Climate stability across the Eocene-Oligocene transition, southern Argentina

Fossilmammal teeth from mid-latitude southern Argentina ( ; 46 8 S) that closely bracket the Eocene-Oligocene transition show no resolvable change in oxygen isotope compositions. In combination with paleoﬂoral observations and geographic considerations, this ﬁnding implies not only that climate was essentially constant, despite interpretations else-where for major mid-and high-latitude cooling, but also that evolution of hypsodonty did not coincide with climate change during the Eocene-Oligocene transition. One possible explanation for Eocene-Oligocene transition climatic stability is that southern high-latitude cooling increased latitudinal temperature gradients and strengthened ocean circulation gyres, including the southward-ﬂowing Brazil Current in the western South Atlantic. Re-gionally increased heat transport in the western Atlantic offset global cooling, producing a nearly constant temperature in southern South America. A more radical interpretation, supported by some marine data, is that the paradigm of major global cooling at the Eocene-Oligocene transition is largely false, in that mean sea-surface temperatures changed very little.


INTRODUCTION
Many paleoclimate studies focus on benthic marine records because isotopic compositions of deep-sea foraminifera provide a quasicontinuous record of ocean temperatures and/ or compositions over many tens of millions of years (e.g., Miller et al., 1987;Zachos et al., 2001).These records delineate gradual cooling since ca.50 Ma, punctuated by distinct cooling and warming events.One of the biggest cooling events-a switch from so-called ''greenhouse'' to ''icehouse'' conditionsoccurred across the Eocene-Oligocene transition ca.33.5 Ma (Miller et al., 1987).The Eocene-Oligocene transition is thought to have been attended by major continental cooling at northern middle and high latitudes and the development of an Antarctic ice cap approximately half its present size (Miller et al., 1987;Wolfe, 1992;Zachos et al., 1994Zachos et al., , 2001)).A popular view of the causes of Eocene-Oligocene transition cooling ascribes a singular role to the development of the Antarctic Circumpolar Current via separation of Tasmania and South America from Antarctica and consequent reorganization of the world's ocean currents (Kennett, 1977).Deep-water circulation through the Tasmanian Seaway appears to have begun at 33.5 Ma (Exon et al., 2002), although there is no conclusive evidence that the Drake Passage was open to deep-water cir-culation until later (32.8 to ca. 30 Ma; Lawver and Gahagan, 1998;Latimer and Filippelli, 2002).If the Antarctic Circumpolar Current was important for Eocene-Oligocene transition cooling, then regions proximal to Antarctica should show the strongest evidence for a temperature decrease.
Gran Barranca (S458429490, W688449160; Fig. 1), southern Argentina, contains one of the few terrestrial sequences in the Southern Hemisphere that tightly brackets the Eocene-Oligocene transition.These rocks are fossiliferous, and so provide a unique opportunity to quantitatively investigate southern continental climate changes attending the Eocene-Oligocene transition via oxygen isotope analysis of fossil teeth.An increase in tooth crown height (hypsodonty) has also been observed in fossils from strata near the Eocene-Oligocene transition (Kay et al., 1999), so we evaluated whether evolution of tooth crown height (hypsodonty) there corresponded with this particular climate event.A similar approach for studying paleoclimate was pioneered by Bryant et al. (1996) in the northern Great Plains of the United States.

BACKGROUND AND SAMPLES
Work on stable isotopes in teeth (reviewed by Kohn and Cerling, 2002; see also Koch, 1998) indicates that the oxygen isotope composition of mammalian phosphatic tissue correlates strongly with rainwater composition, but weakly and negatively with relative humidity, potentially complicating paleoclimatic interpretation of isotope compositions.However, previous study of Eocene and Oligocene phytoliths from Gran Barranca (Mazzoni, 1979) plus geographic considerations imply that relative humidity could not have changed drastically (Appendix DR11 ).
If relative humidity was essentially constant, then tooth d 18 O values at Gran Barranca during this time interval must principally reflect rainwater composition.Rainwater isotope composition in turn depends on temperature and/or precipitation amount, as well as the source and movement of moisture (e.g., Dansgaard, 1964;Rozanski et al., 1993;Friedman et al., 2002).However, there is no evidence for geographic changes in moisture sources or for topographic influences and rain shadows.In particular, a transcontinental seaway east of the Andes appears to have been present throughout the time period (Romero, 1986;Smith et al., 1994;Fig. 1), negating any Andean topographic influence on isotope compositions (and buffering relative humidity).Furthermore, the northward movement of South America has been so slow (Shaw and Cande, 1990) that any latitudinal changes in moisture patterns must have been negligible.Modern precipitation in the region shows a strong influence of temperature on stable isotope compositions (Appendix DR2; see footnote 1).Therefore, tooth isotopic compositions should provide a robust proxy for paleotemperature.
Teeth in large herbivores require a few months to .1 yr to form.For enamel, mineralization begins at the crown of the tooth and sweeps toward the root, producing a thin shell at a rate of ;30-60 mm/yr (Passey and Cerling, 2002; see also summary by Kohn, 2004).Because climate and environmental d 18 O values in most areas vary seasonally, tooth enamel becomes isotopically zoned so that it encodes seasonality over the time interval of formation.However, not all tooth positions form simultaneously, so different teeth record different parts of the yearly seasonality.By analyzing several teeth from a single clade from one locality, focusing on the longest teeth, the range of compositions (i.e., isotopic seasonality) and mean composition can be identified (Kohn and Cerling, 2002).
Fossil teeth were analyzed from samples provided by the Museo de La Plata, La Plata, Argentina.Large notoungulates with typical body sizes .100kg were targeted to minimize species-dependent differences in d 18 O values and because some studies have suggested that the isotope compositions of larger animals are more directly linked to water compositional changes (Bryant and Froelich, 1995).Most teeth were obtained from two related families, Isotemnidae and Leontiniidae.Interfamily compositional differences were not resolvable.Sampling and analytical methods (Kohn et al., 2002; Appendix DR3; see footnote 1) involve isolation and analysis of the PO 4 component of enamel for d 18 O, which minimizes potential diagenetic and/or bacterial alteration (Kohn and Cerling, 2002).Bacterial alteration of biogenic phosphates is attended by precipitation of anomalous phosphate balls (e.g., Hirschler et al., 1990), and the excellent preservation of original biogenic structures in the enamel we analyzed also suggests minimal alteration.
Higher strata include basalts dated as 28.8 Ma.Thus, these samples are certainly earliest Oligocene and in comparison with highresolution marine records (Salamy and Zachos, 1999) postdate the Eocene-Oligocene transition.An important faunal distinction is that teeth from the three younger levels have major hypsodont (high-crowned tooth) elements, whereas teeth from the oldest level do not.

ISOTOPE RESULTS AND MODELS
Zoning profiles for enamel d 18 O (Appendix DR4; see footnote 1) from the occlusal (or wear) surface to the base of each tooth show isotope variability as expected for preserved seasonal isotope variations.Average isotope values (Fig. 2) reflect average yearly climate and show a small systematic increase: 15.2‰ (39.3 Ma, n 5 8), 15.9‰ (38.0 Ma, n 5 21), 16.5‰ (33.3 Ma, n 5 41), and 16.6‰ (30 Ma, n 5 9).Changes in median compositions are smaller (Fig. 2), and there are no apparent systematic shifts in maximum and minimum d 18 O values, which would correspond to summer and winter conditions, respectively.Although the data set is small, in comparing preand post-Eocene-Oligocene transition samples, it is extremely unlikely that the similarity of average compositions is a sampling artifact.There are at least 40 months represented from $5 yr (or partial years) each before and after the Eocene-Oligocene transition.Even if climate anomalies occurred on average in one-third of all years, the likelihood is ,1% that all teeth from before or after the Eocene-Oligocene transition are fortuitously from anomalous years.Anomalous months are even less likely to influence isotopic means.Ranges of composition are similar for three of the four different times: 5.5‰ (39.3 Ma), 4.5‰ (38.0 Ma), 5.5‰ (33.3 Ma), and 2.1‰ (30 Ma).
Tooth compositional changes corresponding to changes in climate can be modeled theoretically if the dependencies among mammal d 18 O values, meteoric water d 18 O values, temperature, humidity, and global ice volume are known (Kohn et al., 2002; Appendix DR5; see footnote 1).For these predictions, we make five assumptions.(1) The dependence of tooth enamel d 18 O on meteoric water d 18 O has a slope of 0.85‰/‰, intermediate among theoretical, global average, and taxon-specific observations (Kohn and Cerling, 2002).( 2) The correlation between the d 18 O of surface water and temperature is 0.3‰/8C, as indicated by mid-latitude records (Kohn et al., 2002).This value is well below the modern temperature dependence along coastal southern South America (;0.5‰/8C;Appendix DR2; see footnote 1), but more consistent with models of how global climate change affects isotopic compositions of precipitation (Boyle, 1997;Hendricks et al., 2000).(3) The dependence of tooth enamel d 18 O on relative humidity is ;20.15‰/%,intermediate between theoretical and empirical estimates (Kohn et al., 2002).( 4) Increased ice volume attending Downloaded from http://pubs.geoscienceworld.org/gsa/geology/article-pdf/32/7/621/3529543/i0091-7613-32-7-621.pdf by Universidad Nacional de La Plata user Eocene-Oligocene transition cooling caused a 0.5‰ increase in the d 18 O of seawater (Zachos et al., 1994(Zachos et al., , 2001)), which would cause a 0.5‰ increase in tooth d 18 O. ( 5) We assume that notoungulates in the clades investigated had sufficiently similar physiology and feeding behavior that their isotope compositions can be compared.For temperate conditions, the teeth of disparate modern herbivores have very similar isotope compositions (Kohn and Cerling, 2002), so this assumption would be potentially problematic only if relative humidities were low (,60%).Such low humidities are unsupported by paleofloral observations.
To calculate a d 18 O shift in the teeth preserved across the Eocene-Oligocene transition, estimates for the change in relative humidity and temperature are needed.We have argued for nearly constant relative humidity, consistent with the persistence of nearby regional water sources-the Atlantic Ocean to the east and the inland waterway to the west and southwest (Romero, 1986;Smith et al., 1994).However, a maximal change in relative humidity of 25% could contribute as much as a 10.75‰ change to tooth compositions.Temperature changes are more important.A general circulation model for the opening of the Drake Passage (Nong et al., 2000) suggests that the temperature decrease at lat 458S during the Eocene-Oligocene transition could have been ;15 8C.In comparison, (1) for continental North America, the temperature decrease across the Eocene-Oligocene transition at mid-latitudes has been estimated at ;10 8C (Wolfe, 1992); (2) the isotope shift exhibited by diagenetic hematite from the northern Great Plains, United States (Bao and Koch, 1998), combined with typical temperature dependencies of isotopes in precipitation, implies an at least transitory temperature decrease of $15 8C; and (3) paleoflora from the southern tip of South America (lat 52-548S) restrict the temperature change to #15 8C (Romero, 1986).Considering the proximity of our study area to the Antarctic Circumpolar Current and paleoclimate records at similar (but northern) latitudes, we therefore modeled a temperature shift across the Eocene-Oligocene transition of 210 8C, although specific geographic locations could have undergone greater or lesser temperature changes.For example, data from Ivany et al. (2000) suggest a smaller shift in sea-surface temperatures (SSTs) (;2 8C), but their samples are also from the much lower latitude Gulf of Mexico, which is expected to be less sensitive to Eocene-Oligocene transition cooling.
A 10 8C decrease in temperature would produce at least an ;2.5‰ decrease in tooth d 18 O values (0.3‰/8C • 210 8C • 0.85).Even accounting for effects of ice volume (10.5‰) and possible decreases in humidity (0 to 10.75‰), the expected decrease in tooth d 18 O values should still be .1‰.A stronger dependence of precipitation composition on temperature (0.5‰/8C; Appendix DR5; see footnote 1) coupled with no change in relative humidity implies an even larger predicted shift, i.e., a decrease possibly as large as ;4‰.Instead, tooth mean and median d 18 O values increased by ;0.5‰.In fact, the observed isotopic shift, the likely ice-volume effect, and the range of permissible changes to humidity together imply that the temperature change was between 12 and 23 8C.Although it may not be possible to rule out abrupt and brief cooling episodes, the similar mean isotope compositions as well as similar ranges of isotopic composition support climatic constancy.

INTERPRETATIONS
An essentially constant climate in southern Argentina across the Eocene-Oligocene transition requires that either major global cooling was somehow regionally suppressed at Gran Barranca (i.e., Gran Barranca was climatically anomalous) or the global cooling paradigm derived principally from northern continental records is false (i.e., Gran Barranca is ''normal'' and other continental records are anomalous).There is support for both possibilities.
Considering regional issues first, major cooling at high (southern) latitudes would have profound consequences for Atlantic circulation patterns.At the present time, the Malvinas and Brazil Currents converge north of the latitude of Gran Barranca (Fig 1;Webb, 1996).However, the Brazil Current was likely a major warm-water source to southern South America as far south as the latitude of Gran Barranca prior to opening of the Drake Passage (Kennett, 1983).Stronger latitudinal temperature gradients due to cooling at high latitudes would enhance gyral circulation, strengthening the warm Brazil Current.Such a process has been documented in the western Pacific, where local warming at lat ;208S occurred during the (global) Little Ice Age of the 1700s and 1800s, because of a coolinginduced strengthening of the South Pacific gyre (Hendy et al., 2002).If cooling at higher latitudes can cause local warming at lower latitudes, then a quasi-constant temperature must occur locally at some intermediate latitudepossibly the latitude of Gran Barranca.The most conclusive test of this hypothesis lies in marine sediment cores from middle and low latitudes, but preservation of planktonic foraminifera for these cores unfortunately is rather poor.Nonetheless, in principle, a strengthened South Atlantic gyre could be evidenced in paleontological, chemical, and/or isotopic rec-ords immediately offshore Africa and South America.
The more radical alternative is that the paradigm of profound mid-to high-latitudinal cooling of the continents is false.The records of SST that exist for the Eocene-Oligocene transition across the globe are actually much more consistent with essentially constant SSTs, but increased temperature seasonality (Zachos et al., 1994;Ivany et al., 2000), rather than globally widespread lower temperatures.Specifically, the compilation of SSTs for the late Eocene versus early Oligocene (Zachos et al., 1994) shows a 0 6 1 8C temperature shift at all latitudes, whereas seasonality analysis of otoliths from the U.S. Gulf Coast (Ivany et al., 2000) shows evidence for somewhat intensified winters in the early Oligocene, with only a small decrease in mean temperature.Paleontological evidence from the northern Great Plains of the United States shows virtually no change in mammalian taxa (Prothero and Heaton, 1996), but major losses of snails (Evanoff et al., 1992) and aquatic reptiles and amphibians (Hutchison, 1992).Oxygen isotopes in equid teeth (Bryant et al., 1996) show no systematic changes across the boundary, but interpretations were compromised by unexpected compositional variability, now understood to be the result of progressive precipitation of enamel and isotopic seasonality.Possibly the Eocene-Oligocene transition in the Great Plains was accompanied by increased seasonal temperature ranges, rather than a systematic temperature drop.In principle, this proposal could be validated (or refuted) by measuring isotope seasonality preserved in fossil teeth across the interval.Although we found no obvious evidence for increased seasonality in South America, a large change could have occurred in North America, particularly in the center of the continent where turnover of temperature-sensitive fauna is best documented.
In southern Argentina, the initiation of hypsodonty occurred between 39.3 and 38.0 Ma, ;5 m.y.prior to the Eocene-Oligocene transition.The increase in grass abundance (Mazzoni, 1979) and possible small increase in d 18 O values in teeth at that time suggest that a small decrease in relative humidity (;5%) under uniformly warm conditions could have caused the isotopic shift and stabilized grasslands, which in turn could have driven hypsodonty.Thus, the rise of South American hypsodonty could reflect climate change and its impact on floras, but started prior to the Eocene-Oligocene transition, which seems insignificant climatically.

Figure 2 .
Figure 2. Summary of isotope ranges and values through time.Thin vertical lines delineate Eocene-Oligocene transition (EOT).Small circles-individual measured compositions; stars-mean compositions; small boxes span central three analyses (medians); thick vertical black lines-total compositional ranges; short horizontal linesage uncertainties, which are smaller than symbol size for data at 38.0 and 33.3 Ma.Broken lines show range of predicted isotope trends for mean compositions relative to late Eocene, assuming 10 8C cooling at Eocene-Oligocene transition and 0.5‰ increase in d 18 O due to ice volume.Minimum shift assumes 0.3‰/8C temperature dependence and 5% decrease in relative humidity; maximum shift assumes 0.5‰/8C temperature dependence and no change in relative humidity.Disparity of these predictions compared to data illustrates magnitude of discordance between regional and global climates.Prediction assuming constant regional climate (solid line) is in better agreement with observed isotopic compositions.Deep-sea foraminiferal isotope curve (Zachos et al., 2001) is shown for reference.SMOW-standard mean ocean water; PDB-Peedee belemnite.