The relative roles of tectonics and global climate in forming the hydroclimate for widespread eolian deposition remain controversial. Oligocene loess has been previously documented in the interior of western United States, but its spatiotemporal pattern and causes remain undetermined. Through new stratigraphic record documentation and data compilation, we reveal the time transgressive occurrence of loess beginning in the latest Eocene in the central Rocky Mountains, that expands eastward to the Great Plains across the Eocene-Oligocene transition (EOT). Our climate simulations show that moderate uplift of the southern North America Cordillera initiated drying in the Cordilleran hinterland and immediate foreland, forming a potential dust source and sink, and global cooling at the EOT expanded the drying and eolian deposition eastward by causing retreat of the North American Monsoon. Therefore, the eolian deposition reflects continental aridification induced both by regional tectonism and global climate change during the late Paleogene.

The roles of global climate change and tectonics in driving eolian deposition and associated continental aridification remain debated. The best-known example is the Chinese Loess Plateau in central Asia. Loess deposition in the region initiated in the late Eocene or during the Eocene-Oligocene transition (EOT) (Licht et al., 2014; Sun and Windley, 2015), concurrent with global cooling, retreat of the Paratethys epicontinental sea, and uplift of the Tibetan Plateau, and these changes have been variously attributed as the causes of loess deposition (e.g., Ramstein et al., 1997; Guo et al., 2002; Dupont-Nivet et al., 2007; Wang et al., 2008). Oligocene eolian deposits have previously been documented in the western United States (Evanoff et al., 1992; LaGarry, 1998; Cather et al., 2008), presenting a new opportunity to test the roles of global climate and tectonism. Loess deposition also signals key changes in the regional ecosystem (e.g., Guo et al., 2002; Caves et al., 2016), and dust plays a feedback role on global climate by altering the radiation budget and nutrient cycling (e.g., Carslaw et al., 2010). Therefore, the timing of deposition, spatiotemporal pattern, and characteristics of the loess have profound influences on late Paleogene ecosystem evolution and the feedback of dust in global cooling.

Previous studies of eolian deposits in the western United States have placed the initiation of deposition in the early Oligocene in the central Rocky Mountains (Rockies) and at the EOT on the Colorado Plateau (Evanoff et al., 1992; Cather et al., 2008). Evanoff et al. (1992) suggested an eastward younging trend based on fossil ages and attributed this to global cooling. However, there is considerable evidence that the southern North American Cordillera experienced renewed uplift during the late Eocene–Oligocene (Mix et al., 2011; Chamberlain et al., 2012), post-dating the protracted Jurassic–early Paleogene crustal shortening and thickening. Mid-Cenozoic uplift may have forced development of a regional rain shadow and caused regional aridification.

Here, we report a new loess record (site WS) in the central Rockies, and new detailed sedimentologic characterization of three records (sites LTG, EF, LG) where Oligocene loess has been previously identified in the western United States (Fig. 1). Our study presents several sets of data that suggest loess deposition initiated earlier, and was more broadly distributed, than previously thought. We use a series of climate model simulations that test scenarios of global cooling, sea-level drop, and topographic change to investigate the roles of tectonics and global climate on eolian deposition.

We studied the late Eocene–Oligocene White River Formation and White River Group in the central Rockies and the adjacent Great Plains (Fig. 1). The Rockies are an extensive region of high mountains and intermontane basins on the east side of the North American Cordillera. The high-elevation, high-relief (1.5–4 km) topography gradually diminishes eastward toward the Great Plains, at ∼1.0 km elevation (Fig. 1). The region is arid to semi-arid, with annual precipitation amounts increasing eastward from ∼15 cm to ∼60 cm (Bryson and Hare, 1974). Precipitation across the Great Plains is from summer and fall moisture mainly originating from the Gulf of Mexico, from winter storms in the Arctic and the Pacific Ocean, and from recycled continental vapor (Zhu et al., 2018). These moisture contributions may have varied as a function of surface topography and climate state in the past (Feng et al., 2013).

We conducted analyses of lithofacies, grain-size, and rock magnetic properties to characterize past changes in depositional environment (see details in the GSA Data Repository1). At all four study sites, tan, massive eolian siltstone overlie deposits of interbedded gray to red mudstone and gray to brown lenticular sandstone and pebble conglomerate, recording fluvial environments (Figs. 1 and 2; Tables DR1 and DR2 in the Data Repository). Contacts between the younger massive strata and underlying fluvial rocks are conformable (Figs. 1B and 1C). The fluvial lithofacies has variable mean grain size and poor sorting (Fig. 2; Fig. DR1), with grain-size distributions varying between unimodal and multimodal (Fig. 3A), which reflect variations in water hydraulic properties, distance of sediment transport, and sediment mixing (Bui et al., 1989; Sun et al., 2002). Scanning electron microscope (SEM) observations show that fluvial lithofacies quartz grains are angular to subangular, and grain surfaces contain irregular breakage blocks and broken cleavage surfaces (Fig. 3B; Fig. DR2). These features are indicative of mechanical breakdown of grains in subaqueous conditions (Krinsley and Doornkamp, 1973).

The overlying massive deposits have better sorting than the fluvial lithofacies (Fig. 2), with bimodal grain-size distributions showing a major peak at 30–130 μm and a minor peak at 2–10 μm (Fig. 3A). The bimodal grain-size distributions are similar to those of the mid- and upper Cenozoic eolian loess deposits in central Asia (Sun et al., 2002; Pye and Tsoar, 2009; Vandenberghe, 2013; Licht et al., 2014; Sun and Windley, 2015) and Quaternary loess in the Midwest of the United States (Muhs et al., 2008). The major, coarser-grained population is interpreted to have been transported by saltation or near-surface suspension. The minor, finer-grained population represents the background supply of suspended particles, fine particles adhered to coarse grains during transport, or it formed by in situ weathering of the coarse grains (Sun et al., 2002; Pye and Tsoar, 2009; Vandenberghe, 2013). SEM inspection shows that quartz grains are subangular to subrounded (Fig. 3C), with surface textures that include dish-shaped depressions, flat cleavage faces and planes, adhesive clay particles, upturned plates, and smooth precipitation surfaces (Fig. 3D; Fig. DR2). These features are typical of eolian deposition, suggesting mechanical impact during powerful sand storms, and adhesion of small particles to coarse grains during eolian transport (Krinsley and Doornkamp, 1973; Pye and Tsoar, 2009).

Sediment provenance based on detrital zircon U-Pb geochronology data shows that the eolian lithofacies contain a cluster of late Eocene–earliest Oligocene (44–32 Ma) zircons derived from syndepositional ignimbrite eruptions, and abundant Archean to middle Eocene (3314–45 Ma) zircons recycled from Cambrian–lower Cenozoic sedimentary rocks in the western United States (Rowley and Fan, 2016) (Fig. 3E). To varying degrees, the rock magnetic parameters of magnetic susceptibility (MS), anhysteretic remanent magnetization (ARM) intensity, and saturation isothermal remanent magnetization (SIRM) intensity all show appreciable enhancement in the eolian sequence, as best demonstrated by results from the WS section (Fig. 2). The relatively low intensity of magnetic parameters for the fluvial facies likely reflects dissolution of ferrimagnetic grains in reducing water-logged conditions (Ding et al., 1999). In all four study sections, the ARM and MS intensity of the eolian sediment is enhanced to a greater degree than is the SIRM, and the increase in MS generally tracks the ARM increase. These data show that the loess features an increase in volume percent of finer-grained magnetite and/or maghemite (Yamazaki and Ioka, 1997), which likely reflects increasing clay skins and Fe-Mn films caused by longer pedogenic time due to a slower sedimentation rate or enhanced pedogenesis in the wetter climate (Ding et al., 1999).

After placing the initiation of eolian deposition at the four sites into three independent groups of age constraints, our records show eolian deposition initiated during the latest Eocene to early Oligocene and expanded from west to east (Fig. 2). Maximum depositional ages at the four sites, based on U-Pb ages of syndepositional volcanic zircons collected from the stratigraphic levels of eolian deposition initiation, are, from west to east, 36.0 ± 0.3 Ma, 35.3 ± 1.0 Ma, 33.0 ± 0.4 Ma, and 31.6 ± 0.5 Ma, respectively (Rowley and Fan, 2016). High-resolution zircon U-Pb ages of volcanic ashes have been published for three of the four study sites. In the WS section in western Wyoming, initiation occurred in the early Chadronian (North America Land Mammal Ages) based on the occurrence of an entelodont artiodactyl, Brachyhyops (van Houten, 1964). In the LTG section, the initiation occurred stratigraphically between ashes F (35.33 ± 0.02 Ma) and J (34.40 ± 0.02 Ma) (Sahy et al., 2015), and in the middle Chadronian stage based on the first occurrence of Leptomeryx mammifer (Emry, 1973; Sahy et al., 2015). In the EF section, the initiation occurred between ash 7 (32.9 ± 0.2 Ma) and ash 5 (34.0 ± 0.2 Ma) (Scott and Bowring, 2000), and in the late Chadronian–early Orellan stage based on the first occurrence of Leptomeryx evansi (Evanoff et al., 1992). In the LG section in Nebraska, the initiation is younger than 33 Ma based on the new age-depth model determined from zircon U-Pb ages of ashes LWA (31.78 ± 0.01 Ma) and SDP (33.41 ± 0.04 Ma) (Sahy et al., 2015), and in the Orellan stage based on the first occurrences of oreodonts Miniochoerus affinis and Miniochoerus gracilis, and rodent Eumys elegans (Zanazzi et al., 2009; Sahy et al., 2015). It is highly unlikely that this younging trend was caused by river erosion, because no major sandstone units with erosional bases are present immediately below the initiation in the LTG and EF sections (Emry, 1973; Evanoff et al., 1992). Although a multi-story sandstone channel complex incised ∼20 m into the middle part of the LG section, the incision occurred ∼1.5 m.y. before the eolian initiation (LaGarry, 1998).

We studied the independent and combined influences of changes in topography in the western United States and global cooling at the EOT on regional hydroclimate using the ECHAM5 atmospheric general circulation model (Roeckner et al., 2003) (Fig. 4). EOT glaciation caused an ∼55 m sea-level drop, in association with pCO2 drawdown (Miller et al., 2005), leading to southward shoreline regression in the Gulf of Mexico. Given that the history of middle Eocene–Oligocene topography of the North America Cordillera remains controversial, we consider two topographic scenarios (see the Data Repository for details). The first scenario assumes that the high Cordilleran hinterland continued into the middle Eocene (Fig. DR3), and subsequent crustal extension and erosion lowered the topography (Wernicke et al., 1987). The second scenario assumes that the southern Cordillera experienced rejuvenated surface uplift during the late Eocene and Oligocene (Mix et al., 2011; Chamberlain et al., 2012) (Fig. DR3). We prescribed EOT cooling, shoreline regression, and the filling of sedimentary basins in each scenario to distinguish hydroclimate change independent of topography (Figs. 4C and 4F).

We use the simulated precipitation minus evaporation (P–E) changes at a regional scale to determine the most likely scenario that can explain the cause of the eastward expansion of eolian deposition. The two scenarios lead to contrasting changes in regional atmospheric circulation and surface water budget (measured by P–E). Increasing elevation of the southern Cordillera reduces P–E in the eastern Cordilleran hinterland by blocking the Westerlies, and increases P–E across the western Great Plains and the central Rockies front by strengthening the proto-North American monsoon (Feng et al., 2013) (Fig. 4E). The monsoonal circulation is characterized by the northward intrusion of tropical moisture toward the central Rockies from the Gulf of Mexico. The intrusion was caused by the barrier of the high Cordillera that blocks the mid-latitude Westerlies during the warm season (May to September), and insulates the warm and moist tropical air masses with high moist enthalpy from the colder and drier mid-latitude air masses with low moist enthalpy (Fig. DR6). Regardless of topography, EOT cooling and shoreline regression decrease P-E in the central Rockies front range and the Great Plains (Figs. 4C and 4F), owing to the retreat and loss of buoyancy (reduction of moist enthalpy) of tropical air masses (Fig. DR6). In contrast, cordilleran elevation loss induces drying in both the central Rockies and the Great Plains, and wetting in the eastern hinterland (Fig. 4B). Although an increase in precipitation may promote eolian deposition by increasing river erosion and detritus available for wind transport (Nie et al., 2018), this effect can be compensated by the increase in soil water content, which limits the mobility of grains (Fig. DR7). More importantly, the concurrent, similar magnitude of drying in both the central Rockies and the Great Plains cannot explain the eastward expansion of eolian deposition.

Uplift of the southern Cordillera causes aridification [Δ(P–E) < 0] in the eastern Cordilleran hinterland and central Rockies and wetting [Δ(P–E) > 0] in the Great Plains. This pattern may explain the earlier initiation of eolian deposition near the Cordilleran front, with the dry Cordilleran hinterland and the Rockies being the likely sources of the sediment. This inference is supported by detrital zircon U-Pb ages, which indicate the eolian grains were recycled from Phanerozoic strata in the western United States (Rowley and Fan, 2016). EOT cooling and shoreline regression extended the drying eastward to the Great Plains, which may explain the later initiation of eolian deposition in this region (Figs. 4E and 4F).

The late Eocene loess deposits documented in this study are likely part of a regional loess plateau in the interior of the western United States. Late Paleogene loess has also been briefly described in northwest Colorado and the southwest Dakotas (Benton et al., 2015; Hembree and Hasiotis, 2007), and is a possible component of the lower Oligocene Renova Formation in southwest Montana (Robinson, 1963). If our hypothesis is valid, late Paleogene eolian deposition in both central Asia and west-central North America suggests continental-scale development of dust provinces coincidentally driven by both tectonism and global climate change. In addition, because the initiation of aridification and dust availability on both continents occurred during the late Eocene cooling that led to the greenhouse-icehouse shift at the EOT, increases in aridity and dust availability may have contributed to enhancing planetary albedo through the replacement of dark forest canopy with reflective semi-arid to arid landscapes, thereby increasing suspending dust aerosols in the atmosphere, and an enhanced ocean biological pump (Carslaw et al., 2010; Caves et al., 2016). The feedback role of eolain dusts in tectonic-climate interaction may have contributed to the transition to an icehouse climate during the late Eocene.

We thank B.H. Hough, S. Allen, S. Ayyash, and E. Evanoff for assistance in the field; C. Godfray, M. Gao, W. Hoffman, J. Jackson, D. Kirkwood, and I. Pujana for facilitating analyses; and Guillaume Dupont-Nivet, Steve Cather, and several anonymous reviewers for their constructive comments. This research was funded by National Science Foundation grants EAR-1119005, 1454802 and 0949191.

1GSA Data Repository item 2020075, details of sample analysis and model simulation setup, is available online at, or on request from
Gold Open Access: This paper is published under the terms of the CC-BY license.