Climate & Ecology in the Rocky Mountain Interior After the Early Eocene Climatic Optimum

As atmospheric carbon dioxide (CO2) and temperatures increase with modern climate change, ancient hothouse periods become a focal point for understanding ecosystem function under similar conditions The early Eocene exhibited high temperatures, high CO2 levels, and similar tectonic plate configuration to today, so it has been invoked as an analog to modern climate change. During the early Eocene, the greater Green River Basin (GGRB) of southwest Wyoming was covered by an ancient hypersaline lake (Lake Gosiute; Green River Formation) and associated fluvial and floodplain systems (Wasatch and 15 Bridger Formations). The volcaniclastic Bridger Formation was deposited by an inland delta that drained from the northwest into freshwater Lake Gosiute and is known for its vast paleontological assemblages. Using this well-preserved basin deposited during a period of tectonic and paleoclimatic interest, we employ multiple proxies to study trends in provenance, parent material, weathering and climate throughout one million years. The Blue Rim escarpment exposes approximately 100 meters of the lower Bridger Formation, which includes plant and mammal fossils, solitary paleosol profiles and organic remains 20 suitable for geochemical analyses, as well as ash beds and volcaniclastic sandstone beds suitable for radioisotopic dating. New 40Ar/39Ar ages from the middle and top of the Blue Rim escarpment constrain age of its strata to ~49.5–48.5 Ma ago, during the “falling limb” of the early Eocene climatic optimum. We used several geochemical tools to study provenance and parent material in both the paleosols and the associated sediments and found no change in sediment input source despite significant variation in sedimentary facies and organic carbon burial. We also reconstructed environmental conditions, including 25 temperature and precipitation (from paleosols) and the isotopic composition of atmospheric CO2 from plants found in the floral assemblages. Results from paleosol-based reconstructions were compared to semi-co-temporal reconstructions made using


The Eocene Period as an Analog for a Future Warm World
The anthropogenic release of fossil fuels drives a rapid and sustained increase in atmospheric carbon dioxide (CO 2 ) 45 that is coupled with climate change (IPCC 2007). To understand the effects of elevated CO 2 on the Earth (e.g., Cotton et al., 2013), we seek out geological periods with high temperatures and high atmospheric CO 2 for comparison. The early Eocene climatic optimum (EECO) has been invoked as a climate analog for our projected future (e.g., Zhu et al., 2019). This warming during the EECO occurred 53.26-49.14 million years ago (Cramwinckel et al., 2018;Westerhold et al., 2018), with peak warming from 51.5-50.9 Ma; this period consisted of long-term global temperature maxima and high CO 2 levels but was 50 tectonically comparable to today West et al., 2020). From the Paleocene to early Eocene, it has been inferred that there were extensive temperate forests dispersed throughout North America (Smith et al., 2012; Breedlovestrout the early Eocene (~41 °N; Wolfe et al., 1998;) is thought to be relatively comparable to its position today, thus changes in the climate of this region are not related to changes in latitude. 80

Using multiple proxies to characterize an environment
High resolution, well-informed snapshots in time of individual regions with thorough climate reconstructions help us to understand climate change and subsequent ecosystem dynamics (Guiot et al., 2009;Li et al., 2010;Shala et al., 2017). There are several terrestrial proxies that can be used to reconstruct paleoclimate and paleoecology, e.g., pedogenic carbonate isotope values (Cerling 1992), floral assemblages (Wilf 2000;Spicer et al., 2009), and stomatal density (Royer 1999). The preservation 85 of abundant, high-quality organic and inorganic specimens and samples due to the tectonic assemblage of the basin, makes the GGRB an excellent location for a multi-proxy approach. Organic geochemistry, specifically, isotopic values of plants (δ 13 C plant ) have been used to reconstruct the value of the isotopic composition of the atmosphere -termed δ 13 C atm (Arens et al., 2000;Stein et al., 2019). δ 13 C atm reflects sources of CO 2 gas to the atmosphere (Keeling 1979;Boutton 1991); for example, mantle degassed values of δ 13 C atm tend to be around −5.4 ‰ (Deines 1992). Identifiable organic fossils with individually compressed 90 leaves can be used to reconstruct the value of the atmosphere. Multiple inorganic geochemistry proxies (see methods) can provide context for the depositional environment (e.g., Sheldon et al., 2006), climate (e.g., Gallagher & Sheldon 2013), hydroclimate (e.g., Sheldon et al., 2002), age (e.g., Turner 1971), and origin of sediments (e.g., Sheldon & Tabor 2009).
Together, these proxies can inform us on ecosystem and depositional dynamics in the context of Eocene climate. This study seeks to combine these proxies to create a holistic reconstruction of the greater Green River Basin (GGRB) during the early 95 Eocene using the extensive deposits of the Blue Rim escarpment.

Description of locality
The Bridger Formation is an approximately 842 m thick series of tuffaceous deltaic-alluvial and minor lacustrine sedimentary strata which overlie and interfinger with the Green River Formation (Koenig 1960;Kistner 1973;Murphey & 100 Evanoff 2001;Clyde et al., 2001). Vertebrate fossils collected from the Bridger Formation formed the basis for defining the Bridgerian North American Land Mammal 'age' ("NALMA"; e.g., Osborn 1909;Wood et al., 1941;Van Houten 1944;Gingerich et al., 2003;Robinson et al., 2004). Mapping and biostratigraphy have permitted further subdivision of the Bridger Formation into intervals A-E (Matthew 1909;Murphey et al., 2011). Several volcanic ash horizons within the Bridger Formation and the underlying Green River Formation have been radioisotopically dated using 40 Ar/ 39 Ar 105 geochronology to have accumulated between approximately 50 Ma and 46 Ma ( Fig. 1; Table S2; Smith et al., 2008). Radioisotopic ages reported or discussed in this contribution have all been calculated using the 28.201 Ma age for the Fish Canyon tuff sanidine standard, and are thus comparable with modern U-Pb geochronology (Kuiper et al., 2008;). Strata exposed along the Blue Rim study area represent the oldest exposed portion of the Bridger Formation.
Unlike much of the Bridger Formation, these 'Bridger A' deposits have not been radioisotopically dated, but regional mapping 110 and correlations suggests they were deposited above the Sand Butte bed of the Laney Member of the Green River Formation, and likely occur beneath the more extensively mapped Bridger B interval , bracketing a depositional age between the ca. 50 Ma '6 th tuff' of the Green River Formation and the ca. 49 Ma age for the Church Butte tuff, which occurs within Bridger B (Fig. 1).
Several potential sediment sources may have contributed to the Bridger Formation: including siliciclastic material 115 of a variety of compositions derived from Phanerozoic strata and underlying basement exposed by Laramide uplifts which surround the basin (Smith et al., 2015); volcaniclastic and siliciclastic sediment delivered by the Idaho paleoriver (Chetel et al., 2011); volcanic ashfall from the Challis and Absaroka volcanic fields; and autochthonous lacustrine carbonates Fig 1a). Evidence supporting these as potential sources include common detrital feldspar ages in the sand grains that are similar to depositional ages (i.e., recently erupted volcanic grains), and many euhedral volcanic biotite and felsic 120 volcanic lithic grains in the Bridger Formation sandstones (Chetel et al., 2011). Whereas the older, and in part coeval Laney Member of Green River Formation is composed primarily of carbonate lacustrine sediments, the Bridger Formation is composed principally of siliciclastic sediment, with several minor intervals of lacustrine carbonate . Sediment accumulation during Bridger Formation deposition appears to have been relatively continuous in the basin center based on existing age control . Because of this, the Bridger Formation 125 has pristine faunal and floral preservation, making it an excellent candidate for understanding ecosystem function Allen et al., 2015;Allen 2017b). The Bridger Formation is exposed laterally over 12 kilometers at Blue Rim and is locally approximately 100 m thick (Kistner 1973). The flora at Blue Rim has been collected and described in great depth and is known for excellent plant preservation including leaves, flowers, fruits, seeds, wood, pollen, and spores (Allen 2017a/b). Eocene floral assemblages at 130 Blue Rim occupy warm, wet biomes not unlike modern subtropical ecosystems; angiosperms including palms were abundant and dicotyledonous taxa were up to 28 m tall (Allen 2015; Allen 2017a/b), antithetical to the dry scrub desert found at Blue Rim escarpment today. Of the multiple quarries created for plant fossil excavation (see Allen 2017a/b), the lower horizon (older) floral assemblage consists of taxa such as the abundant climbing fern Lygodium kaulfussi (fern, Lygodiaceae), "dicots" like "Serjania" rara (soapberry, Sapindaceae), Populus cinnamomoides (poplar, Salicaceae) and Landeenia arailiodes 135 (Sapindales , Goweria bluerimensis (Icacinaceae), and Phoenix windmillis (palm; Arecaceae; Allen 2015; Allen et al., 2015; Allen 2017b). The upper horizon preserves taxa such as Macginitiea wyomingensis (sycamore; Platanaceae), Populus cinnamomoides (poplar, Salicaceae), Cedrelospermum nervosum (elm; Ulmaceae), "Serjania" rara (soapberry, Sapindaceae), and many more (Allen 2017b). 140 3 Methods

Geochronology
Volcaniclastic beds were sampled from the Blue Rim escarpment for 40 Ar/ 39 Ar geochronological dating (e.g., Turner 1971): two samples from the prominent 'blue-green marker' unit (Fig. 3, that occurs approximately halfway up the section; and two sand beds that crop out near the top of the exposure. To prepare sanidine phenocrysts for analysis, 145 samples were crushed, leached in dilute hydrochloric acid (HCl) and hydrofluoric acid (HF) prior to hand-picking of sanidine in refractive index oils using a petrographic microscope, and then ultrasonic cleaning in acetone and ethanol. Sanidine phenocrysts from sampled ash beds were irradiated adjacent to standard sanidine crystals from Fish Canyon tuff (FCs) in cadmium shielding within the TRIGA (Training, Research, Isotopes, General Atomics) water-cooled, low-enriched uranium/zirconium fuel reactor at Oregon State University. Single sanidine crystals were fused using a 25W CO 2 laser and 150 then analyzed for 40 Ar/ 39 Ar composition using a MAP 215-50 mass spectrometer attached to a metal ultra-high vacuum (UHV) gas extraction and clean-up line at the University of Wisconsin Madison WiscAr laboratory. A 28.201 Ma age for Fish Canyon sanidine standard (FCs; Kuiper et al., 2008) was used to calculate apparent ages for each laser fusion analysis, and weighted mean ages were calculated for the youngest coherent population of sanidine dates from each sample. For populations that exceed an MSWD of 1, uncertainties in the weighted mean were multiplied by the inverse of the square root of the MSWD to 155 reflect the additional uncertainty implied by the associated age scatter.

Stratigraphy and fossils
Starting at the base of the Blue Rim escarpment, adjacent to the first sampled paleosol (Fig. 2), an updated 160 stratigraphic column to that found in Allen's (2017b) dissertation was measured (41.7985 °N, -109.5856 °E) in September 2019 (Fig. 3,4,5,6,7). This 67 m stratigraphic column traced the flanks of the escarpment to the top of the badlands. The stratigraphic column was sampled every 3 m (approximately the height of two Jacob-staffs) or at every interval of visual change (color, texture). In addition, plant fossils and plant hash were quarried at two locations approximately halfway (26 m, 33 m) and close to the top of the stratigraphic section (51.5 m, 52 m) for organic-rich fossil samples for isotope and bulk 165 chemistry analysis.
Fresh rock material was excavated by digging at least 20 cm into the surface, avoiding all traces of modern pedogenesis or surficial climate influence (i.e., modern roots or carbonate nodules). One distinctive, laterally continuous paleosol at the base of the section was sampled in five individual profiles over 440 m ( Fig. 3; Fig. S1, S2, 19BRWY2-6). Samples were collected from each horizon present, with no fewer than three samples per paleosol profile. At each location, every present horizon (A, 175 B, C) was sampled. Epipedons were present for all paleosols except 19BRWY4. For horizons that had color, texture, and/or other physical intra-horizonal changes, multiple samples were collected.

Isotope analyses
Paleosol samples were ground to 70 μm in a shatterbox. Approximately 10 g aliquots of samples were weighed out 180 and then acidified in 5% hydrochloric acid (HCl) for 30 minutes to remove carbonate and leave behind total organic carbon of bulk sample. After 30 minutes, these samples were decanted, then re-acidified for a total of three times (and/or until solution stopped bubbling). After these acid washes, they were rinsed with deionized water three times (or more, if given >3 acid washes). Samples were then dried in an oven at 50 °C for 72 hours.

Weathering indices and leaching 210
Weathering was quantified using Chemical Index of Alteration of B horizons (CIA; Equation 1; Nesbitt & Young 1982), a feldspar weathering index based on the discrepancy in ion mobility during weathering.
To test for alteration and expected pedogenic elemental trends, changes in individual element mobility and strain 215 were explored using mass balance (Equations 2-3; Chadwick et al., 1990), where ε represents the strain on an immobile element like Ti or Zr and τ represents the relative gain or loss of a mobile element relative to the paleosol's parent material.
The Paleosol Weathering Index (PWI; Equation 4; Gallagher & Sheldon 2013), which is based on differential bond strengths in cation oxides, provided additional means for examining weathering. Molar concentrations are used to make calculations in Equations 1-4, rather than elemental concentrations. 225 = ((4.20 * ) + (1.66 * ) + (5.54 * ) + (2.05 * )) * 100 (4) We used several geochemical proxies to examine intensity of leaching, including the ratio of barium to strontium (Ba/Sr), which is higher with more leaching and lower with less leaching (Sheldon 2006;Retallack 2001) due to differential solubility; Sr is more soluble than Ba (Vinogradov 1959). The ratio of the sum of base cations to titanium is another metric for 230 leaching, under the assumption that titanium is immobile, while other bases are mobile (Sheldon & Tabor 2009). The ratio of the sum of base cations to aluminum has been used as a metric for hydrolysis (Retallack 1999;Bestland 2000;Sayyed & Hundekari 2006).

Provenance and Parent Material 235
The molar ratios of titanium to aluminum (Ti/Al) and zirconium to aluminum (Zr/Al) was used to screen for consistency in sediment source in soils; direction of change in Ti/Al ratios is related to differences in chemical weathering, while Zr/Al ratios are related to changes in physical weathering (Sheldon & Tabor 2009). The molar ratios of uranium to thorium (U/Th), and lanthanum to cerium (La/Ce) were used to trace potential changes in parent material composition through the stratigraphic unit, where a constant down-profile U/Th and La/Ce ratios reflect single-parent source (Sheldon 2006;240 Sheldon & Tabor 2009). Absolute parent material values for each of these ratios are not well calibrated, but direction of change observed at any site indicates a change in parent material. The U/Th ratio is redox-sensitive, so La/Ce ratios are used as a comparative point in case of highly-reduced environments, resulting in skewed U/Th ratios.
The molar ratios of titanium to aluminum (Ti/Al) and zirconium to aluminum (Zr/Al) were used to screen for consistency in sediment source in soils; direction of change in Ti/Al ratios is related to differences in chemical weathering, 245 while Zr/Al ratios is related to changes in physical weathering (Sheldon & Tabor 2009). The molar ratios of uranium to thorium (U/Th), and lanthanum to cerium (La/Ce) were used to trace potential changes in parent material composition through the stratigraphic unit, where a constant down-profile U/Th and La/Ce ratios reflect single-parent source (Sheldon 2006;Sheldon & Tabor 2009). Absolute parent material values for each of these ratios are not well-calibrated, but direction of change observed at any site indicates a change in parent material, U/Th is redox-sensitive, so La/Ce ratios are used as a comparative 250 point in case of highly reduced environments.

Paleoclimate reconstructions using foliar assemblages
Before and up to 2017, co-author SAE surveyed and described 69 leaf morphotypes collected from multiple quarries at Blue Rim. As per Allen's (2017b) dissertation, two techniques were used to reconstruct precipitation from foliar 255 assemblages: a univariate approach, leaf area analysis (LAA; Wilf et al., 1998) and a multivariate approach, Climate Leaf Analysis Multivariate Program (CLAMP; Wolfe 1993). LAA is based on the correlation between mean leaf area and annual precipitation, related to transpiration. Leaves with higher surface area to volume ratios transpire more during gas exchange; these larger leaves are typically found in wet areas (Wilf et al., 1998). Leaves in drier climates have smaller leaf area to volume ratios, as they do not have as much plant available water accessible to transpire (Wilf et al., 1998). CLAMP uses 31 260 morphological characters on at least 20 species of woody "dicots" from any given site to reconstruct eleven aspects of climate, including mean annual precipitation, as well as mean annual temperature (MAT, comparable to mean annual air temperature -MAAT -discussed in this study ;Wolfe 1993;Spicer et al., 2009). This method is premised on the relationship between these morphological characters in modern flora and corresponding climate parameters.
Physiognomic techniques including CLAMP and leaf margin analysis (LMA; Wilf 1997) were used to calculate mean 265 annual air temperature. LMA uses the correlation between MAAT and the proportion of untoothed to total (untoothed + toothed) species in a local flora (Wolfe 1979;Wilf 1997;Wing & Greenwood 1993;Peppe et al., 2011). See Allen (2017b) for additional reconstruction details based on floral assemblages.

Paleoclimate and Paleoenvironmental Reconstructions using Organic and Inorganic Geochemistry 270
Mean annual precipitation was reconstructed using Chemical Index of Alteration minus potassium (CIA-K; Equation 5; 6; Sheldon et al., 2002; error ± 182 mm yr -1 ), modified from CIA to control for potassium metasomatism in paleosols (Maynard 1992;Ennis et al., 2000;Sheldon et al., 2002). Mean annual air temperature was calculated using PWI (Equation 7; error of ± 2.1 °C; Gallagher & Sheldon 2013). We applied the empirical relationship between δ 13 C plant and δ 13 C atm values found by Arens et al. (2000;Equation 8; R 2 = 0.34, p < 0.001) and used δ 13 C plant values of all individual fossils to reconstruct 275 generalized, non-taxon-specific δ 13 C atm values. We compared this reconstructed value based on a generalized equation with reconstructed values based on species-specific carbon isotope discrimination values (as measured in Cornwell et al., 2018), using fossil Lygodium and Populus to reconstruct δ 13 C atm values based on taxon-specific parameters (e.g., Stein et al., 2019;Stein et al., 2021).
Holdridge life zones are ecoregions classified by water availability and temperature, that can be further subdivided into successional stages reflecting land use, disturbance history, latitude, altitude (Holdridge 1967;Lugo et al., 1999). The parameters for each life zone are calculated based on potential evapotranspiration and humidity provinces (Holdridge 1967; see Appendix D). Similar metrics that use evapotranspiration and precipitation to quantify ecosystems into "floral humidity provinces" based on paleosol measurements, have been established more recently by Gulbranson et al. (2011;see Appendix 290 D). See supplemental materials for methodology used to determine Holdridge life zones and Floral Humidity Provinces for paleosols (this study) and previously published floras (Leopold & MacGinitie 1972;Roehler 1993;Wing et al., 2005;Smith et al., 2008;Wing & Currano 2013; Allen 2017a/b).

Geochronology
Single crystal sanidine 40 Ar/ 39 Ar analyses of four sampled beds yielded ages for the middle and top of the Blue Rim escarpment that are broadly consistent with deposition during the early Eocene ( Fig. 1; Table 1; Table S2). Two samples (BR-3 and BR-4) of a horizon containing pumice clasts and biotite grains taken from the base and middle of the 'blue-green marker' bed yielded similar coherent young populations of Eocene apparent ages (Figs. 4a; 8a), which are mixed with a 300 subsidiary population of older, presumably detrital or xenocrystic grains (Fig. 8b). Sample BR-3 was collected at the base of the main blue-green marker layer, just above the UF 15761S plant quarry (elevation 2053 m; Allen 2017b), whereas sample BR-4 was collected from the lower part of the blue-green layer in the 2014/UF 19297 stratigraphic section at 2056 m (Allen 2017b). Twenty seven of forty fusions of sanidine from these beds form a coherent population that yields a weighted mean age of 49.29 ± 0.18 Ma (MSWD = 1.08; Table 1), which we interpret to reflect the best estimate of the age of deposition. 305 This age is consistent with the Blue Rim being coeval with Bridger A/1b (Fig. 1).
Single fusions of sanidine from two sand beds (samples BR-5 and BR-6) collected near the top of the escarpment (elevation 2091 m; Allen 2017b) yielded a greater proportion (31 of 38) of older detrital and/or xenocrystic ages than occur in samples of the blue-green marker. Nevertheless, the seven youngest grains from the uppermost sample BR-6 form a coherent population that yields a weighted mean age of 48.48 ± 0.60 Ma (MSWD = 0.60), which can be interpreted to be a 310 maximum depositional age. This imprecise age suggests that the uppermost Blue Rim could be as old as Bridger B/B-2, or as young as Bridger D/Br-3 (Fig. 1). Altogether, new geochronology indicates that the stratigraphy between the blue-green marker bed and sand beds likely spans Bridger B, with the uppermost part of the Blue Rim Escarpment being time equivalent to Bridger C or D/3 (Fig4, 5, 6). This constrains the age of the lower plant horizon to be slightly older than 49.29 Ma (Allen 2017b), whereas the upper plant horizon is likely equivalent to Bridger Br-2. Detrital or xenocrystic grains not 315 included in weighted means discussed above yield a combination of Phanerozoic and Proterozoic apparent ages which are consistent with detrital feldspar grains sampled from volcaniclastic strata of the Sand Butte bed of the Laney Member (Green River Formation, Smith et al., 2008;Chetel et al., 2011). The Sand Butte bed forms thick (> 40 m) deltaic foresets that prograded into and filled Lake Gosiute and parts of Lake Uinta from northwest to southeast (Surdam & Stanley, 1980;Smith et al., 2008;. These deposits have been hypothesized to represent the inland delta or megafan of the 320 Idaho paleoriver which drained areas as far west as central Idaho (Chetel et al., 2011).

Paleosol descriptions
Six profiles were sampled laterally from a single paleosol at the base of the Blue Rim escarpment (located at 1 m in the stratigraphic column; Figs. 4-7). Paleosol profiles (Fig. 7, S1) typically consisted of a silty and/or sandy brown, yellow A-325 horizon over a parent material C-horizon of green-grey silty mudstone from ~20 to 112 cm below the surface. Paleosol #1 was missing a B-horizon due to erosion, while paleosol #4 was missing an A horizon, likely truncated during burial. Typically, each profile was lighter colored in upper horizons and darker in lower B-and C-horizons. In the paleosol profiles sampled, every A-horizon but one, and several upper B-horizons, had root traces, kerogenized roots, and/or rhizoliths. We observed drab-haloed roots in paleosols #1 and 4. Paleosol #2 had vertical burrows of up to 1 cm diameter and ~3 cm length, and paleosol 330 #1 had visible peds (Table S1; Fig. 3a). These soils were well-developed Inceptisols based on features, textures, and extrapolation from the local flora ( Fig. 2a; Fig. S1; S5; Table S1).  Table S2).

Paleosol geochemistry
Tau (used to measure mobile element transport) was calculated for soil profiles following Chadwick et al., (1990;340 Equations 2 and 3) and displayed in Supplemental Fig. S2(a-f). Overall, tau values for K, Mg and Na all ranged from 0.0 to -0.5, and tau values for Ca ranged from 0.0 to -1.0, except in paleosol #2 (which was extremely high in Ca), as is typical for Inceptisols. Tau values for Rb and Fe were generally also between 0.0 and -0.5, though this was less consistent between profiles. To note, paleosol #1 (19BRWY1; Fig. 7) was excluded for paleoclimate reconstructions due to the lack of presence of the B-horizon (we identified this soil as an Entisol, which cannot be used for climate reconstructions; see Fig. 7 for 345 location). Paleosol #2 (19BRWY2) was also excluded for climate reconstructions, due to the high % Ca, likely of carbonate origin as this site was reactive to HCl. Likewise, the CIA-K values for 19BRWY1 and 19BRWY2 B-horizon specimens were not reasonably different than the parent material, indicating they were not in equilibrium with the environment (e.g., Sheldon & Tabor 2009). Based on both field taxonomy and these geochemical results (see Supplemental Table S3), paleosols #3-#6 are identified as Inceptisols (Soil Survey Staff, 2014). 350

Sedimentary geochemistry
All Blue Rim escarpment geochemical data has been published at Mendeley Data Repository doi: 10.17632/z6twpstz4r.3. With a few individual outlier analyses, proxies for leaching intensity (Ba/Sr, and sum bases/Ti; Fig.   4b,c), hydrolysis (sum bases/Al; Fig. 4d), and measurements of weathering (CIA; Fig. 4e) are consistent throughout the 355 section. Proxies for provenance also remained nearly constant (Fig. 5a,b) as did proxies for parent material (Fig. 5c,d).
Organic carbon weight % was high in the same locations throughout the section as CIA, and % N was low throughout the section. δ 13 C org values were most depleted in the sections with highest % C and N.

Flora 360
The identifiable fossils from the 2019 field excursion sampled specifically for organic isotope analyses included multiple compression fossils of Lygodium kaulfussi (climbing fern, family Lygodiaceae; Manchester & Zavada 1987), as well as one specimen of Asplenium sp. (fern, family Aspleniaceae; as described in Allen 2017b), an example of cf. Populus cinnamomoides (poplar, family Salicaceae; Manchester et al., 2006), one specimen assigned to cf. Cedrela, (undefined species; mahogany, family Meliaceae; Fig. S5), several dense leaf mats, and assorted twig and branchlet fossils were 365 recovered. These specimens were collected from the same strata as the lower horizon (e.g., UF 15761N, Allen 2017b), located 26 m on the stratigraphic column included in this study . There were also fragments of several unidentified monocots preserved, though no isotope analyses were run on these fossils.

Climate 370
Mean annual precipitation (MAP) values reconstructed using CIA without potash (CIA-K) on paleosol B-horizons (Equation 6) ranged from 608-1167 mm yr -1 , with an average of 845 mm yr -1 (±181 mm yr -1 ; (n = 6) (Equation 6; Fig. 6; S6). The lowest estimated MAP value (288 mm yr -1 ) was excluded due to high % Ca (10.25%) in the B horizon of paleosol 2, skewing CIA-K calculations by artificially minimizing the ratio of Al to other metals in the calculation (see Dzombak et al., in review for discussion of this issue). Mean annual air temperature values (MAAT) reconstructed using PWI on paleosol 375 B-horizons (Equation 7; standard error of (± 2.1 °C)) ranged from 10.4 to 12.0 °C (± 0.7 °C standard deviation of all values from the same paleosol, or temperature reproducibility), with an average of 11.0 °C (n = 6 profiles). Temperature reproducibility from these paleosols falls within the standard error of the paleothermometer model.
A wide range of δ 13 C atm values were reconstructed from δ 13 C leaf from the 34 individual leaf fossils (2019 collection) using a generalized relationship (Arens et al., 2000). Reconstructions using the generalized Arens et al., (2000) model were 380 done on all 34 individual fossil leaves, even though many of these were unidentified. Additional species-specific tests were done on all samples of Lygodium, cf. Cedrela and Populus fossils using isotope discrimination values from extant plants of these genera. Thirty eight percent (n = 13) of the 34 δ 13 C atm values reconstructed using the generalized Arens et al., (2000) model suggested a δ 13 C atm value of between −5.32 and −5.82‰. Fifty six percent of these reconstructed values were between −5.0 and −6.0‰ (n = 19; Equation 8; Fig. 9). Limitations on this reconstruction are include species-specific isotopic C 385 isotope discrimination behaviour (Beerling & Royer 2002;Stein et al., 2019;Sheldon et al., 2020), and values outside the range of most samples (44%) were more extreme potentially preferential diagenesis of certain compounds (Beerling & Royer 2002;Tu et al., 2004). Using identified Lygodium, cf. Cedrela and Populus fossils (n = 8 total), we applied the taxon-specific isotope discrimination principle and reconstructed an average value of −4.40‰ (minimum value of −5.23‰ and maximum value of −3.83‰). These reconstructions were based on isotope discrimination values of 19.99‰ for Lygodium and 20.05‰ 390 for Populus (as reported in modern isotope analyses by Cornwell et al., 2018).

Geochronology
The new 40 Ar/ 39 Ar dates presented here for the blue-green marker bed and sand beds above the floral quarries suggest 395 that the Blue Rim section likely spans ca. 49.5 to 48.5 Ma, making it slightly younger than the EECO. Assuming constant sedimentation rates based on a linear interpolation between ages of 1 m per ~23 ka (or ~44 μm yr -1 , slightly slower than accumulation rates of 65 μm yr -1 +19/-12 in the Laney Member below (e.g., , the newly described paleosols are roughly 684,300 years older than the blue-green marker bed, or 49.97 Ma. The Wilkins Peak Member, and part of the Laney Member of the Green River Formation underlie the Blue Rim strata; thus, the approximate age for the paleosols based 400 on constant sedimentation rate is consistent with Blue Rim being younger than ~50 Ma, when the Wilkins Peak Member transitioned to the Laney Member (Smith et al., 2015;see Fig. 1).

Stratigraphy, provenance, and weathering
Detrital feldspar geochronology and geochemistry strongly suggest that provenance remained constant throughout 405 accumulation of the stratigraphic section, which can be interpreted to mean that the material did not systemically change across deposition, and geochemical proxies used to reconstruct climate are not affected by provenance shifts. A sudden 6 ‰ decrease in δ 18 O was previously observed in micritic lacustrine carbonates within sections of the Green River Formation, and it was hypothesized that this could be due to a sudden change in river capture to include a river with isotopically depleted waters from different headwaters (e.g., Norris et al. 1996;Norris et al., 2000;Doebbert et al., 2010). With new river 410 catchments, there is a potential change in sedimentological inputs that could overwrite climate signals, so these geochemical proxies provide assurance that the climate interpretations based on geochemical proxies are not actually changes in allochthonous materials. Assumptions about provenance were supported by parent material data, which showed that all sources were primarily sedimentary. The consistent Ti/Al, U/Th, and La/Ce ratios correspond to constant parent material throughout the one million years section covered by the Blue Rim stratigraphic column, demonstrating that basin-scale 415 hydrology was likely not reorganized during this time. The one exception to constant parent material and provenance ratios is the anomalously high U/Th ratio in the blue-green marker bed (0.87). This proxy is redox-sensitive, so this anomalously high U/Th ratio is due to the preferential redistribution and accumulation of U in this section, as Th is insoluble and immobile (Pett-Ridge et al., 2007;Sheldon & Tabor 2009), intuitive with the Blue-green marker color. Weathering and leaching were highest in the sections where there was high carbon content; this correlation could be due to organic acids 420 produced by plants in the ecosystem (as represented by % C present) that contribute to chemical weathering, or differences in taphonomic histories of samples, though more compound specific analyses would be needed to determine what drives this trend ( Fig. S4; R 2 = 0.20; p-value = 0.01; Ong et al., 1970;Berner 1992).

Global and regional climate 425
Generally, the early Eocene of North America had widespread wet forests comparable to modern temperate and subtropical forests due to global warmer, wetter conditions (Leopold & MacGinitie 1972;Wing & Greenwood 1993;Greenwood & Wing 1995;Inglis et al., 2017;Murphey et al., 2017). Depending on latitude of site, other studies indicate slightly to moderately warmer conditions temperature reconstructions from Blue Rim (Allen 2017a/b). Generally, mean annual air temperatures reconstructed from other mid to high-latitude sites between 36 and 80 °N have ranges from 35°C (36 °N) to 430 8°C (80 °N; using leaf margin analyses and oxygen isotopes; Fricke & Wing 2004).
The location of Blue Rim appears to have been a "wet forest" at the time of deposition, as calculated from temperature and precipitation estimates from paleosols and using leaf physiognomic techniques (see supplemental methods for details; Fig. 10 These collected fossils (sampled at 26 m on the stratigraphic section and housed at ESS laboratory at the University of Michigan) are comparable to fossils found by Allen (2017b;Bridger Formation) and MacGinitie (1969;Green River Formation) and are characteristic of mesic, forested environment (e.g., wet forest; Hamzeh & Dayanandan 2004;Hamzeh et al., 2006). 450 While the leaf physiognomic and paleosol-based estimates are modestly different, we cannot differentiate between whether there was a slight increase in both MAAT and MAP or whether climate was generally steady throughout the reconstructed portions of the sections, because calculated values were within error. One possible explanation for the slight discrepancy between leaf physiognomic and soil-based temperature reconstructions at Blue Rim could be related to the soil taxonomy; the PWI tool used to reconstruct temperature was calibrated for Inceptisols, Alfisols and Ultisols. However, none 455 of the Inceptisols sampled to calibrate this proxy were from mean annual air temperatures >12 °C (Gallagher & Sheldon 2013).
Therefore, it is possible that temperature reconstructions based on PWI for these Inceptisols are underestimates. Due to the complexities in the formation of soil related to seasonality of precipitation and temperature, seasonal bias has been found in paleosol-reconstructions based on carbonates (e.g., Kelson et al., 2020) but B-horizon bulk geochemical data is in equilibrium with the environment and takes so long to form, therefore is not resolved enough to be affected by seasonality. Seasonal bias 460 that would not be seen in leaf physiognomic techniques, either.
Leaf physiognomic proxies could contribute to the discrepancy as well; CLAMP has been cited as often producing overestimates of precipitation (Wilf et al., 1998;Allen 2017b). This is exacerbated when there are fewer than 25 morphotypes available, which was the case at Blue Rim (Wolfe 1993;Spicer et al., 2009;Allen 2017b). CLAMP estimates may be less accurate due to the threshold number of morphotypes used at Blue Rim escarpment (20); the number of 465 morphotypes used for CLAMP temperature reconstructions was exactly the minimum recommended value (Allen 2017b).
Regardless of the cause of discrepancy or if it represents modest actual change versus stasis, this study demonstrates the importance of the holistic approach that combines both types of proxies. However, based upon their close statistical agreement ( Figure 6) and implications for the overall ecosystem ( Figure 10), we interpret the overall climate as relatively steady (± 5 °C) on 100,000 year or more timescales during this interval. 470

Maintaining Warmth in the Paleogene
The maintenance of extended warmth and its relation to elevated CO 2 during the early Eocene is debated Anagnostou et al., 2016;Gutjahr et al., 2017;Cramwinckel et al., 2018), which emphasizes the importance of δ 13 C atm values that can help to contextualize potential CO 2 sources. Some scientists invoke the destabilization of deep-sea 475 methane hydrates as the mechanism for CO 2 increase (e.g., Dickens 2011), while others pinpoint volcanic emissions (Reagan et al., 2013;Gutjahr et al., 2017;Jones et al., 2019) and reduced silica weathering (Zachos et al., 2008;Lunt et al., 2011). These mechanisms occur on very different timescales (e.g., methane has a lifetime of 12 years in the atmosphere; Schiermeier 2020), and thus require vastly different environmental processes and landscapes. It is possible to determine changes in the source of atmospheric CO 2 by examining the isotopic signature of the atmosphere; volcanic CO 2 has an 480 isotopic composition of −5.4 ‰ (Deines 1992) while methane hydrates are far more depleted (e.g., −64.5 to −67.5 ‰ off the coast of central Oregon; Kastner et al., 1998). As mentioned above, based on the comparative Arens et al., (2000) model, δ 13 C leaf values can be used to infer δ 13 C atm values. These values are also important parameters in models that reconstruct additional environmental variables, such as the concentration of atmospheric CO 2 using paleosol carbonates (Cerling et al., 1991;Cerling 1992) or using stomatal parameters (Franks et al.. 2014). Our large sample size of distinguishable leaf fossils 485 from a single horizon (n = 34) allowed us to statistically determine the most likely δ 13 C atm value based on the general relationship proposed by Arens et al., (2000) to be between −5.32 and −5.82 ‰. Analyses with taxon-specific reconstructions (n =9 total specimens of these genera: Lygodium, cf. Cedrela and Populus fossils), using extant members of these genera to determine isotope discrimination and reconstruct the atmosphere (as published in Cornwell et al., 2018) had an average value of −4.82 ‰ (±0.92 ‰ standard deviation). Both generalized and taxon-specific reconstructions were within 490 error of the isotopic value of mantle CO 2 (−5.4 ‰;Deines 1992), and comparable to the value reconstructed for ~49 Ma using benthic and planktonic foraminifera, −5.4 ‰ (Tipple et al., 2010). Although we cannot rule out short term perturbations of other C pools (e.g., methane hydrates during the PETM, Foster et al., 2018; shift in carbon from the deep ocean due to changes in the carbon compensation depth; Zeebe et al., 2009;Pälike et al., 2012;, these results support previous findings indicating a long-term volcanic source of CO 2 that drove long-term warmth in the Paleogene. 495 However, this observation is incomplete without full constraints regarding the calcite compensation depth and the distribution of carbon isotopes in carbon ocean chemistry (e.g., Komar et al., 2013), both of which are not constrained in the scope of this study. As per other studies, the early Cenozoic period of elevated rates of volcanism (locally in the Rocky Mountains; e.g., Challis volcanics; Fig. 3a; Chetel et al., 2011; as well as globally with the North American Igneous Province; ~61 to ~50 Ma; e.g., Meyer et al., 2007;Storey et al., 2007) can account for this. 500

Conclusions
The age findings contained in this study constrain the time for the Blue Rim wet forests to be slightly younger than previous estimates, with the upper half of the section clearly deposited after the EECO. Based on dating and sedimentation rates, the lower part of the section could overlap with the end of the EECO, but there is an absence of observable changes in 505 proxy records. During the EECO, the Blue Rim escarpment received between 608-1167 mm yr -1 of precipitation, likely related to different moisture regimes, and was a productive paratropical (non-equatorial tropical) forest. Although reconstructed temperature and precipitation values using paleosol and sedimentary geochemistry are lower than published values reconstructed from flora, the values from all the proxies all fall within error of one another (see supplemental materials). Furthermore, the Holdridge life zones and floral humidity provinces calculated for both leaf physiognomic-based 510 and geochemistry-based reconstructions are comparable, pinpointing this region as a warm, wet forest at ~49 Ma.
The new age constraints, evidence from leaf fossils, and inorganic and organic geochemical proxies at Blue Rim escarpment make it possible to reconstruct the depositional environment of the central region of Lake Gosiute in an unprecedented way. The consistent Ti/Al, U/Th, and La/Ce ratios to determine provenance and parent material throughout the Blue Rim stratigraphic column demonstrate that the hydrological and sedimentological inputs remained stable for this 515 location throughout the ~one million years spanned by this section. The use of multiple proxies to cross-compare sites is under-utilized in paleoclimate reconstructions but allows for an improved understanding of regional and more broad-scale climate regimes. Based on ample floral and geochemical data, ~49.5-48.5 million years ago -after the peak of the EECOsouthwest Wyoming was a warm, wet forest atop deltaic deposition ( Fig. 3a; Chetel et al., 2011) with little to no frost and mild temperatures, in agreement with previous work based only on fossil floras (e.g., MacGinitie 1969; Wilf 2000) or 520 paleosols (Hyland et al., 2018) individually.

Author contributions
NDS and RAS conceived and designed the study. RMD and RAS collected the data in the field. MES, SEA and BRJ contributed data. NDS and MES were responsible for funding acquisition, and RAS acquired funding for field work. NDS was 525 responsible for project administration, provision of resources and supervision to RAS. RAS was primarily responsible for the investigation and performed many of the visualizations. RMD, MES, and SEA contributed visualizations as well. RAS and NDS were responsible for the original draft, and all authors (RMD, RAS, MES, SEA, BRJ) were responsible for review and editing.

Acknowledgments
We thank Nikolas C. Midttun for assistance measuring, creating, and sampling the stratigraphic column at the Blue Rim escarpment in June 2019. We thank Steven R. Manchester for personal communications regarding the location of floral fossil quarries at Blue Rim. We thank Selena Y. Smith for personal communications and consultation regarding fossils and for access to her camera and camera stand for fossil photographs. We acknowledge Naomi E. Levin, Christopher J. Poulsen and 535 Gretchen Keppel-Aleks for their feedback on this manuscript. This work was partially funded by NSF Award #1812949 to   using the ratio of the sum of bases to Titanium, (d) hydrolysis, calculated using the ratio of the sum of bases to Aluminum, (e) chemical index of alteration to measure weathering (Equation 1), (f) weight % Carbon (black circles) and % Nitrogen (white circles), (g) δ 13 C org values. Light green transparent areas show stratigraphic levels containing plant fossils. provenance using molar ratios of Zr/Al, (c) parent material using molar ratios of U/Th, (d) parent material using molar ratios of La/Ce. annual air temperature (°C). The boxes are ranges reconstructed for physiognomic proxies, and the circles are values reconstructed using paleosols. The lines for both are established errors for the proxies used, based on the initial proxy calibrations. Errors were calculated for each estimate, and the error bars span the total range of calculated error. More information on error can be found within these proxy calibrations (e.g., Wilf 2000;Sheldon et al., 2002;Spicer et al., 2009;Gallagher & Sheldon 2013)    Reconstructions utilized the generalized relationship between δ 13 C atm and δ 13 C plant (Arens et al. 2000;n = 34;Equation 8). The species-specific mean and standard deviation are shown (based on Lygodium and Populus fossils, n = 8), as are the generalized plant mean and standard 995 deviation. The Tipple et al. (2010) foraminiferal reconstruction is denoted in a star, and the δ 13 C value of the mantle is shown above (Deines 1992).  Smith et al., 2008). Comparative studies for this region based on nearby temperature and precipitation reconstructions are shown in yellow, red, and blue circles, while the climate of the present (as measured in Rock Springs, Wyoming, U.S.A.) is shown in 1005 light blue squares. Average values from Allen (2017b) using leaf margin analysis to reconstruct MAAT and leaf area analysis to reconstruct MAP is shown in red stars.