Edited by Paul E. Olsen, Columbia University, Palisades, NY, and approved May 15, 2015 (received for review March 25, 2015)
This is, to our knowledge, the first multiproxy study of climate and associated faunal change for an early Mesozoic terrestrial ecosystem containing an extensive vertebrate fossil record, including early dinosaurs. Our detailed and coupled high-resolution records allow us to sensitively examine the interplay between climate change and ecosystem evolution at low paleolatitudes during this critical interval of Earth's history when modern terrestrial ecosystems first evolved against a backdrop of high CO2 in a hothouse world. We demonstrate that these terrestrial ecosystems evolved within a generally arid but strongly fluctuating paleoclimate that was subject to pervasive wildfires, and that these environmental conditions in the early Mesozoic prevented large active warm-blooded herbivorous dinosaurs from becoming established in subtropical low latitudes until later in the Mesozoic.
A major unresolved aspect of the rise of dinosaurs is why early dinosaurs and their relatives were rare and species-poor at low paleolatitudes throughout the Late Triassic Period, a pattern persisting 30 million years after their origin and 10–15 million years after they became abundant and speciose at higher latitudes. New palynological, wildfire, organic carbon isotope, and atmospheric pCO2 data from early dinosaur-bearing strata of low paleolatitudes in western North America show that large, high-frequency, tightly correlated variations in δ13Corg and palynomorph ecotypes occurred within a context of elevated and increasing pCO2 and pervasive wildfires. Whereas pseudosuchian archosaur-dominated communities were able to persist in these same regions under rapidly fluctuating extreme climatic conditions until the end-Triassic, large-bodied, fast-growing tachymetabolic dinosaurian herbivores requiring greater resources were unable to adapt to unstable high CO2 environmental conditions of the Late Triassic.
One of the major predictions of models of elevated atmospheric CO2 is the increased frequency and magnitude of events comprising very high temperatures, an enhanced hydrological cycle, and increased precipitation extremes (1, 2). Because such environmental extremes act as limitations on organisms, past time intervals of elevated CO2 and associated climate extremes might be expected to profoundly influence biogeographic patterns, especially on land, which is relatively unbuffered climatically compared with the oceans. One such time of elevated CO2 was the Triassic Period, during which both dinosaurs and mammals first appeared. In particular, it has remained an open question why the global ecological dominance of dinosaurs was delayed in the tropics for at least 30 million years after their first appearance and diversification into the three major clades Sauropodomorpha, Theropoda, and Ornithischia (3, 4). Hypotheses proposed to explain this lag have focused largely on competition (or lack thereof) with nondinosaurian archosaurs, principally those on the line to crocodylians (pseudosuchians), but none provide a clear explanation for this unusual and persistent biogeographic pattern.
The rise of dinosaurs to ecological dominance was a diachronous evolutionary event (5⇓⇓–8). Small carnivorous early theropod dinosaurs were widespread at low paleolatitudes, whereas evidence for Triassic herbivorous dinosaurs (i.e., sauropodomorphs and ornithischians) in the tropics is completely absent (6, 7, 9, 10) (Fig. 1). In addition, tropical North American theropod dinosaurs were rare and species-poor (5, 7, 10) compared with higher-latitude assemblages. These patterns have been hypothesized to track largely zonal climatic conditions across Pangaea (6, 8, 11, 12) (Fig. 1), but detailed paleoclimatic data and mechanistic explanations have been lacking. Here, we argue these biogeographic patterns are a result of extreme environmental fluctuations in the tropics enhanced by high atmospheric CO2, which suppressed large-bodied herbivorous dinosaurs until after the end-Triassic mass extinction.
Late Triassic Pangean map showing latitudinal climate zones (11, 12) and the distribution of major dinosaur clades. See SI Appendix for occurrence data. Question marks indicate geochronologic uncertainty for the Thailand sauropodomorph and Argentine Laguna Colorada heterodontosaurid occurrences (i.e., they may be Early Jurassic in age). Each dinosaur symbol in most cases represents a region with multiple fossiliferous localities containing the illustrated clades.
We present, to our knowledge, the first high-resolution paleoenvironmental multiproxy record from the same sedimentary sequences that produce abundant early dinosaur and other vertebrate fossils (6, 8, 10). Specifically, we sampled fluvial and overbank sediments of the Upper Triassic Chinle Formation of the Chama Basin in north central New Mexico (13, 14). This nonmarine succession from low-paleolatitude Pangaea moved from ∼10°N to 14°N during the late Norian and Rhaetian (15), suggesting that this area would have experienced a semiarid climate through the entire sequence (11) (Fig. 1). The formation in this region contains exceptionally diverse and abundant vertebrate assemblages, which allows the early evolution of dinosaurs, their contemporaneous flora, and their paleoenvironment to be examined through time. Furthermore, tight age control is provided by a recent U–Pb radioisotopic age of 211.9 ± 0.7 Ma from the Hayden Quarry (HQ) in the lower portion of the Petrified Forest Member of the Chama Basin (7), and magnetostratigraphic data (16) that are consistent with a late Norian to Rhaetian age for the sequence.
We recovered abundant woody charcoal, coalified wood, palynomorphs, and other organic material from samples within greenish gray-colored reduced intervals, particularly those preserved within paleochannels (Fig. 2). Palynomorph data show a major change from a seed fern-dominated (Alisporites) assemblage with accessory gymnosperms (Patinasporites/Enzonalasporites, Camerosporites, and Monosulcites) in the Poleo Sandstone to one dominated by conifers and seed ferns (Patinasporites/Enzonalasporites, Alisporites, and/or Froelichsporites) in the lower portion of the Petrified Forest Member (Fig. 2). No palynomorphs were recovered from the upper portion of the Petrified Forest and lower portion of the siltstone members, but assemblages from the upper portion of the siltstone member exhibit a lower diversity, with the loss of Camerosporites and Froelichsporites, and much rarer Protodiploxypinus (Fig. 2). The relative abundance of these palynomorph taxa (Camerosporites, Alisporites, Froelichsporites, and Protodiploxypinus) varies inversely with that of Patinasporites/Enzonalasporites (Fig. 2). The abundance of taxa also changes rapidly within short stratigraphic intervals; Alisporites, Froelichsporites, and Patinasporites/Enzonalasporites vary by an order of magnitude through stratigraphically adjacent samples within the HQ interval. Fern spores are locally abundant in isolated samples from the lower portion of the Petrified Forest Member and upper portion of the siltstone member. These fluctuations suggest rapid changes in local floral assemblages.
Organic carbon isotope (δ13Corg), palynomorph, and vertebrate fossil records from the Upper Triassic Chinle Formation of the Chama Basin, New Mexico. Stratigraphic section modified from ref. 6; see SI Appendix for detailed HQ sections and vertebrate biostratigraphic data. Radioisotopic age from ref. 7. Blue δ13Corg data are carbon isotopic composition of bulk organic matter, red are charcoal, and green are vitrinized wood. The pCO2 was calculated following the methodology of Schaller et al. (27, 35, 40); error bars represent S(z) = 3,000 ± 1,000 ppm (see SI Appendix, SI Text, for details). Samples from the Hartford Basin (purple squares) were aligned using U–Pb dates from pedogenic carbonates (29) and the Newark age model [further correlation to Newark pCO2 data (35) will require additional geochronologic constraints for the Ghost Ranch section]. Vertebrate fossil localities: CaQ, Canjilon Quarry; CoQ, Coelophysis Quarry; H2, H3, and H4, HQ paleochannels; SQ, Snyder Quarry. Vertebrate taxa: A, Drepanosauridae (nonarchosauriform archosauromorph); B, Vancleavea campi (nonarchosaur archosauriform); C, Phytosauria; D, Aetosauria; E, Shuvosauridae; F, noncrocodylomorph Loricata; G, Crocodylomorpha; H, Dromomeron romeri (lagerpetid dinosauromorph); I, Silesauridae (nondinosaurian dinosauriform); J, nonneotheropod Theropoda; K, Neotheropoda.
The carbon isotopic composition of bulk organic matter (δ13Corg) covaries with palynomorph relative abundance, with the former displaying multiple positive/negative fluctuations (1–2‰) in the Petrified Forest and siltstone members (Fig. 2). The upper portion of the section is less densely sampled because of a rarity of organic material in these red, oxidized sediments, but still displays continued variability of similar magnitude, particularly at the Coelophysis Quarry (Fig. 2). These redbed samples have total organic concentrations (TOC) comparable to many samples from reduced layers lower in section (SI Appendix, Table S1), and there is no statistically significant relationship between TOC and δ13Corg values, so it is unlikely these fluctuations represent variable TOC. Although the Poleo Sandstone at the base of the stratigraphic section is not densely sampled because of its predominantly coarse grain size, data from organic-rich siltstone lenses yield more-constant isotopic values. Similar variation in carbon values through the section are also exhibited in wood and charcoal isotopic records, indicating a consistent signal that cannot be explained simply by changing source composition or taphonomy. Furthermore, each negative or positive shift of this magnitude occurs in a stratigraphic interval of 1 m or less within conformable sequences.
Comparison of this δ13Corg record with environmentally diagnostic palynomorphs suggests that δ13Corg is likely influenced directly or indirectly by environmental, plausibly climatic, changes. Both Alisporites and the Patinasporites/Enzonalasporites pollen complex are generally considered indicators of relatively arid Triassic environments (17). However, within this larger climatic context, Alisporites is more abundant in more humid environments and the Patinasporites/Enzonalasporites vesicate pollen complex is more abundant in drier environments, based on lacustrine strata of similar age in eastern North America (18). The relative abundance of both Camerosporites (R2 = 0.68) and Alisporites (R2 = 0.63) negatively correlates with δ13Corg values, whereas the Patinasporites/Enzonalasporites pollen complex (R2 = 0.75) correlates positively with δ13Corg values, suggesting that humidity-dependent plant 13C discrimination may be responsible for the isotopic shifts, with more 13C-depleted values being associated with distinctly more humid climates (e.g., ref. 19). Both the palynomorphs and covarying δ13Corg values therefore reflect strong environmental fluctuations, all potentially related to high atmospheric pCO2 (Fig. 2), and thus unusually strong environmental pressures associated with climatic change.
The extensive charcoal record in the Petrified Forest Member provides additional evidence of paleoenvironmental variability and aridity. Abundant charred wood (ranging from <1 mm to >8 cm in maximum diameter) occurs in a minimum of four stratigraphically different paleochannels (Fig. 2), suggesting a pervasive influence of paleowildfire on the landscape. Reflected light microscope data from the lower Petrified Forest Member in the HQ paleochannels (Fig. 2) indicate significant differences in the intensity of these fires. Woody charcoal from the HQ 2 paleochannel formed at elevated temperatures of at least 680 °C, whereas in the slightly stratigraphically higher HQ 3 paleochannel, the sampled charcoal was exclusively semifusinitic (only partially charred) and generated by low temperature (320 °C) combustion (see SI Appendix, SI Text and Table S3). These values bracket those previously reported (350–450 °C) from the nearby Snyder Quarry (20). Widely variable paleowildfire burn temperatures reflect an arid but fluctuating environment that allowed variable accumulation of combustible organic matter.
Wildfire activity likely stimulated plant community change. The abundance of large macroscopic woody charcoal indicates these were probably surface fires (21), which have greater potential to enhance both vegetation mortality and postfire runoff and erosion (21). Based on these conditions, we hypothesize a scenario where environmental changes led to impaired ecosystem function, resulting in fuel buildup due to increased floral mortality, and hotter fires, as seen at HQ 2. A hot fire would have enhanced vegetation death, damaging soils and increasing erosion, and causing positive feedback effects that accelerated floral change. Alternatively, increased aridity may have accomplished the same by reducing fuel moisture content. This would have also promoted fire ignition and facilitated extensive burns at high temperatures.
These wildfire data are consistent with the linkage between xerophytic palynomorph abundance and bulk organic carbon isotope values (Figs. 2 and 3), which we hypothesize represents a rapidly and dramatically changing climate signal that is likely an indirect indicator of fluctuations in aridity. Changes in isotopic values and palynomorph composition can reflect direct effects of climate (temperature, precipitation, atmospheric gas inventories), more indirect effects through plant community change or plant physiology, or a combination of these factors. Specifically, organic carbon isotopic values can directly reflect changes in aridity by way of decreased primary productivity, as observed in modern vascular plants (22). Alternatively, climate shifts almost certainly changed the taxonomic composition of plant communities, which would lead to community-induced changes in carbon isotope value, because both Triassic and modern C3 plants display taxonomic differences in carbon isotope discrimination (23⇓–25). Although low organic contents in some samples resulted in larger relative uncertainties in a few instances, the large isotopic excursions exceed these uncertainties, and many of the shifts are recorded by samples with precise values (see SI Appendix, Table S1 and Fig. S1). Similarly, although we cannot exclude the possibility that the excursions in the bulk organic samples partly reflect differential concentration of charcoal and coalified wood versus other organic matter because of relative enrichment during diagenesis (26), when we restrict consideration to just charcoal and coalified wood, large excursions in isotopic composition are still evident, indicating that they are not a taphonomic effect of differential preservation. Furthermore, the observed excursions are larger than differences caused by variable burning temperatures (cf. ref. 26), indicating that wildfires cannot be the sole source of the variations.
Linear regression of δ13Corg values versus the relative abundance of select palynomorph taxa.
The observed relationship between δ13C and floral change, wildfires, and overall climate variability is set against the backdrop of unusually high atmospheric CO2 over the interval we sampled (27). We calculated atmospheric pCO2 from multiple stable carbon isotope analyses of soil carbonate and preserved organic matter, using a standard diffusion model (28) (see SI Appendix, SI Text and Table S2), as ∼1,200 ± 400 ppm at the base of the section, increasing to ∼2,400 ± 800 ppm near the top. The minimum pCO2 levels at ∼212 Ma are recorded in both the Chinle Formation data presented here and studies of contemporaneous exposures in the Hartford Basin of eastern North America (27) (see ref. 29 for geologic age), providing strong evidence for the global relevance of these results. The relatively small offsets between these data are likely the result of unconstrained differences in paleoaltitude at these two widely separated sites, assumptions in the S(z) parameter, or slight taxonomic differences in plant fractionation (see SI Appendix, SI Text). The observed secular pCO2 increase toward the Rhaetian could be a result of increased midocean ridge degassing or, more plausibly, a decrease in the rate of continental weathering, driven by the establishment of weathering-limited environments in Pangaea’s interior (30); the δ13C record of marine carbonates indicates no decrease in the rate of organic carbon burial during this time (31). Regardless of the source of this change, the overall very high pCO2 levels are expected to be associated with a higher frequency of seasonal extremes, consistent with the evidence presented here for large-magnitude climate fluctuations. In contrast, the overall evidence of increasing aridity from the Petrified Forest Member into the siltstone member is most simply attributed to the northward drift of central Pangaea from the humid tropics into the more arid subtropics (12).
Our data are consistent with previous estimates of precipitation and temperature (32, 33), and suggest Pangaea’s low-paleolatitude continental interior experienced strong environmental fluctuations in the Late Triassic. These conditions, accentuated by frequent wildfire activity, brought about vegetational change over both long-term and potentially seasonal timescales (34). Despite (or perhaps because of) these major and repeated environmental fluctuations, vertebrate assemblages changed little through the succession. Members of the same major clades are preserved through the section, with only species-level turnover apparent at some intervals (Fig. 2). This suggests that the vertebrate biota were resistant to the fluctuating environment and the changing plant communities associated with it. Even with a well-sampled fossil record, it is clear that dinosaurs and their close relatives remained rare components of the fauna, comprising less than 15% of specimens except for the taphonomically aberrant Coelophysis Quarry (5, 7), and were dominated in diversity, abundance, and body size by pseudosuchian archosaurs (7). Furthermore, large-bodied sauropodomorphs are completely absent from not only the Chinle Formation but from all low-latitude localities in Triassic Pangaea (Fig. 1).
Although our paleoenvironmental results are, to our knowledge, the first quantitative multiproxy record from sites containing a rich early dinosaur fossil record, they support cyclostratigraphic and qualitative sedimentological evidence of hyperseasonality and extremely high CO2 during the Late Triassic throughout the tropics (e.g., refs. 11, 14, and 32⇓⇓–35). Dinosaurs are generally rare in these areas, and there is no evidence for the large herbivorous forms found at higher latitudes (5, 7, 8, 36). In contrast, climate modeling, sedimentological evidence (e.g., coals and paleosols), and stable isotope data provide evidence for cooler and more humid conditions in the temperate high latitudes of both Laurasia and Gondwana (37⇓⇓–39), where large herbivorous dinosaurs are common. Thus, all available data suggest that the climate extremes we observe in the western Pangaea record and their ecological consequences were a feature of the Late Triassic tropics in general.
Our data demonstrate that a generally stable vertebrate community with a rarity of dinosaurs (especially large-bodied forms) coexisted with dramatically fluctuating plant communities, the latter reflecting highly variable environmental conditions enabled by high atmospheric pCO2. We propose that this unstable environment, characterized by recurring arid/humid extremes, prevented the establishment of the types of dinosaur-dominated faunas that are observed in coeval but much higher-latitude records from South America, Europe, and southern Africa (5), where aridity and temperatures were less extreme. These faunal contrasts support the idea that early dinosaurs were latitudinally sorted (6, 8). Our results suggest that fluctuating aridity in tropical and subtropical Pangea may have been an important driver in this sorting, resulting in resource-limited conditions that could not support a diverse community of fast-growing tachymetabolic large dinosaurs, which required a particularly verdant and stable environment to thrive. This could explain why Triassic dinosaur faunas at low latitudes are restricted to small, slower-growing carnivorous forms, whereas large-bodied herbivores, including sauropodomorph dinosaurs, are absent at low paleolatitudes during the Late Triassic “hothouse.” The unpredictable availability and composition of plant food and water resources in this environment would have been a challenge to the high metabolic requirement of large-bodied herbivorous dinosaurs, whereas carnivorous small theropods and other small-bodied dinosauromorphs were not so dependent. Likewise, diverse and abundant smaller-bodied herbivorous, omnivorous, and carnivorous pseudosuchians, with their lower resource requirements (41, 42), could withstand the dramatic climatic fluctuations.
Thus, we provide, to our knowledge, the first mechanistic explanation for the diachronous rise of dinosaurs during the Late Triassic Period. Furthermore, these data demonstrate the long-term ecological consequences of high CO2 and climate extremes over geological timescales and provide context for predictions of anthropogenic climate change.
All samples (isotopic, palynological, and charcoal) were taken from fresh, unweathered rock and placed in precise measured stratigraphic sections (see SI Appendix, SI Text). These sections were correlated to each other using both the top of the Poleo Sandstone and base of the Entrada Sandstone as a datum. Stratigraphic placement of vertebrate fossil localities followed the same method. Isotopic samples were prepared using standard methods (SI Appendix, SI Text). Charcoal samples were embedded in epoxy resin, polished, and analyzed under a reflected light microscope with an oil immersion objective; paleowildfire temperatures were calculated based on reflectance values from published experimental data (see SI Appendix, SI Text). The carbon isotopic composition of bulk organic matter of sediment and wood was analyzed by mass spectrometry (see SI Appendix, SI Text). Organic and inorganic carbon isotope measurements from paleosols were used to estimate the concentration of atmospheric CO2, according to the soil diffusion model of Cerling (see SI Appendix, SI Text).
We thank A. Downs, A. Kasprak, D. Musher, and R. Price-Waldman for assistance with fieldwork and C. Johnson for assistance with δ13Corg analysis. Fieldwork and research were funded by the US National Science Foundation (EAR 0801138 to J.H.W. and EAR 1349650, 1349554, 1349667, and 1349654 to R.B.I., J.H.W., N.D.S., S.J.N., and A.H.T.), Richard Salomon Foundation (J.H.W.), National Geographic Society Research & Exploration Grant 8014-06 (to K. Padian, R.B.I., S.J.N., N.D.S., and A.H.T.), University of California Museum of Paleontology (R.B.I.), University of Utah (R.B.I.), the Grainger Foundation (I.J.G.), the Dyson Foundation (M.F.S.), Field Museum of Natural History Women's Board (N.D.S.), and Geocenter Denmark (S.L.). S.L. publishes with the permission of the director of the Geological Survey of Denmark and Greenland. Fieldwork was conducted with the permission and support of Ghost Ranch Conference Center.
↵2Present address: The Dinosaur Institute, Natural History Museum of Los Angeles County, Los Angeles, CA 90007.
Author contributions: J.H.W. and R.B.I. designed research; J.H.W., S.L., R.B.I., I.J.G., M.F.S., and M.D. performed research; J.H.W., S.L., R.B.I., I.J.G., M.F.S., S.J.N., N.D.S., and A.H.T. analyzed data; and J.H.W., S.L., R.B.I., I.J.G., S.J.N., N.D.S., and A.H.T. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1505252112/-/DCSupplemental.
