New pollen data from four Oligocene floras in volcanic landscapes of Colorado record important climatic shifts that reshaped the local flora and promoted the development of sub-arid vegetation types. We combined new pollen data with previous megafossil evidence to assess vegetation changes during the Eocene–Oligocene Transition (EOT). Pollen data are the basis for updating the list of flora identified at Creede. Local extinctions in response to lower summer rainfall abruptly removed many of the exotic woody taxa of eastern North American and Asian affinity. This loss was followed by the appearance of xeric shrubland taxa of the Ponderosa pine-fir woodland and sagebrush flora that characterize the Colorado area today. Tell-tale genera appear and suggest an understory of shrubs such as Artemisia, Elaeagnus, Ribes, Ephedra, Jamesia, and Shepherdia. Poaceae are also present. Significantly, herbaceous taxa of the Asteraceae, Rosaceae, Cucurbitaceae, Euphorbiaceae, and Caryophyllaceae make their first appearances in the fossil record of Colorado here.

The new Oligocene pollen data record a significant drop in summer rainfall and a climatic cooling at 33.8 Ma of several degrees that relates to the Oi-1 glaciation in Antarctica. The particular taxa that continued after the EOT were a basis for estimating changes in soil moisture during this time. The conditions in Colorado are reminiscent of Wolfe’s “terminal Eocene event.” This remarkable shift precipitated the development of a local pollen and megafossil flora more “modern” in aspect (e.g., a larger proportion of extant local genera are present). The impressive floristic turnover was probably a response to the increasingly continental climate that embraced the area during the Oligocene.

The global climate change at the Eocene–Oligocene transition (EOT) (33.4 Ma; Prothero and Berggren, 1992) is the largest abrupt cooling of the Cenozoic (Zachos et al., 2001). During the EOT, changes in circulation resulting from the development of deep-water passages separating Antarctica from South America (beginning ca. 33 Ma) (Graham, 2011; Lawver et al., 2011) may have contributed to the major worldwide cooling and associated aridity related to Antarctic glaciation Oi-1 (Cather et al., 2008). This post-Eocene cooling and aridity resulted in local and regional extinctions of plant, vertebrate, and invertebrate taxa (Prothero, 1985a, 1985b; Hutchison, 1992; Legendre and Hartenberger, 1992; Prothero and Berggren, 1992; Wolfe, 1992; Meyer and Manchester, 1997). In Colorado, the EOT is associated with a period of cooling and severe aridity lasting most of the Oligocene. One of our strongest contributions is the addition of the pollen record from four Oligocene floras, which had not been reported previously. This pollen record is an important aspect of this paper for assessing the total flora. The emphasis of this study is on the evolutionary and climatic significance of the floristic changes during the EOT.

Five Colorado floras of late Eocene through Oligocene age lie within the geographic mix of caldera and volcanic settings in the central Colorado volcanic area and in the San Juan volcanic field in southwestern Colorado (Fig. 1). The floras are, in ascending order: the Florissant (34.1 Ma), Antero (33.8 Ma), Pitch-Pinnacle (between 32.9 and 29 Ma), Platoro (28.0 Ma), and Creede (26.9 Ma) assemblages (Fig. 2; Table 1). The purpose of this report is to supplement previous work on plant megafossils of the Oligocene in Colorado with our new work on fossil pollen. We characterize the pollen and megafossil floras from just before and after the EOT and interpret the climatic and floristic shifts that they demonstrate with emphasis on the following questions:

  • (1) What was the scale of the climatic shift based on the floras, and how does it relate to possible uplift of the region?

  • (2) What was the evolutionary significance of the floristic changes during the Colorado EOT?

  • (3) What estimates of paleoelevation may be based on these floras and the local geology of the sites?

  • (4) How do climate tools such as coexistence approach, detrended correspondence analysis, the Sørensen index, and nearest living relatives analysis assist the interpretation of the floras in terms of inferred mean annual temperature (MAT) and precipitation (PPT)?

  • (5) How do the inferred climate shifts compare to conditions in other parts of the world during the EOT?

Our fossil records span the latest Eocene Florissant flora (34.1 Ma) and four progressively younger Oligocene floras (33.8 Ma–26.9 Ma; Table 1). This interval is closely synchronous with large-scale explosive volcanism in the southern Rocky Mountains (mainly 37–27 Ma), the development of the Oi-1 glaciation and the “Ice House,” and the period of great aeolian activity along the Arizona–New Mexico border (Zachos et al., 2001; Cather et al., 2008).

2.1. Regional Geological Setting of the Five Paleogene Floras in Colorado

The Florissant fossil beds formed after a comparatively small volcanic event that dammed a lake at the end of the Eocene. In contrast, the Oligocene floras that immediately followed in southwestern Colorado developed in local volcanic sediments. The San Juan volcanic field produced 28 large-volume ignimbrite eruptions between 36.9 Ma and 26.9 Ma (Lipman and McIntosh, 2008; Lipman et al., 2015). Ponding of distal ash from these and the formation of lakes within source calderas are the main sites for the floras discussed here.

The Florissant Formation formed when a lahar blocked the Florissant Valley in the Thirtynine Mile volcanic area (36.9 Ma to 34.3 Ma), leading to the development of Lake Florissant (McIntosh and Chapin, 2004; Fig. 1). Distal ash associated with Mount Aetna nearby accumulated in another intermontane lake to form the Oligocene Antero Formation (33.8 Ma). In addition, lake deposits of the Pitch-Pinnacle flora accumulated within and near the Marshall caldera at ca. 33.8 Ma (Lipman et al., 2013, 2015). Shortly after this, a dune field of aeolian dust and sand in the Chuska Mountains of northwestern New Mexico and Arizona (Fig. 1) formed at ca. 33.5 Ma and lasted until 28–26 Ma (Cather et al., 2008, 2012). Called the Chuska erg, or sand sea, these deposits (as thick as 535 m) resulted from large accumulations of airborne and fluvial material in the central and southern Colorado Plateau (McIntosh et al., 1992).

The most intense volcanic activity was focused in the San Juan volcanic field from 34 to ca. 26 Ma (Lipman, 2007; Lipman et al., 2015). Multiple ignimbrites of the Treasure Mountain Group erupted between 31 and 28.6 Ma, creating the Platoro caldera complex (Lipman et al., 2013). Lake deposits in the caldera contain the Platoro flora. Among the San Juan eruptions was one of the largest single volcanic events in Earth history: the 5000 km3 Fish Canyon Tuff erupted from La Garita caldera in the central San Juan volcanic field at 28.02 Ma (Mason et al., 2004; Lipman, 2007). The Creede flora was deposited within the Creede caldera, the youngest of seven calderas nested within La Garita (Lipman, 2000). The history of the Oligocene Lake Creede (Bethke and Hay, 2000; Larsen and Lipman, 2016) provides data on paleolimnology based in part on two drill cores penetrating >700 m through the caldera lake deposits (Larsen and Nelson, 2000). The depth of the lake fluctuated at times by tens of meters, with waters variously saturated with carbonates and saline mixtures (Larsen and Lipman, 2016).

Aerosols from these multiple and sizeable volcanic events probably had a cooling effect on the mean annual temperature of the region, similar to the aerosols from recent eruptive events (such as Krakatau, Tambora, and Pinatubo) that have influenced climate by blocking solar energy (Williams, 2012). Given their magnitude, such large-scale eruptive events in the San Juan volcanic field could potentially have influenced not just regional but global climate. Cather et al. (2009) postulated that enormous amounts of volcanogenic silicic dust generated episodically in this region may have contributed locally to Oligocene global cooling by intercepting solar radiation and enhancing oxygen production via iron fertilization of the oceans. Because of the close temporal relationship between the Oi-1 event of Zachos et al. (2001) and the onset of aeolian deposition and increasing aridity of the region, Cather et al. (2008) suggested that the Oi-1 event helped to drive the Oligocene aridification of the Colorado Plateau.

2.1.1. Paleoelevation

An important question in the geologic literature deals with the elevational history of the Southern Rocky Mountain region relative to the high elevations of the fossil plant localities today. A wide range of methods and results have been presented, predicting extremely high (>3000 m) or extremely low (<500 m) paleoelevations for certain floras at their time of deposition (Fig. 3; Tables 1 and 2); e.g., multiple methods examined by Gregory and McIntosh (1996), the enthalpy-based method of Wolfe et al. (1998), the lapse rate method based on Axelrod and Bailey (1976), and Wolfe’s (1994) paleobotanical method, which combined lapse rates with MAT of a flora and MAT of an isochronous flora at sea level to calculate paleoelevation. As a result, two constraints in particular must be recognized in order to come up with a reasonable estimate of paleoelevation for the five floras: the initial Paleogene elevation, plus any potential subsequent tectonic activity in the Colorado Plateau and southern Rocky Mountains.

2.2. Previous Paleobotanical Descriptions

In-depth analyses of the pollen data from the Oligocene floras have not been previously published. The report on the Antero megafossil flora was by Durden (1966). Gregory and McIntosh (1996) described the Pitch-Pinnacle flora. Lipman (1975) and Meyer (1986) each published details on the Platoro flora. After the early work on the Creede flora by Knowlton (1923), Axelrod (1987) completed a detailed revision and geologic description of the flora. Steven and Ratté (1965) mapped and described the physical stratigraphy of the Creede area and included a few identifications of pollen and megafossils. Howard Schorn undertook a taxonomic revision of the Creede megafossil flora (Schorn and Wolfe, 1989), part of which was then published by Wolfe and Schorn (1989) on the evolutionary significance of the flora. Wolfe and Schorn (1990) followed with a major taxonomic revision of the Creede flora, which included a critical review of the study by Axelrod (1987). The earlier reports show that the Creede flora was warmer and more arid in aspect than the present local vegetation, indicating that a post-Eocene climate cooling and/or uplift had occurred (Axelrod, 1987; Wolfe and Shorn, 1989). These Eocene and Oligocene floras are a window into the history of the Cordilleran (Axelrod and Raven, 1985; Benedict, 1991) flora of the southern Rocky Mountains.

In this paper, the isotopic dates are rounded to the nearest first decimal (tenth). Dates of the fossil-bearing sediments are given in Table 1. A list of collections consulted and slides made from pollen and/or spore samples collected at the Eocene and Oligocene sites used in this study are listed in Table S11 of the Supplemental Items. Pollen and/or spore slides bearing paleobotany “D” and “W-1” locality numbers are reposited at the U.S. Geological Survey (USGS) Core Research Center at the Denver Federal Center. Slides from the Antero and Pitch-Pinnacle pollen floras were kindly made available by Dena Smith at the University of Colorado Museum and Herb Meyer of the National Park Service, and collections from the Platoro caldera (D4772) were provided by Peter Lipman. Plant megafossils and pollen samples were collected by Leopold and Steven at several localities in the Creede caldera: at Five-Mile Bridge, about five miles southwest of Creede, Colorado, and near Birdseye Gulch, one mile directly south of Creede (Fig. 2). Both areas are also collection sites of Axelrod (1987) and Knowlton (1923). Leopold’s other Creede collections are from a site north of Seven Mile Bridge (D4293) and another lakeside site (D1811) in the caldera (Fig. 2). These Creede collections were closely associated with a prominent 5-m-thick white volcanic ash described by Axelrod (1987) along the north rim of the Creede crater. Sediments from locality D1811 were sampled as paper shale laminae containing pollen but no megafossils. Among our pollen and spore collections from seven Creede Formation localities, four samples yielded preservation adequate for confident identification of the grains. Pollen of Ephedra, which is a shrub, is counted with the non-arboreal (non-tree) pollen (NAP).

Pollen preparation of USGS samples followed Doher (1980) and involved treatment with KOH, HF, and acetolysis. Residues were mounted in glycerin jelly and cured and sealed with clear plastic (Doher, 1980; Traverse, 2007). Some preparations from the University of Colorado Museum collections (some of which were collected by Herb Meyer) were done at Global Geolab Limited in Medicine Hat, Alberta, Canada, where the Creede pollen was mounted in clear plastic (polyvinyl alcohol). Photomicroscopy was performed using a Leica DFC295 camera running the Leica Application Suite version 4.0 software package. Pollen counts were conducted using 13 samples from the five Paleogene sites; these are recorded in Table 3 and are listed in chronological order. Megafossil and pollen compositions of the five Paleogene floras, along with their sources, are compiled in Table 4.

Using PAST version 3.0.4 (Hammer et al., 2001), a detrended correspondence analysis (DCA) was performed in 2016 by Cindy Looy of University of California Berkeley to evaluate ecological trends among the five floral communities (Fig. 4). The pollen counts in Table 3 were used for this purpose, omitting those group taxa described as “undetermined” or not assigned to a living plant group. The information for a taxon’s abundance, presence, or absence in each sample can determine whether the taxa have a similar distribution on the DCA plot.

Moisture availability in the soil is a critical factor controlling the success of plant life in semi-arid environments. Soil moisture indexes of particular taxa provide one way to estimate the balance between actual evaporation (AE) versus potential evaporation (PE) and represent an important environmental feature. The annual rainfall and evaporation conditions expressed by this ratio (AE/PE) demonstrate a relationship critical to plant survival in many western (continental) environments (Thompson et al., 2012) and can be used to reveal local and regional patterns of influx and loss of floral taxa.

Our interpretation of paleoclimate and environments here used nearest living relatives (NLR; Mosbrugger and Utescher, 1997) as well as the coexistence approach (CA) method (Utescher et al., 2014; Zaborac-Reed and Leopold, 2016), by which generic identification of the nearest living relatives are a basis for estimating temperature tolerances for the entire flora. These represent rough approximations of temperature conditions for the floras and are not adjusted according to the estimated elevations of the sites. We used the CA method to apply the ranges of MAT tolerances on the pollen and megafossil genera of the various floras based on values from Thompson et al. (2015), Fang et al. (2011; see Zaborac-Reed and Leopold, 2016), and others (Table 5). Taxa previously identified as outliers by Utescher et al. (2014), monotypic taxa, uncertain identifications, and taxa identified by singleton pollen grains were excluded from CA analysis. Mean annual temperature (MAT) and cold-month mean temperature (CMMT) values were examined (Table 5).

Table 6 includes estimates of MAT and annual rainfall (PPT) by Axelrod (1987), MacGinitie (1953), Gregory and McIntosh (1996; Climate Leaf Analysis Multivariate Program [CLAMP] data), and Wolfe and Schorn (1989). Evaluating precipitation using the CA method has not proven feasible here (Zaborac-Reed and Leopold, 2016).

The Sørensen index of similarity was used to evaluate similarities between the five floras (in particular, the Eocene Florissant versus the Oligocene floras). The Sørensen index is weighted to recognize similarity between binary data sets, and it is therefore helpful for evaluating differences between ecological communities (Looman and Campbell, 1960). The formula used for the Sørensen index is:
graphic
where “A” is the number of taxa found in both data sets, “B” is the number of taxa unique to data set 1, and “C” is the number of taxa unique to data set 2. Family-level ids, conifers, monocots, and dicots identified in Table 4 were used for these statistical analyses. Fern allies, extinct taxa, and the IDs “small pine” and “legume” were not evaluated. The results appear as percentage values for ease of discussion (% = S*100).

The floral lists for the five floras include megafossil, pollen, and spore identifications obtained from various samples (Tables 1 and 4). Because a number of the taxa in these floras are wind pollinated, it is important to note when pollen identifications were also confirmed by megafossil evidence. Relative abundance of pollen types in sequential samples is portrayed in the histogram (Fig. 5), which serves to provide a snapshot of the overall vegetation change during the EOT.

4.1. Floral Composition

4.1.1. Late Eocene (Florissant)

The Florissant megafossil and pollen flora is extremely diverse and warm-temperate in character, including more than 100 taxa (MacGinitie, 1953 as modified by Manchester, 2001; Leopold and Clay-Poole, 2001; Bouchal, 2013; Bouchal et al., 2016; Zaborac-Reed and Leopold, 2016; Table 4). Rosaceae is the main family (eight genera).

The pollen flora demonstrates a great diversity of arboreal dicots, including a few East Asian genera. Pollen counts record a flora rich in arboreal pollen (AP) types, especially woody broad-leaved dicots (10%–55% in abundance); some 63 taxa include 22 summer-moist tree taxa (Tables 3 and 4). An earlier pollen assignment identified as Pteroceltis (Leopold et al., 2008) is incorrect and is now rejected.

The pollen histogram (Fig. 5; Table 3) shows a relatively important role for Pinaceae (12%–30%) and other conifers, as well as Taxaceae/Cupressaceae/Taxodium-type (TCT) pollen (8%–65%), presumably representing pollen of Sequoia affinis, Juniperus, or Chamaecyparis. At Florissant, cool indicators such as Picea and Abietineae are rare (∼5%). The non-tree pollen (NAP) such as Ephedra and other dryland shrubs are a minor group (2%–12%).

A considerable number of families and genera found at Florissant are known for their tropical or subtropical distribution today (Table 7). Some, such as Platycarya and Eucommia, have East Asian affinities.

4.1.2. Oligocene (Antero, Pitch-Pinnacle, Platoro, and Creede)

4.1.2.1. Antero

The pollen and spores of the Antero flora are figured in Plate I. Importantly, the Antero flora contains cool indicators, such as megafossils and abundant pollen of Picea (up to 30%) and Abietineae (Fig. 5; Tables 3 and 4). Other taxa include Pinus pollen (∼25%), along with megafossils and pollen of Juniperus, Sequoia, and some broad-leaved trees common in nearby Florissant (e.g., Juglandaceae, Ostrya-Carpinus, and Eucommia). Other woody taxa include Betula, Crataegus, and mesic streamside types such as Ulmus and Salix (Table 4).

4.1.2.2. Pitch-Pinnacle

The pollen and spores of the Pitch-Pinnacle flora are figured in Plate I. Megafossils and pollen of Pinus, Abies, Picea, and Juniperus are present, as well as non-tree types such as Cnidoscolus, Rhus, Ribes, Rosa, and streamside types (Salix and Ulmus) (Table 4).

4.1.2.3. Platoro

The pollen and spores of the Platoro flora are figured in Plates II, III, and IV. Some Platoro samples contain only burnt and charred plant tissue, pollen, and spore specimens (Plate IV) (Lipman, 1975). Megafossils and pollen of Pinus, Picea, Abies cf. bracteata, and Juniperus are present. NAP types are diverse and include pollen of Artemisia, Ephedra, Mahonia, Elaeagnus, Shepherdia, Ribes, and Poaceae. Megafossils of Berberis, Cercocarpus, Mahonia, and Ribes are also present (Table 4). Streamside taxa included Populus and Salix.

4.1.2.4. Revised list of Creede

We present a revised complete flora of the Creede Formation (Table 8), which includes megafossil identifications by Axelrod (1987), with revisions and additions by Howard Schorn (Schorn and Wolfe, 1989), later published as Wolfe and Schorn (1989, 1990). Our pollen identifications (Table 8) expand the Creede flora and indicate that Ulmaceae and the following genera rejected by Wolfe and Schorn (1989, 1990) were indeed present in the flora: Shepherdia, Fraxinus, Acer, Larix/Pseudotsuga-type, Tsuga, Alnus, and Quercus. Streamside shrubs (e.g., Acer, Betula, and Ulmus-type) are also present. The pollen and spore flora of the Creede Formation are shown in Plates V and VI. The fossil pollen and spores in some samples are very well preserved.

The pollen histogram (Fig. 5; Table 3) shows that at Creede, cool indicators include Picea pollen, which was consistently present (∼8% in three samples), and Abietineae pollen, which was significant in two samples (∼28%; Fig 5). Pinus is an important element. Low biomass types lumped as NAP expanded at Creede; these include Ephedra, Artemisia, Ambrosia-type, Sarcobatus, and Amaranthaceae.

Remarkable megafossil specimens of Abies cf. bracteata have been figured (Axelrod, 1987). Leopold and Zaborac-Reed (2014) also identified its distinctive pollen. Axelrod (1987) found abundant leaves of Cercocarpus at each Creede locality and reported that Crataegus creedensis was a widespread rosaceous tree or shrub, because its leaves were found at several Oligocene sites (e.g., in Montana, Oregon, and Colorado). Wolfe and Schorn (1989) and Axelrod (1987) identified the following Rosaceous megafossil genera at Creede: Crataegus, Prunus creedensis, and Potentilla (2 spp.), Sorbus (mountain ash), and Holodiscus stevenii Schorn and Wolfe.

Extinct taxa of the Rosaceae include Eleopoldia (identified at Pitch-Pinnacle and Creede by Wolfe and Schorn, 1990), an herb with highly lobed leaves related to Geum. Another extinct (megafossil) taxon of Rosaceae identified at Creede was Eleiosina praeconciana (Cockr.) Schorn and Wolfe, an herb with alternate leaves. Stockeya creedensis (R.W. Br.) Wolfe and Schorn (1990, p. 23), another Rosaceae element, was suggested to be ancestral to the genus Chamaebatiaria. One megafossil specimen from Creede currently assigned to the genus Eleopoldia (Wolfe and Schorn, 1990, their plate 13 and their figure 3) has been examined by Adams and is being reassigned to Juniperus cf. californica (Robert Adams, 2013, written commun.).

4.1.2.5. Comparison of the four Oligocene floras

The Oligocene floras of Colorado were less diverse than at Florissant (Table 4) and had a great deal in common. The gymnosperms in particular were very similar in composition and included many genera of Pinaceae. Pinus megafossils and pollen were found in each of the floras studied (Table 4). Pinus was consistently the most abundant pollen in the Oligocene (yellow bars, Fig. 5; Table 3). Pinus aristata-type pollen (which bears a distinctive lacy frill around the equator of the central cell, a feature shared with Pinus contorta) was present and may have been related to the megafossil Pinus crossii (of Axelrod, 1987; Plate V, 10). Picea pollen and megafossils were also present in all the Paleogene floras; its pollen (see blue bars in Fig. 5; Plate V, 9) was common in all but the Platoro samples. Abietineae pollen (Abies plus Keteleeria-type) (green bars, Fig. 5; Table 3), was abundant in the Creede flora, reaching 25%. Abies pollen was found in all floras, while megafossils were confirmed in Antero, Pitch-Pinnacle, and Creede.

In particular, megafossils identified as Abies rigida (which have diagnostic cone scales and foliage) and pollen of the Eocene species A. cf. bracteata (which is morphologically unique) were represented at Creede (Leopold and Zaborac-Reed, 2014, Plate 1, G), Antero, and Platoro. Keteleeria-type pollen, which is similar to Abies pollen (Zanni and Ravazzi, 2007; Leopold and Zaborac-Reed, 2014), was seen in Antero, Pitch-Pinnacle, and Creede. Keteleeria megafossils have been reported from the Pacific NW in the Latah flora near Spokane, Washington (Brown, in LaMotte, 1952), the Bridge Creek flora of Oregon (Meyer and Manchester, 1997), and the Quilchena locality of British Columbia (Mathewes et al., 2016). We found pollen unmistakably of Tsuga, though rare and poorly preserved, at Antero, Platoro, and Creede (Plate V, 12), which may substantiate Axelrod’s Tsuga petranensis at Creede. Both the pollen and megafossil records of the Creede specimens suggest the Tsuga cf. caroliniana type.

Taxaceae/Cupressaceae/Taxodium-type (TCT) pollen was rare at the four Oligocene floras (Fig. 5), with its typical form resembling Taxodioipollenites hiatus of Potoniè (1951), which may have represented Juniperus (Plate V, 17–19). However, leafy Juniperus remains were found at the four Oligocene floras and were also abundant at two of Axelrod’s (1987) Creede sites.

Ephedra pollen was present in all four Oligocene floras (Table 4). Ephedra pollen was represented by three species, following characters documented by Steeves and Barghoorn (1959; Plate I, 1; Plate II, 20–27; Plate V, 20–22). The two medium-sized forms we clearly identified were the first- and second-order branch-furrowed pollen of E. nevadensis and the smooth-furrowed pollen of E. torreyana (Davis, 2001; Bolinder et al., 2016; Bouchal et al., 2016). These were accompanied by a third species differentiated from above-mentioned by its small size and distinctive sculpture, short axis, and first- and second-order branched furrows (Plate I, 1; Plate II, 24; Plate V, 22). This pollen morphology is visually distinct from other Ephedra pollen, because it is more breviaxial, having a relatively high polar/equator (P/E) ratio (∼0.63) and small size (∼35μm). The high P/E ratio, small size, and secondary branching furrows combine to indicate that this pollen form may be related to either Ephedra exiguua Fredrickson or to Ephedripites lusaticus Krutzsch, which are similar in appearance and found in Eocene floras (Fredrickson, 1981; Garcia et al., 2016). This form of Ephedra pollen was found at Florissant (Wingate and Nichols, 2001, their plate 3 and their figure 9) as well as the Oligocene floras. On the histogram, pollen of Ephedra was grouped among the NAP types (purple bars, Fig. 5) because of its shrubby growth habit. Ephedra pollen is a minor element at Florissant, Antero, and Pitch-Pinnacle (up to 2.6%) but becomes more common (up to 7%) in the Platoro and Creede floras (Table 3).

In the Oligocene floras (as well as at Florissant), we have recorded a tiny form of bisaccate Pinaceae pollen that we call “small pine” hereafter. While some of these grains may have been aberrant and/or shrunken forms of the genus Pinus, they were somewhat degraded specimens, as seen in Plate II (33–36) and Plate V (15 and 16). Some could represent Cathaya (see Saito et al., 2000), a member of the Pinaceae, whose pollen resembles that of Pinus but is slightly smaller (Ying et al., 1993; T. Saito, 2013, personal commun.). However, many specimens of “small pine” were quite weathered and very small (less than half the size of Pinus) and did not have characteristics to assign them to Cathaya.

Larger bisaccate grains strongly resembling Podocarpus (Erdtmann, 1957) (Plate II, 17; Plate V, 13 and 14) were found as occasional in all five floras and designated as such in Table 3. While there is some disagreement among others as to this identification, these specimens had the characteristically robust, thick-walled, rugulate cells with over-sized sacci typically seen in Podocarpus (Sivak, 1975; Liu and Basinger, 2000). Well-preserved wood found in Brown’s Canyon (Oligocene) in Colorado was identified by R.A. Scott (in Van Alstine, 1969) as probable wood of Podocarpoxylon. These grains appeared as trace elements in most of the study samples (Table 3). Previous records demonstrate clearly that Podocarpus pollen was frequent in the late Cretaceous of Alabama (Leopold and Pakiser, 1964, their plates 4 and 7). Dilcher (1969) described Podocarpus megafossils from an Eocene flora in Tennessee, and Jarzen and Dilcher (2006) identified Podocarpaceae pollen in an Eocene deposit in Florida. Reinink-Smith and Leopold (2005) and Morley (2011) report Podocarpaceae pollen as consistent minor elements in the Neogene of Alaska. Other Alaskan trace occurrences include White and Ager (1994, plate 2, p. 68) and Reinink-Smith et al. (2017). Modern species of Podocarpus grow today in the subtropical lower montane wet climate of Tamaulipas, Mexico (Boyle et al., 2008; Reinink-Smith et al., 2017).

Although pollen of broad-leaved arboreal dicots and Sequoia-type (TCT) pollen dominated in the Eocene Florissant flora (∼50%) (Fig. 5), the composition of the dicot taxa changed greatly in the Oligocene; their numbers became much reduced (∼20% in Antero and 1%–2% by the late Oligocene) (Fig. 5). By the end of the Oligocene, summer-moist hardwood trees, particularly those with an East Asian affinity, had largely disappeared from the megafossil record (Table 4).

The angiosperm records in the four Oligocene floras included only a few families (Table 4). Rosaceae was the main dicot family of the Paleogene Colorado floras and included several tree, shrub, and herb genera represented in both the megafossil and pollen floras. Because pollen of rosaceous taxa was typically difficult to identify to genus, we have placed a number of simple tricolpate pollen of rosaceous structure in a group as Rosaceae undet. (Plate I, 30 and 31; Plate VI, 14–16). The Oligocene Rosaceae included the megafossil record of Cercocarpus (mountain mahogany) at Antero, Platoro, and Creede. Mahonia (Oregon grape and Berberidaceae) megafossils tended to be abundant at all Axelrod’s (1987) Oligocene sites as well as at Florissant.

The tree dicots found in the Oligocene floras included mainly streamside types such as Salix (Plate III, 4; Plate VI, 7), Populus (Plate I, 13; Plate VI, 5 and 6), and Betula (Plate I, 12; Plate VI, 21). Minor elements of the Oligocene floras included pollen and megafossils, such as Acer-type tricolpate pollen (Plate I, 8–10, 27; Plate VI, 8), Alnus (Antero and Creede, Table 3; Plate VI, 21), Quercus (Plate I, 7; Plate III, 5; Plate VI, 11 and 12), and Fraxinus (Plate VI, 19). Acer-type pollen found at Creede (Plate VI, 8) supports the identification of the megafossil Acer riogrande of Axelrod (1987), which was dismissed by Wolfe and Schorn (1989, 1990) as Ribes lacustroides. Mahonia megafossils were found at all four Oligocene floras as well as at Florissant. Many of the woody dicots found in the Oligocene were microphyllous and have relatives living in southwestern United States and Mexico today. However, a very large cordate-shaped entire leaf was identified at Creede as Catalpa by Wolfe and Schorn (1990), who said it might be related to an Asian species, C. ovata. Alternatively, this may be a leaf of Nuphar, which was identified at Creede by Axelrod at several field localities (1987, p. 27). This is especially likely in a flora chiefly of microphyllous leaves. Pollen of Nuphar has not yet been found.

Ulmaceae pollen, which was important in the late Eocene, comprised a trace element (1%–4%; Table 3) of the four Oligocene floras (Plate I, 15, 16, 24, 25; Plate III, 23–27; Plate VI, 31 and 32). This pollen reaffirms the presence of the Ulmus type leaves in megafossils at Creede as suggested by Knowlton (1923) and Brown (in Steven and Ratté, 1965). Ulmaceae taxa may, like Ulmus, thrive along lake-shore environments. Both Knowlton (1923) and Brown (in Steven and Ratté, 1965) identified Planera leaves at Creede. However, Wolfe and Schorn (1990) re-identified some Planera specimens as Cercocarpus (1990).

Eucommia, Carya, and Juglans pollen, so obvious in Florissant, were sparse minor elements (1%–6%) in the early to mid-Oligocene (Table 4). These pollen are wind-blown types that dropped out before the Creede flora and did not appear in the Oligocene megafossil records. These pollen records might therefore be a product of long-distance wind transport or redeposition.

The role of NAP increased significantly from 15% to 40% in the younger samples, especially at Creede (Fig. 5). Ribes pollen (identifiable to genus) is present at Pitch-Pinnacle, Platoro, and Creede, affirming the leafy megafossil remains of seven Ribes species at Creede (Plate VI, 22–24; Table 8). The pollen is distinctive, with few large, simple pores (from seven to nine in number).

Sarcobatus (Family Sarcobataceae) pollen was a common (2%–20%) and consistent shrub element in most Oligocene pollen samples in all four Oligocene floras (Table 4; Plate I, 17; Plate III, 32–34; Plate IV, 15 and 16; Plate VI, 27–29). The pollen count indicates that it played a significant role at Creede (Table 3). This periporate form was unmistakable: its 12–20 pores have thickened annulae that help to distinguish them from similar grains of Amaranthaceae/Chenopodiaceae. The pollen is almost identical to that of the extant salt-tolerant shrub Sarcobatus vermiculatus (greasewood) of the Great Basin—a plant found in all desert communities of the western United States.

Various diverse examples of Elaeagnaceae pollen (Plate I, 14; Plate III, 10–17; Plate VI, 33), and Shepherdia cf. argentea (buffalo berry) (Table 3) pollen were seen in Platoro and Creede samples. Megafossils of Shepherdia are also recorded at Creede by Axelrod (1987). This hardy shrub has thrived in desert environments of the western United States.

Asteraceae types were represented in the four Oligocene floras by megafossils and pollen. These represent early Rocky Mountain records of the family Asteraceae (Leopold and MacGinitie, 1972; Graham, 1996). R.W. Brown (in Durden, 1966) reported Artemisia? megafossil remains at Antero. We found Artemisia pollen at Antero, Pitch-Pinnacle, and Creede (Plate I, 3 and 23; Plate VI, 17 and 18). Axelrod (1987) described specimens in plate 29 as Fallugia, which Wolfe and Schorn (1989, 1990) later described as Eleopoldia. Upon further examination, we concluded that at least some of these leaf specimens that Axelrod assigned to Fallugia (Axelrod, 1987, his plate 29 and figures 2 and 3) are extremely similar to the living species Artemisia rigida and/or A. scopulorum, which might match with our finding of Artemisia pollen there. R.W. Brown (in Durden, 1966) reported Artemisia leaves and an Asteraceae fruit identified as Bidens sp. (beggar’s tick) at Antero. Along with this, spiny tricolpate pollen of Asteraceae (Asteroideae-type) occurred at Platoro and Creede (Plate III, 20–22; Plate VI, 30). Finally, Cockerell (1933) reported a Solidago praecoccinea megafossil herb at Creede, although this was rejected by Axelrod (1987).

These are tantalizing records of some importance because unequivocal pollen of this highly derived family (Asteraceae) was present as rare elements at Florissant (Wingate and Nichols, 2001; Bouchal et al., 2016) as well as at all four of these Oligocene sites but are not characteristic of earlier Cenozoic floras of the United States.

4.2. Pattern of Floristic Turnover: Change in Composition of the Eocene and Oligocene Floras

4.2.1. Sørensen Index of Similarity

The Sørensen index was used to evaluate the similarity of the five Paleogene floras. As shown in Table 9, the Florissant and the four floras evaluated together as a single unit (done in order to increase the likelihood of similarity) only yield a 57.3% similarity—a moderate but not strong result. The highest similarity value observed was when the older Oligocene floras (Antero and Pitch-Pinnacle) were evaluated against the younger floras (Platoro and Creede), yielding a value of 73.6%—an indication of moderately strong similarity among these floras. One striking result was the similarity index of 50.4% between the Florissant flora with the nearby Antero flora (which followed only 0.3 m.y. later).

4.2.2. Dropouts (Outgoing Types)

A moderate number of tree dicots were prominent at Florissant, including hardwoods such as Carya, Carpinus, Juglans, Ostrya, Platanus, Tilia, and a number of Quercus species, essentially “dropped out” of the Colorado megafossil record immediately after the Eocene (Table 4). In contrast, the number of taxa represented by a shrub growth habit increased (Table 4). In the histogram, the overall abundance of arboreal dicot pollen dropped from ∼45% in some Florissant samples and 11% –18% in Antero, to a trace 2% in the Pitch-Pinnacle, Platoro, and Creede samples (Fig. 5). A number of the broad-leaved dicot dropouts seemed to have no further record in the Colorado Oligocene, although some mesic types did reappear in the Miocene Troublesome Formation farther north near Grand Lake in central Colorado (Leopold, 1969, p. 409). Many of these outgoing taxa prefer summer-moist soil conditions that typify hardwood forests of eastern United States and East Asia.

The modern soil-moisture index values of AE/PE for the outgoing hardwood taxa found at Florissant, such as Carya, Juglans, and Tilia, are high and narrow, between 0.8 and 1.0 (Fig. 6A). The disappearance of these taxa may mark local extinctions or extirpations through aridity in the Oligocene floras of southern Colorado. As Wolfe (1987) reported, many of these taxa dispersed and appeared in other western floras (see also Wing, 1987).

4.2.3. Continuing (Western Shrubs and Streamside Types)

Another group represented taxa that were minor elements in the Florissant and continued on through the Oligocene floras. These include “western” shrubs such as Shepherdia, Sarcobatus, Cercocarpus, and Ephedra. Immediately after the Florissant samples, the Antero flora contained a number of the Florissant genera (Table 4). However, some of these did not persist beyond Antero. Instead, streamside taxa that frequent moist habitats persisted in the Oligocene, including Alnus, Salix, Populus, and Ulmus. These taxa are adapted to broad, variable soil-moisture conditions (AE/PE = 0.3–0.9) and thrive in streamside environments during severe drought conditions (Fig. 6B).

4.2.4. Newcomers (Incoming Types)

This group includes newly evolved taxa of dryland environments. Most are members of highly evolved and herbaceous or shrub groups not typical before the Oligocene in Colorado; examples are Asteraceae, Caryophyllaceae, Cucurbitaceae, and Potentilla (Table 4). Some new shrubs included Artemisia, Jamesia, Berberis, Arceuthobium, Halesia?, and Heteromeles. This group of incoming taxa was of particular interest because they are members of today’s regional sub-arid cool flora of the southern Rocky Mountains. Benedict (1991) has listed these genera as current natives of the Ponderosa pine shrubland community of southern Colorado: Artemisia, Sarcobatus, Jamesia, and others. They tolerate low soil-moisture conditions with AE/PE values generally between 0.2 and 0.6 (Fig. 6C). Hence their appearance here is ecologically significant.

4.3. Results of the Coexistence Approach (CA)

Climate tolerances for the taxa used in CA analyses of the four Oligocene floras studied herein are found in Table 5. Using CA, Baumgartner and Meyer (2014) evaluated the macrofossil-only assemblage of Florissant to initially conclude a MAT of between 10 °C and 18 °C for the local climate. Further analysis using the Palaeoflora database (http://www.palaeoflora.de/) and supplemental sources narrowed the range to between ∼10 °C and 13 °C. Zaborac-Reed and Leopold (2016) evaluated the megafossil and pollen data together and concluded that the MAT range was between 14.3 °C and 18.2 °C. Baumgartner and Meyer (2014) pointed out that the climate signal based only on megafossils probably gives a better reading of the local climate, while pollen might give a more regional climate signal since some grains could potentially be blown in from a distance. Differences in site elevation may also play a part in the interpretations of MAT; Zaborac-Reed and Leopold (2016) suggested a lower paleoelevation for Florissant, along with a slightly higher MAT.

The ranges of MAT estimated for the four Oligocene floras were not as well constrained as those of Florissant. The Genus Abies set the upper limit for all four floras at 18.2 °C, but the lower constraints resulted in broad ranges for MAT (Table 10). Even so, the results for the Creede flora are supported by results from clumped isotope Δ47 thermography based on original lacustrine carbonates, which suggest a MAT of 9 °C ± 2 °C (Hyland and Huntington, 2015). The total cooling between the late Eocene Florissant flora and the Oligocene Creede flora is estimated at between 5 °C and 9 °C.

4.4. Detrended Correspondence Analysis (DCA)

One presentation that highlights the contrast between the late Eocene and the Oligocene floras is the detrended correspondence analysis (DCA) plot (Fig. 4). Eocene samples fell to the right of the plot, while Oligocene samples primarily fell to the left. Samples from the Antero Formation fell in the middle between Florissant samples and the Oligocene samples. Xeric type of genera such as Juniperus, Sarcobatus, Artemisia, and Ephedra fell on the left, while moisture-loving hardwoods such as Carya, Juglans, Engelhardia, and others fall on the right along the X axis. This suggests the taxa became arranged here according to their adaptations to mesic growing season conditions on the right side, and arid-tolerant taxa fell on the left side of the plot. Axis 1 accounted for 60% of the variance in the data, while Axis 2 accounted for another 10% of the variance in the data (Fig. 4) (C. Looy, 2016, personal commun.).

4.5. Other

We provide estimated PPT values suggested by Axelrod (1987), MacGinitie (1953), and Wolfe and Schorn (1989) (Table 6). The relatively small average leaf size (microphylls) and the prevalence of teeth on leaf margins suggested that the precipitation for the four Oligocene floras was somewhat limited (MacGinitie, 1953; Axelrod, 1987; see Wright et al., 2017). Additionally, the proportion of summer-moist dicots to total dicots dropped from 36% at Florissant to 13% by Creede (Table 4). The loss of dicots requiring summer-moist conditions (Fig. 7A) clearly suggests a decline in summer rainfall.

While reviewing the results of the CA analyses, we observed an apparent trend after Florissant in the decreasing role of “warm-adapted” taxa; these are taxa requiring MAT ranges above 0 °C (e.g., Carya) (MAT Low column on Table 5). In comparison, Juniperus and Artemisia are more “cool adapted” because they grow in areas where MAT values are at or below 0 °C (MAT Low column on Table 5). About 53% of the taxa at Florissant are warm adapted and require MAT ranges above freezing (Fig. 7B), but Antero records an apparent shift to majority of cool-adapted taxa (∼60%). The diminishing role of warm-adapted taxa continues through the Oligocene so that by the time of the Creede flora (27 Ma), the warm-adapted taxa represented only ∼32% of the floral list. This trend exemplifies the onset of cooler conditions during the Oligocene.

The CA method was used to analyze cold-month mean temperature (CMMT) values as well to see if trends in winter temperatures could be determined (Table 5). The CMMT values at Florissant were well constrained to between 6.0 °C to 8.6 °C based on the presence of Vauquelinia and Picea (Table 10). However, the ranges of CMMT for three of the Oligocene floras were in excess of 10 °C. Picea set the upper end of the range of CMMT at 8.6 °C for all the Oligocene floras (Table 10). The lower value for Antero was set by either Eucommia (a pollen ID) at –6.0 °C or by Mahonia (a megafossil ID) at –7.0 °C (a difference of at least 14.6 °C). The lower value for Pitch-Pinnacle was set by Heteromeles at 0.9 °C, which is better constrained than Antero. At –7.0 °C, Mahonia set the lower threshold for Platoro and Creede. These very broad ranges made interpreting the role of CMMT difficult.

5.1. Vegetation at Creede

At Creede, the generic composition of our pollen data suggests the development of a mixed pine woodland with understory shrubs such as Artemisia, Elaeagnus, Shepherdia, and Ribes (gooseberry), along with Juniperus and Poaceae (Benedict, 1991). These taxa are common in the understory of the Ponderosa pine and juniper woodland today. With the megafossil evidence provided by the reports of Wolfe and Schorn (1989) and Axelrod (1987), the vegetation represented by the Creede flora clearly establishes a type of Pinus ponderosa and Juniperus woodland. Axelrod reported that the megafossil species of shrubs at Creede are slightly different but closely related to the modern species of the area. The common plants of that woodland include the monotypic genus Jamesia, along with Ribes, Juniperus, and Cercocarpus. We record Quercus pollen, which supports the record of Quercus at Creede in Axelrod’s (1987) report. The Oligocene Creede flora seemed to be at the ecotone between forest and woodland: it resembled the mixed Pinus ponderosa and Juniperus woodland but also included megafossil elements of the present-day mixed conifer forest zone with Abies, Picea, and Pinus cf. aristata (Axelrod, 1987). Pollen from the cooler spruce and fir forest zone appears at Antero, Pitch-Pinnacle, and Creede (Fig. 5).

5.2. Vegetation Changes: Three Trends during the EOT

MacGinitie (1953) spoke of the Florissant flora as having been very like the semi-deciduous vegetation type of San Luis Potosi Province of Mexico (Rzedowski, 1966; Rzedowski, 1978). Boyle et al. (2008) called the Florissant vegetation subtropical to warm temperate in terms of modern counterparts. MacGinitie (1953) further described the terminal Eocene floristic trend as indicating increased seasonality of precipitation.

During the Oligocene, three important changes in the Paleogene floras occurred, and they demonstrate an increasingly continental climate: the loss of the warm-adapted, summer-moist arboreal dicots that typify the Eocene; the influx of new taxa, including sub-arid shrubs such as Artemisia; and the shift to the low-biomass shrubs and herbs, such as Bidens, Potentilla, Caryophyllaceae, Cucurbitaceae, and Sarcobatus, which typify the Oligocene of Colorado. It should be noted that grasses played a regular but minor role in the Eocene and Oligocene pollen floras of Colorado. Later, pollen data at some Miocene sites in Idaho record open grassland conditions (Leopold and Wright, 1985; Leopold and Denton, 1987).

5.2.1. Loss of Arboreal Dicots

While the gymnosperm taxa were similar in the late Eocene and Oligocene floras here, the composition of the angiosperms changed greatly during the EOT. The first shift was the loss by local extinction of many exotic taxa of eastern North American and East Asian affinity. These were warm-temperate, broad-leaved hardwood trees from the late Eocene flora (Wing, 1987). Because most genera of that group require summer-moist conditions, this loss of mesophytic hardwoods has distinct climatic overtones. The change suggests that a major drop of summer rainfall occurred. The graph in Figure 7A shows the severity of this abrupt loss at 33.8 Ma in the Antero flora. Many arboreal dicots were lost at the EOT; these include exotic taxa such as Platycarya, Carya, Eucommia, and Bombacoideae, all of which prefer fully humid summer-moist habitats. This trend in increasing aridity continued through the Oligocene so that by 27 Ma, all megafossil records of the summer-moist, broad-leaved taxa, including East Asian genera (e.g., Platycarya and Eucommia) were missing (Fig. 7A; Table 4). The number of woody dicots rebounded slightly, but taxa with a tree growth habit did not dominate the flora the way they did at Florissant (Table 4). Axelrod (1987) observed that the Creede climate was too cool to include the rich diversity of oaks found in Florissant.

Because the Creede flora had many generic similarities to the present local flora, botanists H.D. MacGinitie (1953) and D.I. Axelrod (1987) both declared that the Creede flora looked so “modern” (young) in aspect that it was probably late Miocene or Pliocene in age. However, the isotopic age of 26.9 Ma is well established (Lipman, 2007). The increasing similarity with the modern local flora has been used as a rough index of geologic age as expressed by Barghoorn (1951) and Wolfe and Barghoorn (1960) (see also Leopold, 1967, p. 226) for this feature in Rocky Mountain floras.

5.2.2. Incoming New Taxa

A second feature, along with the loss of broad-leaved trees, was the replacement by new incoming taxa. These included low-biomass, arid-adapted shrubs such as Artemisia, Shepherdia, Chamaebatiaria, Ribes, and Berberis. They represent some of the first appearances of understory plants that now characterize the native Ponderosa pine woodland and juniper shrublands of the southern Rockies (Benedict, 1991) not previously present in these Colorado floras.

5.2.3. Herbaceous Habit

The third development we see among the Oligocene samples is the appearance of low biomass plants with an herbaceous habit. Previous to the Oligocene, records of herbaceous Poaceae go back to the Cretaceous (Strömberg, 2011), but the typical herbs of the early Tertiary are aquatics such as Potamogeton or Nymphaea (Leopold and MacGinitie, 1972). In the Oligocene sediments, as an example, we see a record of perennial herbs such as the unique Potentilla, the pollen of which corroborates the herbaceous megafossil of that genus at Creede (Wolfe and Schorn, 1989). Potentilla is a widespread perennial herb or shrub of the Rosaceae. Poaceae are present in the Oligocene floras, as evidenced by regular minor appearances of grass pollen. Pollen and megafossils of Artemisia and other Asteraceae, including a megafossil of Bidens, are on record in the Oligocene floras; these are significant because pollen of the highly evolved family Asteraceae becomes more common in the Oligocene of Colorado. These are evolutionary jumps in the plant world (Axelrod, 1987; Graham, 1996), leading to the singular development in the Neogene of herbaceous groups such as Caryophyllaceae, Violaceae, Cucurbitaceae, Euphorbiaceae, Chenopodiineae, Amaranthaceae, and others, especially in the Asteraceae. Similar pollen records of new herbaceous and shrub taxa occur in the Neogene of Alaska (Leopold and Liu, 1994; Reinink-Smith et al., 2017) and concurrently in northwest China (Liu, 1988). All these features mark the Oligocene of Colorado as a special opening door for plants of a modern low-biomass landscape.

5.3. Evaluation of Climate Change

The Oligocene floras demonstrate an increase in taxa that are frost adapted and tolerate a cool climate. This is first seen at Antero (33.8 Ma), with a decline in MAT of ∼3.3 °C. This condition continued through Creede time (Table 6). The main cooling took place abruptly prior to ca. 33.8 Ma. More than half of the taxa at Florissant require MAT values above freezing, while the number of taxa at the four Oligocene floras were increasingly cool adapted, tolerating MATs at or below freezing (Fig. 7B; Table 5). Antero records a shift to majority of cool-adapted taxa (58%), and over two-thirds of the taxa at Creede were cool adapted (66%) (Fig. 7B; Table S2 [footnote 1]).

Axelrod (1987) determined the paleoMAT by the modern range of several species that are related to Creede taxa. These include Pinus chihuahuana (Pinus coloradensis fossil), Pinus montezumae (Pinus riogrande fossil), and Pinus engelmanii (Pinus engelmannoides fossil). These taxa now occur at elevations from 2.0 to 2.3 km and have a MAT of 11 °C to 12 °C. Axelrod reported that a wide number (about a third) of the taxa at Creede are clearly related to living species at the forest-woodland ecotone. Axelrod suggested therefore that the MAT at Creede was ∼11.5 °C (Axelrod, 1987, p. 57) as compared to the 1.9 °C MAT at Creede today (Axelrod, 1987; Wolfe and Schorn, 1989) and that the CMMT was 3.5 °C. In contrast, Wolfe and Schorn (1989) suggest a very low paleoMAT of <2.1 °C.

The trends across the EOT record a definite decline in winter temperatures, corresponding to the “terminal Eocene event,” a cold period described for other western U.S. floras by Wolfe (1978, 1981, 1995). The EOT is marked by major environmental change, namely the development of a cooler climate overall and in the trend toward colder winters (CMMT; noted by Wolfe and Leopold, 1967; Axelrod, 1987; Wolfe and Schorn, 1989, p. 192) (Fig. 7B; Table 10), as well as a definite drop of estimated PPT. The increase in aridity is suggested by several authors, the DCA findings, and by our moisture-index data (Fig. 6; Wolfe, 1981; Axelrod, 1987, p. 57; Wolfe and Schorn, 1989; Wolfe, 1992).

The Sørensen index results indicate that the Florissant similarity to the Oligocene floras is only 57% (when all the Oligocene floras were combined into a single unit; Table 9). Even allowing for taphonomic bias (Pitch-Pinnacle and Platoro being small floras), there is no strong similarity between the Florissant and the Oligocene communities. The similarity of Florissant to Creede was only 47%, which supports the notion that the differences in these communities were becoming increasingly pronounced over time. These results are consistent with shifts in the ecological communities, beginning at the EOT, as shown by the disappearance of many warm-adapted taxa after Florissant and the influx of new, derived families that appear in the Oligocene floras (Table 4). The similarity between the older versus younger Oligocene data sets is moderate at 74%. This, along with the similarity between Antero and Creede (70%), underscores the influence of the new families and low biomass taxa in defining the Oligocene communities; the shift emphasizes the climatic changes that these new families and taxa were able to exploit.

5.3.1. Climatic Cooling

A trend across the Eocene–Oligocene floral transition is marked by striking climatic changes. These include (1) an apparent cooling seen as a decline in MAT first seen in the Antero flora; (2) evidence of clear differences between the Oligocene floras and the late Eocene Florissant flora; and (3) a growing aridity, as demonstrated by the DCA graph (Fig. 4). Though the Oligocene floras are less diverse, they contain an increasing number of taxa that tolerate cool to cold conditions. We infer that the MAT probably declined to ∼9 °C by Creede time (26.9 Ma), a total cooling of ∼6 °C, over ∼7 m.y. (Table 6).

5.3.2. Evidence of Drying: DCA and Growing Aridity

The DCA pattern reinforces the ecological pattern (moist-loving taxa of Florissant versus many sub-arid Oligocene types) that we recorded among the fossil taxa. When the DCA method was applied to these floras, a change was obvious immediately at Antero. The results of DCA demonstrate clearly a fundamental shift from the summer-moist elements of the Florissant flora on the right (e.g., FLO-a and FLO-b, Fig. 4) to the transitioning Antero and FLO-c samples in the middle, and the semi-arid Pitch-Pinnacle, Platoro, and Creede floral data together on the left. The component genera of each flora cluster closely together (Fig. 4; Table 6); the taxa themselves suggest that the data along the horizontal axis represent a sharply declining precipitation regime during the early Oligocene.

5.3.3. Moisture Index Values: AE/PE—Interpreting Aridity

Water stress is one of the most important physiological challenges for tropical tree species (Brenes-Arguedas et al., 2011). In addition, seasonal rainfall clearly determines the geographical distribution of warm-temperate forest ecotypes in eastern United States (Thompson et al., 2012; Fig. 1). Precipitation gradients correlate with patterns of species richness at macroecological scales (Clinebell et al., 1995). Because of this relationship, we would expect that a progressive turnover of species identities would occur due to changes in precipitation. The transition in composition of the floras during the EOT demonstrates turnover of species relating to moisture availability, as indicated by inferred soil-moisture indices.

Figure 6 shows a plot of the values for actual evaporation (AE) rate divided by the potential evaporation (PE), on the horizontal axis. These are some examples of soil-moisture data (in different taxa) that represent three groups of taxa considered as “outgoing,” “continuing,” or “incoming” types during the EOT. The data were based on biogeographic range of taxa presented by Thompson et al. (2012), e.g., from summer-moist hardwood forest and xeric southwestern regions that exemplify the changing phases of the EOT with respect to soil-moisture availability (Fig. 6).

5.3.3.1. Group One (Fig. 6A: Outgoing taxa)

The broad-leaved hardwood taxa (e.g., megafossils of Carya, Juglans, Tilia, Ostrya, Carpinus, Castanea, and Diospyros) require high moisture-index values between 0.8 and 1.0. These are the outgoing arboreal genera that have distributions in summer-moist areas of eastern United States or East Asia, with a preference for moist growing-season conditions in July (Thompson et al., 2012). These elements were present in the Florissant flora but are absent in the Oligocene floras. Summer-moisture availability in the area undoubtedly dropped below AE/PE = 0.8 after the EOT. Axelrod came to a similar conclusion (1987, p. 62), as did Wolfe and Schorn (1989, p. 191).

5.3.3.2. Group Two (Fig. 6B: Continuing or streamside taxa)

The continuing taxa are streamside forms (e.g., species of Betula, Alnus, Populus, and Salix) that have AE/PE moisture-index values that run the gamut from low to high between 0.2 and 1.0 on the horizontal axis. Their tolerance of moisture conditions is wide ranging. These taxa can succeed by occupying streamside positions in spite of drought. Ulmus may be one of these taxa because it thrives on stream margins. Most of these continue in time through the Creede flora.

5.3.3.3. Group Three (Fig. 6C: Incoming taxa)

Within the incoming taxa group are the now-native western shrubs and trees well adapted to high evaporation values in the environment. They grow in arid areas of interior United States (see Barnosky, 1984). We have called these “newcomers,” because many of these first appear in the Oligocene records in Colorado. They are highly adaptable, with low AE/PE values of soil moisture ranging from 0.2 to <0.8. These moisture-index values do not overlap with those of the deciduous hardwood taxa, Group A (Fig. 7A). The incoming new taxa (e.g., Artemisia, Quercus cf. grisea, and Shepherdia cf. argentea) have smaller leaves and can compete in a high-stress (high-evaporation rate) environment. Along with these were the herbs that, in the United States, appear as new derived families mainly known from Miocene and younger floras (e.g., Cucurbitaceae and Caryophyllaceae).

5.3.4. Evaluation of Growing Aridity and Floristic Consequences

For many plant taxa, data are available that suggest soil-moisture conditions (Thompson et al., 2012). Soil moisture represents a limiting factor that relates strongly to the groups of taxa that responded to climate change at the EOT (Fig. 6). The Florissant flora embraced a few genera in these three groups (e.g., Carya, Betula, and Quercus) at a time when the evaporation rates were low enough in relation to summer rainfall to permit the broad-leaved hardwoods to exist (typically >100 mm July average precipitation). In contrast, the smaller Oligocene assemblages that followed shifted in aspect from summer-moist to summer-dry conditions. Clearly the factor that was critical in the loss of the broad-leaved trees was the impact of increased evaporation rate (relative to precipitation; AE/PE <0.8), resulting in the lower diversity observed in the late Oligocene floras. Significantly the streamside and lakeside taxa (Fig. 6B) were able to persist during the EOT on the margins of wetland areas in spite of this change. The incoming new families and taxa of which we have record had the advantage of tolerating high evaporation rates (AE/PE <0.8; Fig. 6C). Their influx demonstrates that they succeeded in the increasingly continental climate. Finally, the loss of summer-moist dicots, seen in the drop from ∼36% at Florissant to almost 13% by Creede (Fig. 7A; Table 4), supports our conclusion of a decline in summer PPT.

5.4. Local Environments of the EOT in Colorado

The plants record the temperature trends and demonstrate that the climate at Florissant and the Oligocene floras were warmer than they are today. For example, many species identified by Axelrod (1987) and Wolfe and Schorn (1989) now grow in regions farther south, in New Mexico and Mexico.

The final record of the latest Eocene local environment is well recorded at Florissant. The lakeside slopes were characterized by a mesic forest dominated by Sequoia affinis. Broad-leaved deciduous hardwood trees with diverse members of Juglandaceae, Tilia, and such flourished, along with Sequoia in the Florissant lowland forest. A rich mixture of Ulmaceae, Salix, and Fagopsis covered the shorelines. Herbs were a minor element. The climate was warm (MAT between ∼14 and 18 °C; Zaborac-Reed and Leopold, 2016; Table 6) with semi-arid conditions marked by gypsum accumulations occurring on certain highland pond areas (MacGinitie, 1953, p. 56). Occasional light frosts may have occurred (MacGinitie, 1953).

In contrast, the younger floras that followed Florissant record simpler flora with smaller leaves, an increasingly drier environment (evaporation increased relative to rainfall), a lower average mean annual temperature of 12.7 °C (Antero and Pitch-Pinnacle), and later ∼9 °C at Creede (Table 6), and a growing role of cool-adapted taxa. The samples from Antero showed a smaller flora in transition, suggesting that vegetation changes were likely due to one large climatic shift immediately after Florissant. The region was increasingly raked by fire and ash fall from nearby volcanoes (as indicated by the burnt plant tissue and pollen grains at Platoro). Extremes in aridity occurred under a cool climate driven at least in part by the global cooling associated with Antarctic glaciation (Oi-1; Cather et al., 2008). By Creede time, the vegetation became open woodland with an assortment of rosaceous shrubs, Artemisia (sagebrush), Elaeagnaceae, and grass (Table 4). Pine and fir forests were dominant at higher levels on the Creede crater walls, while at lower woodland sites, the dominant conifer was Juniperus with largely drought-tolerant shrub genera native to Colorado today (Wolfe and Schorn, 1989). The dry, cool conditions with frosts apparently continued to the end of the Oligocene at Creede.

Our data suggest three broad changes in temperature may have occurred in Colorado during the Oligocene (Table 6). The first change was the abrupt difference in MAT (from Florissant to Antero) from an average of ∼16 °C to 12.7 °C in 0.3 m.y. The second and more moderate drop (from Pitch-Pinnacle to Platoro) was ∼6.5 °C (over 5.3 m.y.), which is significantly slower. The third change suggests a difference in MAT of ∼2.8 °C (between Platoro and the Lake Creede), which may be due in part to elevational differences between the two sites (Table 2). Overall the botanical data support a regional cooling during the EOT that probably began very rapidly. A trend of continuing and increasing aridity is displayed by the vegetation and flora.

5.5. How Do the Following Tools Assist in Interpretation?

The coexistence approach serves to quantify estimates of the main temperature features. The detrended correspondence analysis documents the basic shift toward arid-adapted taxa after Florissant. As a statistical tool, the Sørensen index of similarity quantifies similarities between floras. The use of NLR analysis gives a very direct way to quantify the significance of floral changes, e.g., of temperature and/or aridity of the environment. Soil-moisture indexes provide a link between the identified taxa and their requirements.

5.6. Development of the Cordilleran Flora and Creede

The combined loss of exotic taxa, along with the replacement of these by newer plant groups important in the modern vegetation at Creede, are evidence that the Cordilleran flora of southern Colorado became more modern in aspect at this time. Axelrod and Raven (1985) wrote very clearly that the present-day Cordilleran flora probably originated back in 27 Ma with the Creede flora; they suggested that this was the time for the origin and development of the present southern Rocky Mountain flora.

5.7. What Estimates of Paleoelevation Can Be Made Based on These Floras?

Evidence of erosional events brought forth by Cather et al. (2012) and Karlstrom et al. (2012) supports low to moderate paleoelevations for the five floras examined in this study but not the high (or in some cases extremely high) paleoelevations reported by some authors (Fig. 3; Tables 1 and 2). Cather et al. (2012) concluded that the average elevation of the Rocky Mountains at ca. 45 Ma was ∼350 m above sea level and that the Florissant area paleoelevation may have been ∼1.5 km at the time of deposition. MacGinitie (1953, p. 53) indicated the paleoelevation of Florissant may have been between 0.3 and 0.9 km.

Karlstrom et al. (2012) concluded that the region experienced 0.5–1.0 km of surface uplift in the Neogene. The uplift in elevation since Creede deposition inferred by Steven and Ratté (1965), using structural evidence, provides a presumed uplift of ∼1.5 km. These results constrain the paleoelevation of the floras of this study to low to mid elevations at the time of deposition, allowing for Neogene uplift of between 0.5 and 1.5 km to their subsequent modern elevations (Fig. 3; Table 2). This gives a rough estimate for paleoelevation prior to the Neogene (Fig. 3). It should be noted, however, that elevation estimates can vary due to many factors. Differences in elevation between sites give the potential for a great deal of complexity in how temperature differences might be calculated and interpreted.

Wolfe (1992) stated that the paleoelevation of a floral site may also be estimated appropriately by using paleotemperature—especially MAT. There are various techniques for calculating MAT: e.g., CA, NLR, and plant physiognomic methods such as CLAMP (Wolfe, 1993). However, CLAMP was considered problematic because the inferred temperature results tended to be consistently cooler than expected (e.g., by 2 °C; Peppe et al., 2010), producing paleoelevations near modern levels and with large margins for error (Fig. 3). Furthermore, Cather et al. (2012) stipulated that high paleoelevational results predicted by the use of leaf physiognomic techniques (such as CLAMP) are very unlikely, given the results of their mapping study.

Finding paleoelevation using methods that depend on modern lapse rates of a region can be problematic. Axelrod and Bailey (1976) suggested many reasons why lapse rates may have been different in past climatic regimes. Local lapse rates reflect local topography; however, lapse rate patterns from one side of a range can be completely different from the other side, especially in complex terrain (Minder et al., 2010). For instance, applying a normal lapse rate of 183m/°C to the projected uplift of between 500 and 1500 m, a cooling of ∼2°–8 °C might have resulted from the Neogene uplift. When topography is unknown, assumptions about a site invariably lead to such widely varying results (e.g., Gregory and McIntosh, 1996; Cather et al., 2012; Table 2; Fig. 3). The paleoelevations derived from using botanical assemblage-based methods such as NLR and CA fit well with additional support from studies using vesicle paleoaltimetry, oxygen-isotope geochemistry, and carbonate-clumped isotope thermometry. These are all generally compatible with episodes of uplift and erosion observed in Cather et al.’s (2012) and Karlstrom et al.’s (2012) summaries.

5.8. Regional and Global Correlation

In northern Europe during the EOT, a sharp cooling lowered the MAT by ∼5°–8 °C (Mosbrugger et al., 2005). Sea-surface temperature dropped 9°–10 °C around the globe (Zachos et al., 2001) causing “icehouse conditions.” An extinction of ∼60% of woody plant taxa occurred in Oregon, where the MAT dropped ∼10 °C (Retallack et al., 2004; Dunn, 2007; Table 11). However, the EOT resulted not so much in the extinction of plant taxa as in a local extirpation and dispersal to lower elevations and latitudes (Wolfe, 1987). A decline in aquatic amphibians and some reptiles in North America occurred, related to an increase in aridity as well as a decrease in the number of rivers, streams, and other water sources (Hutchison, 1992, in Prothero and Berggren, 1992). In Nebraska, the physical differences in soil type between the Brule and the Chadron formations demonstrate a clear trend toward aridity at the Eocene–Oligocene boundary (Terry, 2001).

In the early Oligocene of northeastern China, the southward spread and increasing importance of shrubs such as Ephedra and cf. Nitraria is evident. The setting in China is interpreted as a low-biomass, woody savannah with arid-adapted shrubs or open deciduous forest with temperate hardwoods by the end of the Oligocene (Liu, 1988; Leopold et al., 1992). A sharp cooling in Argentina accompanied by evidence of fire and burnt phytoliths portray environments of severe conditions (Selkin et al., 2015) that mirror volcanic conditions at the Platoro site (Plate V), where some foliage, pollen, and spores appear badly burnt. A wave of extinctions off the coast of Tanzania occurred at this time (Pearson et al., 2008).

In Colorado and New Mexico, the EOT was a turbulent period. Violent eruptions in the Southern Rocky Mountain volcanic field of Colorado and the Mogollon-Datil volcanic field of New Mexico were accompanied by large caldera collapses. Enormous ash falls created dunes covering an area the size of Wisconsin (Cather et al., 2012). Called the aeolian Chuska erg (Fig. 1), these deposits had a thickness of as much as ∼535 m in the Chuska Mountains. Their aerosols may have been a local factor in cooling the climate by blocking solar radiation.

The initial drop in temperature of ∼3.3 °C in 0.3 m.y. (34.1 Ma to 33.8 Ma) was so fast that it seems unlikely to have resulted solely from uplift. It was probably related to the global cooling and Antarctic glaciation Oi-1 event as recorded by Zachos et al. (2001). We interpret that the subsequent slower cooling (between 32.9–29 Ma and 28 Ma) may also be related to global cooling. At the end, the increase of almost 3 °C per m.y. (between 28 Ma and 27 Ma) might reflect the combination of elevational differences between Platoro and the Lake Creede as well as a slight warming near the end of the Oligocene (see Zachos et al., 2001). The total MAT decline in the region of between 5 °C and 9 °C during the Oligocene may have involved a complex interaction between global cooling, minor surface uplift at local volcanoes, and some local cooling due to volcanic aerosols. During the Miocene, the flora at the Troublesome Formation of Central Colorado records a number of mesic hardwoods that indicate their continuance and a return to somewhat more summer-moist growing conditions (Leopold, in Tschudy and Scott, 1969, p. 409). Independent evidence demonstrates that major surface uplift in the southern Rocky Mountains occurred during the past 10 m.y. (Cater, 1966; Cather et al., 2012; Karlstrom et al., 2012).

The Eocene–Oligocene transition (EOT) was marked globally by one of the most abrupt and severe environmental changes in the Cenozoic record (see Zachos et al., 2001). The present report emphasizes the significance of floristic change during the Oligocene in Colorado. We document a major floristic turnover based on five local floras that span the late Eocene through most of the Oligocene. The late Eocene Florissant and the Oligocene Antero, Pitch-Pinnacle, Platoro, and Creede floras are a basis for estimating mean annual temperature (MAT) changes and developing aridity through the EOT and the Oligocene interval. In addition, we combined the new Oligocene pollen data (not previously published) with the megafossil identifications at Creede in order to access the whole flora. Based on these data, we prepared a corrected list of the Creede flora.

Based on recent studies, the southern Rocky Mountains probably lay at ∼1.5 km in elevation, though growth of volcanoes may have raised local areas to as high as ∼1.8 km during the Oligocene (Meyer, 1986; Axelrod, 1987; Cather et al., 2012; Table 2). Major uplift of the Colorado Plateau began in the late Miocene and may have continued until the Holocene, raising these sites by ∼0.5 km to 1.5 km in elevation after the Oligocene floras were deposited (Steven and Ratté, 1965; Karlstrom et al., 2012).

Climate and Floristic Change

The four Oligocene floras of south-central Colorado formed during a period of super-active volcanism and heavy ash fall. The Oligocene cooling and increased aridity in Colorado correlate with the Oi-1 glaciation in Antarctica. The EOT data reveal a severe drop in floral diversity at 33.8 Ma immediately after the summer-moist, warm Florissant flora. This change records an abrupt shift to a cool, semi-arid climate that lasted through most of the Oligocene interval (to 27 Ma). The Oligocene floras appear to represent a cool period that Jack Wolfe called the “terminal Eocene event” (1978). Because the Florissant and Antero floras are close in terms of age, location, and elevation, the floristic differences between the two floras observed at the EOT are likely related to global cooling and increasing aridity rather than surface uplift.

The estimated decline in mean annual temperature of between 5° and 9 °C during the Colorado Oligocene involved a complex interaction between global cooling and volcanic aerosols. This cooling compares with the global EOT cooling of ∼5 °C. Indications from Europe, Africa, northern China, Argentina, and other sites, including global marine records (Table 11), record remarkable changes in world climate starting at the end of the Eocene or earliest Oligocene.

In Colorado, many exotic plant genera of eastern North American and East Asian affinity that depend on July rainfall became locally extinct, and were replaced by taxa tolerant of semi-arid conditions. The lost genera are a basis for estimating the decline of summer precipitation leading to growing aridity during the Oligocene. Physical changes in soil profiles are noted in the EOT Chadron and Brule formations.

Evolutionary Significance

This important fossil record demonstrates an abrupt floral shift during the EOT from a hardwood forest flora rich in exotic mesic hardwoods to a modern, pine-fir woodland, with an understory of arid-adapted shrubs such as Artemisia, Cercocarpus, and low biomass herbs. This shift features the appearance of several endemic and small genera that now characterize the Cordilleran flora (e.g., Jamesia and Chamaebatiaria). Additional Oligocene taxa include the appearance of several highly evolved “new” plant groups (e.g., Asteraceae, Caryophyllaceae, and Elaeagnaceae) not typical of the early Cenozoic in the Rocky Mountains. Such taxa expanded regionally during the Oligocene. Our data lend support to the theme of Axelrod and Raven (1985), who described the Oligocene interval in Colorado as being the time when the Cordilleran flora of the southern Rocky Mountains first appeared.

The developing cooling and aridity caused a loss of broad-leaved dicots that were replaced by drought-resistant and cold-tolerant conifers, shrubs, and new herbs during the Oligocene in Colorado. The environmental changes favored understory plants of low biomass resistant to drought.

In summary, the reshaping of the montane flora in southern Colorado during the Eocene–Oligocene transition was a product of interactions between several environmental factors: active volcanism; global cooling; the Antarctic glaciation of Oi-1; and losses in warm season precipitation. The developing aridity and cooling caused a loss of broad-leaved dicots that were replaced by drought- and cool-tolerant conifers, shrubs, and herbs during the Oligocene in Colorado. This floristic shift may illustrate a cause and effect, namely that cooling and aridity were the main stimulus that forced evolutionary changes in the regional vegetation.

The authors wish to express their gratitude for the editorial comments and support of Herb Meyer, Karl Karlstrom, Peter Lipman, Cindy Looy, Jeff Benca, Caroline Strömberg, Robyn Burnham, Steve Manchester, David W. Love, and Peter Wilf. The authors also wish to thank the University of Washington Biology Department and the Expo ’90 Foundation and International Cosmos Prize for their support. We thank Shanaka de Silva, Rhawn Denniston, and the anonymous reviewers at Geosphere for their thoughtful comments.

1Supplemental Items. Table S1: Slides used in this study and their collection locations. Table S2: warm-adapted versus cool-adapted taxa (support for Figure 7B). Please visit https://doi.org/10.1130/GES01689.S1 or access the full-text article on www.gsapubs.org to view the Supplemental Items.
Science Editor: Shanaka de Silva
Associate Editor: Rhawn Denniston
Gold Open Access: This paper is published under the terms of the CC-BY-NC license.