The Phanerozoic's longest-lived and most widespread glaciation, the late Paleozoic ice age, is undergoing a resurgence in interest. Long-held models of the timing, duration, and magnitude of glaciation are being reevaluated due to emerging evidence from former high latitudes, evidence that the late Paleozoic ice age was punctuated by long-lived glacial minima or possibly ice-free times.

The history of the late Paleozoic ice age is archived within the biostratigraphically well-constrained, carbonate-dominated succession of Arrow Canyon, Nevada, United States. In this paleo-tropical succession, the distribution of lithofacies, flooding surfaces, and subaerial exposure horizons and their stacking into meter-scale cycles record a detailed climate history. The onset of this phase of glaciation during the middle Mississippian was followed by a dynamic evolution of glacioeustasy through the late Mississippian to late Pennsylvanian. Moderate- to high-amplitude glacioeustasy was likely interrupted by an earliest Pennsylvanian short-lived glacial minimum, but otherwise appears to have persisted through the middle Pennsylvanian.

Upper Pennsylvanian strata record low- to moderate-amplitude relative sea-level changes, suggesting a long-lived interval of diminished ice volume. This proposed glacial minimum is coincident with a notable minimum in glaciogenic sedimentation near the former southern pole, aridification across paleo-tropical Pangea, and significant floral and faunal turnover, suggesting a link between tropical environmental change and high-latitude glaciation. These conclusions, however, are at odds with those traditionally inferred from Euramerican cyclothems, i.e., persistent high-amplitude glacioeustasy driven by a single, large ice sheet. Rather, the Arrow Canyon archive of varying depositional facies and cycle stacking patterns records major changes in the magnitude of short-term glacioeustasy. This finding contributes to recent and growing near- and far-field evidence for a more dynamic glaciation history than previously inferred from the classic Euramerican cyclothems.

Much of our understanding of the late Paleozoic ice age (LPIA) has been built on stratigraphic records from the low-latitude tropics, typified by Euramerican cyclothems. These far-field records preserve high-frequency sequences of open-marine, paralic, and terrestrial environments. Cyclothems have long been considered the most sensitive proxy of high-frequency (105 yr) glacioeustasy, driven by the waxing and waning of expansive ice sheets in Southern Hemisphere Gondwana (e.g., Wanless and Weller, 1932; Wanless and Shepard, 1936; Heckel, 1977). The persistence of cyclothems in upper Mississippian through lower Permian successions has been argued as evidence for repeated short-term (20–400 k.y.), high-amplitude (30 to >150 m) sea-level changes, and in turn the persistence of large-scale continental glaciation throughout an ∼50 m.y period of the Carboniferous–Permian (Heckel, 1977, 1986, 1994, 2002; Veevers and Powell, 1987; Horbury, 1989; Frakes et al., 1992; Soreghan and Giles, 1999a; Smith and Read, 2000; Cook et al., 2002; Zempolich et al., 2002; Heckel et al., 2007).

This paradigm has been challenged by emerging near-field records. In particular, recent compilations of paleo–high-latitude Gondwanan records document discreet intervals of widespread glaciogenic sedimentation, punctuated by periods of normal marine or fluviodeltaic sedimentation, or long-lived pedogenesis (e.g., Isbell et al., 2003a, 2003b, 2008a, 2008b; Fielding et al., 2008a, 2008b, 2008c, and references therein). These records argue for intervals of glacial minima during the late Paleozoic ice age, with geographically restricted ice centers or possibly ice-free conditions near the South Pole for periods of several million years. This conclusion calls into question how fluctuating continental ice sheets could have sustained large (>50 m) sea-level changes throughout the late Paleozoic. Moreover, recent studies of some far-field cyclothem records also suggest varying characteristics of cyclothems, with intervals of potential lower amplitude glacioeustatic forcing (e.g., West et al., 1997; Smith and Read, 2000; Wright and Vanstone, 2001; Gibling and Giles, 2005; Heckel, 2008). Accordingly, climate simulations indicate that the response of ice sheets to orbital forcing was quite sensitive to variations in pCO2 and overall ice sheet size, leading to a range of possible glacioeustatic magnitudes (Horton et al., 2007). Thus, changes in the frequency and amplitude of Carboniferous–Permian glacioeustasy should be expected during the course of the late Paleozoic ice age if the extent and volume of stable continental ice sheets varied. This hypothesis, however, has been minimally tested in the low-latitude paleo-tropical basins.

This study documents lithologic, pedogenic, and early diagenetic facies and their stratigraphic stacking into meter-scale cycles in the Carboniferous succession of Arrow Canyon, Nevada, United States. This carbonate-dominated succession has nearly complete exposure, relatively high subsidence rates, and a very well constrained biostratigraphy; in addition, carbonates tend to fill accommodation space and are thus highly sensitive to climate change and siliciclastic input. Our approach is to use the changes in facies and cycle types to reconstruct ∼30 m.y. of Carboniferous relative sea-level and climate history on the paleo-tropical western margin of Euramerica. Though we show that the long-term accommodation record is dominated by subsidence variations, the short-term high-amplitude relative sea-level record is inferred to reflect the repeated waxing and waning of Gondwanan ice sheets. A tectonic explanation for such moderate- to high-amplitude relative sea-level changes is ruled out by their character, longevity, bounding surfaces, and the record of coeval glaciogenic sediments in high latitudes. Thus, the presence and magnitude of such short-term changes are taken as proxy for minimum possible ice volume (cf. Read, 1995; Smith and Read, 2000; Wright and Vanstone, 2001). In particular, in this paper we note the absence of such moderate- to high-amplitude relative sea-level changes during intervals traditionally ascribed to peak glacioeustasy (e.g., late Pennsylvanian), an absence that cannot be explained by tectonic drivers. Because the lack of high-amplitude glacioeustasy might be due to small ice sheets or the presence of a very large, stable ice sheet (DeConto and Pollard, 2003; Horton et al., 2007), we look to the high-latitude record of glaciogenic sedimentation to distinguish large, stable ice sheets from small, feckless ones.

Different climate modes are suggested for the Arrow Canyon succession by changes in patterns of meter-scale cyclicity and facies distributions. These changes delineate intervals of distinct short-term relative sea-level fluctuations and regional climate. These intervals are ultimately linked to changing glaciation by ties between ice volume and eustasy (e.g., Read, 1995; Smith and Read, 2000), and between continental ice sheets and low-latitude precipitation (Cecil et al., 2003; Poulsen et al., 2007). The Arrow Canyon succession records the onset of glacioeustasy in basal Chesterian time, a short-lived glacial minimum in earliest Morrowan time, and a long-lived, significantly drier glacial minimum during later Desmoinesian through early Virgilian time.

The Arrow Canyon Range is in southeastern Nevada, in the eastern Great Basin province. During the middle to late Paleozoic, Arrow Canyon was situated in the tropics on the west coast of the North American craton, near the seaward margin of an interior seaway that at times extended from California to Alaska (Poole and Sandberg, 1991; Ross, 1991) (Fig. 1). Southeastern Nevada was in the foreland basin inherited from the Devonian–early Mississippian Antler orogeny (Dickinson, 2006). During middle Mississippian to early Permian time, the former foreland evolved as a series of basins and uplifts. Tectonic controls on this structural landscape are not well constrained and may relate to distal effects of the Ouachita-Marathon orogeny and thus be linked to the Ancestral Rocky Mountains (Kluth and Coney, 1981; Kluth, 1986; Dickinson, 2006). Alternatively, they may relate to a continuing post-Antler tectonic evolution of the western margin of North America (Trexler et al., 1991, 2003, 2004; Stevens and Stone, 2007).

Platform geometries evolved during the course of Carboniferous sedimentation. We use “platform” in the sense of Read (1985) to encompass both ramps and shelves. Middle to late Mississippian Arrow Canyon strata accumulated on a carbonate or mixed carbonate-siliciclastic ramp, with correlative basinal facies to the west in western Nevada and Death Valley (Stevens et al., 1991; Poole and Sandberg, 1991; Trexler et al., 1996). During Pennsylvanian time, the study area was on the northeast Bird Spring platform of the Keeler Basin (also known as Bird Spring Basin; Kluth, 1986) that initiated as a ramp during the early Pennsylvanian (Morrowan; Stevens et al., 1991, 2001; Stevens and Stone, 2007) but evolved into an attached shelf by late Pennsylvanian time. During Morrowan, Atokan, and possibly early Desmoinesian time, the Bird Spring platform was a distally steepened ramp, indicated by rare sediment gravity flows in the basin (Yose and Heller, 1989; Miller and Heller; 1994), unrestricted storm-dominated strata in Arrow Canyon, and the absence of evidence for any margin (e.g., biohermal buildups or grainstone shoal complexes at a break in slope). By contrast, in upper Desmoinesian through Wolfcampian time, the Bird Spring platform prograded westward and evolved into a (locally) rimmed shelf, as evidenced by abundant turbidites, debris flows, and megabreccias in basinal settings (Yose and Heller, 1989; Miller and Heller; 1994), restricted platform interior facies in Arrow Canyon, and phyloid algal bioherms with bypass channels, which mark the shelf margin in the Nevada Test Site and Death Valley (Miller and Heller, 1994; Stevens et al., 2001).

Nearly continuous, high subsidence in the Arrow Canyon region promoted a thick, relatively stratigraphically complete Carboniferous succession. Long-term (second- to third-order) accommodation in this setting was driven by regional tectonics; however, this study examines the high-frequency stratigraphic record to extract climatic signatures during the Carboniferous interval. Most of the succeeding Permian and Mesozoic strata were removed by a regional unconformity caused by the Cretaceous Sevier orogeny (Page and Dixon, 1997). During Neogene time, basin and range extension uplifted the Arrow Canyon Range in a north-south–trending block, and canyons cutting through the range now expose a nearly complete Famennian to Wolfcampian carbonate-dominated succession.

A composite section was measured on both sides of Arrow Canyon, encompassing Osagean through lower Virgilian (Visean through Gzhelian Russian Platform stages) strata of the Yellowpine, Battleship Wash, Indian Springs, and Bird Spring Formations. Upper Virgilian and Wolfcampian strata are heavily recrystallized, obscuring many depositional and early diagenetic features; these strata are not discussed here. Above the base of the Indian Springs Formation, the section was tied to a standard canyon section, measured by Amoco Oil Company as part of an extensive biostratigraphic research program in Arrow Canyon, and along which brass tags and metal spikes were affixed every 1.5 m (stratigraphic locations in this paper are labeled with an “A” prefix, corresponding to this standard measured section). To refine lithofacies interpretations, identify faunal assemblages, and describe early diagenetic features, 663 samples were collected, thin-sectioned, stained with Alzarin red and potassium ferricyanide (Dickson, 1965), and examined under transmitted and cathodoluminescent light.

Biostratigraphic control is provided by conodonts, calcareous algae, foraminifera, and fusulinids ( Appendix A). The thick, relatively complete Arrow Canyon succession includes the Global Stratotype Section and Point for the Mississippian-Pennsylvanian boundary (Brenckle et al., 1997; Lane et al., 1999; Richards et al., 2002; Ellwood et al., 2007; Barnett and Wright, 2008; Bishop et al., 2009), and has been a focus of biostratigraphic research for nearly 50 yr (Cassity and Langenheim, 1966; Webster, 1969; Pierce and Langenheim, 1972; Rice and Langenheim, 1974a, 1974b; Langenheim et al., 1984; Baesemann and Lane, 1985; Brenckle, 1997; Stamm and Wardlaw, 2003). Chronostratigraphy for this study is based primarily on extensive Amoco biostratigraphic data, collected over several decades, and subsequently donated to the Universities of Iowa (conodonts) and Kansas (foraminifera) (Groves and Miller, 2000). This biostratigraphic control begins below the Osagean Yellowpine Formation and continues through the Chesterian to lower Wolfcampian Bird Spring Formation. Many of the range charts and original thin sections have since been reinterpreted (Leavitt, 2002; Vladimir Davydov, 2007, personal commun.; Greg Wahlman, 2006–2007, personal communs.; Bishop et al., 2009) and are used here.  Appendix A presents the preferred biostratigraphic zonation tied to meterage, A numbers, and the bed numbers of Cassity and Langenheim (1966) where appropriate.

Carboniferous strata in the study area were deposited on a westward-facing platform that we divide into broad depositional settings: outer, middle, and inner ramp (Mississippian to middle Pennsylvanian) and shelf interior (upper Pennsylvanian). Lithofacies descriptions for each are provided in Tables 1–4,234. In the following we highlight salient features of each setting and offer an environmental interpretation. Lithofacies are defined and interpreted using their Dunham textures, sedimentary structures, and photozoan versus heterozoan grain types. Heterozoan fossil assemblages consist primarily of echinoderms, bryozoa, brachiopods, foraminifera, solitary corals, calcareous red algae, sponge spicules, and trilobites. Photozoan assemblages consist primarily of coated grains, peloidal lime mud, fully micritized grains, stromatolites, thrombolites, fusulinids, colonial corals, and dasyclad and other green algae, as well as most grains common to the heterozoan assemblage (e.g., James, 1997).

Outer-ramp lithofacies consist of mud- and calcisilt-rich heterozoan carbonates that were deposited largely below the influence of storm waves. Middle-ramp lithofacies consist of heterozoan and photozoan wacke-packstones that were deposited from storm wave base (SWB) to fair weather wave base (FWWB). Inner-ramp facies consist of pack-grainstone shoals and the variably restricted photozoan wacke-packstones, laminites, and restricted siliciclastic facies deposited in their lee. By late Pennsylvanian time, the Bird Spring platform had evolved into a shelf, and shelf interior facies were deposited in Arrow Canyon. These facies consist of more open photozoan pack-grainstones and more restricted, variably dolomitized wacke-packstones, evaporative dolomudstones, and laminites. They reflect deposition in shelf interiors that were episodically restricted, highly evaporative, and aggraded to sea level.

Outer-Ramp Facies

Outer-ramp lithofacies (Table 1) consist of fine-grained carbonates common in uppermost Chesterian to Desmoinesian (Serpukhovian to Moscovian) strata, and siliciclastic mudstones common in upper Chesterian (Serpukhovian) strata. Outer-ramp facies include marls (OR1), wavy interbedded cherts and calcisiltites (OR2), and massive to laminated calcisiltites (OR3) (Fig. 2). These fine-grained carbonates have a calcisiltite or mud matrix, common spicule molds, and where identifiable a heterozoan assemblage of fossils.

Outer-ramp lithofacies were deposited below SWB, delineated by lime mud- and silt-sized particles, the absence of photozoan grain types, and an abundance of planar lamination. SWB in this interior seaway is estimated as between 30 and 60 m, based on modern analogs (e.g., Logan et al., 1969, 1970; Purser and Evans, 1973). An abundance of chert and spicule molds in these facies suggests that they formed in nutrient-rich waters, possibly associated with upwelling from the Keeler Basin (see following).

Outer-ramp siliciclastic lithofacies (Table 1) consist of millimeter- to centimeter-scale, laminated, quartz silt- and clay-rich mudstones, with rare marine fossils (IR4). Siliciclastic-rich deposits occur primarily in upper Chesterian strata and likely record regional siliciclastic influx derived from the Antler highlands to the west or the craton to the east (Poole and Sandberg, 1991; Trexler et al., 2004). The absence of cross-stratification suggests that these facies were deposited in quiet water, below FWWB and likely below SWB, possibly on a distal delta front (Bishop et al., 2009).

Middle-Ramp Facies

Middle-ramp lithofacies (Table 2) are divided into gently cross-stratified calcisiltites (MR5), heterozoan wacke-packstones (MR6), and photozoan wacke-packstones (MR7). These facies are abundant in Morrowan to Desmoinesian strata. Cross-stratified calcisiltites (MR5) have gentle truncation surfaces that locally approach hummocky cross-stratification. Heterozoan wacke-packstones (MR6) have a heterozoan assemblage of fossils and a calcisiltite and/or muddy matrix. Photozoan wacke-packstones (MR7) have photozoan grain types, a matrix composed of locally peloidal lime mud, and lack the evaporites and early dolomitization common to inner-ramp and shelf interior photozoan wacke-packstones.

Gently cross-stratified calcisiltites are interpreted to reflect deposition at or near SWB. Wacke-packstones are interpreted to have been deposited below FWWB (7–20 m, Logan et al., 1969, 1970; Purser and Evans, 1973; Gischler and Lomando, 2005), where mud was not winnowed by waves or currents, and above SWB (30–60 m; Logan et al., 1969; Purser and Seibold, 1973), where laminae of sand-sized grains were deposited but subsequently largely obscured by bioturbation. Photozoan lithofacies are also constrained by the depth of the photic zone, estimated to be ∼30 m in this muddy interior seaway (cf. Purser and Seibold, 1973; Foster et al., 1997). All of these sediments were deposited in unrestricted waters, seaward of skeletal banks and coated grain shoals, as suggested by their normal marine flora and fauna, general lack of early dolomitization, absence of evaporite pseudomorphs, and the progressive shoaling of these muddy carbonates into grainy ones within generally regressive meter-scale packages (see following cycle discussion).

Inner-Ramp Facies

Inner-ramp facies consist of higher-energy facies (pack-grainstones, thrombolites, metazoan framestones, cross-bedded sandstones) deposited above FWWB, as well as the lagoonal restricted facies deposited in their lee (lagoonal wacke-packstones, tidal laminites, laminated siltstones). They are common throughout Osagean to Desmoinesian strata.

Higher-energy facies are subdivided into heterozoan, photozoan, and siliciclastic facies (Table 3). Heterozoan pack-grainstones (IR9) have a heterozoan assemblage of fossils, locally a calcisilt or muddy matrix, and are common in Osagean to Desmoinesian strata. High-energy photozoan facies consist of skeletal, oolitic and/or oncolitic and/or peloidal pack-grainstones (IR10), thrombolitic boundstones (IR10b), and metazoan framestones (IR10c). Metazoan framestones are typically surrounded by grainstones or less commonly wackestones, implying baffling of wave energy. Thrombolites are cut by channels filled with grainstones, suggesting energetic environments, likely proximal to sea level (e.g., Osleger and Read, 1991). High-energy inner-ramp siliciclastic lithofacies consist of planar tabular to trough cross-stratified sandstones (IR8) that form thin-bedded to massive units, up to 5 m thick, in the upper Chesterian.

Low-energy inner-ramp facies consist of lagoonal wacke-packstones, cryptmicrobial to thick-laminites (IR13), and various siliciclastic facies (IR15). Lagoonal wacke-packstones rarely have a heterozoan fauna (IR11) and more commonly a photozoan (IR12) one, which can be partially dolomitized (IR12b).

Peritidal facies consist of millimeter-scale cryptmicrobial laminites and millimeter- to centimeter-scale thick-laminites (IR13). Cryptmicrobial laminites consist of even to crinkly millimeter-scale laminae of variably peloidal dolomudstone, with sparse mudcracks and laterally linked hemispherical stromatolites. Thick-laminites consist of millimeter- to centimeter-scale graded laminations of peloids and lime mud with sparse mudcracks. Both cryptmicrobial and centimeter-scale thick-laminites are locally cut by decimeter-scale channels with coarse, intraclast-rich fill (cut-and-fill structures).

Inner-ramp siliciclastic lithofacies consist of irregular beds and lenses of variably calcareous, quartz sand and silt (IR15) that commonly occur above flooding surfaces and exposure horizons. The thickest, in Atokan strata (A220), is ∼4 m thick, consists of planar to low-angle cross-laminated calcareous siltstone, and occurs between a caliche and an outer-platform sulfidic marl. An earliest Morrowan siliciclastic sandstone also occurs as an ∼1-m-thick massive siltstone to very fine sandstone with root haloes developed in its upper surface (A57).

High-energy inner-ramp facies are interpreted as shoals, skeletal banks, patch reefs, and more rarely, grainy foreshore deposits. Photozoan and heterozoan pack-grainstones were deposited near and above FWWB (∼7–20 m; Logan et al., 1969, 1970; Purser and Evans, 1973; Gischler and Lomando, 2005), as indicated by a common lack of carbonate mud and, in many photozoan facies, coated grains. These grainstone shoals, framestones, and boundstones typically formed breaks behind which low-energy inner-ramp facies accumulated. Heterozoan and photozoan wacke- packstones are interpreted as lagoonal deposits based on a diverse flora and fauna and their position between high-energy facies and peritidal facies in generally regressive meter-scale cycles (see following). Thick and cryptmicrobial laminites (F14) were deposited on lower and upper intertidal flats, respectively (e.g., Hardie and Shinn, 1986; Elrick and Read, 1991).

Siliciclastic lenses (IR15) are common at cycle bases and are interpreted as lowstand deposits reworked during transgressions in the platform interior. Their occurrence above subtidal flooding surfaces as well as exposure horizons, coupled with their sedimentary structures, suggests a variety of transport mechanisms (eolian, fluvial, shallow marine). Local pin-stripe lamination with subtle truncation in the thick lower Atokan example (A220) suggests the local preservation of primary eolian sedimentary structures. The concentration of siliciclastic sediments at cycle bases suggests, for lowstands and early transgressive phases, either a proximal source only available episodically, systematic changes in (eolian?) transport direction or strength, or less dilution due to a diminished carbonate factory. The early Morrowan massive siltstone to very fine sandstone (A57) is interpreted as an eolian deposit due to its uniform grain size and root haloes developed in its upper surface.

Shelf Interior Facies

In mid-Desmoinesian through early Virgilian time, shelf interior facies were deposited in Arrow Canyon (Table 4). Shelf interior lithofacies consisted of variably coated photozoan pack-grainstones, thrombolites, and metazoan framestones (S16), undolomitized (S17) to dolomitized (S17b) photozoan wacke-packstones, dolomudstones (S19), cryptmicrobial to thick-laminites (S18), and various siliciclastic facies (S21, S22). These facies are generally similar to those of the older inner ramp, described above. Notable distinctions are generally more abundant synsedimentary dolomitization, especially muddy sediments, and evaporite pseudomorphs, particularly in massive (up to 15 m thick) outcrop exposures of homogeneous, peloidal, very poorly fossiliferous dolomudstone (S19), with common to abundant cauliflower- and popcorn-chert (Fig. 3B). Replacement dolomites consist of fine to very fine crystalline (15–60 μm), subhedral to euhedral turbid crystals with planar boundaries. Cauliflower- or popcorn-cherts are centimeter-scale nodules that consist of chert, quartz, blocky calcite, dolomite, and felted fabrics of silicified former anhydrite as well as encapsulated dolomite rhombs that predate compaction (Fig. 3B). Swallowtail pseudomorphs of blocky calcite after gypsum are also rarely present.

Siliciclastic facies occur as thin irregular beds and lenses, which occur at flooding surfaces (S21). In addition, in Desmoinesian strata (A380), a 12-m-thick calcareous sandstone occurs (the Tungsten Gap chert); it has local planar lamination and rare phosphatic brachiopods, and is used as a marker bed across the Arrow Canyon Range.

Grain-dominated shelf-interior lithofacies (S16) were deposited in shallow, high-energy settings on the shelf, where tidal or wave energy winnowed mud. Muddy shelf interior lithofacies (S17, S17b, S18, S19) were deposited in shallow, generally restricted waters, as suggested by a low-diversity, restricted fauna, large fusulinids, thin-walled brachiopods, peloidal muddy matrix, common evaporite pseudomorphs, and the succession of dolomitized muddy facies above high-diversity grainy limestones within regressive meter-scale cycles (see following). Thick packages of homogeneous dolomudstone (S19), lacking bedding surfaces and with common to abundant cauliflower chert (some with encapsulated aphanocrystalline to fine crystalline dolomite rhombs) indicate long-lived concurrent dolomitization and displacive anhydrite growth under unchanging depositional conditions. We interpret this facies to have been deposited in highly restricted shallow lagoons under an arid climate regime, where postdepositional dolomitization was driven by seepage from overlying lagoonal waters or reflux from adjacent sabkah environments (e.g., Sears and Lucia, 1980; Kendall and Skipwith, 1969; Montañez and Read, 1992a, 1992b; Montañez and Osleger, 1993, 1996; Lehmann et al., 1998, 2000).

Silty carbonate lenses (S21) preserved at cycle bases are interpreted as lowstand deposits reworked during ensuing transgressions. The thick Desmoinesian calcareous sandstone (Tungsten Gap Chert) likely formed in estuarine environments, on the basis of rare phosphatic brachiopods, the paucity of unabraded normal marine fossils, local flat lamination, and the absence of indicators for subaerial exposure.

Significant Surfaces

Significant surfaces are used to delineate meter-scale cycles. Five types of surfaces are observed in the Arrow Canyon succession, including flooding surfaces (1) and four types of subaerial exposure horizons: paleo-karst surfaces (2), meteorically altered surfaces (3), pedogenically altered surfaces (4), and deflation surfaces (5). Subaerial exposure horizons are common (n = 52) and are associated with extensive moldic porosity in the underlying limestones (though this feature is not considered unequivocally diagnostic of exposure; Melim et al., 2002). Subaerial exposure horizons are abundant in Chesterian and Morrowan (Serpukhovian and early Bashkirian) strata, common in Atokan and mid-Desmoinesian (upper Bashkirian and Moscovian) strata, and rare in upper Desmoinesian to lower Virgilian (Kasimovian and Gzhelian) strata. Flooding surfaces are common throughout Meramecian to Virgilian (upper Visean to Gzhelian) strata.

Flooding Surfaces

Flooding surfaces are recognized as abrupt contacts where more shoreward facies are retrogradationally overlain by more basinward facies. These surfaces can be sharp, erosive, or, more rarely, gradational. Facies above flooding surfaces are commonly enriched in insoluble material and locally have phosphatized grains. Flooding surfaces represent transgressions of the shoreline and mark cycle tops in the absence of subaerial exposure.

Karsted Horizons

Microkarst and macrokarst horizons are developed on subtidal limestones in lower Chesterian and mid-Desmoinesian strata. Microkarst surfaces are recognized as scalloped, pitted bedding planes with less than a few centimeters relief, whereas macrokarst is associated with centimeter- to decimeter-scale potholes, fissures, and dissolution pipes (Fig. 4). Karst surfaces are commonly filled by sandy palimpsest carbonate, terra rosa, sand- to cobble-sized lithorelicts, and/or millimeter- to meter-scale rhizoliths (e.g., Bishop et al., 2009). Roots, pebbles, and skeletal grains are commonly chertified and/or blackened with Fe and Mn oxides. Beneath karst surfaces, limestones have significant moldic porosity, which is commonly filled by Fe-poor (pink staining), nonluminescent meteoric cements. At a very well developed lower Chesterian karst horizon, several episodes of meteoric cementation and reworking occur; these have been interpreted to reflect missed sea-level beats and the landward merging of unconformities (Bishop et al., 2009).

Horizons of Meteoric Cementation

In Chesterian to Virgilian strata, meteoric cements fill primary and secondary porosity and are commonly associated with extensive moldic porosity; many of these surfaces do not exhibit karst or pedogenic features. Calcite cements are all nonferroan (stain pink with Dickson's solution) and nonluminescent to dully luminescent. They occur as micritic meniscus cements in primary pores (Fig. 4D), as very fine to fine crystalline equant calcites (crystal silt) that geopetally fill primary and secondary pores (commonly within rhizoliths; Fig. 4C), and as circumgranular isopachous cements with scalenohedral terminations and/or syntaxial overgrowths on echinoderm ossicles.

Meniscus cements and geopetal crystal silt are interpreted as vadose cements that formed during subaerial exposure (Dunham, 1969, 1971). Individual generations of vadose cements are not observed in successive cycles, suggesting that these cements are related directly to overlying cycle bounding surfaces. Nonferroan, nonluminescent circumgranular cements and syntaxial overgrowths on echinoderm ossicles are interpreted as phreatic cements that precipitated from oxidized meteoric waters (Bishop et al., 2009). Unless phreatic meteoric cements appear reworked, they are not considered diagnostic of subaerial exposure (i.e., they might have formed during subsequent unconformities). Vadose meteoric cements are considered diagnostic of subaerial exposure and can be the only indicator of exposure, or can be associated with karsted, rooted, and pedogenically altered horizons.

Deflation Surfaces

Four truncation surfaces are developed on highly restricted shelf interior dolomudstones in Missourian and Virgilian strata. These sharp surfaces exhibit as much as 20 cm of relief, place less restricted photozoan grainstone and/or wacke-packstones over highly restricted dolomudstones, and are not associated with secondary porosity, meteoric cements, or karst features.

Truncation surfaces are interpreted as possible deflation surfaces that formed during sea-level falls when highly restricted, dolomitized, evaporite-rich sediments were exposed and deflated by winds, as noted on modern (Shinn, 1983; Hardie and Shinn, 1986) and ancient (Montañez and Read, 1992a, 1992b) arid tidal flats. The absence of an overlying lag, however, suggests that these horizons could also have formed by subaqueous erosion.

Pedogenically Altered Surfaces

Pedogenic features are abundant in upper Chesterian strata, common in Morrowan strata, and rare in Atokan, Desmoinesian, and Missourian strata. Pedogenic alteration occurs on both carbonate and siliciclastic substrates and ranges from rooted horizons indicative of ephemeral plant colonization to mature paleosol profiles.

Rooted horizons are typically developed on limestones and contain millimeter- to decimeter-scale roots that bifurcate downward. Roots are preserved as redoximorphic root haloes, carbonate and Fe-Mn oxide rhizoliths, and cherty compressions or adpressions. Root walls are commonly lined by carbonate and/or clay, and root casts are locally filled by meteoric cements (Fig. 4B). Rooted horizons commonly also contain grains and clasts blackened with Fe and Mn oxides.

Paleosols are distinguished and classified according to Mack et al. (1993) on the basis of their soil (ped) structure (fine granular to angular blocky, platy, rhomboid), horizonation, slickensides, extent of development of redoximorphic features, and evidence for translocation and accumulation of Fe and Mn oxides, phyllosilicates, carbonates, and organic matter. Chesterian paleosols in the Indian Springs and lowermost Bird Spring Formations include Argillisols, Calcisols, Vertisols, and Protosols (locally gleyed, ochric, or plinthitic), and were described at length in Bishop et al. (2009) (see also Richards et al., 2002; Barnett and Wright, 2008). Paleosol development in the overlying Pennsylvanian interval of the Arrow Canyon succession is limited to Calcisols (n = 3) and Protosols (n = 3).

Three paleosols occur in Atokan and Missourian strata and are characterized by CaCO3 accumulation, occurring as irregular 1–20-cm-thick locally blackened crusts that veneer bedding planes. These crusts effervesce in H2O2, indicating a concentration of Fe and Mn oxides. They contain teardrop-shaped pisoids and laminations of silty carbonate (Fig. 4E). Carbonate rhizoliths are common and typically have geopetal fills of crystal silt. One of these paleosols crops out in lower Atokan strata as a white micritic bed, with local brown discoloration. In thin section, this horizon contains glaebules, defined by abundant circumgranular cracks and rhizoliths (Fig. 4F).

These paleosols are interpreted as Calcisols because their most prominent feature is the accumulation of pedogenic CaCO3. Blackened crusts record vadose meteoric processes and are distinguished from modern outcrop patina by the teardrop-shaped pisoids, which are oriented perpendicular to bedding (rather than modern-day “down”). Although the accumulation of pedogenic carbonate on limestone parent material must be interpreted climatically with caution, these Calcisols likely reflect a subhumid to semiarid climate regime (Royer, 1999; Gong et al., 2005).

One interval of upper Morrowan strata contains three stacked, weakly developed brown to gray horizons with redoximorphic root haloes and platy ped structure; they pass downward into olive-brown wackestones with marine fossils. These are classified as Protosols due to weakly developed soil features. No climate interpretation can be made on the basis of these immature soil profiles.

Facies are distributed stratigraphically into meter-scale progradational-retrogradational packages, bounded by disconformities or flooding surfaces. These packages are 15 cm to 19.5 m thick, averaging 3.5 m; they are referred to as cycles because each represents one cycle of retrogradation and progradation, despite not always exhibiting a predictable succession of facies (cf. Wilkinson et al., 1997, 1998). These cycles are classified according to their bounding surface and uppermost facies (i.e., disconformities developed on peritidal sediments, disconformities developed on subtidal sediments, and conformable flooding surfaces developed on subtidal sediments), and are further characterized by the dominant position of constituent facies on the platform (i.e., inner, middle, and outer ramp, or shelf interior). Representative facies belts and cycles for different intervals of time are shown in Figures 5–7,67. Figures 8–15,9101112131415 show representative windows of Arrow Canyon strata with outcrop photos and stratigraphic logs, and illustrate how facies and significant surfaces stack to form meter-scale cycles. (For a complete Arrow Canyon log, see Bishop, 2008.)

Peritidal Cycles

Peritidal cycles consist of subtidal lithofacies capped by intertidal facies with sharp to erosive bounding surfaces. Peritidal cycles are common in Meramecian and early Morrowan inner-ramp settings, and in late Desmoinesian to Virgilian shelf interior settings (Figs. 5–7,67, 9, and 15).

Peritidal cycles in Meramecian and lower Morrowan rocks were deposited in inner-ramp settings and consist of heterozoan (IR9) and photozoan (IR10) pack-grainstones that commonly pass upward into lagoonal wacke-packstones (IR11, IR12, IR12b) and are capped by tidal laminites (IR13) (Figs. 2D and 9). Generally, middle- to inner-platform peritidal cycles reflect shallowing from near FWWB to sea level.

In late Pennsylvanian strata, peritidal cycles are preserved in shelf interior settings and are capped by tidal laminites (S18), with cycle bases of lagoonal photozoan wacke-packstones (S17, S17b) that increase in restriction upward (increasing mud, dolomite, and evaporite pseudomorphs, and decreasing faunal diversity and abundance) (Figs. 7C and 15). Rarely, cycle bases consist of photozoan pack-grainstones (S16) that pass upward into restricted lagoonal facies (S17b) and tidal laminites (S18). These cycles reflect increasing restriction due to aggradation and/or progradation and the filling of accommodation, from above FWWB to sea level.

Disconformable Subtidal Cycles

Disconformable subtidal cycles contain exclusively subtidal facies and are bounded on their upper surface by a subaerial exposure horizon. These cycles occur throughout the Arrow Canyon succession, but are particularly abundant in Chesterian and Morrowan strata (e.g., Figs. 5–7,67).

In deposits that formed in outer-ramp settings, disconformable subtidal cycles contain marls (OR1) that pass upward into wavy interbedded chert and calcisiltite (OR2), massive to laminated calcisiltite (OR3), and cross-stratified calcisiltite (MR5), and are capped by subaerial exposure horizons. Exposure surfaces are commonly overlain by photozoan or heterozoan pack-grainstone transgressive deposits (IR9, IR10) (Fig. 6). Individual disconformable subtidal cycles commonly lack one or more of the lithofacies in these ideal cycles (Fig. 11). In upper Chesterian strata of the Indian Springs Formation, exposure-capped cycles consist of outer-ramp laminated siliciclastic mudstones (OR4) and inner-ramp heterozoan pack-grainstones (IR9); rooted horizons and paleosols may be developed on either of the constituent facies (Fig. 8). In Morrowan strata, very thin (<0.5 m) cycles consist of marls (OR1) overlain by heterozoan pack-grainstones (IR9) with rooting structures (Fig. 10). Outer-ramp exposure-capped cycles indicate shallowing from below SWB to above sea level (i.e., exposure).

In middle-ramp settings, disconformable subtidal cycles commonly contain either heterozoan (MR6) or photozoan (MR7) wacke-packstones that pass upward into pack-grainstones (IR9, IR10) and are capped by exposure horizons (Figs. 5 and 6). Disconformable subtidal cycles in middle-ramp settings reflect shallowing from below FWWB to above sea level.

In upper Pennsylvanian shelf interior settings, disconformable subtidal cycles are bounded by (possible) deflation surfaces, and in one instance a caliche. In these cycles, more open-marine photozoan pack-grainstones (S16) are overlain by muddy lagoonal facies (S17, S17b, S19) with stratigraphically upward increasing dolomite and evaporite pseudomorphs, and decreasing faunal diversity and abundance (Fig. 7). These facies transitions reflect increasing restriction across the shelf, culminating in truncation surfaces, which may represent deflation. In shelf interior settings, exposure-capped cycles reflect shallowing from above FWWB to above sea level.

Conformable Subtidal Cycles

Conformable subtidal cycles are bounded by flooding surfaces and are common throughout Meramecian to Virgilian strata in Arrow Canyon. Typically these cycles are composed of the same internal lithofacies as disconformable subtidal or peritidal cycles, but they lack subaerial exposure horizons and tidal flat facies.

In outer-ramp settings, conformable subtidal cycles contain marls (OR1) overlain by wavy interbedded chert and calcisiltite (OR2), massive to laminated calcisiltite (OR3), and cross-stratified calcisiltite (MR5). Cycle tops commonly contain a pack-grainstone deposit (IR9 or IR10; similar to the transgressive deposit in disconformable subtidal cycles) that likely formed during the sea-level lowstand or early transgression (Fig. 11). Such cycles reflect shallowing from below SWB to near FWWB. A few lower Desmoinesian cycles deepen upward and consist of basal photozoan grainstones (IR10) that pass upward into outer-ramp marls (OR1) and spiculitic calcisiltite (OR3), and are bounded at their tops by hardgrounds (i.e., omission surfaces) (A288–A296; Fig. 13). These are analogous to a combination of the give-up and catch-down cycles of Soreghan and Dickinson (1994), in that they reflect the temporary drowning of the ramp before relative sea-level fall brings the seafloor back within the zone of maximum production, resuscitating the carbonate factory.

In middle- to inner-ramp settings, subtidal cycles contain heterozoan (MR6) or photozoan (MR7) wack-packstones that coarsen upward into pack-grainstones (IR9, IR10; Figs. 10 and 14). These cycles also locally contain thin outer-ramp facies at their cycle bases (Fig. 12). In middle-ramp settings, conformable subtidal cycles reflect changes in relative sea level from near SWB to near to above FWWB.

In shelf interior settings, conformable subtidal cycles have photozoan lime pack-grainstones (S16) at their bases, passing upward into increasingly restricted, dolomitized lagoonal photozoan wack-packstones (S17, S17b, S19; Fig. 15). These cycles are bounded by flooding surfaces and contain facies that increase in their degree of restriction upward, yet did not shallow completely to intertidal depths.

Noncyclic Interval

Osagean (lower Visean) strata in the Yellowpine Formation contain a noncyclic interval consisting of ∼60 m of massive, poorly bedded echinoderm pack-grainstones (IR9). These carbonates are consistent with deposition adjacent to stable echinoderm banks, which were common during Mississippian time (Kammer and Ausich, 2006). This interval contains no facies changes or exposure horizons, suggesting continuous deposition between SWB and FWWB (Bishop et al., 2009).

Long-term Accommodation History

Long-term sedimentary accumulation rates for Arrow Canyon demonstrate that local variations in accommodation from middle Mississippian to end-Pennsylvanian time overwhelmed the global eustatic signal in this depositional setting. However, this locally controlled long-term accommodation trend provides a context in which to assess the nature of short-term relative sea-level fluctuations. Long-term accumulation rates were calculated based on rock thickness per stage, constrained by the integrated biostratigraphic zonation ( Appendix A) and calibrated to the Gradstein et al. (2004) time scale. Most cycles shoal above SWB, so changes in this long-term accumulation plot primarily reflect changes in the creation of accommodation space, rather than how efficiently it was filled (Fig. 16C).

Figure 16C presents two accumulation plots for Arrow Canyon, determined using the North American and Russian Platform (global) stages, respectively. The differences in accumulation rates derive from the fact that, although in Arrow Canyon the North American stages are well constrained biostratigraphically, these stage boundaries are relatively poorly dated, relying on biostratigraphic correlations to better-dated Eurasian sections (e.g., Gradstein et al., 2004). In contrast, though global stages are more precisely dated, they are only loosely defined in Arrow Canyon, and North America in general. Thus, both age models have considerable uncertainty. However, despite the two plots differing in detail, they clearly document a similar long-term history of variations in accommodation space in the study area during the Carboniferous.

Estimated long-term sediment accumulation rates (Fig. 16C) indicate that accommodation was twofold to tenfold higher during the Morrowan through Missourian (20–70 m/m.y.) than during the Chesterian (7 m/m.y.) or Virgilian (12.6 m/m.y.), with greatest accommodation during the Atokan (68 m/m.y.) and Desmoinesian (59 m/m.y.). This long-term accommodation history likely reflects the combined effects of changes in tectonically driven subsidence in the Keeler Basin and second- and third-order (>1 m.y.) eustatic fluctuations. Notably, the Atokan maximum in accommodation occurs during a eustatic second-order regression and/or lowstand (Ross and Ross, 1987; Galonka and Kiessling, 2002). This suggests that long-term eustatic changes in this region were largely superseded by local subsidence. It is this local increase in subsidence that allows for the preservation of mostly complete high-amplitude, high-frequency cycles during the global sea-level lowstand. Similarly, during the late Pennsylvanian, diminished subsidence sensitizes the Arrow Canyon record to even moderate amplitude high-frequency changes in relative sea level, which should cause exposure of the shallow subtidal facies. Thus, the Arrow Canyon section provides a window of opportunity for capturing the full range of amplitudes of relative sea-level change when the creation of accommodation space is high (early to mid-Pennsylvanian), and constrains the maximum that sea level could have changed when accommodation creation was lowered (late Pennsylvanian), given that at such times accumulating carbonates would be very sensitive to any high-amplitude relative sea-level changes (i.e., tidal flats would be stranded landward and syndepositional dolomite would be limited; Read et al., 1986; Montañez and Read, 1992a).

Relative Sea-Level History

The history of short-term relative sea-level fluctuations is reconstructed using water depths inferred from the lithofacies in each cycle and the distribution of subaerial exposure horizons (Fig. 16). Ranges of water depths for each facies are calibrated to inferred depths for FWWB, SWB, and the photic zone based on sedimentary structures, grain types, and using modern and ancient facies analogs (Tables 1–4,234, and references therein). Comparisons between the deepest and shallowest facies in each cycle provide an estimated range of relative sea-level change within that cycle. These depth ranges are highly sensitive to inferred SWB and FWWB, which can vary significantly depending on geographic setting. For this Carboniferous shallow interior seaway, SWB is inferred to be 40 m and FWWB to be ∼20 m, although FWWB can be considerably shallower (∼8 m in the Persian Gulf; Purser and Evans, 1973; Gischler and Lomando, 2005) and SWB considerably deeper (∼60 m in the Yucatan; Logan et al., 1969) on modern carbonate platforms. These values provide a relatively conservative estimate of sea-level changes. Amplitudes of sea-level change estimated beyond these wave-base thresholds are less well constrained; the maximum water depth for deep subtidal sediments is arbitrarily defined as 100 m but could have been greater. Overall, the magnitudes of sea-level falls may be underestimated where cycles are exposure capped and/or contain facies that formed below SWB. However, for conformable subtidal or peritidal cycles with facies deposited entirely above SWB, these estimates capture the maximum relative sea-level change, given the depths inferred for each facies.

Estimates of short-term relative sea-level change, in concert with variations in the types of cycle bounding surfaces, are indicative of different climate states and glacioeustatic forcing (Fig. 16G). In general, during times of low-amplitude glacioeustasy, basinward settings develop amalgamated conformable subtidal cycles that show minimal facies changes or may be noncyclic. In shoreward settings, carbonates usually fill accommodation space, creating flat-topped platforms with peritidal cycles that prograde extensively across the platform (Read et al., 1986; Read, 1995). Moderate- to high-amplitude relative sea-level changes should be recorded by (1) the juxtaposition of facies from markedly different water depths in the same cycle, and/or (2) subaerial exposure horizons developed on subtidal sediments. Basinward settings may record more conformable subtidal cycles with juxtaposed outer- and middle- to inner-platform facies, or disconformable subtidal cycles. In contrast, shoreward settings should record cycles capped by subaerial exposure horizons developed on inner-platform facies due to overall lowered accommodation coupled with forced regressions.

The record of relative sea-level change and cycle bounding surfaces is depicted in Figure 16. Four features are prominent and indicate shifts between intervals of high-amplitude and low-amplitude short-term relative sea-level fluctuations, including (1) a Chesterian (late Visean) onset of relative sea-level fluctuations of moderate to high amplitude; (2) a short-lived (earliest Morrowan or Bashkirian) dampening of relative sea-level changes and the development of peritidal cycle caps; (3) a return to high-amplitude, high-frequency relative sea-level changes in the mid-Morrowan (early Bashkirian) to mid-Desmoinesian (late Moscovian); and (4) a marked decrease in amplitude of late Desmoinesian to early Virgilian (Kasimovian to early Gzelian) relative sea-level fluctuations and a concomitant return to dominantly peritidal cycles.

Temporal Distribution of Climatic Signatures

Osagean and Meramecian (Visean) age strata contain ∼65 m of noncyclic middle- to inner-ramp encrinites and ∼15 m of photozoan inner-ramp facies. Low-amplitude relative sea-level changes are inferred because these strata contain only middle- to inner-ramp facies, with a long-lived noncyclic interval and conformable subtidal and peritidal cycles, lacking evidence for exposure of subtidal sediments. Meramecian strata likely exhibit autocyclic behavior because peritidal flat facies do not correlate regionally between Arrow Canyon and neighboring Battleship Wash (Bishop et al., 2009). Fenestral tidal flat facies suggest a subhumid climate for this time (e.g., Bova and Read, 1987; Hardie and Shinn, 1986).

The onset of moderate- to high-amplitude relative sea-level fluctuations occurs 1 m above the Meramecian-Chesterian (late Visean) boundary (Bishop et al., 2009), where subaerial exposure horizons are developed on subtidal encrinites and (stratigraphically higher) on middle- to outer-platform carbonates and siliciclastics (Figs. 5 and 8). In upper Chesterian (Serpukhovian) cycles, exposure horizons developed on middle- to outer-ramp siliciclastic mudstones record changes in relative water depth at least equivalent to the depth of FWWB and likely SWB (>20–40 m; Fig. 16E). Notably, Chesterian Vertisols and calcic paleosols suggest that a seasonal dry subhumid to semiarid climate existed minimally by this time (Bishop et al., 2009).

Earliest Morrowan (earliest Bashkirian) strata exhibit evidence for a short-lived minimum in magnitudes of relative sea-level change (Figs. 16E, 16G). The lower ∼20 m of Morrowan strata consist of mixed heterozoan-photozoan inner-ramp facies capped by thin tidal flat laminites; they contain only one thin (50 cm) incursion of outer-ramp facies (Fig. 9). The presence of tidal flats and rarity of outer-ramp facies suggest low-amplitude relative sea-level oscillations, during which the carbonate factory kept pace with transgressions, excluding outer-ramp facies, and tidal flat progradation kept pace with short-term regressions (Read et al., 1986; Read, 1995, 1998).

Mid-Morrowan to mid-Desmoinesian (early Bashkirian to Moscovian) strata record significantly higher-amplitude short-term sea-level changes (Figs. 16E, 16G). High-amplitude relative sea-level fluctuations are demonstrated where exposure horizons are developed on these dominantly outer-ramp cycles (Figs. 10–13,111213). In addition, rapid sea-level changes are required by the short cycle durations in this interval. A tenfold decrease in average cycle duration occurs during the Atokan maximum in accommodation (Fig. 16C). In part, this decline in cycle duration reflects the ability of deeper water facies to record most high-amplitude sea-level changes, and thus minimize missed beats. However, some of the change in cycle duration likely also reflects changes in sea-level forcing (e.g., from the 100 k.y. eccentricity to the 40 k.y. obliquity band). Atokan strata include two caliches (Figs. 4F and 11) and a thick, reworked siliciclastic eolianite (at approximately A220), suggesting continuation of an overall dry subhumid to semiarid climate regime.

Relative sea-level changes diminished in amplitude during the late Pennsylvanian (Fig. 16). During a mid-Desmoinesian (upper Moscovian) transitional interval, amplitudes of relative water-depth change decrease markedly, yet middle- to inner-ramp cycles are bounded by subaerial exposure horizons (Fig. 14). These exposure-capped cycles suggest forced regressions under at least moderate-amplitude relative sea-level fluctuations. In addition, this interval likely corresponds to the development of a (rimmed) shelf to the west. This transition is recorded by the loss of storm-dominated outer- to middle-ramp facies, increasingly restricted, evaporative facies, common metazoan bioherms, and the slightly younger development of Missourian phyloid algal bioherms in western Nevada and Death Valley (Miller and Heller, 1994).

Upper Desmoinesian, Missourian, and lower Virgilian (uppermost Moscovian to lower Gzhelian) strata record a long-lived interval of low- to possibly moderate-amplitude relative sea-level fluctuations. These uniformly shallow-water, photozoan facies accumulated under approximately fourfold diminished accommodation. The low accommodation and shallow-water deposition require that any moderate or high-amplitude relative sea-level fluctuations would generate exposure-capped cycles. However, exposure horizons are rare—between 1 and 5 of 32 cycles, depending on whether truncation surfaces are interpreted as deflation surfaces. This suggests a long-lived interval of low-amplitude high-frequency sea-level changes, punctuated by rare moderate-amplitude sea-level changes. Given the rarity of exposure horizons and uniformly shallow, subtidal to peritidal facies, the range of relative sea-level change indicated in Figure 16 thus represents a maximum. The reappearance of common tidal flat-capped cycles further supports generally low-amplitude fluctuations, because even moderate sea-level falls would have caused shorelines to rapidly regress across the low-relief platform interior, stranding any tidal flats that might have formed during sea-level stillstands (e.g., Read et al., 1986; Read, 1995, 1998). The presence of cycles with thick (15 m) accumulations of highly restricted lagoonal dolomudstones at other times during this interval suggests the occurrence of sea-level changes of low to moderate amplitude; large enough to generate (possible) deflation surfaces in some cycles and to prevent tidal flat complexes from prograding across the highly restricted, shallow lagoons, but low enough to allow long-lived accumulation of these restricted facies without triggering facies changes. In addition, a Missourian calcisol, abundant concurrent dolomitization and evaporite emplacement (Fig. 3B), and a paucity of meteorically altered surfaces attest to a marked aridification during this time.

Upper Virgilian and Wolfcampian strata are heavily recrystallized, masking depositional and early diagenetic textures, and obscuring the relative sea-level record for this interval; therefore, environmental, sea-level, and climatic conditions are not interpreted from these strata.

Tectonic Control of Meter-Scale Cycles

It has been suggested that “jerky” subsidence or “yo-yo” tectonics might produce meter-scale peritidal cycles in proximity to faults (e.g., Cisne, 1986; Hardie et al., 1991). In particular, the forward stratigraphic models of De Bennedictis et al. (2007) suggest that in extensional settings, peritidal cycles might be produced in the hanging wall or graben by episodic downdrop (causing rapid deepening) followed by aggradation due to carbonate productivity. And disconformable subtidal cycles might be produced by episodic uplift of the footwall or horst (causing rapid shallowing and subaerial exposure horizons) followed by slow background subsidence, allowing the return of carbonate production. De Bennedictis et al. (2007) documented rates, frequencies, and magnitudes of slip that are consistent with many meter-scale carbonate cycles.

The Arrow Canyon record of high-amplitude relative sea-level change (disconformable subtidal cycles) would require high-frequency normal faulting with large slips (>40–50 m) that persisted over a very long period of mid-Mississippian to mid-Pennsylvanian time (∼25 m.y.). However, this interval brackets several different tectonic regimes on the west coast of North America (Trexler et al., 1991, 2003, 2004; Stevens and Stone, 2007). In addition, such fault-controlled meter-scale cycles require depressed carbonate productivity to prevent “keep-up” peritidal conditions (during the interval of slow background subsidence), which would preclude deep subtidal facies. Otherwise, the resulting disconformable cycles would consist entirely of peritidal carbonate facies. Most important, Arrow Canyon disconformable subtidal cycles require both abrupt sea-level rises (commonly >40 m at the base of each cycle) and abrupt sea-level falls (because cycle thickness is much less than maximum recorded water depth), culminating in exposure of the platform. Such a relative sea-level history requires both forced transgressions and regressions, inconsistent with the structural model presented here. Moreover, the absence of seismites and growth faults and the relatively large (∼200 km) distance from Arrow Canyon to the major Keeler basin fault system (Stevens and Stone, 2007) do not support a structural origin for these cycles. Rather, abundant evidence for coeval glaciation in Gondwanan records (see following discussion) implicates glacioeustatic forcing for these medium- to high-amplitude relative sea-level changes.

In contrast, intervals of low-amplitude relative sea-level change (dominantly peritidal and conformable subtidal cycles) are consistent with structural, autocyclic, or eustatic origins (e.g., Hardie et al., 1991; Burgess, 2001; De Bennedictis et al., 2007; Bosence et al., 2009). Regardless of the predominant cause, however, such intervals cannot have been characterized by high amplitudes of glacioeustasy (see following).

Implications for Reconstructing the Late Paleozoic Record of Ice Volume

The paleo-tropical mixed carbonate-siliciclastic succession at Arrow Canyon provides substantial insights into Carboniferous low-latitude climate dynamics (Fig. 17). In other low-latitude settings, persistent high-amplitude (30 to >150 m) glacioeustasy throughout middle Mississippian to early Permian time has been extrapolated from the relief on unconformities and juxtaposed facies in cyclothems, and the geochemical records of their biotic components (e.g., Heckel, 1977, 1986, 1994; Adlis et al., 1988; Horbury, 1989; Soreghan and Giles, 1999a; Smith and Read, 2000; Wright and Vanstone, 2001; Cook et al., 2002; Joachimski et al., 2006). In contrast, high-latitude sedimentary records suggest multiple, smaller ice sheets that waxed and waned to varying degrees at different times, leading to alternating long-lived intervals of glacial maxima and minima (Isbell et al., 2003a, 2003b, 2008a, 2008b; Montañez et al., 2007; Fielding et al., 2008a, 2008b; Caputo et al., 2008). The low-latitude Arrow Canyon succession preserves a record of shifting modes of glacioeustasy that suggests a dynamic ice-volume history, consistent with the variable glaciation inferred from high latitudes. This conclusion demands a more nuanced view than the prevailing low-latitude model of persistent high-amplitude glacioeustasy caused by the protracted waxing and waning of a long-lived, geographically expansive late Paleozoic ice sheet.

In Arrow Canyon, basal Chesterian strata record the onset of high-amplitude relative sea-level fluctuations and inferred glacioeustasy (Bishop et al., 2009). This is slightly later than the onset of glaciogenic sedimentation in South America (Caputo et al., 2008), but predates evidence of glacial sedimentation elsewhere in Gondwana (Isbell et al., 2003a; Fielding et al., 2008a, 2008b). Tropical estimates for the onset of glacioeustasy range from late Meramecian (mid-Visean) to late Chesterian (late Serpukhovian) time but tend to converge near the base of the Chesterian (upper Visean) (Bishop et al., 2009). Variability in these records is likely due to differing subsidence regimes and positions on platforms, causing different sensitivities to eustatic sea-level changes, as well as possible limitations in biostratigraphic constraints and intercontinental correlation. Soon after the onset of glacioeustasy, repeated fluvial incision in the midcontinent suggests high-amplitude (>30–85 m) relative sea-level fluctuations (Smith and Read, 2000). In Arrow Canyon, a potentially long-lived disconformity occurs immediately above the Meramecian-Chesterian boundary (Bishop et al., 2009); this disconformity is interpreted to mask several of the high-amplitude glacioeustatic events inferred from the midcontinent (Bishop et al., 2009). Thus, the onset of glacioeustasy inferred from the Arrow Canyon succession is consistent with previously published near- and far-field records. This high-amplitude glacioeustasy continues until just above the Mississippian-Pennsylvanian boundary.

In lowermost Morrowan (lowermost Bashkirian) strata, common tidal flat cycle caps and a paucity of intercalated outer-ramp facies suggest a minor short-lived glacial minimum, or a large but stable ice sheet (DeConto and Pollard, 2003). This time period corresponds with an interval of normal fluvial and/or lacustrine sedimentation across eastern Australia (Fielding et al., 2008a, 2008b), which was near the polar circle at the time, more consistent with a time of limited glacial extent. Such short-lived minima have been attributed to fourth-order sea-level cycles superimposed on third-order sea-level highstands (Read, 1995), consistent with diminished glacioeustatic forcing during times of smaller ice sheets.

The mid-Morrowan to mid-Desmoinesian interval of inferred high-amplitude glacioeustasy is consistent with both low-latitude and high-latitude records, which suggest this as a time of peak Carboniferous ice (e.g., Veevers and Powell, 1987; Frakes et al., 1992; Isbell et al., 2003a; Fielding et al., 2008a). In addition, the abundance of cherty heterozoan limestones with molds after sponge spicules, horizons with scattered phosphatized grains, and a subhumid to semiarid climate (see following) suggests upwelling of cool, nutrient-rich waters onto the Bird Spring platform during this interval (cf. Pope and Steffen, 2003).

Relative sea-level changes in upper Desmoinesian through lower Virgilian strata suggest an interval of significantly diminished magnitudes of relative sea-level fluctuations. It is during this interval, one traditionally attributed to peak icehouse conditions, that the greatest disparity exists between many near- and far-field records. In eastern Australia, late Pennsylvanian strata contain none of the glaciogenic sediments or evidence of high-amplitude relative sea-level fluctuations that typify other glacial intervals (Fielding et al., 2008a, 2008b, 2008c). Rather, these strata record normal marine and fluvial conditions, with only modest relative sea-level changes. Similarly, Antarctic records indicate nonglaciated basement highs on which thick paleosols developed and that were onlapped by postglacial strata (yet were not eroded by subsequent ice sheets), precluding the existence of a large ice sheet over Antarctica (Isbell et al., 2003a, 2003b, 2008b). These ice-free basins were polar, indicating a largely ice-free southern pole during this interval (Montañez et al., 2007; Fielding et al., 2008a, 2008b, 2008c; Isbell et al., 2008b). In subpolar settings, Desmoinesian to Virgilian ice is currently reported from several basins: the Paraná, Karoo-Kalahari, the Arabian Peninsula, and possibly the Congo (Rocha-Campos et al., 2008; Holtz et al., 2008; Isbell et al., 2003a, 2008a; Stollhofen et al., 2008; Martin et al., 2008). However, ice flow directions in these basins require multiple ice sheets (e.g., Isbell et al., 2008a), likely at least five to seven (J. Isbell, 2008, personal commun.). The absence of an ice sheet over Antarctica and eastern Australia and the presence of multiple smaller ice sheets in lower latitude basins place important constraints on continental ice sheet volume and possible amplitudes of glacioeustasy. Given the same aerial extent, multiple smaller ice centers can lock up significantly less water than one large ice sheet (Isbell et al., 2003a), leading to greatly dampened glacioeustasy. Thus, the late Pennsylvanian glacioeustatic minimum inferred from Arrow Canyon strata corresponds with a time of limited glacial extent in high-latitude Gondwana.

During the inferred upper Desmoinesian to lower Virgilian glacial minimum, other low-latitude records (e.g., midcontinent cyclothems) have traditionally been interpreted to reflect high-amplitude glacioeustasy. However, few of these interpretations are based on quantifiable evidence for sea-level change (e.g., erosional relief at lowstands). Rather, in upper Missourian to lower Virgilian strata, erosional relief on disconformities demonstrates only low- to moderate-amplitude (20–40 m) incision events (Feldman et al., 2005). Only two records suggest high-amplitude (60–70 m) sea-level fluctuations, based on fluvial incision at lowstands. Both intervals, however, occur just below the Desmoinesian-Missourian boundary (Schenk, 1967; Heckel et al., 1998). Furthermore, reexamination of some analogous later Virgilian “classic” incised valleys has resolved multiple superimposed lower-amplitude incision events, reducing inferred eustatic magnitudes from >60 m to <30 m (Fischbein et al., 2009).

The paradigm of high-amplitude late Desmoinesian to early Virgilian glacioeustasy is derived primarily from facies changes in Kansas-type cyclothems (Heckel, 1977, 1986, 1994; Boardman and Heckel, 1989). In these cyclothems, exposure horizons are juxtaposed against phosphatic black shales that accumulated below a pycnocline. The inferred depth of the pycnocline (>50–90 m) provides the fundamental constraint on relative sea-level change, leading to glacioeustatic estimates of >100 m (Heckel, 1977, 1994). This model is based on thermocline depths estimated from modern ocean basins, and requires the absence of any haloclines across the much more restricted midcontinent epeiric sea. However, such tropical seas would be highly sensitive to changes in runoff and the creation of shallow haloclines (e.g., Algeo and Heckel, 2008; Algeo et al., 2008). Indeed, the organic matter content and geochemistry of “core” shales in midcontinent cyclothems now support such a shallow (15–30 m) halocline (Algeo and Heckel, 2008). In addition, in the Visean paleo-equatorial Mount Head Group of Canada, glendonites (pseudomorphs after thinnolite) record near-freezing temperatures in shallow-marine facies, suggesting very shallow thermoclines in this tropical interior seaway (Brandley and Krause, 1994, 1997). In concert, this evidence for shallow (15–30 m) Carboniferous epeiric sea pycnoclines indicates that “core” shales might have formed at shallow depths and magnitudes of glacioeustasy may have been significantly less (∼15–30 m) than commonly inferred.

In a similar fashion, high-amplitude (>70–120 m) glacioeustatic fluctuations have been reconstructed from oxygen isotopic changes within cyclothems (Adlis et al., 1988; Joachimski et al., 2006). Such estimates apply Quaternary open-ocean models of gentle thermoclines, absent haloclines, and buffered tropical climates (glacial-interglacial changes in sea-surface temperature of 1–4 °C) to Pennsylvanian shallow epeiric seas. However, these epeiric seas would have been especially prone to sluggish mixing of water masses (e.g., Holmden et al., 1998; Panchuk et al., 2005, 2006), large salinity fluctuations (e.g., >8 ppt across the Bahamas platforms; Patterson and Walter, 1994), shallow pycnoclines (Algeo and Heckel, 2008), and the development of subtle intrashelf depressions (e.g., where black shales form in shallow Mesozoic basins; Immenhauser and Scott, 2002; Homewood et al., 2008). For example, depleted δ18O in conodont apatite deposited during highstands is interpreted to represent significantly diminished ice caps (Joachimski et al., 2006). However, interglacial temperature changes beyond the 2–4 °C assumed by Joachimski et al. (2006) would have a significant effect on glacioeustatic estimates; e.g., an extra ∼2 °C cooling would offset 40–50 m of inferred sea-level change. Indeed, to the degree that such epeiric seas buffered continental temperatures, the seas’ temperatures would have been perturbed (Stanley, 2006). Accordingly, evidence for near-freezing tropical surface waters (Brandley and Krause, 1994, 1997) and low-latitude low-altitude alpine glaciation (Soreghan et al., 2008) suggests a dynamic tropical climate with episodically cold, shallow waters. Joachimski et al. (2006) utilized a static salinity structure for the midcontinent sea, assuming no change between glacial and interglacial settings. This premise would be remarkable given the episodically stratified water column (punctuated by black shale deposition). Specifically, Joachimski et al. (2006) compared conodonts deposited nearshore during lowstands with those deposited offshore during highstands. Algeo and Heckel (2008) inferred salinity gradients of 10‰ within the stratified midcontinent epeiric sea, with highstand surface waters diluted by increased continental runoff. Using a modern equatorial Atlantic analog (Fairbanks et al., 1992), this salinity change would cause a 1‰ δ18O depletion, offsetting ∼100 m of inferred sea-level change (Joachimski et al., 2006). It might be argued that lateral salinity gradients would lead to lower salinity during lowstands due to deposition more proximal to the shoreline and presumed freshwater input. However, it is only during highstands in the midcontinent that evidence exists for stratified water columns (black shale deposition, rather than the stenohaline fauna and oxygenated bottom waters of transgressive and regressive limestones; Algeo and Heckel, 2008). Furthermore, midcontinent paleosols and coal deposits imply a semiarid to subhumid climate during lowstands and a humid to subhumid climate during early transgressions, possibly sustained through highstands (e.g., DiMichelle et al., 2010, and references therein). Joachimski et al. (2006) assumed that the conodont-bearing organism did not respond to changing environmental conditions; however, there is much uncertainty regarding where in the water column conodont-hosting organisms lived and how they adapted to changing temperature, salinity, nutrient inputs, and oxygenation. Thus, the assumptions that modern thermocline profiles would be operable in late Pennsylvanian seas (Adlis et al., 1988), that glacial-interglacial tropical temperature changes in the Pennsylvanian were comparable to those of the Pleistocene (Joachimski et al., 2006), and that haloclines would not be operable, may be grossly oversimplified in these geochemical models.

Even within the Euramerican cyclothem record, a more nuanced interpretation is emerging. Rygel et al. (2008) compiled published estimates of glacioeustatic magnitudes; they showed that these reported estimates (of varying vintage and veracity) delineate broad patterns of fluctuating magnitudes through the late Paleozoic. More specifically, Heckel (2008) revisited his classic cyclothem interpretation to conclude that the majority of the late Pennsylvanian cyclothem record actually reflects highstands in long-term sea level and thus corresponds to diminished ice volume over Gondwana: these intervals include many of the major and intermediate cyclothems, for which the largest glacioeustatic magnitudes have been inferred. This analysis again reinforces that the cyclothems of the midcontinent, and the meter-scale cycles of Arrow Canyon, are not one size fits all, a simple repeated motif. Rather, they vary in character, and this variability reflects the dynamic nature of late Paleozoic glaciation.

Numerous authors have interpreted incision of 30–50 m in upper Virgilian carbonates (Wilson, 1967; Goldstein, 1988; Rankey et al., 1999), and Soreghan and Giles (1999a, 1999b) documented >30–85 m of relief on some unconformities. However, in Arrow Canyon, late Virgilian relative sea-level changes are difficult to interpret due to pervasive recrystallization, which obscures many depositional characteristics.

Aridification

The Arrow Canyon record indicates a marked aridification during the proposed late Pennsylvanian glacial minimum, a signature recorded elsewhere across equatorial western Pangea (Rankey, 1997; West et al., 1997; Olszewski and Patzkowski, 2003), equatorial eastern Pangea (DiMichele et al., 2009, and references therein), and high-latitude western Gondwana (Gulbranson et al., 2010). This late Pennsylvanian aridification is coincident with mass extinctions of tropical peat-forming lycopsids (DiMichelle and Phillips, 1996) and a concomitant change in the quality of Appalachian coal (Cecil et al., 2003).

It is commonly held that arid tropical conditions were coincident with icehouse intervals during the Paleozoic. This conclusion is largely based on extrapolating climate changes on short time scales (<105 yr; glacial-interglacial) based on Pleistocene analogues to longer time scales (106 yr; icehouse-greenhouse). Thus, evidence for arid glacials and humid interglacials has been extrapolated to require arid icehouse intervals and humid greenhouse intervals. In Permian time, however, aridification of the tropics was coincident with long-lived greenhouse intervals during the final phase of the late Paleozoic ice age (Montañez et al., 2007). During the late Pennsylvanian, aridification of western equatorial Pangea (and withdrawal of the epicontinental seas) has been linked to the final assembly of Pangea and thermotectonic buoyancy of the supercontinent, mechanisms independent of ice sheet dynamics (Veevers, 1994; Ziegler et al., 2002). However, this explanation is confounded by a marked aridification east of the main Pangean tropical mountain belt (DiMichele et al., 2009). Climate models suggest that aridification during glacial minima was caused by a weakening of Hadley cell convection and southward drift of the Intertropical Convergence Zone, leading to decreased precipitation levels and intensified monsoonal circulation across western equatorial Pangea (Poulsen et al., 2007; Montañez et al., 2007). In the Arrow Canyon record, the coincidence of significantly more arid conditions with the decline in amplitude of glacioeustasy argues for a strong component of climate forcing, suggesting a mechanistic link between pantropical aridification and the retreat of high-latitude Gondwanan ice sheets.

The record of glacioeustasy archived in Arrow Canyon requires a much more dynamic Carboniferous glaciation than commonly perceived. This record is at odds with the prevailing view that cyclothems require high-amplitude glacioeustasy throughout the ∼50 m.y. interval of their deposition. The Arrow Canyon record is more consistent with high-latitude records that suggest alternating long-lived intervals of glacial maxima and minima, including a late Desmoinesian–early Virgilian glacial minimum. This late Pennsylvanian minimum was coincident with a marked aridification in Arrow Canyon and across the tropics, supporting a link between high-latitude ice sheet extent and stability and ocean-atmospheric dynamics in the tropics (e.g., Poulsen et al., 2007).

These climatic oscillations had a significant impact on the facies, cyclicity, and stacking patterns in Arrow Canyon. The recognition of a dynamic late Paleozoic ice age suggests that not all cyclothems formed under the same glacioeustatic forcing, and should lead to a much more nuanced understanding of late Paleozoic tropical sedimentation.

We thank Mike Eros, Neil Kelley, and Erik Gulbranson for field assistance, and Paul Brenckle, Greg Wahlman, and Vladimir Davydov for biostratigraphic expertise. Blaine Cecil, Mitch Harris, Scott Ritter, and Lynn Soreghan provided advice and discussion. Scot Franklin, Tom Brady, James Sippell, Mark Boatwright, and Cathy Wiley all helped arrange access and sampling permits. We also thank Norm Winter and Greg Baxter for analytical and laboratory support. The clarity and focus of this manuscript were improved considerably by detailed reviews from Tom Algeo, Andrew Barnett, and associate editor Kate Giles. This work was supported by National Science Foundation grant EAR-0545654 (Sedimentary Geology and Paleontology Program) to Montañez.