, Arbour & Jackson, 2012 and references therein). The completed dataset was then subjected to a separate PCA using a correlation matrix for principal component axis construction to reconstruct a theoretical Late Jurassic—Late Cretaceous ecomorphospace.

Convex hulls were drawn around iguanodontians (basal Iguanodontia, Hadrosauroidea, Hadrosauridae), diplodocids, non-diplodocid sauropods (Camarasauridae, Brachiosauridae), stegosaurs, ceratopsids and ankylosaurs as these were the major subgroupings of interest. The hulls were drawn around all data points for each group, ignoring time differences to see how individuals from different time periods plotted relative to their respective groups as a whole (see ‘Assumptions’). It is also possible to make comparisons between clades that would have had coeval representatives but lack enough representative material to be included in the analysis (e.g., Late Jurassic sauropods and Early Cretaceous iguanodontians).

To determine if groups from the Late Jurassic/Early Cretaceous showed significant overlap, a non-parametric analysis of variance (NPMANOVA) was conducted on a rarified subset (n = 6 per group) of PC scores for all axes individually accounting for >5% of the total variation (Code S1, File S1). This process was repeated over 1,000 iterations and the resulting p-values were combined using the harmonic mean p method, which is designed to combine multiple p-values from non-independent tests into a single value (Wilson, 2019). Two specific tests were conducted, one comparing stegosaur PC scores to scores from other clades (e.g., ceratopsids, ankylosaurs) showing overlap with the stegosaurs both within reconstructed ecomorphospace and in reconstructed feeding height (see ‘Assumtions’), and another comparing sauropods to iguanodontians (assuming overlap is observed). Conducting follow-up analyses in this manner increases statistical power by limiting the number of pairwise comparisons being made and therefore reducing the degree of correction needed to control for familywise error rate (Hammer & Harper, 2006). If the omnibus test comparing all sauropods to all iguanodontians was significant (α = 0.05), follow-up pairwise comparisons were conducted between diplodocids, non-diplodocid sauropods and iguanodontians using rarified datasets over 1,000 iterations, combined using the harmonic mean p method, and then corrected for multiple comparisons with Holm-correction (Holm, 1978). Pairwise comparisons were not conducted between all taxa as differences in feeding height (see ‘Assumptions’) would have likely facilitated niche partitioning between certain clades and increasing the number of pairwise comparisons reduces statistical power.

Phylogenetic considerations

Phylogenetic similarity between ankylosaurs, stegosaurs, ceratopsids and iguanodontians may also influence the results as more closely related groups will be more similar to each other due to shorter divergence times (Pagel, 2002; Symonds & Blomberg, 2014). Although phylogenetic similarity is certainly important in the assembly of animal bauplans, I have opted to not apply phylogenetic corrections to the data presented here for several reasons. First, there is some evidence to suggest that certain traits evolve in animals consuming the same resource, even in distantly related clades. For example, herbivores consuming mechanically resistant foodstuffs will tend to evolve short skulls and deep jaws regardless of phylogenetic relatedness (Bowman, 1961; Anapol & Lee, 1994; Nogueira et al., 2005; Mallon & Anderson, 2013). This is because short skulls position the bite point closer to the adductor musculature increasing mechanical advantage, and subsequently bite force, whereas deeper jaws are well suited for resisting high resistive forces generated when hard materials are bitten into (Greaves, 1974; Janis, 1990; Janis, 1995; Greaves & Thomason, 1995; Spencer, 1995; Mendoza, Janis & Palmqvist, 2002; Metzger & Herrel, 2005; Herrel et al., 2006; Mitchell et al., 2018). Selecting skull measurements shown to be applicable across a wide range of taxa (including Mammalia and Reptilia) and are informed by such form-function relationships helps to mediate the influence of phylogeny on ecological inferences. Second, application of a phylogenetic correction to the data, such as a phylogenetic principal components analysis (pPCA), does not necessarily correct phylogenetic signal and third, applying a phylogenetic correction can introduce new issues. For example, pPCA does not entirely remove phylogenetic signal from the dataset as the ‘corrected’ values produced are still plotted within the original shape space (Polly et al., 2013). Additionally, pPCA produces axes that are not independent from each other adding computational issues to subsequent statistical analyses (e.g., NPMANOVA). For these reasons, and for simplicity, I chosen to employ traditional analytical techniques that do not apply phylogenetic correction and so the results presented here should be interpreted accordingly.

Body size distribution through time

Body size is an important trait to consider in ecological research as it has been tied to numerous other variables such as species abundance, metabolic rate, energy requirements, reproductive mode and resource use (Colinvaux, 1979; Peters & Wassenberg, 1983; LaBarbera, 1989; Griffiths, 1992; Blackburn & Gaston, 1994; Brown, 1995; Gaston & Blackburn, 2000; Ernest, 2005; White et al., 2007). Several studies have noted that (at least some) dinosaurian ecosystems appear to exhibit species-richness body size distributions that bias towards larger species, contrary to those observed in mammalian communities. Whether this signal is more a reflection of differences in community structure resulting from differences in life history (e.g., Codron et al., 2012; Codron, Carbone & Clauss, 2013; Schroeder, Lyons & Smith, 2021; Holtz, 2021) or more of an artefact of taphonomic biases (e.g., Brown et al., 2013a; Brown et al., 2013b), is still debated and beyond the scope of the present study. However, because body size is such an important ecological factor, even preliminary consideration into potential differences in body size-species richness distributions between dinosaur assemblages through time is necessary to further elucidate ecological interactions leading up to, during, and after the ‘sauropod decline’ in North America.

Although other analyses such as an “ecological structure analysis” (see Noto & Grossman, 2010) or browse profile analysis (Coe et al., 1987; Wyenberg-Henzler, Patterson & Mallon, 2021) would provide more information than simple body size species-richness distributions would, given uncertainties associated with feeding heights and the incompleteness of taxonomic assemblages from key time intervals (see ‘Assumptions’), it is presently not feasible to conduct such analysis—although it would be prudent to do so in the future. Body mass data for ornithischians and sauropods from select Late Jurassic—Late Cretaceous formations were amassed from the literature (see Tables S3 and S4). Pre-Late Cretaceous North American formations were selected if there were multiple (n > 5) taxa known with available body mass data. To make comparisons between time periods easier, only a handful of ‘representative’ formations from Late Cretaceous North America were selected because they have been well-sampled (e.g., Dinosaur Park, Two Medicine, Horseshoe Canyon, Hell Creek formations), or are some of the few Late Cretaceous North American formations not from the north/north-western portion of the continent that meet the minimum number of taxa threshold (e.g., Kaiparowits, Aguja, Kirtland, Ojo Alamo formations). Several of the better sampled non-North American formations from the Middle Jurassic—Late Cretaceous (e.g., Shaximiao, Yixian, Nemegt formations) were also included for comparison as assemblages from other continents do not have appeared to undergo a ‘sauropod decline’. Body masses were log₁₀-transformed, with the type of skewness evaluated and kernel density distribution overlain on a histogram. Skewness type was determined according to the categories provided in Bulmer (1979) where skewness (s), is considered strong when —s— ≥ 1, moderate when 0.5 ≤ —s— < 1, and weak when —s— < 0.5.

Broad relationships between North American clades

The first four PCs account for 89.5% of the variation within the North American dataset (PC1 = 58.5%, PC2 = 15.1%, PC3 = 9.4%, PC4 = 6.5%) (Fig. 2, Fig. S1). All variables considered weight positively with increasing values along PC1. Along PC2 distal tooth row length and snout width increases while paroccipital process breadth decreases. Higher values on PC2 also reflect narrower, more pointed snouts (i.e., lower SSI values). Ignoring time intervals, most taxa overlap on PC1 and PC2, with the exception of ceratopsids which do not overlap with stegosaurs and diplodocid sauropods (Fig. 2, Fig. S2). Along PC3 across all time bins most clades show overlap except for ankylosaurs which show some separation from other taxa (mainly stegosaurs), reflecting wide snouts and posterior skull widths and short occiputs of the former. Stegosaurs, diplodocids and to some extent non-diplodocid sauropods show separation from ankylosaurs and iguanodontians along PC4 reflecting relatively long distal tooth row lengths, narrower and less square snouts, and shorter occiputs observed in the former set of clades (Figs. S1 and S2). Ceratopsids plot more centrally along the axis and overlap with all taxa along PC4.

Comparison of North American clade distribution through time

From the Late Jurassic to the Late Cretaceous, North American ankylosaurs exhibit a shift towards the ecomorphospace occupied by Late Jurassic stegosaurs (potentially due to increased sampling of ankylosaurs through time) as reflected by increases in overall skull dimensions along PC1 and increases in distal tooth row length and snout width, decreases in paroccipital process breadth, and widening and squaring of the snout along PC2. However, ankylosaurs continue to remain separate from stegosaurs along PCs 3 and 4 due to differences in posterior skull width, snout width and occiput height.

North American iguanodontians expand their occupied ecomorphospace during the Late Cretaceous primarily along PCs 1 and 3 reflecting increases in overall skull dimensions and shortening occiput height, decreases in the distance between the coronoid process and jaw joint and distance between quadrates. The original ecospace occupied by more basal representatives are later occupied by immature hadrosaurids in the Late Cretaceous. Early Cretaceous iguanodontians almost completely overlap with time-averaged non-diplodocid sauropod ecospace along the first three PC axes and show minor overlap along PC4. Overlap along PCs 1 through 3 is even more pronounced between sauropods across all time bins and Late Cretaceous iguanodontians (especially Late Cretaceous hadrosauroids and immature hadrosaurids).

influence of the inclusion of non-North American taxa on PCA results

The first four PCs for the analysis of the second dataset account for 89.9% of the variation within the dataset (PC1 = 59.7%, PC2 = 14.6%, PC3 = 9.4%, PC4 = 6.2%) (Fig. 3, Fig. S3). Variable loadings along PC axes remain unchanged with the inclusion of non-North American taxa in the analysis (Fig. S4). Relationships between clades remain largely similar to relationships observed between clades for the North American PCA. Non-North American ankylosaurs and iguanodontians plot between or very close to points representing North American members of the same clade (Figs. 2 and 3, Figs. S1 and S3). The main difference between the North American PCA (Fig. 2, Fig. S1) and the all-inclusive PCA (Fig. 3, Fig. S3) is the ecospace occupied by non-diplodocid sauropods. Two Brazilian specimens and one Mongolian specimen expand the occupied ecospace of non-diplodocid sauropods along PCs 2, 3 and 4. This expansion of occupied ecospace of non-diplodocid sauropods along PCs 2, 3 and 4 increase the relative amount of overlap with iguanodontians.

Results of NPMANOVA

The degrees of freedom, test-statistic and raw-p-values for each test and comparisons conducted for the North American dataset are available in Tables S6–S11 and for the combined non-North American and North American dataset are available Supplementary Tables S12–S17. Comparisons between stegosaurs and ankylosaurs were significant (harmonic mean p = 0.003) even when Late Cretaceous ankylosaurs were excluded (harmonic mean p = 0.002) regardless of which dataset was used. Omnibus tests comparing iguanodontians to sauropods were also significant (harmonic mean p = 0.003) even when adult Late Cretaceous hadrosaurids were excluded (harmonic mean p = 0.002) to assess if their absolutely larger size was impacting the amount of overlap between groups, particularly along PC1 regardless of which dataset was used. Follow-up pairwise-comparisons between North American diplodocids, non-diplodocid sauropods and iguanodontians indicated significant differences among all groups regardless of whether or not adult Late Cretaceous hadrosaurids were included as would be expected for taxa occupying distinct areas of morphospace (Tables 2 and 3). These pairwise comparisons were also significant for the second dataset containing both North American and non-North American taxa (Tables 4 and 5).

Body size distributions

No clear relationship between the shape of the body size species-richness distribution and presence/absence of sauropods, time or location is readily apparent from the body size distributions plots (Fig. 4). Through time, formations show variation in skewness ranging from weakly to strongly skewed. Both sauropod-bearing and non-sauropod-bearing formations show skewness ranging from weak to strong. Sauropod-bearing formations have a greater maximum body size than non-sauropod-bearing formations.

Inclusion of non-North American taxa in analyses

The results of analyses conducted for the dataset of North American taxa and North American + non-North American taxa indicate that the inclusion of non-North American taxa have very little influence on the ecological inferences and conclusions drawn from the data. Therefore, all interpretations given below can be considered applicable to both sets of analyses unless otherwise noted.

Plausibility for competition from a dietary perspective

Results of the NPMANOVAs suggest that Late Jurassic taxa were not competitively excluded by the ancestors of Late Cretaceous ceratopsids, ornithopods and ankylosaurs. These results are perhaps unsurprising as the competitive replacement hypothesis was originally dismissed on the grounds of morphological dissimilarity (Lucas & Hunt, 1989) meaning significant differences in overall shape should be expected. It is also possible that the medians of compared clades do significantly differ despite there being meaningful overlap present as an NPMANOVA only compares the medians (Hammer & Harper, 2006). Making an accurate assessment of competitive potential may therefore require a more nuanced approach wherein morphometric variables weighting on PC axes are interpreted within an ecological context rather than just simply in terms of separation versus overlap.

Body size and ecospace

The lack of a clear distinct pattern in species-richness body size distributions between sauropod- and non-sauropod- bearing formations suggests that the presence/absence of sauropods does not dramatically alter the shape of the distribution. Thus, sauropods may have merely been “filling in” the large-bodied end of the size spectrum in iguanodontian-bearing communities. While sauropods may have influenced the distributions of Late Jurassic formations (e.g., Morrison), iguanodontians (and to some extent ceratopsids) filled most of these larger size (>1,000 kg) niches during the Late Cretaceous—at least in North America. This replacement of sauropods with large-bodied ornithischians is also potentially reflected in feeding heights of these taxa. Although it is difficult to more precisely define the nature of sauropod feeding envelopes due to uncertainties surrounding neck posture, all reconstructions agree that sauropods would have fed at heights above ∼3 m (Foster, 2003; Mallon et al., 2013; Barrett, 2014; Stevens & Parrish, 1999; Stevens & Parrish, 2005; Woodruff, 2017). Iguanodontians, particularly the Late Cretaceous hadrosaurids, are reconstructed as feeding between 0 and 4 m. Thus, both sauropods and iguanodontians would have both filled higher-browsing niches within their respective ecosystems. From an ecomorphological perspective, sauropods and iguanodontians also appear to show niche overlap. In combination, these similarities in cranial ecomorphology, feeding height (at least in a broad sense) and body size category, would suggest that iguanodontians were potentially capable of opportunistically filling sauropod niches. However, such an assertion would require further testing involving other lines of evidence such as dinosaur feeding heights and consideration of plant heights through time.

Caveats and future research

Caution must be applied to the interpretation of data points (representing taxa collected from several geological stages), especially when some of those stages lack more than a few data points for a given clade (e.g., Early Cretaceous non-diplodocid sauropods). With the discovery of more material from before the Late Cretaceous of North America, it will be important to conduct follow-up ecomorphological studies to assess how ecospace distribution of the more poorly sampled clades changes with increased sampling. These future studies will also help to further assess whether competitive replacement or opportunistic niche invasion was more likely. For instance, the use of only one ankylosaur specimen from the Late Jurassic is also problematic, as it is impossible to truly assess how much overlap between ankylosaurs and stegosaurs was actually present prior to the extinction of stegosaurs in North America.

Another consideration is that the niches described here are constructed based on multiple taxa from each clade. This means the niche breadth of an individual taxon within a certain clade was not assessed. Thus, species-level interactions were not characterized and niche occupation may have differed between taxon within the same clade. However, from the relative stability of Late Cretaceous ecosystems as commented on by Mallon (2019), it would seem plausible that species from the same clade would occupy relatively similar niches to one another. This said, ecomorphological investigations into niche occupation of different species within a given clade with multiple specimens per taxon would need to be conducted for there to be any certainty in the accuracy of this claim. Such an analysis is presently not possible for all the taxa considered here as some are only represented by a single specimen available for study.

Conflict between the interpretations based on the results of NPMANOVAs (that there is no overlap) and interpretations based on observed overlap within ecomorphospace is another potential problem. This disagreement may relate to the nature of the NPMANOVA which compares the centroids of each group rather than the overall distributions themselves (Hammer & Harper, 2006). As mentioned earlier, this disagreement may also simply reflect differences in morphology that did not result in differences in ecological niche between groups. Dental microwear analyses comparing ankylosaurs and stegosaurs, and sauropods and iguanodontians, may prove a fruitful line of inquiry in the future, as the relative frequencies of pits and scratches on teeth have been shown to reflect differences in feeding style and the mechanical resistance of consumed plant materials (Fiorillo, 1998; Semprebon et al., 2004; Semprebon, Sise & Coombs, 2010; Rivals & Semprebon, 2011; Whitlock, 2011; Mallon & Anderson, 2014). Similar proportions of microwear features on the teeth of stegosaurs/ankylosaurs and sauropods/iguanodontians of the same time interval would further support competitive replacement as a potential explanation for megaherbivore turnover in North America between the Late Jurassic and Late Cretaceous especially if microwear can be sampled and compared between co-occurring taxa from these groups. Stable isotope analyses may also provide useful insights into dietary and habitat preferences of co-occurring taxa and have been used to make dietary inferences in other studies (e.g., Tütken et al., 2004; Fricke & Pearson, 2008; Cullen et al., 2020; Frederickson, Engel & Cifelli, 2020).

As mentioned previously, there are several requirements in addition to similarities in dietary niches that must also be met for competitive exclusion to have been possible (see Methods). This includes resource limitation, overlap between taxa in time (e.g., geological time, seasonality, diel activity), overlap in space, and overlap in feeding envelopes (both between adults and across different ontogenetic stages). Establishing overlap in feeding envelopes may prove especially challenging for some clades. In sauropods, disagreement over neck posture cause result in maximum feeding height estimates differing as much as 4.4 m (Camarasaurus; Foster, 2003; Barrett, 2014; Stevens & Parrish, 1999; Stevens & Parrish, 2005; Woodruff, 2017). Stegosaur feeding envelopes may have extended up to 3.3 m above ground level, 2.3 m greater than the estimate put forth by Foster (2003), as Mallison (2010) has suggested a tripodal feeding posture was feasible based on mechanical computer modelling in Kentrosaurus. Although more definitive feeding height estimates are needed for these clades, should future research reinforce or significantly increase differences in feeding envelopes between the taxa assumed to have overlapping feeding envelops in the present study, this would provide a potential niche dimension for resource partitioning to have occurred along. Future work focusing on ecological interactions during the ‘sauropod decline’ should focus on assessing resource availability and investigating ecological interactions at narrower geographic and temporal scales in addition to considering the impacts of ontogeny on ecological interactions.

It is also important to investigate the potential for other factors as contributors and/or causes behind the ‘sauropod decline’. Even though establishing environmental and/or floral changes as potential factors can be difficult, it is not entirely impossible. Although some research has been conducted on these alternative explanations, as we continue to sample sediments from the Early Cretaceous and the early Late Cretaceous, follow-up assessments of these hypotheses will be necessary. Insight into potential driver(s) behind the ‘sauropod decline’ may also provide information into the factor(s) behind turnover observed in other dinosaur groups, such as large carnivorous theropods (Nesbitt et al., 2019; Holtz, 2021). Similarities in the positioning of clades for analyses with and without non-North American representatives despite differences in community composition between North American ecosystems and those on other landmasses further suggests that non-competitive factors were more important in the North American ‘sauropod decline’. Assessing this will require consideration of other niche dimensions such as feeding height and environmental tolerances to investigate if partitioning is occurring along other niche dimensions.