Magnetic stratigraphy of the Eocene-Oligocene floral transition in western North America
Published:January 01, 2008
- PDF LinkChapter PDF
Donald R. Prothero, 2008. "Magnetic stratigraphy of the Eocene-Oligocene floral transition in western North America", Paleontology of the Upper Eocene Florissant Formation, Colorado, Herbert W. Meyer, Dena M. Smith
Download citation file:
Eocene and Oligocene floras of the western United States show a climatic deterioration from warmer conditions to much cooler and drier conditions. Recent 40Ar/39Ar dates and magnetic stratigraphy have greatly improved their correlation. In this study, the uppermost Eocene Antero Formation, Colorado, is entirely reversed in polarity, and is correlated with late Chron C13r, based on 40Ar/39Ar dates of 33.77–33.89 Ma. The early Oligocene Pitch-Pinnacle flora of Colorado is within rocks of normal polarity, and best correlated with Chron C12n (30.5–31.0 Ma), based on 40Ar/39Ar dates of 32.9–29.0 Ma (although correlation with Chron C11n is also possible). The late Oligocene (40Ar/39Ar dated 26.26–26.92 Ma) Creede flora of southwestern Colorado is correlated with Chron C8r. The early Oligocene (40Ar/39Ar dated at 31.5 Ma) Granger Canyon flora in the Warner Mountains, near Cedarville, northeastern California, is correlated with Chron C12r. These results are compiled with previously published dates and magnetic stratigraphy of the Eugene-Fisher floral sequence in western Oregon, the Bridge Creek floras in central Oregon, other floras in the Warner Mountains of northeast California, and the Florissant flora of central Colorado. In Colorado, the climatic change seems to have occurred between the Florissant and Antero floras, and is dated between 33.89 and 34.07 Ma, or latest Eocene in age, although the Pitch-Pinnacle flora suggests that the deterioration was less severe and took place in the early Oligocene. In northeast California, the dating is not as precise, so the climatic change could have occurred between 31.5 and 34.0 Ma (probably early Oligocene). In western Oregon (Eugene and Fisher Formations), the change occurs between the early Oligocene Goshen flora (33.4 Ma) and the early Oligocene Rujada flora (31.5 Ma). In the John Day region of Oregon, it occurs before the oldest Bridge Creek flora, dated at 33.62 Ma (right after the Eocene-Oligocene boundary). Thus, only two of these four floral sequences (Eugene, Oregon, and Cedarville, California) clearly show the early Oligocene climatic change consistent with that documented in the global marine record, whereas the climatic change was seemingly abrupt in the late Eocene in Colorado between 33.89 and 34.07 Ma, and also sometime during the late Eocene (before 33.62 Ma) in central Oregon.
The Florissant flora of Colorado provides one of our best windows on the fauna and flora of the late Eocene, the last pulse of warming before the “Oligocene deterioration” that was first documented by Wolfe (1971, 1978). Both the floras and the faunas provide evidence of relatively warmer and wetter conditions than prevailed in the Oligocene, just a million years later. However, to place this evidence in context, it is helpful to examine floras and faunas that occurred before and after Florissant, and also to look at regions outside the Colorado Rockies.
In recent years, the application of magnetic stratigraphy to the calibration of the faunal changes in Eocene-Oligocene land mammals (Prothero and Emry, 1996) and marine invertebrates and microfossils (Prothero, 2001) has made great strides. In most cases, we now have the stratigraphic resolution and chronostratigraphic precision to date most faunas to the nearest 100,000 yr, and correlate them all to a global time scale, allowing direct correlation of the global deep-marine climatic record with the land record and shallow-marine record of North America. However, fewer magnetostratigraphic studies have been conducted on important fossil plant localities, especially those near the Eocene-Oligocene transition. Important magnetostratigraphic studies of floras from the San Diego middle and late Eocene (Walsh et al., 1996), the Eocene-Oligocene sequence near Eugene, Oregon (Retallack et al., 2004), and the upper Eocene Florissant Formation in Colorado (Prothero and Sanchez, 2004) have now been published, but those results have not been summarized in one place. In addition, many new floras have been sampled, and two of those are reported in this paper. Studies such as those by Myers (2003) have updated the chronology of many of those floras, but the new results have refined the correlations even further.
Such a refined correlation is highly desirable, because the sequence of floras in western North America strikingly documents the details of climatic change in this region through the Eocene and Oligocene (MacGinitie, 1953; Wolfe and Hopkins, 1967; Axelrod and Bailey, 1969; Wolfe, 1971, 1978, 1992, 1994; Wing, 1987; Myers, 2003). Indeed, the dramatic cooling and drying near the Eocene-Oligocene boundary was the original basis for the term “Oligocene deterioration” of Wolfe (1971), renamed the “Terminal Eocene Event” by Wolfe (1978). According to Wolfe (1978), this floral change was the most dramatic of the entire Cenozoic, yet later work by Wolfe (1994) showed that the climatic change was not as extreme as originally suggested, and this has been corroborated by Retallack et al. (2004). In addition, there was confusion about dating in the earlier analyses published by Wolfe (1971, 1978). Most of those problems have now been resolved (Wolfe, 1992; Myers, 2003), but further refinement of dating is always valuable to test previous hypotheses.
This paper will summarize previous magnetostratigraphic chronologies of important middle to late Eocene and Oligocene floras in western North America, and provide new data for four important plant localities that have been recently sampled and analyzed. More stratigraphic sections and floras need to be studied, of course, but these results already point out some interesting problems that have not yet been resolved. The precise calibration of these floras, in turn, is important to assessing the magnitude and timing of paleotemperature change in North America during the Eocene-Oligocene transition.
Four new sections were sampled for this study. The first was the classic upper Eocene Antero Formation and its floras from the South Park Basin, Park County, central Colorado (Figs. 1A, 1B). This formation was first named by Johnson (1937), and fully described by Stark et al. (1949, p. 63–66) and Epis and Chapin (1975) based on exposures north of Hartsel, Colorado. Although Stark et al. (1949) reported an aggregate thickness of the Antero Formation of over 200 m in some places, currently there are no continuous surface exposures that would allow collection of that much section. Instead, sampling was conducted in several isolated roadcuts northeast of Hartsel (sites 1–4, locations given in Table 1) and in the southern part of the South Park Basin, southwest of Hartsel (three sites spanning ∼10 m of section, locations given in Table 1).
The second section was collected in the Pitch-Pinnacle Formation, on the western flank of the Sawatch Range just southwest of Monarch Pass, Saguache County, Colorado (Fig. 1B). Described in detail by Gregory and McIntosh (1996), our sampling followed the main section in their paper (Gregory and McIntosh, 1996, their Figure 5), concentrating on a large mine excavation (“Pitch-Pinnacle quarry” of Gregory and McIntosh, 1996, their Figure 2) which yields the main flora (lat 38°23.158′N, long 106°19.284′W), and on roadcuts uphill and down section from the mine pit. Seven sites, spanning ∼14 m of section, were taken wherever exposures allowed (locations given in Table 1).
The third section studied was from the late Oligocene Creede flora (Axelrod, 1987; Wolfe and Schorn, 1989), in the Creede Formation in Mineral County, south-central Colorado (Fig. 1B). Reynolds et al. (2000) did an extensive paleomagnetic study of the subsurface portion of this formation obtained by drilling cores up to 703 m long, but they did not sample the surface sections of the uppermost part of the Creede Formation, which are the source of the well-known plant and insect fossils (Axelrod, 1987; Wolfe and Schorn, 1989, 1990). The best and most accessible exposure, which was sampled for this study, is the classic Bridsey Gulch section (UCMP PA 574) of Axelrod (1987), just north of Highway 149 (SE 1/4 SW 1/4 NW 1/4 Sec. 6, T. 41N., R. 1E., Creede 7.5 min quadrangle, Mineral County, Colorado; lat 37°49.769′N, long 106°55.087′W). Six sites spanning ∼25 m of section were sampled, up to the point where the exposures were completely covered as the steep slopes gave way to a shallow bench. Fossil plants, insects, and other organisms are found in many levels in this section (Axelrod, 1987), although they are most abundant at the road level and at several quarries that were encountered during the traverse.
The fourth section was taken above and below the Granger Canyon plant locality (UCMP PA 760), in the Warner Range (Fig. 1A) near Cedarville, in northeast California (Myers, 1998, 2003). It is located at 1440 m elevation on the north slope of Granger Canyon (SE 1/4 SE 1/4 section 30, T. 42N., R. 16E., Warren Peak 7.5 min quadrangle, Modoc County, California; lat 41°28.470′N, long 120°11.207′W). The Warner Range contains a thick sequence of mostly volcanic rocks spanning a long interval of the Cenozoic (Duffield and McKee, 1986; Carmichael et al., 2006), but this particular fossil locality is a fine-grained lacustrine shale sandwiched between ignimbrites and volcanic agglomerates that yields an early Oligocene flora (Myers, 1998, 2003). Other important floras are known from the Cedarville area in the Warner Mountains, but they proved to be less accessible. I hope to return and sample those floras at a future date.
Samples were collected as oriented block samples in the field and subsampled with a drill press in the lab to produce a core. At least three samples were collected at each site, although for many sites where sampling was easier, four to six samples were obtained (Table 1). Poorly consolidated samples were hardened with sodium silicate in the field. Samples that were too fragile to drill were placed in plastic ring molds, into which I poured Zircar aluminum ceramic and which I then dried in the magnetically shielded room, to form coresized cylinders. Samples were measured on the 2G Enterprises cryogenic magnetometer at Occidental College, using a Caltech-style automatic sample changer. After measuring natural remanent magnetization (NRM), each sample was treated with alternating field (AF) demagnetization at 2.0, 4.0, 6.0, 8.0, and 10.0 millitesla (mT) to determine the coercivity behavior of the sample, and to demagnetize any multi-domain grains before their remanence was baked in. Every sample was then thermally demagnetized at 50 °C steps from 200° to 680 °C to examine the demagnetization behavior in detail. This process removes any chemical remanent overprints due to iron hydroxides such as goethite (which dehydrates at 200 °C), and shows how the samples behaved as the Curie temperature of magnetite (578 °C) and the Neel temperature of hematite (680 °C) were approached.
Results were graphed on orthogonal demagnetization (“Zijderveld”) plots (Fig. 2), and average directions of each sample were determined by the least-squares method of Kirschvink (1980). Mean directions for each site were then analyzed using Fisher (1953) statistics, and classified according to the scheme of Opdyke et al. (1977).
Antero Formation, Colorado
A representative orthogonal demagnetization plot of the Antero Formation is shown in Figure 2A. All the sites taken in this study (Table 1) were reversed in polarity, consistent with results from the Antero tuff reported by McIntosh and Chapin (2004, Table 3, p. 227). As shown by Figure 2A, there was little overprinting of the magnetic signal, and only a single component of remanence appears to be present. Judging from the large response to AF demagnetization, the remanence must be held in a low-coercivity mineral such as magnetite. This is confirmed by the fact that the remanence completely disappeared above the Curie point of magnetite, 578 °C. The mean direction for all Antero samples was D = 184.2, I = −34.8, k = 6.5, α95 = 12.1, n = 26. The 40Ar/39Ar date of 33.77 ± 0.10 Ma (McIntosh and Chapin, 2004) for the Antero Tuff, or the 40Ar/39Ar date of 33.89 ± 0.25 Ma (Obradovich, 2003, personal commun.) suggest that these reversed rocks correlate with late Chron C13r, or latest Eocene in age.
Pitch-Pinnacle Formation, Colorado
A representative orthogonal demagnetization plot of the Pitch-Pinnacle rocks is shown in Figure 2B. Like most Pitch-Pinnacle sites, the sample shown in Figure 2B had a slight overprint to the northwest, which was removed by the first thermal step at 100 °C. The remanence of the sample then decayed steadily to the origin, and was almost completely gone by 600 °C. This suggests that it is largely held in magnetite, an interpretation that is corroborated by the large response to AF demagnetization (consistent with the remanence being held in a low-coercivity mineral). All the sites sampled through Gregory and McIntosh's (1996, Fig. 5) sequence were normal in polarity (see Fig. 3), except for the lowest site, up the road from the main locality and just outside the mine property entrance (near the “12” notation on the map of Gregory and McIntosh, 1996, Fig. 2; exact location given in Table 1). They were all Class I sites of Opdyke et al. (1977), so they were statistically removed from a random distribution at the 95% confidence interval. The mean of all normal sites was D = 20.7, I = 50.7, k = 6.5, α95 = 13.5, n = 21, and the mean of the single reversed site was D = 178.1, I = −47.4, k = 5.4, α95 = 59.2, n = 3. These directions are antipodal within error estimates, so this positive reversal test shows that the characteristic remanence has been obtained and overprinting has been removed. Inverting the reversed directions 180° and averaging them with the normal directions, the formational mean direction was D = 17.7, I = 50.5, k = 6.5, α95 = 12.6, n = 24 (Table 1).
Gregory and McIntosh (1996) interpreted 40Ar/39Ar dates from epiclastic sanidines in these rocks to indicate an age of deposition between 32.9 Ma and 29.0 Ma. These dates indicate that the best correlation of this normal polarity sequence is with earliest Oligocene Chron C12n (30.5–31 Ma), with the lowest reversed site (site 7) possibly correlating with latest Chron C12r. However, the imprecision of the younger age estimate allows the possibility of correlation with Chron C11n. Given that the flora seems to be intermediate in temperature estimate between Florissant and Creede, and most similar to that of Florissant, it seems more likely that the correlation with Chron C12n-C12r is correct.
Creede Formation, Colorado
Representative orthogonal demagnetization plots of the Creede Formation are shown in Figures 4A and 4B. Reynolds et al. (2000) obtained mixed results for their samples from the Creede drill cores, with some samples yielding stable results and others giving directions that were unstable or difficult to interpret. My analysis from the surface exposures of the uppermost Creede Formation at Bridsey Gulch produced both normal and reversed polarities (Table 1), with most samples yielding a stable interpretable result, typically with only a single component and no overprinting (Figs. 4A–4B). Nearly all samples showed minimal decreases in intensity during AF demagnetization, suggesting that a high-coercivity mineral overprinting (probably from iron hydroxides such as goethite) is present. However, samples decreased in intensity rapidly through thermal demagnetization above 200 °C (the dehydration threshold for goethite), and almost all remanence was gone by 600 °C, suggesting that most of the remanence is held in magnetite and not in hematite or some other high-coercivity mineral.
Averaging all the stable normal and reversed directions produced a normal mean of D = 17.1, I = 55.6, k = 12.5, α95 = 17.7, n = 7, and a reversed mean of D = 197.2, I = −42.7, k = 4.4, α95 = 22.2, n = 13. These directions are antipodal within error estimates (Table 1), yielding a positive reversal test and suggesting that the overprinting has been removed and the directions are primary (Fig. 5). Inverting the reversed directions through 180° and averaging all directions produces a formational mean of D = 17.4, I = 49.5, k = 5.7, α95 = 15.4, n = 20.
Magnetostratigraphic results of the Bridsey Gulch section are shown in Figure 6. All sites were Class I sites of Opdyke et al. (1977), meaning that they were significantly different from a random distribution at the 95% confidence level. Two sites spanning the lower 5 m of section are normal in polarity. The remaining four sites, spanning the upper 20 m of section, are all reversed in polarity.
Correlation of the Creede flora section is shown in Figure 7. Lanphere (2000) and Reynolds et al. (2000) indicated that the 40Ar/39Ar dates bracketing the Creede Formation range from 26.26 ± 0.04 Ma on the highest Fisher Dacite above the Creede Formation to 26.92 ± 0.07 Ma for the Snowshoe Mountain Tuff underlying the Creede Formation, suggesting that the entire 700+ m of the Creede Formation was deposited in 550,000–770,000 yr or less. The age of 26.26–26.92 Ma on the Creede Formation suggests that the reversed-normal sequence at Bridsey Gulch correlates with some part of Chron C8r–C9n, although the many polarity changes documented by Reynolds et al. (2000) for the lower 700 m of the section in the core samples apparently must also be accommodated within the narrow window of time of 550,000–770,000 yr spanning Chrons C8n–C9n (Fig. 7). Whatever interpretation is adopted, the tight age constraints from the 40Ar/39Ar dates clearly indicate that the Creede flora is earliest late Oligocene on the standard time scale.
Granger Canyon, Northeast California
The section at Granger Canyon (Fig. 8) was much shorter, because most of the exposures were thick ignimbrites and volcanic agglomerates that were very brittle and resisted hand sampling, and consequently most sites had only a few samples that survived the sampling process. Interbedded within this sequence is the thin shaly unit that produces the Granger Canyon flora (Fig. 8). Most of the sites produced an apparent reversed polarity (Fig. 4C) with a high-coercivity overprint probably held in goethite or other iron hydroxides. However, the remanence was completely gone by 600 °C, suggesting that magnetite was the primary carrier of the remanence, and little hematite was present. The single site in the lower agglomerate (Fig. 4D) produced a normal polarity, which seems to be held in both magnetite and hematite, given that significant remanence was present even at 680 °C.
Averaging all the stable normal and reversed directions produced a normal mean of D = 22.8, I = 11.2, k = 7.8, α95 = 107.8, n = 2, and a reversed mean of D = 193.5, I = −27.6, k = 8.1, α95 = 19.3, n = 10. These directions are antipodal within error estimates, yielding a positive reversal test and suggesting that the overprinting has been removed and the directions are primary. Inverting the reversed directions by180° and averaging all directions produces a formational mean of D = 15.4, I = 24.6, k = 8.3, α95= 16.8, n = 12.
The magnetostratigraphic interpretation of the Granger Canyon is shown in Figure 8. As stated earlier, the lower ignimbrite-agglomerate apparently is normal in polarity, although this is based on a single site with only two surviving samples. The fossiliferous shale interval and the upper ignimbrite are entirely of reversed polarity, spanning at least 20 m of section sampled for this study. The total available section is much thicker in the field, because there are dozens of separate ignimbrites spanning hundreds of meters of section in the Warner Mountains.
Based on the 40Ar/39Ar date of 31.5 ± 0.4 Ma from a pebble breccia immediately below the floral locality (Myers, 1998, 2003), the best correlation seems to be with Chron C12r, or early Oligocene (Fig. 7). The single site of normal polarity in the lower unit does not appear to match the long interval of Chron C12r, which is reversed in polarity from 31 to 33 Ma (Fig. 7). However, during the early Oligocene there were numerous short events of normal polarity in Chron C12r (Hartl et al., 1993) that might explain this behavior. Alternatively, the normal polarity is based on a single site with only two stable samples, so it is possible that normal overprinting was not removed from this high-temperature ignimbrite, and that the normal polarity is not a primary or characteristic direction. In any case, the reversed portion of the section (containing the flora) is consistent with the correlation with Chron C12r and with the 31.5 ± 0.4 Ma date just below the flora.
Correlation of some important Eocene-Oligocene floral localities is shown in Figure 7. In southwestern Colorado, four important floras are now calibrated by magnetic stratigraphy. The upper Eocene Florissant Formation (Evanoff et al., 2001; Prothero and Sanchez, 2004) is correlated with early Chron C13r, based on its reversed polarity and a 40Ar/39Ar date of 34.07 ± 0.10 Ma on the upper part of the section (Evanoff et al., 2001).
The Florissant flora (Meyer, 2003) is famous for its warm temperate characteristics, consistent with a warming trend in the late Eocene. A variety of temperature estimates (Table 2) have been calculated for this flora over the years (summarized by Wolfe, 1994, Table 2; Meyer, 2001, p. 210). Meyer (2001, p. 211) suggested that the mean annual temperature was ∼13 °C, whereas Wolfe (1992) suggested 12.0–12.5 °C and later (1994) gave an estimate of only 10.8 °C. Gregory and McIntosh (1996, Table 7) gave estimates of mean annual temperature from 12.8 to 13.9 °C, with a warm-month mean of 30.0–33.3 °C. Gregory and McIntosh calculated a cold-month mean temperature of −2.0 to 1.3 °C, and a mean annual range of temperature of 27.7–29.1 °C. However, these values for the cold-month and warm-month means are inconsistent with the values for mean annual temperature and mean annual range of temperature, and the mean warm-month temperature appears to be overestimated (Meyer, 2007, personal commun.).
The mammalian fauna of the Antero Formation (Stark et al., 1949, p.66) included large brontotheres, the rhinoceroses Trigonias, Subhyracodon, and Hyracodon, the horse Mesohippus, the oreodont Agriochoerus, the deerlike Hypisodus, plus the rabbit Palaeolagus. Although most of these mammal specimens are too fragmentary to be diagnostic of age, both brontotheres and Trigonias are known to have vanished at the end of the Chadronian (late Eocene). MacGinitie (1953) suggested a possibly Orellan or “middle Oligocene” (in the terminology of the time) age for the Antero, which would make the formation earliest Oligocene, according to modern mammalian biochronology. However, Kron and de Toledo (1994) confirmed the presence of brontotheres and other Chadronian mammals from the Antero Formation, making it latest Eocene. While sampling was being conducted for this study, additional brontothere lower teeth were observed at the outcrops that yielded paleomagnetic sites 5–7 (Table 1).
MacGinitie (1953, p. 75) and Wolfe (1992, p. 425) provided a very brief mention of the Antero flora (including Pinus cf. crossii, Quercus sp., Vaccinium sp., Cercocarpus cf. C. henricksonii, Eleopoldia cf. E. lipmannii), but no formal description was published. The original Wolfe collection was very limited and fragmentary and was thought to be lost, but it may be stored in the Smithsonian (G. Retallack, 2006, personal commun.). According to Wolfe (1992), the Antero flora is a post-deterioration flora representing cooler and drier climates and is very similar to the late Oligocene Creede flora. However, this interpretation is based on such a small sample of taxa (only six taxa and 16 specimens, according to Wolfe, 1992, p. 425) that it must be viewed with caution (H. Meyer, 2006, personal commun.).
Durden (1966) provided a slightly longer list of plant taxa, including Pinus, Quercus, Ulmus, Cercocarpus, Sequoia affinis, Chamaecyparis linguaefolia, Salix spp., Bidens sp., “Betula florissanti,” Mahonia marginata, Cardiospermum terminalis, and Abies or Pseudotsuga sp. Several of these occur at Florissant and not at Creede, and this list is based on a larger sample than Wolfe (1992) reported. Durden (1966) deposited this collection at Yale. The paleotemperatures for the Antero flora shown in Figure 9 indicate the approximate conditions for both a “Creede-like” and “Florissant-like” interpretation, but these estimates have not been derived using a quantitative methodology applied specifically to Antero. During our sampling, it was apparent that many additional good plant fossils could be collected, so the flora really needs to be recollected on a larger scale and reanalyzed. If the new dates are correct and Wolfe's (1992) paleobotanical interpretation is valid, then this suggests that the climate change in the southern Rocky Mountains (Fig. 9) was very rapid (between 33.89 and 34.07 Ma) and occurred in the latest Eocene, not in the early Oligocene as all other global climatic records show (Miller et al., 1987; Miller, 1992; Prothero, 1994; Zachos et al., 1992, 2001). This is true whether one adopts the Eocene-Oligocene boundary age of 33.7 Ma (Berggren et al., 1995) or 33.9 Ma (Luterbacher et al., 2004). By contrast, Wing (1987) showed that the Chadronian and Orellan plant assemblages of southwest Montana showed no effect of the Eocene-Oligocene climatic change. However, if Durden's (1966) interpretation is correct, the Antero flora is only slightly different from that of Florissant. Clearly, the Antero flora requires further study to draw any confident conclusions.
According to Gregory and McIntosh (1996), the Pitch-Pinnacle flora yields a mean annual temperature of 12.7 ± 1.5 °C (compared to 10.8–13.9 °C for Florissant but only 0–4.2 °C for Creede; Antero is not mentioned). The cold-month mean temperature is estimated at 4.5 °C (warmer than Florissant) and the warm-month mean temperature is 20.4 °C (whereas Florissant gives values of 30.0–33.3 °C, according to Gregory and McIntosh, 1996), and the mean annual range of temperature is estimated at 18.5 °C (compared to 27–29 °C for Florissant, according to Gregory and McIntosh, 1996). The flora has angiosperm genera that are common to both Creede and Florissant, and conifers that are similar to those of Creede (Gregory and McIntosh, 1996). Clearly this is a warmer-climate flora only slightly cooler than Florissant, yet it is younger than the supposedly cool-climate (like Creede) Antero flora. Gregory and McIntosh (1996) argued that either it grew before the Oligocene deterioration at an elevation of 2–3 km or it could be a post-deterioration flora from an elevation of 1 km. However, because Antero and Florissant are from similar elevations, this does not explain why the cooler Antero flora occurs chronologically between warm-climate Florissant and Pitch-Pinnacle floras (Fig. 9). As mentioned above, the Antero flora is too poorly known to place much confidence in this conclusion. Because of this uncertainty, the deterioration may have occurred in the early Oligocene in Colorado, as it does in most other regions.
In the Warner Range of northeast California, Myers (1998, 2003) listed the pre-deterioration Steamboat and Badger's Nose floras of the Cedarville Flora as 34–35 Ma in age, or latest Eocene. The Steamboat floras produce mean annual temperature estimates of 17.1 °C, with a warm-month mean temperature of 24.3 °C and a cold-month mean temperature of 5.6 °C, and ∼234 cm of annual precipitation. The Badger's Nose floras yield mean annual temperature estimates of 12.5 °C, with a warm-month mean temperature of 21.1 °C and a cold-month mean temperature of 2.6 °C, and ∼178 cm of annual precipitation (Myers, 1998, 2003). The temperature values of the Badger's Nose flora are similar to those of the warm temperate late Eocene floras, such as Florissant (Wolfe, 1978, 1992). The early Oligocene (31.5 ± 0.4 Ma, Chron C12r) Granger Canyon flora (discussed above) yields a mean annual temperature of only 9.6 °C, with a warm-month mean temperature of 21.1° and a cold-month mean temperature of −0.4 °C, but with almost 260 cm of precipitation, so it appears to be a post-deterioration flora (Myers, 2003). However, the age constraints (Fig. 7) are insufficient to evaluate how much time elapsed for the climatic deterioration in this region. If the current dates are correct, it occurred between 34 and 31.5 Ma, so it could have happened any time in the latest Eocene or earliest Oligocene (Fig. 9).
The longest and most complete sequence of floras spanning the Eocene-Oligocene transition occurs in the Eugene and Fisher Formations near Eugene, Oregon. Recent high-resolution lithostratigraphy, biostratigraphy, tephrostratigraphy, 40Ar/39Ar dating, and magnetostratigraphy by Retallack et al. (2004) allowed fine-scale dating and correlation of the classic floras and marine invertebrate faunas in this sequence (Figs. 7 and 9). The middle Eocene Comstock flora was correlated with Chron C18n.2n (39.7 Ma), and the flora suggests a mean annual temperature of 22.4 °C with a warm-month mean of 26.5 °C, a cold-month mean of 7.0 °C, and a mean annual range of temperature of 19.5 °C, all suggestive of warm paratropical conditions. The late Eocene subtropical Goshen flora was one of the original bases for Wolfe's (1978) concept of a late Eocene warming event. It is now dated to occur within early Oligocene Chron C13n (33.4 Ma), and yields a mean annual temperature of 19.7 °C with a warm-month mean of 25.1 °C and a cold-month mean of 6.8 °C, and a mean annual range of temperature of 18.3 °C, only slightly cooler than the warm paratropical conditions of the Comstock flora. It should be noted, however, that under normal climatologic conditions, these estimates for Goshen and Comstock for mean warm-month temperature and mean cold-month temperature are not possible in combination with the mean annual temperatures that are estimated (H. Meyer, March 2007, personal commun.). The Rujada flora, dated at 31.5 Ma (Retallack et al., 2004), shows temperature estimates consistent with those of a post-deterioration flora (Fig. 9), with a mean annual temperature of 13.0 °C, a cold-month mean temperature of 2.4 °C, a warm-month mean temperature of 23.6 °C, and a mean annual range of temperature of 21.2 °C (Retallack et al., 2004). Stratigraphically above the Rujada flora are the early Oligocene Coburg and Willamette floras. They are dated at Chron C12n (30.9 Ma) and Chron C11r (30.1 Ma), respectively. The Willamette flora yields a mean annual temperature of 13.2 °C with a warm-month mean of 20.8 °C and a cold-month mean of 6.2 °C, and a mean annual range of temperature of 14.6 °C, and the much less well known Coburg flora is very similar (Retallack et al., 2004). These floras clearly suggest that the Oligocene deterioration had taken place by 31.5 Ma (Fig. 9) but was not as severe and abrupt as once suggested by Wolfe and Hopkins (1967) and Wolfe (1978). Thus, the Eugene-Fisher floral sequence places the climatic change in the earliest Oligocene (consistent with the global record), somewhere between 33.4 and 31.5 Ma. This 2-million-year gap between floras does not allow us to assess the rapidity of the change as well as for the closely spaced Antero and Florissant floras in Colorado, but the results are consistent with previous ideas that the climatic change occurred in the earliest Oligocene.
In the John Day region of central Oregon, the John Day Formation yields floras relevant to this discussion. The Iron Mountain assemblage of the Bridge Creek flora (Meyer and Manchester, 1997; Myers, 2003) is 40Ar/39Ar dated at 33.62 Ma. This is slightly younger than the Eocene-Oligocene boundary of 33.7 Ma according to Berggren et al. (1995), or 33.9 Ma according to Luterbacher et al. (2004). It yields a post-deterioration mean annual temperature of 10.3 °C, with comparable results for cold-month mean annual temperature of 0.9 °C and warm-month mean annual temperature of 17.7 °C (Myers, 2003). Just above this flora is the Fossil assemblage of the Bridge Creek flora, which is 40Ar/39Ar dated at 32.58 Ma, and yields a mean annual temperature of 12.1 °C with a warm-month mean of 19.0 °C and a cold-month mean of 3.1 °C, also suggesting a post-deterioration flora (Myers, 2003). These dates are from ashes directly inter-bedded with the floras (Retallack et al., 2004). Thus, the climatic event in central Oregon seems to have happened before 33.62 Ma (a date that is right after the Eocene-Oligocene boundary), not coincident with the early Oligocene Oi1 event of Zachos et al. (1992, 2001), which is dated at 33.0 Ma.
Available data suggest that the climatic cooling documented during the earliest Oligocene in the global climatic record seems to occur at different times in the floral records of the western United States (Fig. 9). In south-central Colorado, the age difference between the Florissant and Antero floras suggests that it occurred in the late Eocene, between 33.89 and 34.07 Ma, and was very abrupt (less than 200,000 yr). However, the occurrence of the 30.5–31.0 Ma (based on its preferred correlation with Chron C12n) Pitch-Pinnacle flora in Colorado suggests that the deterioration occurred in the early Oligocene; this in turn indicates that the Antero flora is anomalous and needs further analysis. In northeast California, the deterioration appears to have occurred in the early Oligocene, sometime between 34 and 31.5 Ma. In coastal Oregon, it occurred in the early Oligocene between the Goshen and Rujada floras, or between 33.4 and 31.5 Ma. In the Clarno Formation Bridge Creek floras of central Oregon, the available dates suggest that the deterioration was already in place by the beginning of the Oligocene, 33.62 Ma. Thus, the floras suggest some diachroneity of the climatic transition, with at least two regions (Colorado, central Oregon) providing possible evidence that the cooling had occurred before the global event in the early Oligocene (at 33.0 Ma), whereas two other regions (northeast California, coastal Oregon) document an early Oligocene deterioration that is consistent with the global record. Clearly, further work needs to be done on the paleobotany, geochronology, and chronostratigraphy of these floras, especially Antero and the Bridge Creek sequence, to test this hypothesis.
I thank A. Buller, K. Fjeld, and D. Weiser for help with sampling in northeast California, and my family for support during both field seasons. I thank S. Bogue and J. Kirschvink for help building and maintaining the Occidental paleomagnetics laboratory. I thank Jeff Myers for help with the Cedarville localities, and Sara Gooch for guiding us to the Granger Canyon locality. I thank Chuck Chapin, Herb Meyer, Jeff Myers, Greg Retallack, and an anonymous reviewer for helpful comments on this paper. This research was supported by National Science Foundation grant EAR03–09538, and by a grant from the Donors of the Petroleum Research Fund, administered by the American Chemical Society. This paper is dedicated to the memory of Jack A. Wolfe, who was a pioneer in so many areas of paleobotany.
Figures & Tables
Paleontology of the Upper Eocene Florissant Formation, Colorado
- absolute age
- Creede Formation
- Florissant Lake Beds
- Mineral County Colorado
- Modoc County California
- Park County Colorado
- Saguache County Colorado
- United States
- Antero Formation
- Pitch-Pinnacle Formation