Understanding how biodiversity has changed through time and space is a central aim of paleobiology. To elucidate accurate biodiversity patterns in deep time, regional case studies, where sampling biases can be minimized, are needed. The Upper Jurassic Morrison Formation of the western USA crops out over 1.2 million km² and covers 12 degrees of latitude. It was deposited over a ~9-million-year time period and was home to some of the most iconic dinosaurs. Utilizing a new, high-resolution chronostratigraphic framework for the formation, tetrapod occurrences from the Paleobiology Database were temporally and spatially mapped to examine patterns of diversity change through time and space, and the geographic ranges of taxa were examined to shed light on niche partitioning. Latitudinally, diversity was found to peak in the center of the basin, perhaps due to the availability of water resources. Diversity increased over time in the Morrison Formation, and there is no evidence to indicate a decline in diversity prior to the extinction of the fauna at the end of the Jurassic. There appears to be some degree of geographic separation of faunas in the Morrison basin, with southeastern and northwestern fauna, albeit with a number of overlapping taxa. High-resolution climate models paired with detailed sedimentological analysis could help to elucidate the drivers of the patterns observed here.

All vertebrate occurrences in the Morrison Formation were downloaded from the Paleobiology Database (PBDB; paleobiodb.org; accessed 23/12/2022). The data were visually inspected and occurrences related to eggshells or tracks were removed, leaving only those pertaining to body fossils. This resulted in 1397 occurrences. Taxonomy was cleansed following the recent literature. Occurrences were manually attributed to systems tracts described in Maidment & Muxworthy (2019) based on stratigraphic logs or descriptions in the literature for each locality and supplemented with first-hand observations of a number of quarries. A full list of quarries, systems tracts, and references for the stratigraphic location are provided in the spreadsheet “Quarry data.csv” in the Online Supplementary Material available with the manuscript. As not all references provided stratigraphic logs or descriptions, it was not always possible to attribute quarries to stratigraphic locations, but 1144 occurrences (82%) could be attributed to a systems tract. The occurrences represent 300 discrete collections, for which stratigraphic data is known for 182 (60%). 957 occurrences are identified to the generic level or better, of which 799 could be assigned stratigraphic data (83%). These data are available in this data package in the spreadsheet “Occurrence data with STs.csv”.  Diversity analyses were carried out in R ver. 4.0.4 (R Core Team, 2021) using the Tidyverse package (Wickham et al., 2019) and all code is available in this data package.

Latitudinal biodiversity

Raw diversity—In order to assess how biodiversity changed with latitude in the Morrison Formation, two measures were used. The first measure was diversity, which herein equates to generic richness. Generic occurrence data (available in this data package in the spreadsheet “Genera_with_latitude.xlsx”) were binned per degree of latitude and the number of distinct genera in each latitudinal bin was summed. This was carried out for the total dataset and for data within each system tract. The second measure was abundance. An occurrence in the PBDB is the presence of a taxon within a collection; however, for some collections, there were multiple occurrences of the same taxon, and that is signified in the PBDB using abundance data. Abundance was calculated for each collection based on the “abund_value” column in the PBDB data. Where no abundance was specified for an occurrence, the abundance was assumed to be equal to one. Not all abundances are equal: a single abundance datapoint might indicate a single, more-or-less complete articulated sauropod skeleton or might refer to a single isolated fish scale. Microsites and bone beds are therefore heavily over-represented in the abundance data, while sites with articulated skeletons may be under-represented. Abundance data was binned per degree of latitude. This data is available in the spreadsheet “corrected abundance with latitude.xlsx” in this data package.  

In order to assess whether the raw diversity patterns observed were influenced by sampling bias, the number of collections per degree of latitude was calculated from the PBDB occurrence data. This data is available in the spreadsheet “Collections_with_latitude.xlsx” in this data package. Diversity, abundance, and collections were plotted against latitude in R, and correlations between the curves were investigated using Spearman’s Rho, Kendall’s Tau, and generalized least squares regression using a first-order autoregressive model (corARMA). The latter was carried out because it reduces the chances of overestimating the statistical significance of regression lines due to serial correlation in the latitudinal series. Data was naturally log-transformed prior to GLS regression, which was carried out using the gls() function in the R package nlme (Pinheiro et al., 2018). Code for these analyses is available in this data package as “diversity_analysis_code.R”.

Subsampled diversity—In order to account for the strong degree of sampling bias observed in the data (see Results), shareholder quorum sub-sampling (SQS; Alroy 2010) was carried out on the whole dataset using the ‘estimateD’ command and a confidence interval of 0.95 in the R package iNEXT (Hsieh et al. 2016). The analysis was carried out in R ver. 4.0.4 (R Core Team, 2021) and the code is available in this data package, “iNext_code.R”. Cleansed generic occurrence data from the PBDB was used, and abundances of specific taxa were calculated for each degree of latitude. Investigation of the corrected abundance data (see above) indicated that it was overwhelmed with occurrences from two sites: specimens of Diplodocus from the Mother’s Day Quarry in southern Montana (1483 specimens recorded), and specimens of Allosaurus from the Dry Mesa Quarry of Utah (200 specimens recorded). These quarries are bone beds and the abundance values most likely relate to the number of individual bones found, rather than the number of individuals that were actually present. Using these abundance data when sample-standardizing is therefore problematic, and consequently, it was not used. To investigate whether latitudinal bins with very low sample sizes and a limited number of generic occurrences were impacting the results of the analysis, the latitudinal data were examined and latitudinal bins with fewer than 10 occurrences were removed. The analysis was re-run. SQS was carried out at quorum levels from 0.7 to 0.3. Sub-sampled diversity analyses were also attempted for each system tract, but there was too little data to provide meaningful results.

Temporal diversity

Raw diversity—In order to assess how diversity changed through time in the Morrison Formation, diversity (=generic richness) and abundance were again used. Cleansed occurrences and abundance data from the PBDB were binned by systems tract; those for which no systems tract data was known were discarded. Diversity and abundance were plotted against systems tract in R.

Subsampled diversity—To account for different levels of sampling in different systems tracts, shareholder quorum subsampling was carried out on cleansed occurrence data for each systems tract, following the method used for latitudinal diversity. SQS was carried out with quorum levels from 0.7 to 0.3.

Collector curves

In order to assess how well sampled the B4 and C6 systems tracts were relative to each other (see Discussion), collector curves, showing the cumulative number of unique collections and the cumulative number of new taxa identified per year for the B4 and C6 systems tracts were built using the year the occurrence was published, which was provided in the PBDB download. The data is contained in the spreadsheet “Collector_curve_data.xlsx” and the code is provided as “collector_curve_code” in this data package.