Assemblage-level structure in Morrison Formation dinosaurs, Western Interior, USA

­The Upper Jurassic Morrison Formation is both geographically extensive and well-sampled, making it an ideal candidate for biogeographic analysis at both coarse and ‑ner scales. Historically, however, this has not translated into a consensus on patterns of ecological structure and connectivity, particularly with regard to the characteristic dinosaur faunas. Here, we use both traditional (genus richness, alpha and beta diversity) and bipartite network-based (biogeographic connectivity, local endemism, and average occurrence) measures to examine patterns of structure on a per-locality basis. Given the broad geographic range of the formation, we subdivide the Morrison Formation into four discrete regions based roughly on latitude and lithology—north (Montana, South Dakota, and northern Wyoming), west (Utah and western Colorado), east (central and eastern Colorado and southern Wyoming), and south (Arizona, New Mexico and Oklahoma). Further investigation revealed many coeval sites (ca. 152 Ma) in the east and west regions. Presence-absence data were also compared using network analysis to determine the presence and content of discrete subassemblages within the larger region-level assemblages. Based on our results, we favor reconstructions of the Morrison Formation as a ‘mosaic’ type environment over most of its depositional history, with patches of open environments interspersed with more closed, forested regions. ­is is suggested by relatively low rates of local endemism (patches are consistent in plant and animal structure) and connectivity across the majority of the formation, as well as the recovery of three non-overlapping assemblages dominated by dierent guilds of herbivorous dinosaurs.

Become a member of the UGA to help support the work of the Association and receive notices for monthly meetings, annual eld conferences, and new publications. Annual membership is $20 and annual student membership is only $5. Visit the UGA website at www.utahgeology.org for information and membership application. e UGA board is elected annually by a voting process through UGA members. However, the UGA is a volunteer-driven organization, and we welcome your voluntary service. If you would like to participate please contact the current president or committee member corresponding with the area in which you would like to volunteer.

INTRODUCTION
e Late Jurassic Morrison Formation (~157-149 Ma) of North America represents one of the most extensive Mesozoic terrestrial depositional basins yet discovered, covering approximately 1.5 million square miles of territory, although much of the formation remains unexposed (Foster, 2003). As a result, the Morrison Formation has proven to be one of the most pro-ductive sources for dinosaur material, with dozens of named taxa inclusive of a wide range of bauplans, body sizes, and ecological habits. Despite the paleontological riches, however, little agreement on the basic ecosystem has been reached, with authors variously claiming a dominant conifer forest habitat, xeric plains-type habitats, or a "Jurassic savanna" habitat dominated by lush, low-growing vegetation and at least seasonally wet climates, or a mosaic of at least two distinct habitat types.
Here, we show contrasting patterns in distribution between the largest bodied herbivores and all other dinosaurian taxa, and interpret the low degrees of endemism and connectivity as evidence for a 'mosaic' type landscape throughout most of the depositional basin represented by the Morrison Formation.
Prior studies of broad-scale ecological reconstruction in the Morrison Formation have primarily been focused on the abundant macro-and microfossil paleobotanical evidence. Plant fossils representing a diverse ora of cheirolepidiaceans, conifers, ferns, ginkophytes, horsetails, and podocarpaceans are known throughout the formation (Tidwell, 1990;Parrish and others, 2004). Between the abundance of drought intolerant plants (e.g., ferns, horsetails) and the known abundance of large herbivores (with their attendant forage requirements), some of the original interpretations of the Morrison Formation paleoenvironment focused on it as a lush, wet oodplain (Tidwell, 1990;. Additionally, the apparent absence of arrested growth structures in woody remains and the presence of wood-rotting fungi have also been cited as evidence for a generally humid, mesic environment (Tidwell and others, 1998). More recently, however, two separate interpretations have taken root. One camp suggests that the paleoenvironment recorded by the Morrison Formation was xeric or only seasonally wet, with a ora comprised primarily of herbaceous and low-to mid-height woody podocarpaceaean shrubs; taller woody plants would have been restricted to riparian environments (Engelmann and others, 2004;Parrish and others, 2004;Rees and others, 2004). is view interprets the paleoenvironment as roughly equivalent to a modern tropical savanna grassland, with ferns and bryophytes occupying the ground-level browse layer currently dominated by grasses. More recently, some authors have presented evidence for the paleoenvironment of the Morrison Formation as a 'mosaic, ' either with a matrix of conifer forest dominating interspersed open habitats (Gee, 2010;Hotton and Baghai-Riding, 2010), or as a primarily open environment with interspersed woodland environments (Whitlock, 2011). In all cases, it seems unlikely that dry-adapted plants such as cheirolepidiaceans would be restricted to riparian lands with wet soils, and indeed the opposite interpreta-tion is now typically accepted (Hotton and Baghai-Riding, 2010).
ere has been some attempt to segregate the Morrison Formation into ecological regions, such that the northern exposures of the Morrison (particularly in Montana and South Dakota) represent a substantially wetter paleoenvironment than the exposures to the south (Parrish and others, 2004;Noto and Grossman, 2010). Montana, in particular, houses some localities reconstructed as a unique depositional environment, dominated by mires, coal swamps, and peat bogs (Parrish and others, 2004). Palynological evidence also indicates the presence of a more humid environment in the northern extremes with a potentially higher water table as well (Hotton and Baghai-Riding, 2010). is is corroborated by the more frequent occurrence of semi-aquatic reptiles (e.g., crocodilians, turtles; Foster and McMullen, 2017) in the Morrison Formation of Montana, South Dakota, Wyoming, and eastern Colorado. Furthermore, there has been a suggestion of a longitudinal gradient of drainage and precipitation as well (Demko and others, 2004;Turner and Peterson, 2004), with drier margins and a 'center' with the majority of standing water in the forms of lakes and wetlands.
From the earliest stages of ecological assessment in the Morrison Formation, a 'core' fauna of large diplodocids, Stegosaurus, and Allosaurus has been recognized, with other taxa appearing with less regularity others, 1980a, 1980b), although there has been little consensus on how those faunas have been constructed. More recently, Noto and Grossman (2010) analyzed the paleoecology of dinosaurs in the Morrison Formation (and other formations), dividing the localities into six "assemblages" based on geographic proximity and using Ecological Structure Analysis (ESA) to identify any potential relationships between the behavior or ecology of the dinosaurs (e.g., body size, locomotor habit) and the regional environment. Although they did not make explicit statements about generic constituencies of any inferred communities, they did identify what they termed a "semi-arid" set of assemblages from the latitudinal center of the formation, bounded by a more arid set of assemblages from the northern and southern extremes. is distinction was largely recognized in the faunas as a function of body size (smaller overall body size in the semi-arid environments) and locomotor habit (semi-arid environments were more dominated by bipedal organisms; Noto and Grossman, 2010).
Here, we attempt to build on this prior work by examining the Morrison Formation on a quarry level to try and pull out any signal that might indicate the presence of distinct communities at the generic level. Using classical ecological inferences (alpha and beta diversity and shared taxonomic scope) in concert with newer bipartite statistical methods (following Sidor and others, 2013), we characterize the dinosaur assemblages at both regional and individual locality scales, and interpret the results in the context of describing the overarching ecological conditions present at the time of Morrison Formation deposition.

METHODS
Because the Morrison Formation incorporates both a large geographic and temporal range, the 291 localities sampled were divided into four regions: north (65 localities), inclusive of localities in Montana, South Dakota, and northern Wyoming (above 43°N); west (72 localities), inclusive of localities in Utah and Colorado just east of the city of Gunnison, Colorado (Cabin Creek [see Turner and Petersen, 1999] and westward); east (128 localities), inclusive of Colorado east of Cabin Creek and Wyoming below 43°N and the lithologically similar Snowmass locality (Foster and Wedel, 2014); and south (26 localities), inclusive of localities in Arizona, New Mexico and Oklahoma ( gure 1). Di erences in recognized members across the formation as well as limitations on the quality of intraformational correlations hamper further subdivision into more de nitive age units at this time (Trujillo, 2006). New 40 Ar/ 39 Ar and U-Pb isotope ages, however, identify 28 localities across the east and west regions that are roughly coeval (ca. 152 Ma; Trujillo andothers, 2014, 2015;Trujillo and Kowallis, 2015). ese localities are examined independent of the rest of the data set to provide further insight into potential regional di erences independent of time dilation e ects. Because we are predominantly interested in examining regional patterns of taxonomic structure, we refrain from attempts to group localities into bins of equal number or attempts to lter localities by alpha diversity and other indicators.
Taxonomic data was pooled from the Paleobiology Database (http://fossilworks.org; see supplemental information) and supplemented with additional information from Foster (2003). One caveat with using data of this sort is that we are, at best, two steps removed from the original material (collector->compiler->present authors), each step having the potential to introduce errors ranging from incorrect taxonomic assignment to typographic issues. As it is reasonable to assume the general care of our colleagues with regard to data collation, the biggest concern is misidenti cation of the fossils themselves. Dinosaur workers in general and those familiar with the Morrison Formation speci cally will no doubt already be aware of the di culty in correctly assigning many fossil specimens, as well as the potential for "hidden" taxonomic diversity (e.g., Galeamopus, "Elaphrosaurus;" Chure, 2001;Carrano and Sampson, 2008;Tschopp and others, 2015;Tschopp and Mateus, 2017). While recognizing the potential confounding e ects of these issues, there is fundamentally little that can be done to resolve them at this level, and thus we note their existence here and calibrate our interpretations accordingly. Similarly, we note the potential for taphonomic bias towards larger bodied taxa, which we attempt to mitigate again by limiting our scope to presence-absence rather than abundance.
Taxonomically unresolved taxa-e.g., "Elaphrosaurus"-or records without identi cation to the genus level were omitted. e recently proposed generic separation between Apatosaurus and Brontosaurus (Tschopp and others, 2015) is not incorporated into this analysis for reasons of practicality; the records in question predate this proposal, and in lieu of attempting to examine all specimens of Apatosaurus in question (some 75 total records composed of untold numbers of individual specimens), we elect to simply continue use of the genus Apatosaurus. To examine large-scale trends in connectivity and diversity, alpha diversity (using genus richness), beta diversity (as measured by Sørenson's Index, or SI), and shared taxonomic scope were calculated using the pooled assemblage data from each major region. Sørenson's Index was chosen as a measure of beta diversity vis-à-vis continuity (sensu Kole and others, 2003). Importantly, SI satis es the requirement of be- ing sensitive only to proportions of species, not to total number (Janson and Vegelius, 1981), the latter being a potential issue given the unequal areas and taxonomic compositions of the regions under consideration. To characterize locality-level connectivity, bipartite resampling methods (following Sidor and others, 2013) were used to calculate three measures: average occurrences per taxon (interpreted as range size, following Sidor and others, 2013), average endemic taxa per locality, and biogeographic connectivity (the number of taxon linkages as a proportion of the maximum possible). Each analysis presents a distribution of scores obtained via 1000 bootstrap replicates (see supplemental information for discussion of replicate choice). A Welch two-sample unequal variances t-test was used to determine statistical signi cance. To look for potentially distinct assemblages within the taxonomic samples, co-occurrence and edge matrices were computed in R using the cooccur package (Gri th and others, 2016); resultant networks were analyzed for patterns of generic clustering using Gephi (Bastian and others, 2009). For the temporal groupings in the east and west regions, localities with good stratigraphic data were correlated with the nearest precise radiometric age by K.C. Trujillo.

RESULTS
Alpha diversity is consistently high in the north, west, and east regions (21-22 genera) and substantially lower in the south region (10 genera; table 1). Sauropod alpha diversity appears to decrease with decreasing latitude, but this trend is not seen in either theropods or ornithischians. In theropods, there appears to be a 'bulge' in diversity in a central region (west and east regions combined). Beta diversity is highest between the north and east regions (SI = 83% similarity; table 2), but is also notable between the west and east regions (71%). Similarity is lower between the north and west regions (65%), and is lowest between the south region and all others (56-58%), with the north and south being most di erent. is trend is generally conserved when the samples are restricted by taxonomic group, with the exception of sauropods, where similarity is highest between the south region and the west and east regions (83%), although again the north and south are least similar (63%).
Shared taxonomic scope is reasonably high in the Morrison Formation ( gure 2). is is exempli ed by the large-bodied sauropods, where 5 of 12 taxa occur in all four regions and 8 of 12 occur in at least two. e regionally endemic taxa, Dyslocosaurus, Dystrophaeus, Kaatedocus, and Suuwassea, are also known from only a single locality each, which suggests that their rarity is either a function of their rarity in the fauna as a whole or the rate at which they are misidenti ed by collectors, or both. eropod shared taxonomic scope is less robust, with only two of the largest bodied taxa (Allosaurus and Ceratosaurus) occurring in all four regions, and two others (the very large Torvosaurus and the small Coelurus) occurring in three. Nearly half (4/10) of theropod taxa are known from a single region only, and there is the possibility of a great many more singleton taxa that would increase this total (e.g., any potential taxa subdivided from the catch-all genus "Elaphrosaurus," Koparion). ese regionally endemic theropods are invariably small-bodied. Ornithischian-shared taxonomic scope is somewhat intermediate between sauropods and theropods, in that regional endemism is high (4/10 taxa), as in theropods, but as many taxa occur in at least three regions (4/10), a distribution somewhat similar to that seen in sauropods. Two taxa (Camptosaurus and Stegosaurus) are known from all four regions. For the 152 Ma time bin, absolute taxonomic diversity is highest in the east region (18 taxa vs. 12 taxa in the west). Beta diversity is generally high (0.66), but this is largely driven by the sauropods (1.0). eropod (0.5) and ornithischian (0.5) assemblages are both relatively less similar. e di erence in taxonomic scope is largely due to the absence of many relatively small bodied taxa, particularly theropods (e.g., Coelurus) in the west.
Bipartite analyses result in signi cantly di erent (p <0.0001; t and df reported in supplemental information) distributions of all three characteristics (average occurrence, average endemics, biogeographic connectivity) for all regions, excluding only biogeographic connectivity between north, east, and west regions (1 ≤ p ≤ 0.9994). Despite this statistical signi cance, certain characteristics of certain regions appear more similar to each other than others. As noted, biogeographic connectivity is statistically identical in the north, east, and west regions ( gure 3), and is signi cantly lower  than in the south region, where connectivity is highest. Despite the lower connectivity scores, locality-level endemism is essentially zero in the central and north regions, with an order of magnitude higher rates recorded in the south region. Average occurrence (AO) scores do increase as the number of localities sampled increases, but none are directly limited by the number of localities (i.e., all AO scores are less than 1/4 of the number of sampled localities). Time-restricted samples show the same general pattern as the broader samples, with the exception of an unusually high connectivity in the east 152 Ma time bin ( gure 3). Assemblage-level structure in the Morrison Formation is generally built around a consistent "core" of taxa (Allosaurus, Apatosaurus, Camarasaurus, Diplodocus, and Stegosaurus) that are highly connected with the majority of other taxa, and might be considered the 'typical' Morrison fauna others, 1980a, 1980b). Camptosaurus and Ceratosaurus are also commonly found co-occurring with most other taxa. Larger herbivores (i.e., sauropods, Camptosaurus) and some large predators (i.e., Allosaurus, Ceratosaurus) display little preference with regard to co-occurring taxa, although smaller carnivores (Coelurus, Tanycolagreus) and many other large predators (Fosterovenator, Stokesosaurus, Torvosaurus) do appear to show some segregation. ese patterns are most distinct in the central region ( gure 4). e most locality level segregation is found among the medium-to small-sized ornithischians in the central region, which appear to sort themselves into three general groups based upon non-overlapping exemplar taxa: (1) a 'Mymoorapelta' group, which co-occurs with no other smaller herbivores, only large sauropods; (2)   by presence/absence) in the north region and all 15 occurrences in the central region are mutually exclusive.
Othnielosaurus is a relatively rare taxon, but it is substantially more common in the northern Morrison ( ve localities vs. two each in the west and east). Coelurus is found exclusively with the Othnielosaurus assemblage, and Tanycolagreus is found exclusively with the Dryosaurus assemblage, although the two taxa do co-occur (in the absence of both Dryosaurus and Othnielosaurus) in the west region.
Within the 152 Ma time bin, we observe this segregation in the east ( gures 4E and 4F), but not in the west   Assemblage-level Structure in Morrison Formation Dinosaurs, Western Interior, USA Whitlock, J.A., Trujillo, K.C., and Hanik, G.M. ( gure 4D). Furthermore, Dryosaurus does not co-occur with any other taxa in this sample and so provides no information relevant to its participation in discrete assemblages. Aside from a slightly lower taxonomic diversity, the 152 Ma time bin is representative of the regions as a whole with regard to taxonomic assemblages.

DISCUSSION
Diversity metrics (genus richness, SI) generally suggest similar diversity levels in localities in the northern and central parts of the Morrison Formation, with the greatest diversity occurring in the central region. Regional interconnectivity appears to be tied to some combination of body size and ecological habit. Large-bodied herbivores (such as sauropods, and certain ornithiscians like Camptosaurus) tend to occur in more regions overall (lower rates of endemism), and have a wider individual distribution (weighted towards occurrence in four regions over occurrence in two). Camptosaurus is itself potentially cosmopolitan not simply within the Morrison Formation, but also may appear in coeval basins in Europe as well (Foster, 2013). is would be a substantial geographic range for a single taxon, although representatives of other Morrison Formation genera such as Allosaurus, Ceratosaurus, Stegosaurus, and Torvosaurus have been reported from Iberia (Mateus, 2006;Mateus and others, 2006;Escaso and others, 2007;Hendrickx and Mateus, 2014). Whether this is a true distribution or simply a matter of taxonomic 'lumping' (e.g., 'Elaphrosaurus;' Carrano and Sampson 2008) is unclear, however.
ere does not appear to be a clear size-distribution relationship for theropods, as in general large-and small-bodied individuals are recovered as both widely distributed (e.g., Allosaurus, Coelurus) and regionally endemic (e.g., Saurophaganax, Stokesosaurus). is may be due to increased competitive pressures facing carnivores, access to appropriate prey, and/or reinforcing behaviors such as territorialism. We agree with Foster (2000) in that we nd little evidence to suggest drastic di erences in environment with latitude, particularly with regard to the northern Morrison Formation as an environmental outlier, as has been previously suggested (Parrish and others, 2004;Hotton and Baghai-Riding, 2010;Whitlock, 2011). Sauropod diversity does decline with declining latitude, but the di erences between the north and central regions are not substantial and are primarily driven by the presence of two north region endemics, Kaatedocus and Suuwassea. ese two taxa are both very recently described (Harris and Dodson, 2004;Tschopp and Mateus, 2013) and are represented by a single specimen each; consequently, it is not implausible that additional specimens from other regions are either undiscovered or misidenti ed in existing collections.
Bipartite statistics predict greater provincialism when biogeographic connectivity and range sizes are small and average endemism is high (Sidor and others, 2013). In the south region, we nd agreement between range size and endemism-both of which predict a high degree of provincialism-but disagreement with biogeographic connectivity, which would predict a very cosmopolitan distribution of taxa. Similarly, range sizes and endemism predict a more cosmopolitan distribution in the north, east, and west regions, but the comparatively lower biogeographic connectivity in those regions predicts provincialism. Some of these e ects may be attributable to time dilation e ects; i.e., substantial turnover in environmental structure may result in these statistics capturing discordant e ects due to overlaying di erent environmental structures on top of each other, both in physical space (as a function of deposition) and in 'statistical space' (as a function of data pooling). However, we postulate that these results may be still interpretable due to a highly fractured environments. Extremely patchy ecosystems would result in smaller ranges and increased regional endemism. If substantial corridors connecting similar patches did not exist (or were outside the depositional area recorded), these patches would be isolated from each other in the fossil record, which is consistent with a low per-locality connectivity measurement, as observed. We do interpret some degree of connectivity, largely driven by the cosmopolitan sauropod fauna, as complete balkanization of depositional environments has been shown to correspond to even lower degrees of connectivity than observed here (Sidor and others, 2013). In the case of the widely distributed taxa (sauropods, large theropods like Allosaurus), it seems likely that their large size would have made them largely immune to barriers, inclusive of larger patches of less suitable habitat. In modern ecosystems, large rivers present barriers to dispersal for many organisms, including such varied groups as African Bu alo (Naidoo and others, 2012), elephants (Robertson, 2013), lions and other African carnivores (Cozzi andothers, 2013), wolves (Blanco andothers, 2005), passerine birds (Ayres and Clutton-Brock, 1992;Hayes and Sewlal, 2004; but see Gascon and others, 2000), and primates (Goodman and Ganzhorn, 2004;Harcourt and Wood, 2012), although there is some evidence that larger bodied taxa are less a ected, particularly when the river in question is small or seasonally low (Cozzi and others, 2013;Robertson, 2013). e river itself does not need to be impassable to be a barrier; in many cases the ecosystem on either side of the river is signi cantly di erent as to be a barrier unto itself (Oates, 1998;Meijard and Groves, 2006). A more continuous environment would result in a signal with low regional endemism and high connectivity, as there would be no barrier regions where a variety of ecomorphs would be less likely to occur or be preserved.
When restricted to localities deposited at or near 152 Ma, however, our results paint a more consistent picture. e west region scores lower in range size and connectivity and higher in endemism than the east, all of which suggest that the west was more provincial at this time.
is fragmentation supports the idea that there was no 'rain shadow' e ect caused by coeval and preceding upli and erosion of ri shoulders associated with a magmatic arc west of the Morrison Formation's depositional area (Turner and Fishman, 1998;Dickinson, 2006). Such upli could have potentially created a 'rain shadow' on the leeward basin, resulting in an arid region without signi cant permanent riparian structure. Due to the high rate of sediment deposition attributable to erosion from those highlands (Turner and Fishman, 1998), however, it seems unlikely that there was a signi cant rain shadow e ect on the west region at this time (Turner and Peterson, 2004). In addition, a rain shadow would likely have resulted in a relatively homogenous, 'savanna' type environment without major river systems to serve as dispersal barriers for smaller dinosaurs (Ash and Parrish and others, 2004). Instead, the apparently robust riparian structure in that region may well have provided signi cant (if ephemeral on geologic time scales) barriers to dispersal in the form of major river systems and alkaline-saline wetlands (e.g., "Lake" Too' di' chi' in southwest Colorado and southeast Utah; Fishman and others, 1995;Turner and Peterson, 2004), which could potentially have limited the range sizes and regional distribution of smaller dinosaurs (e.g., smaller theropods, ornithischians). e riparian structure would have been less restrictive to the extremely large sauropods and larger theropods like Allosaurus, which do indeed have a more cosmopolitan distribution. Range size is also strongly correlated with body size in extant mammals (McNab, 1966;Swihart and others, 1988), largely as a function of energetics, which would further explain the consistently cosmopolitan distribution of these larger organisms. e larger range sizes and degree of connectivity overall in the east, however, suggest greater environmental homogeneity, consistent with reconstructions of this region dominated by a broad alluvial plain (Turner and Peterson, 2004). erefore, we favor the nuanced interpretation many modern studies have put forward, such that the geographically large region recorded by the Morrison Formation likely included a diverse array of habitats and habitat structures inclusive of both drier, 'savanna'-type environments and wetter, riparian-dominated environments, consistent with regional tectonic and eustatic changes at the time. However, the patchiness (or 'mosaic' distribution) of these environments as well as the e ects of time dilation make direct comparisons to modern ecosystems (e.g., 'savanna') somewhat problematic, particularly as the geographic scale of the study increases. e interpretation of our bipartite analysis results is consistent with the network analysis which recovered three distinct ornithischian assemblages in the central region-a 'Mymoorapelta' assemblage, a 'Dryosaurus' assemblage, and an 'Othnielosaurus' assemblage-as well as with prior work (Noto and Grossman, 2010). e Mymoorapelta group appears to be more 'open' adapted, based upon the feeding preferences of the locally dominant diplodocid sauropods (Stevens and Parrish, 1999;Upchurch and Barrett, 2000;Stevens and Parrish, 2005;Whitlock, 2011), although the Mymoorapelta type locality does feature abundant evidence for conifers (Tidwell and others, 1999;Hotton and Baghai-Riding, 2010). e Dryosaurus and Othnielosaurus assemblages both contain more closed-adapted genera, including a wide variety of other medium-sized ornithischians, but they di er regarding sauropod content. Dryosaurus assemblages include both diplodocids and the mid-to high-browser Camarasaurus (Fiorillo, 1998;Upchurch and Barrett, 2000;Whitlock, 2011); Othnielosaurus assemblages are strongly associated with Camarasaurus.
is suggests the preference of Dryosaurus for edge environments (where open and closed adapted taxa would be expected to intermingle) and closed environments for Othnielosaurus. As Othnielosaurus is reconstructed as substantially smaller than Dryosaurus (e.g., Farlow and others, 2010), this is a reasonable assumption to make. ere is some evidence for habitat preference in theropods as well, with many large-bodied taxa appearing to prefer open environments and medium-to small theropods choosing closed or edge environments. is association is not as strong, however, suggesting that habitat choice is more labile. is is consistent with observations of modern animals, which o en display variance in habitat preference largely dependent on nonbiotic environmental factors such as local temperature (May and others, 2010;van Beest and others, 2012).

CONCLUSIONS
A picture of the Morrison Formation paleoecosystem emerges as a mosaic of diverse communities and environmental structures in a massive, long-lived depositional basin, regionally dominated by distinct populations of relatively small-bodied ornithischian herbivores and theropod carnivores, with widespread and potentially migratory populations of large-bodied sauropods and certain other exemplar taxa such as Allosaurus and Camptosaurus. Where temporal restriction is possible, distinct regional di erences emerge between the western and eastern regions of the basin, consistent with coeval tectonic events, such that upli and extensive riparian development to the west created greater habitat fractionation and provincialization in dinosaurian assemblages. Direct comparisons to modern ecosystems (e.g., use of analogs such as 'savanna') are likely of limited application, at least at this geographic scale. e apparent patchiness of much of the Morrison basin suggests a 'mosaic' type environment, and as such we must look to di erent approaches, such as this one, to gain a better understanding of the community-level dynamics in the megafauna.