The fossil record of Mesozoic polar ecosystems records the existence of diverse, vertebrate-rich terrestrial faunas at high latitudes dating back through at least the Early Cretaceous. These environments are of special interest as they existed within a warmer global climate but at latitudes that were still light-restricted for much of the year. Some of the largest samples from such ecosystems come from deposits of the upper Strzelecki Group and Eumeralla Formation in southern Victoria, Australia, where thousands of vertebrate fossils have been collected over more than 40 years, representing bony fishes, mammals, dinosaurs and other reptiles, and temnospondyls. This paper presents a comprehensive assessment of these Early Cretaceous vertebrate faunas, based on study of all available fossil materials. Although there are some significant differences in content and abundances between the two strata, the greatest distinctions between localities seem to be size-related. Currently there are no records of sauropod dinosaurs, lissamphibians, or lepidosaurs from these beds, suggesting that even in warmer periods these polar landscapes were depauperate in certain groups. Temnospondyls and crocodilians do not seem to have coexisted. However, the high-energy rift-valley setting of these deposits lends a strong taphonomic overprinting, and further work is needed to confirm the true absence of some small-bodied vertebrate groups. The dataset includes locality and measurement data for individual vertebrate specimens, with different iterations related to the distinct analyses performed in this study.

Fossils and Localities

During a four-month visit (May through August 2024) to the Museums Victoria Palaeontology collection (abbreviation: NMV P) housed in the Melbourne Museum, MTC studied all available catalogued vertebrate fossils as well as numerous uncatalogued materials (many not yet fully prepared) from the upper Strzelecki Group and Eumeralla Formation. This was supplemented by a return visit in May 2026. These materials have been collected by THR, PVR, and colleagues since 1979 in an ongoing effort to document the Early Cretaceous vertebrate record of southern Australia. Over more than 40 years, many important discoveries and analyses have emerged from these efforts, including taxonomic (e.g. Flannery et al. 1995; Rich et al. 1997; Warren et al. 1997; Herne et al. 2018, 2019; Poropat et al. 2020; Kotevski et al. 2024), physiological (e.g., Chinsamy et al. 1998; Vickers-Rich et al. 1999; Woodward et al. 2018), paleoecological (Benson et al. 2013), and biogeographic (e.g., Benson et al. 2012; Rich and Vickers-Rich 2012a). A recent summary (Poropat et al. 2018) provides an excellent overview of these accomplishments.

In light of the wealth of previous studies, some significant aspects of the present work should be highlighted. First, we utilized data from every individual specimen in the collection (except for a single locality, noted below), regardless of its taxonomic informativeness, including categories of data (size and shape) not previously reported. We also revised numerous taxonomic and anatomical identifications in the collection. Finally, we report taxonomic abundances for individual localities; these have been previously recorded only as presence/absence data (e.g., Rich and Rich 1989; Rich and Vickers-Rich 2012b; Kool and Lowery 2023).

The well-preserved fossils from the lacustrine beds of Koonwarra in the upper Strzelecki Group were not included in this study, as they derive from a distinct depositional and preservational regime. This fauna is dominated by plants, invertebrates, and bony fishes (e.g., (Jell and Duncan 1986); although dinosaur feathers are known (Kundrát et al. 2020), no other tetrapod remains have been found.

A total of 4287 specimens were studied, 2906 from the upper Strzelecki Group and 1381 from the Eumeralla Formation. The only specimens not examined firsthand were those that had been loaned to other institutions, subjected to destructive sampling, or were otherwise unavailable. These 222 unavailable specimens accounted for 5.2% of the total catalogued sample (166 from the upper Strzelecki Group and 56 from the Eumeralla Formation). We could not directly measure these specimens and so relied on information in the museum’s catalogue for taxonomic, anatomical, and locality data.

This sample represents nearly all the vertebrate materials that have been collected, prepared, and catalogued from these strata as of 2026. However, mechanical preparation of these fossils is difficult and time-consuming, and so many more specimens have been (and are being) collected than can ultimately be prepared. Fortunately, the project has developed a rigorous method for triaging specimens after their initial collection, which has been applied for more than 30 years, largely by the same individuals (Pickering 2015, 2016; White and Pickering 2015, 2016; Kool 2017). We assessed the potential impact of this method using information from detailed collecting and preparation logs. These document the results of sustained excavations at the productive localities of Flat Rocks (from 1992–2023) in the upper Strzelecki Group and Eric the Red West (= ETRW; from 2014–2017 although the first collections were made in 2005) in the Eumeralla Formation. They detail field identifications and preparation decisions, including discarded specimens; although the field identifications of these latter cannot be verified, we view them as sufficient to approximate the taxonomic content of the discarded materials.

Primary locality data, including sedimentological information, was obtained from digital and printed collections records, field notes, preparation notes, scientific papers, and the Dig at Dinosaur Cove (1984–1994) and Dinosaur Dreaming (1994–2025) annual reports. Together, these specify dozens of fossil-producing localities (Fig. 3). Most outcrops of the upper Strzelecki Group and Eumeralla Formation occur in limited areas along the cliffs and shore platforms of the Victorian coast. They are often separated by hundreds or thousands of meters in which exposures are either nonexistent, covered by vegetation, physically inaccessible, or unproductive. As a result, many localities are isolated, whereas others are geographically clustered. Several sites have been extremely productive and often include distinct sublocalities centered on especially rich beds or exposures.

Additional sedimentological data was obtained from personal visits to localities and examination of unprepared fossiliferous blocks in the Melbourne Museum collections. We also made observations on the preservation and condition of individual fossils, which should be considered supplemental to the significant amount of taphonomic information presented by Seegets-Villiers (2012).

Measurements, Identifications, and Statistics

For this study, each specimen was measured, categorized as to shape, and identified taxonomically and anatomically as finely as possible. For the purposes of addressing taphonomic processes, fossils were treated as clasts, and therefore those specimens consisting of multiple articulated parts were measured as a single unit. By contrast, any isolated elements, including small fragments, were measured separately.

Each specimen was measured in three dimensions (length, width, and height), with length arbitrarily assigned to the longest dimension (e.g. Blob and Fiorillo 1996) and the remaining two measurements taken approximately orthogonally to the first. For complete elements, length was aligned to the longest anatomical dimension. Measurements were made with a Mitutoyo 505-647-50 analog caliper or a Mitutoyo 500-173 digital caliper to the nearest tenth of a millimeter. A metric measuring tape was used for measures greater than 200 mm, captured to the nearest millimeter.

Size was calculated in two ways: (1) as equal to the largest single dimension (i.e. length), and (2) as the geometric mean (GM) of all three dimensions (Jungers et al. 1995). Frequency distributions were calculated for both size measures using 5-mm bins up to 220 mm, plus a single bin for all specimens larger than 200 mm. We evaluated frequencies with a one-sample Kolmogorov-Smirnov (K-S) test of normality, and compared size frequencies between the two strata using a two-sample K-S test (e.g. Sokal and Rohlf 1995) in R Studio (Posit Team 2026). Note that we used the Posit Assistant AI tool in R Studio to perform coding error checks and corrections.

Shape categories were intended to capture broad, hydrodynamically relevant shape ‘types’: equidimensional/compact, elongate, plate-like, and conical. We followed previously defined categories (Blob and Fiorillo 1996; Rogers et al. 2017) used for studies of vertebrate microfossil bonebeds (VMBs), but these also approximate the sphere/rod/disc/blade categories used in Corbat et al. (2017). We captured shape data for 2725 (94.8%) upper Strzelecki Group specimens and 1301 (94.2%) Eumeralla Formation specimens. Specimens for which the complete shape could be neither seen nor estimated (usually due to matrix) were not included in these totals. We compared shape distributions between the two strata using a chi-square test (C²; e.g. Sokal and Rohlf 1995) in R Studio (Posit Team 2026).

With a few exceptions, the isolated nature of most fossils precludes genus- or species-level identifications, as a result we opted for higher-level taxonomic categorizations to maximize the numbers of specimens that could be classified. However, any resulting similarities (or dissimilarities) will not necessarily reflect such patterns as they may appear at lower taxonomic levels (Forcino et al. 2012). We assigned specimens to one of 14 non-overlapping higher taxa, as follows: Actinopterygii (i.e., non-sarcopterygian Osteichthyes), Dipnoi, Temnospondyli, Reptilia (i.e., reptilian fossils that could not be more specifically categorized), Testudinata, Plesiosauria, Crocodylia, Pterosauria, Dinosauria (i.e., non-avian dinosaur fossils that could not be more specifically categorized), Theropoda (= non-avian theropods), Ankylosauria, Ornithopoda, Aves, and Mammalia. 2460 specimens from the upper Strzelecki Group (= 84.7% of specimens) and 1153 specimens from the Eumeralla Formation (= 83.5%) could be placed into one of these taxonomic categories (Table 1).

We performed two additional, but separate, taxonomic iterations: (1) we counted all non-avian dinosaur specimens as the sum of Dinosauria, Theropoda, Ankylosauria, and Ornithopoda; and (2) we distributed the indeterminate reptile (Reptilia) and dinosaur (Dinosauria) specimens among their respective lower taxonomic categories in proportion to the previously identified specimens. Note that this iteration required two steps to allot specimens into various dinosaur groups: first, some of the indeterminate reptile specimens were allotted to Dinosauria, and then all Dinosauria specimens were distributed among its constituent groups. We compared taxonomic distributions between these strata using a chi-square test. The complete dataset is included as Supplementary File 1.

We grouped data from all localities within the upper Strzelecki Group and compared them with similarly grouped data from the Eumeralla Formation. We also examined individual localities within each stratum. For taxonomic analyses, we focused on sites with at least 10 taxonomically identifiable specimens; this attempts to maximize the number of localities while reducing the risk of losing any original paleoecological signal (Forcino 2012). For size and shape analyses, we examined those with at least 20 total specimens. In some cases, we grouped smaller individual localities together if they were derived from a relatively restricted geographic and stratigraphic interval (typically, < 200 m and within a single major bed). Conversely, for especially productive localities, we examined individual sublocalities both separately and grouped within their major ‘host’ locality. Table 2 lists the major localities, and sublocalities examined. The datasets used to examine individual localities are included as Supplementary Files 2 and 3.

To assess the likelihood of whether additional taxa are likely to be recovered with additional sampling, we generated a species accumulation curve for each stratum. We assigned specimens to the same 14 higher taxa described above—although not all taxa were present in each stratum (i.e. no crocodylians in the upper Strzelecki Group, no temnospondyls or birds in the Eumeralla Formation)—and included all localities at which at least four distinct taxa had been recovered. This resulted in 16 localities from the upper Strzelecki Group and seven from the Eumeralla Formation. We used the ‘specaccum’ function in the ‘vegan’ package (2.7-5) in R Studio (Oksanen et al. 2026; Posit Team 2026) to calculate these curves. These data are included in Supplementary File 4.

We used non-metric multidimensional scaling (NMDS) (Rohlf 1972; Bush and Brame 2010) to compare the taxonomic contents of individual localities within each stratum. A permutational analysis of variance (PERMANOVA) was used to evaluate whether stratigraphy, time period, or size were factors in determining taxonomic compositions of different sites (Anderson 2001, 2014). Both were performed using R Studio (Posit Team 2026). The locality-diversity dataset used for these analyses is included as Supplementary File 5.