Abstract

Upper Triassic sandstones in diverse locations in eastern California, southern Arizona, and on the Colorado Plateau (USA) yield detrital zircons that are remarkably similar in age and geochemistry, leading to the hypothesis that they are temporally related and were derived from similar sources. Volcaniclastic sandstone from the lowest Vampire Formation in eastern California, the Sonsela Member of the Chinle Formation at Petrified Forest National Park, northeastern Arizona, and the herein-named Waterman formation in southern Arizona yield zircons that range in age from ca. 205 to ca. 235 Ma. Together with the similar range of ages, these zircons uniformly have Th/U ratios between ∼0.2 and 2. In addition, the Waterman formation contains zircon grains with an age range from ca. 225 to 250 Ma, but with markedly lower Th/U ratios of 0.1–0.2, and a distinctively older group with ages to ca. 280 Ma. In a general sense, variations in Hf concentrations and Yb/Gd ratios support the discrimination of grains based on age and Th/U.

We use age and geochemical data from the zircons to infer that these units capture a slice of time during development of the early Mesozoic Cordilleran magmatic arc along western North America. Plutonic rocks that record magmatism in the arc are Permian–Triassic in age, and match zircon ages in the detrital grains, thus providing a view of which parts of the arc were actively eroding into the stream systems that deposited the three units. Streams diverged from a common source that maintained a relatively uniform magma composition over time, as indicated by a narrow range of Th/U values, as well as tapping a somewhat different source evidenced by a grouping in which Th/U ratios are lower. Once the streams left the highlands of the arc and the depocenter of the lowest Vampire Formation, they diverged, such that one flowed to the area of the Colorado Plateau while the second flowed toward southern Arizona. At the same time, a stream system originating in the older, Sonoran part of the arc flowed from the south into southern Arizona.

INTRODUCTION

Much of the understanding of the inception and development of the late Paleozoic–early Mesozoic Cordilleran magmatic arc along southwestern North America is based on plutonic rocks in the Mojave Desert (Miller et al., 1995; Barth and Wooden, 2006) and northwestern Sonora (Riggs et al., 2009, 2010; Arvizu et al., 2009) and on backarc sedimentary strata deposited on the continent (Fig. 1). The nature of deposition as recorded by sedimentary strata in southwestern Laurentia changed dramatically in late Paleozoic time from carbonate platform rocks (Kaibab and Rain Valley Formations: Blakey and Knepp, 1989) to dominantly fluvial and terrestrial environments (Lower Triassic Moenkopi Formation, Upper Triassic Chinle Formation; Stewart et al., 1972a, 1972b). In the absence of volcanic sections in many parts of the early Cordilleran arc, much of the record of magmatism is in backarc sedimentary successions.

Unmetamorphosed Paleozoic sections are exposed on the Colorado Plateau and in southern Arizona, and equivalent strata have long been recognized in deformed sections in eastern California and western Arizona, where they are overlain by strata of the Buckskin and Vampire Formations; these two units are considered to be correlatives of the Triassic Moenkopi and Chinle Formations and parts of the Jurassic Glen Canyon Group (Reynolds and Spencer, 1989). The position of these Mesozoic rocks close to remnants of the Cordilleran magmatic arc suggests that the more proximal setting would be reflected in more arc-derived detritus. Similarly, Paleozoic passive-margin sedimentary strata in southern Arizona locally underlie a sedimentary succession that is in turn overlain by Jurassic volcanic rocks. These post-Paleozoic strata off the Colorado Plateau complement the understanding of the timing of arc magmatism afforded by Triassic units on the Plateau, and add to the expanding knowledge of the drainage systems that flowed from the arc.

We present the results of U-Pb geochronology and geochemistry of zircons from the lowest Vampire Formation in eastern California and from fluvial sedimentary strata from southern Arizona that we name the Waterman formation (Fig. 1). Our results show that both units off the Colorado Plateau are temporally correlated with the Chinle Formation, and that each unit provides a distinct and critical clue about Triassic paleogeography of southwestern Laurentia as well as the growth and erosion of the Cordilleran magmatic arc.

TECTONIC AND MAGMATIC SETTING OF THE EARLY MESOZOIC CORDILLERA

The Mesozoic Cordilleran arc formed along the truncated western coast of Laurentia. Along the southwestern margin, northeast-southwest–trending Paleozoic passive-margin facies were faulted in Pennsylvanian–Permian time by a strike-slip system that accompanied the initiation of subduction of the Pacific oceanic lithosphere under North America (Walker, 1988; Stone and Stevens, 1988; Bateman, 1992; Miller et al., 1992; Saleeby et al., 1992; Dickinson and Lawton, 2001). Plutonic rocks that record subduction are exposed in eastern California and northern Sonora and range in age from ca. 270 to ca. 215 Ma (Fig. 1; Miller et al., 1995; Barth et al., 1997, 2011; Barth and Wooden, 2006; Riggs et al., 2009; Arvizu et al., 2009).

Preserved sections with volcanic rocks recording Permo-Triassic igneous activity are rare in most parts of the arc. The notable exception is in the east-central Sierra Nevada, where Schweickert and Lahren (1993, 1999) documented a caldera complex at Tioga Pass, California, and where dates of 232–218 Ma (Barth et al., 2011) were obtained on volcanic rocks interpreted to be outflow from that caldera complex. Depositional features of these volcanic rocks, including angular fragmentation and fine ash-rich laminations, suggest that they were erupted subaqueously (Douglas et al., 2011). In the Mineral King pendant, a tuff dated as 220 Ma (N. Riggs and C. Busby, 2012, personal observ.) is interstratified with marine sedimentary rocks (Busby-Spera, 1984, 1986). These observations suggest that at least the northern part of the arc, in Late Triassic time, was marine. To the south of the Sierra Nevada, however, early Mesozoic volcanic rocks are rare and poorly dated where present.

TRIASSIC SEDIMENTATION TO THE EAST OF THE CORDILLERAN ARC

Truncation of the Neoproterozoic–Paleozoic margin brought about the end of marine passive-margin sedimentation and coincided with a major change in the nature of continental sedimentation (e.g., Lawton, 1994). The Lower–Middle Triassic Moenkopi Formation records transport from highlands to the east; overall, the depositional setting is attributed to a flexural foreland basin developed behind the Permian–Triassic Sonoma orogen (Dickinson and Gehrels, 2008). The Moenkopi Formation comprises an array of facies, from fluvial to the east to intercalation with marine limestone farther west; paleocurrents indicate flow from the south and southeast (Stewart et al., 1972b; Blakey, 1989; Blakey et al., 1993; Dubiel, 1994). Detrital zircons from one sample of fluvio-deltaic facies in the Moenkopi Formation suggest a maximum depositional age of 235 Ma, and Dickinson and Gehrels (2008) inferred that most or all detrital zircons in the Moenkopi Formation were derived from the southeast, including early Mesozoic zircons from the poorly understood East Mexico arc.

The Upper Triassic Chinle Formation was deposited in a back-bulge basin (Lawton, 1994) or one resulting from dynamic backarc subsidence as Pacific oceanic lithosphere was subducted (Dickinson and Gehrels, 2008). The Chinle Formation comprises fluvial facies that overall record transport from south and southeast toward coastal to shallow-marine environments in Nevada (Stewart et al., 1972a; Blakey and Gubitosa, 1983; Lupe and Silberling, 1985; Lucas and Marzolf, 1993; Riggs et al., 1996), but that locally were derived from the growing magmatic arc to the west (e.g., Howell and Blakey, 2013). The southern and southwestern edges of Chinle Formation outcrops are erosional; to the east and southeast the Chinle Formation correlates with the Dockum Group in New Mexico (e.g., Lucas, 1991).

Shinarump Member

The basal Shinarump Member of the Chinle Formation (Fig. 2) consists of fluvial conglomerate and sandstone (Stewart et al., 1972b; Blakey and Gubitosa, 1984). Volcanic clasts are rare, and throughout the Chinle Formation decrease in size and abundance from south to north (Stewart et al., 1972b), suggesting deposition on a broad alluvial plain linked to highlands to the south termed the Mogollon Slope (Bilodeau, 1986). Triassic igneous sources in these highlands, however, have not been documented, and central Arizona, where they would have been, now consists of uplifted Proterozoic basement. Paleocurrents in the Shinarump Member indicate flow from the south and southeast (Stewart et al., 1972b; Blakey and Gubitosa, 1983).

Dickinson and Gehrels (2008, 2009) suggested that deposition of the basal Shinarump Member began ca. 230 Ma, well after magmatism began in the Mojave Desert and Sonora, Mexico. Volcanic clasts in the Shinarump Member range in age from ca. 220 to 230 Ma (Oberling et al., 2010; our data). Although the source of the clasts is uncertain, they correlate in age, and broadly in chemistry, with Triassic plutons in the Mojave Desert (Oberling et al., 2010).

Blue Mesa, Sonsela, and Petrified Forest Members

The middle and upper Chinle Formation is dominantly mudstone (Fig. 3), with sandstone interbeds throughout and conglomerate horizons that are mostly confined to the Sonsela Member (Fig. 2). Depositional environments are dominantly fluvial and paludal (Stewart et al., 1972a; Blakey and Gubitosa, 1983; Dubiel, 1987, 1989; Dubiel et al., 1991). Paleocurrents are variable but most commonly indicate transport from the south and southwest in the more southern exposures in Petrified Forest National Park (Howell, 2010; Howell and Blakey, 2013), where our samples were collected. The local stratigraphy for the Chinle Formation in Petrified Forest National Park follows Woody (2006) and Martz and Parker (2010). Exposed strata are assigned to five members; from oldest to youngest, these are the Mesa Redondo, Blue Mesa, Sonsela, Petrified Forest (previously the upper Petrified Forest), and Owl Rock Members.

Igneous detritus is present throughout much of the Blue Mesa, Sonsela, and Petrified Forest Members. Volcanic and rare plutonic detritus ranges from fine, altered ash and pyrogenic crystals throughout the three members to granules, pebbles, and cobbles in the Sonsela Member. Ramezani et al. (2011) established maximum ages of deposition of ca. 225 to ca. 208 Ma for these three units based on high-precision CA-TIMS (chemical abrasion-thermal ionization mass spectrometry) dating of detrital zircons. Volcanic clasts range in age from 235 Ma to ca. 217 Ma (Riggs et al., 2012).

TRIASSIC(?) STRATA OFF THE COLORADO PLATEAU

Buckskin and Vampire Formations

The Buckskin Formation (Reynolds and Spencer, 1989) was named to describe greenschist-grade metamorphosed (Stone and Kelly, 1989) sandstone, siltstone, and conglomerate that disconformably overlie the Permian Kaibab Formation in the Mojave Desert of western Arizona and eastern California (Fig. 1). The Buckskin Formation consists of four informal members (Reynolds and Spencer, 1989; Hargrave, 1999) that are correlated with the Moenkopi Formation (Reynolds and Spencer, 1989) or the Moenkopi and Chinle Formations (Hargrave and Reynolds, 1999) on the Colorado Plateau based on lithology and stratigraphic position overlying the Permian Kaibab Formation. The contact between the Buckskin and Vampire Formations is marked by conglomerate (Fig. 4) that is locally interbedded with volcaniclastic sandstone (Volcanic sandstone unit of Stone and Kelly, 1989). Thick quartzite that overlies the conglomerate is well sorted, fine grained to gritty, and in some exposures has large-scale cross-bedding. The basal conglomerate of the Vampire Formation records an early Mesozoic uplift (Reynolds et al., 1989), and has been correlated with the Chinle Formation (Hargrave, 1999; Hargrave and Reynolds, 1999); the formation as a whole is broadly within the Chinle Formation–Glen Canyon Group (i.e., Late Triassic–Middle Jurassic) interval. Although depositional environments of the Vampire Formation are incompletely documented, the unit is inferred to be subaerial based on the presence of cross-bedded (i.e., eolian) sandstone, interbedded conglomerate and sandstone lenses, and its stratigraphic position overlying a regional unconformity.

Waterman Formation

Sedimentary rocks that overlie Permian carbonate and underlie Mesozoic volcanic rocks in the Waterman Mountains (Fig. 1) were first described by Hall (1985). Similar rocks considered to be equivalent are also exposed in the Sierrita, Tucson, and Mustang Mountains (Fig. 1). In the Mustang Mountains, Hayes and Raup (1968) first mapped “Volcanic and sedimentary rocks of Mustang Mountains,” which they assigned a Triassic and Jurassic age. In general the rocks are poorly exposed and facies are discontinuous from outcrop to outcrop. Exposures in the Waterman Mountains are the most complete, and this area is considered the type locality for these strata, which we call the Waterman formation. Strata in the Sierrita Mountains are strongly sheared and altered, in part due to proximity to a Cretaceous intrusion and its mineralizing fluids. The outcrop in the Tucson Mountains is part of a slide block in the Cretaceous Tucson Mountain caldera. These exposures do not provide useful stratigraphic information, but lithologic similarities among scattered outcrops suggest that the formation was originally widespread across the backarc region.

In the Waterman Mountains, the Waterman formation is as much as 20 m thick. Basal conglomerate (Figs. 5A, 5B) that is 2–10 m thick overlies Permian Concha Limestone, a Middle–Late Permian formation equivalent to the Kaibab Formation on the Colorado Plateau (Blakey and Knepp, 1989), on an irregular eroded contact (Hall, 1985). Clasts are as much as 20 cm in diameter, with an average diameter of 4–5 cm, and comprise quartzite, chert, and carbonate commonly armored by chert; the clasts are derived from the underlying Paleozoic section. Dark orange to red, very dense quartzite clasts are likely Cambrian Bolsa Quartzite, which is far more durable than locally derived Permian sandstone (Scherrer Formation). Very rare clasts of conglomerate are present (possibly the Jelly Bean conglomerate of Armin, 1987, which is exposed in the underlying Paleozoic section), but no igneous clasts of any kind were found in this study or reported by Hall (1985). Channel-form coarse sandstone is interbedded locally within the conglomerate (Fig. 5C) in discontinuous lenses and stringers as much as 20 cm thick. Thin sections reveal rounded-quartz-dominated sand grains with as much as 100% monocrystalline and polycrystalline quartz and pseudo matrix; other constituents are 2%–13% feldspar and very rare altered volcanic and metamorphic grains. Conglomerate is overlain in places by structureless red sandstone and siltstone as much as 2 m thick that has identical composition to the sandstone lenses and matrix within the conglomerate.

The Waterman formation in the Mustang Mountains (Figs. 1 and 5A) is poorly exposed, and, although faulted, ∼10 m thick. The formation overlies Permian Concha Limestone (Hayes and Raup, 1968) on an irregular contact with a few meters relief. Facies are very similar to those in the Waterman Mountains. Clasts are angular in places (Fig. 5D), and in general smaller than clasts in the type section; clast types are similar, although white chert is more common than in the Waterman Mountains section. Overall the average clast size is 2 cm; outsize clasts have a maximum diameter of 12 cm, and average ∼6 cm. Red, fine-grained sandstone intervals are 5–8 cm thick; the lateral extent is masked by the poor exposures. In both sections, the relatively poor exposure and structural disruption preclude gathering any paleocurrent data.

Hall (1985) interpreted the depositional environment of the succession in the Waterman Mountains as braided stream or alluvial fan; a similar environment is reasonable for the Mustang Mountains. The angularity of clasts in some places indicates relatively short distance of travel and the nearby topography; the dominant clast types in both sections reflect the immediately underlying rock. Very well-rounded clasts suggest a well-developed fluvial system and long transport distances.

The Waterman formation exposures in the Mustang and Waterman Mountains are considered to be stratigraphically equivalent and the same unit based on composition and stratigraphic position. Both units comprise conglomerate and sandstone with some fine-grained mudstone-siltstone lenses, although some differences exist within conglomerates in each area. Both sections overlie Middle–Upper Permian limestone, and in the Mustang Mountains this erosional surface is part of a deep canyon filled by younger pyroclastic units. The uppermost sandstone in the Waterman Mountains is overlain by a volcaniclastic unit (Fig. 5A) that contains andesitic to dacitic clasts, and by lava flows, one of which yielded a multigrain, three-fraction, TIMS U-Pb zircon age of 176 Ma (our data). In the Mustang Mountains, conglomerate and sandstone are overlain by ignimbrite (Fig. 5A), from which Lawton et al. (2012) obtained a single-crystal SHRIMP (sensitive high-resolution ion microprobe) U-Pb age of 176 ± 2 Ma.

METHODS

We collected one sample of the lowermost Vampire Formation in the “Volcanic Sandstone (Triassic or Jurassic)” unit of Stone and Kelly (1989) in the Palen Mountains (Fig. 1; 030509–3, Supplemental Table1) and two samples of the Waterman formation, one each from the Waterman and Mustang Mountains (Fig. 1; 111308–1, 111408–2, Supplemental Table [see footnote 1]). Two samples of the Sonsela Member of the Chinle Formation are included here, both from Petrified Forest National Park (Figs. 1 and 2; Jim Camp Wash bed sample 050508–1, and Long Logs sandstone, Supplemental Table [see footnote 1]). Although the detrital zircon signature of the Sonsela Member in Petrified Forest National Park is almost certainly not representative of the unit across the entire Colorado Plateau, the detrital zircon signatures of the lowermost Vampire Formation and the Waterman formation, as discussed in the following section, indicate that these units reasonably correlate with the Sonsela or Petrified Forest Members. The provenance and significance of the Sonsela Member in Petrified Forest National Park was described in Howell (2010) and Howell and Blakey (2013).

Samples were crushed and zircons separated according to standard techniques of density separation (e.g., Gehrels, 2000) and only minimal (0.1–0.3a) magnetic separation was done. All samples except Long Logs were then annealed at 875 °C for 48 h, followed by 12 h in an 80 °C oven in a 10:1 HF-HNO3 mixture. Zircons were washed in warm HNO3 and ultrapure H2O, then mounted and polished. Cathodoluminescence imaging was done at Northern Arizona University using a JSM-6480LV scanning electron microscope to identify grain shapes, zoning patterns, and cores. Samples were analyzed for U and Pb isotopes and for trace element concentrations at the University of California, Santa Barbara, Laser Ablation Split-Stream laboratory using a Nu Plasma HR MC-ICP-MS (high resolution multi-collector-inductively coupled plasma–mass spectrometer), a Nu AttoM single collector ICP-MS (Nu Instruments Ltd., Wrexham, UK), and an Analyte 193 excimer ArF laser-ablation system equipped with a HeLex sample cell (Photon Machines, San Diego, USA) using a 24 μm beam. Analytical and procedural details and all analytical data are provided in the table in the Supplemental File2; all age errors reported are 2σ, unless explicitly stated otherwise. Data were reduced using Iolite 2.10 and 2.21 in Igor Pro 6.2 (www .wavemetrics.com). Analyses were evaluated for discordance based on a comparison of 235U/207Pb and 238U/206Pb for Permian and Triassic grains, and 207Pb/206Pb and 238U/206Pb for Proterozoic grains. Grains that were >10% normally discordant (i.e., 235U/207Pb age or 207Pb/206Pb age >10% older than 238U/206Pb age) or 5% reversely discordant (i.e., 238U/206Pb age >5% older than 235U/207Pb age) were not used in interpretations; these are indicated by “discordant” in the Supplemental Table (see footnote 1).

Zircon grain shapes in the lowest Vampire Formation sample are quite different from those in the Sonsela Member of the Chinle Formation and Waterman formation. Zircons in the Vampire Formation sample are euhedral for the most part, and cathodoluminescence images, although generally poor in quality, show chemical zones parallel to grain boundaries, indicating little modification of grain shape. In the Waterman formation, grains range in shape from euhedral to subrounded, and zoning within grains is commonly truncated by the grain boundaries. Grain shapes and zoning-band truncations in the Jim Camp Wash bed zircons (Sonsela Member) are similar to those in the Waterman formation, although a greater percentage of grains are euhedral. Images are not available for the Long Logs sandstone sample. These observations suggest that, as would be expected for a volcanic sandstone, zircons in the lowest Vampire Formation sample underwent little transport prior to deposition. This contrasts with samples from the other two units, in which subrounded grains indicate a degree of transport. Common euhedral grains in the Jim Camp Wash bed suggest original transport by ash clouds.

RESULTS

Density probability plots (Fig. 6) show very distinct similarities and differences between representative samples of the three units. In addition to the density probability plots, an average-age plot shows the range of ages in each sample. Sonsela Member samples from Petrified Forest National Park (Figs. 6A, 6B) are dominated by Triassic grains nearly to the exclusion of all other ages. Of 83 Phanerozoic grains in the Jim Camp Wash bed, 86% have ages in a continuous range from 202 to 223 Ma (Fig. 6A) and only 6 are Proterozoic (7%). Similar to the Jim Camp Wash bed zircons, 77% of grains in the Long Logs sandstone have Phanerozoic ages between 209 and 232 Ma (Fig. 6B). In both units, the Proterozoic grains are spread from ca. 1000 to 1700 Ma, but clusters have too few grains to be statistically significant.

The lowest Vampire Formation (Fig. 6C) grain distribution is nearly identical to that of the Long Logs sandstone, including minimum and maximum ages and the small number (6%) of Proterozoic grains. Phanerozoic grains are all Triassic and range in age smoothly from 215 to 235 Ma. The Waterman formation shows the greatest internal diversity. The grains from the Waterman Mountains sample (Fig. 6D) range in age from ca. 210 to 250 Ma; Proterozoic grains (16%) are ca. 1400 Ma and 1600 Ma populations. The Mustang Mountains sample (Fig. 6E), in contrast, is dominated by ca. 1400 Ma grains. Phanerozoic grains range from ca. 240 to 280 Ma, although few grains are older than ca. 260 Ma, and very few grains overlap in age with the Waterman Mountains sample. The Mustang Mountains sample thus bears little resemblance to the other Triassic units.

The differences and similarities between the units are additionally highlighted by Th/U ratios and trace elements such as Hf, Yb, and Gd (Figs. 7 and 8). Use of these trace elements is based on the assumption that changes in magma chemistry during cooling are reflected in zircon chemistry. In general, the increase in Hf solid solution in zircon that accompanies decreasing temperature is matched by a decrease in Th/U and relative enrichment in heavy rare earth elements (Claiborne et al., 2006, 2010; Fohey-Breting et al., 2010; Barth et al., 2011, 2012). This observation has been successfully used to correlate intrusive and extrusive phases of the early Mesozoic California Cordilleran arc (Barth et al., 2012) and to make provenance inferences about the igneous sources of detrital zircons (Riggs et al., 2012). In this case, we propose that zircons carry distinctive signatures that can be used to isolate groups of cogenetic grains within a detrital sample and enhance the correlation of disparate groups of rocks.

Figure 7 compares Th/U with age and shows that all three units have grains that plot within the age range of ca. 210–235 Ma and Th/U values of 0.3–2; the Sonsela Member from Petrified Forest National Park and the lowest Vampire Formation samples overlap strongly. The Waterman formation, in contrast, shows a tripartite division of ages and Th/U values. A major part of the 210–235 Ma grains plot in the field encompassed by the Sonsela Member and Vampire Formation samples; the field is referred to as Waterman Mountains group 1. Another distinct field comprises grains from the Mustang Mountains sample that range in age from ca. 245 to 280 Ma and have variable Th/U ratios. A few grains from the Mustang Mountains are in a third field that consists of Waterman Mountains grains that range in age from ca. 225 to 250 Ma, thus overlapping in age with the Sonsela Member and Vampire Formation field, but having distinctively lower Th/U ratios that range from 0.1 to ∼0.3 (called group 2).

A plot of Yb/Gd versus Hf ppm (Fig. 8) supports the similarity of the Waterman Mountains group 1 grains to Sonsela Member and Vampire Formation grains and the differences between the Waterman formation groups. The dashed-outline box in Figure 8 demonstrates that the area of overlap between the Waterman Mountains group 1 grains, the lowest Vampire Formation sample, and the Sonsela Member sample is relatively noncoincident with group 2 grains, although the data overlap more than would be inferred by the comparison of Th/U ratios.

DISCUSSION

U-Pb ages, Th/U ratios, and rare earth geochemistry highlight similarities and differences between the three Upper Triassic units. The units are remarkably similar in terms of the age and composition of sources, although zircon geochemistry allows distinctions to be made about magmatic provinces within the arc. In addition, these data allow us to make inferences about the timing of magmatism, topography, and evolution of continental landscape to the east of the arc. Geochemistry of zircons has been used successfully to demonstrate cogenesis of plutonic and volcanic facies (Barth et al., 2012); we suggest here that this geochemistry can also be used to infer cogenesis of suites of detrital zircons, as well as to highlight potential differences in source.

The Cordilleran arc in Late Triassic time is generally envisioned as marine to the north and subaerial to the south, at least in California. The observations that the Chinle Formation contains volcanic detritus as old as 235 Ma and that deposition began ca. 230 Ma (Dickinson and Gehrels, 2008, 2009) indicate that by middle Carnian time, the portion of the arc now exposed in Southern California was a subaerial, eroding feature that shed detritus eastward. Stratigraphic evidence for this is in the Black Mountain region to the north of San Bernardino (Fig. 1), where a ca. 244 Ma pluton is nonconformably overlain by Lower Jurassic strata (Miller, 1981; Stone, 2006), indicating that post–Early Triassic, pre–Early Jurassic uplift brought magmatic arc rocks to the surface. Similarly aged uplift is documented in western Arizona (Reynolds et al., 1989), although in western Arizona the unconformity does not involve magmatic arc rocks. Knowing the precise location of volcanoes that fed Late Triassic streams into the backarc region is difficult (Riggs et al., 2012), but it is easiest to infer that the closest present-day exposures of the arc were the closest sources.

Deciphering paleotopography and uplift history in northern Sonora is also important for understanding the distribution of grains derived from the southern sources. The presence of Permian plutons (Arvizu et al., 2009; Riggs et al., 2009, 2010) indicates that the arc is older in Sonora, and detritus from this part of the arc is represented by zircon grains in the Waterman formation. The Permian–Jurassic Antimonio Group in northwest Sonora records forearc shallow-marine and fluvial sedimentation (González-León, 1997; González-León et al., 2005; Lindner et al., 2012); these strata contain Permian cobbles (280 Ma) in strata as old as Lower Triassic (Riggs et al., 2010; Lindner et al., 2012), indicating that this southern part of the arc was actively eroding long before areas to the north. By Late Triassic time, both volcanic and plutonic parts of the arc at least as far north as the central Mojave Desert were eroding and shedding material to the north-northeast.

Discriminating Source Areas Based on Detrital Zircon Chemistry

The pronounced similarity in age and Th/U ratio between zircons from group 1 of the Waterman formation, the Sonsela Member of the Chinle Formation, and the lowest Vampire Formation strongly suggests that sources were similar in time and space. Grains vary in age over 30 m.y., and volcanic arcs change in chemistry and character over such time spans, but the general nature of the underlying continental crust should remain relatively constant, and all currently known remnants of the arc were emplaced into Mojave crust (Barth and Wooden, 2006). The value of using Th/U ratios to assess magmatic provinces within the arc is supported by the presence of ∼15 grains from the Waterman formation within group 2 that have ages similar to group 1 grains, but much lower Th/U values than group 1 grains (Fig. 7); we infer that these were derived from a very different source from group 1 grains. Likewise, a comparison of Gd/Yb and Hf ppm (Fig. 8) supports a similar chemistry for the sources of group 1 grains and those from the Sonsela Member and the Buckskin Formation. It remains equally likely, however, that the range of values describes a variety of sources, but the drainage systems between arc and backarc were relatively few and entrenched, perhaps analogous to the relatively few major rivers that drain the eastern Cascade Range and the wide dispersal area of these rivers. One indication of this possibility is exemplified by a comparison of detrital grain chemistry to that of the plutonic remnants of the Cordilleran arc in the Mojave Desert.

Permo-Triassic plutons in the Mojave Desert are characterized by low Th/U ratios (Barth and Wooden, 2006); these low ratios are not generally reflected in Chinle Formation zircons from volcanic clasts in the Sonsela Member (Riggs et al., 2012). The Th/U ratios of the sandstones discussed herein are more variable, but in part are within the low Th/U fields of Mojave Desert plutons (Fig. 9). (See the Supplemental File [see footnote 2] for a discussion regarding comparison of data derived from LA-ICP-MS [data in this paper] and those derived by reverse geometry–SHRIMP [Mojave Desert Triassic plutons; Barth and Wooden, 2006].) This suggests that the Mojave Desert plutons were a source of detritus, and that another, as-yet undiscovered source area was also present. Based on the differences in zircon chemistry between volcanic clasts from the Chinle Formation and Triassic plutons, Riggs et al. (2012) speculated that such a source lay near the present Colorado River, where Reynolds et al. (1989) documented uplift and erosion of pre-Triassic strata. Although it is also possible that the differences in Th/U seen between these clasts and plutons can be attributed to chemical stratification with a magma chamber, Barth et al. (2012) have used a strong similarity in whole rock and zircon chemistry, including Th/U, to show that Triassic plutons and ignimbrites in the Sierra Nevada are cogenetic.

The detrital zircon signature of the Chinle Formation varies depending on depositional sites. For example, the Sonsela Member detrital zircon spectrum from a sample northeast of Petrified Forest National Park (Dickinson and Gehrels, 2008) includes Paleozoic and Neoproterozoic zircon populations not found in the Petrified Forest National Park samples. Regardless, we find it significant that (as observed by Howell, 2010) the Sonsela Member within Petrified Forest National Park carries a distinctive detrital zircon signature that reflects a dominant contribution from the Cordilleran arc and that this signature closely parallels that of the lowest Vampire Formation sample, which had a very proximal depocenter with respect to the arc.

The differences between the two Waterman formation samples require explanation. The majority of grains in the Mustang Mountains sample range in age from ca. 245 to 265 Ma; those from the Waterman Mountains are ca. 207–245 Ma. The formation is everywhere underlain by Permian sedimentary rocks and overlain by Jurassic volcanic rocks and thus occupies the same stratigraphic level across southern Arizona. Despite the difference in age implied by the zircons in the two samples, we infer that, similar to the Chinle Formation, this broadly coeval and correlative formation exposed over a wide area had variable source areas over time. Plutons in the Mojave Desert are as old as ca. 250 Ma (Barth and Wooden, 2006), and igneous rocks in northwestern Sonora are as old as ca. 280 Ma (Riggs et al., 2010; Lindner et al., 2012); although Th/U data are relatively sparse, ratios in these igneous rocks are between ∼0.06 and 0.8. A Th/U versus age plot (Fig. 10) shows some overlap between the Mustang Mountains and Sonoran arc igneous rocks from the Los Tanques pluton and volcanic clasts; it is therefore reasonable to infer that grains in the Waterman formation were derived both from the Mojave Desert and from the arc in Sonora, or that Triassic igneous rocks in Sonora are as young as 240 Ma, although ages this young are as yet undocumented.

An alternative explanation for the compositional range in these zircons is that selective diffusion of trace elements took place during transport, yielding a distorted range of Th/U ratios. If this were the case, then measured trace element ratios in zircon would represent an integrated signal of igneous and sedimentary processes, and their ages and compositions would be more difficult to interpret in terms of the range of source areas and the fluvial pathways described here. We consider this explanation unlikely, however, because experimental data show that Pb diffusion rates are very slow, similar to those of smaller trivalent cations, and tetravalent-cation diffusion rates are even slower (e.g., Cherniak and Watson, 2003; Cherniak, 2010). Thus, we assume that ratios of trivalent and tetravalent cations are no more affected by postcrystallization diffusional fractionation than is U/Pb, and that these cation ratios yield insight into magmatic processes at the time of closure of the U/Pb system in these zircons.

Late Triassic Dispersal System

The age equivalence of three geographically disparate units and the correlation of these units based on zircon geochemistry provide the basis for improved understanding of the topography of southwestern North America during a slice of Late Triassic time (Fig. 11). By Late Triassic time, a continental magmatic arc, which may previously have been offshore and thus unavailable to contribute detritus to the continental interior, constituted a subaerial series of volcanoes along the western coast from the central Mojave Desert in California to Sonora, Mexico. In Figure 11, the shoreline is inferred to cross the arc in the southern Sierra Nevada based on marine facies in volcanic and sedimentary rocks in the central and northern Sierra Nevada (Busby-Spera, 1984, 1986; Douglas et al., 2011).

Interpretation of the pathways taken by rivers that drained the arc is constrained by the three major observations and inferences regarding the detrital zircons. (1) The ages, Th/U ratios, and trace element compositions of zircons from the volcanic sandstone in the lowest Vampire Formation are the best possible approximation of the composition of the magmatic arc in that area; in addition, the ubiquitous euhedral shape of grains indicates that for the most part, predepositional transport of these grains was minimal. (2) The similarity in ages, Th/U ratios, and trace element compositions of zircons from the lowest Vampire Formation, from the two Chinle Formation samples, and from group 1 of the Waterman formation suggests that the grains were derived from petrologically similar sources, possibly in a relatively small geographic area, although these areas exposed rocks that ranged in age as much as 30 m.y. (3) The Chinle Formation samples in Petrified Forest National Park do not contain a record of the Sonoran portion of the magmatic arc. In addition, any interpretation needs to accommodate paleocurrent studies that indicated paleoflow to the northeast in Petrified Forest National Park (Howell, 2010; Howell and Blakey, 2013).

A few plausible scenarios accommodate these observations, and Figure 11 presents our preferred interpretation. In all cases, groups 1 and 2 in the Waterman formation must have been deposited by different strands, considering the differences in grain ages. The similarity between Waterman formation group 1 grains and those in the Chinle Formation suggests a connection between these two units. In this case, we envision that the Mustang Mountains section was deposited by a stream system that originated in the Sonoran portion of the arc, to account for the numerous Permian grains. As the Permo-Triassic arc grew in Sonora (cf. Arvizu et al., 2009), younger (Triassic) grains were incorporated into the stream system and the older sources were to the southwest of a divide that kept their detritus moving to the south and west. This younger stream system was the source of group 2 in the Waterman Mountains and would have also tapped a source that was younger and with different chemistry, providing the ca. 205–225 Ma higher Th/U zircons of group 1. This scenario provides a short and direct link between southern and northern Arizona, but it is difficult to envision how ca. 225–245 Ma, low Th/U zircons, which are not present in the Chinle Formation samples, would have been selectively winnowed out of the system between southern and northern Arizona.

Our preferred interpretation (Fig. 11) is that the Waterman formation grains were deposited by two strands, one of which came eastward from the Mojave Desert, and the other derived from sources in the Sonoran arc. The pathways shown in Figure 11 are speculative, but allow the Mojave Desert strand to deposit grains only as old as ca. 250 Ma, whereas the Sonoran strand carried older grains. The depocenter topography, as indicated by deposition in paleochannels in Paleozoic rocks and conglomerate that contains clasts as old as Cambrian, reflects the immediate area, but clast compositions suggest substantial relief at least locally between source and depocenter.

The lowest Vampire Formation sample was likely derived from proximal sources. Low Th/U grains match values of Mojave Desert plutons (Barth and Wooden, 2006), but many grains have Th/U > 1. We surmise that the mountains of the arc contained volcanic and plutonic deposits of varying ages but relatively similar compositions that were tapped by a river. This river then flowed east and north from the Vampire area to deposit grains in the Chinle Formation samples, an interpretation consistent with paleocurrents in the Sonsela Member of the Chinle Formation. The Chinle Formation grains, which range in age over ∼30 m.y. and in Th/U between 0.4 and 2, must have had multiple sources. For example, euhedral grains suggest that ash clouds played a role in carrying zircons from source to depositional site, and it is unlikely that these grains were from the same volcano as zircons carried by the stream system.

Proterozoic grains are very common in the Mustang Mountains sample of the Waterman formation; the majority of these are 1.4–1.46 Ga, with a smaller set of 1.6–1.65 Ga grains. Proterozoic and Paleozoic rocks are not exposed between southern Arizona and Caborca, Sonora (Papago terrane of Haxel et al., 1988, 2005; Stewart and Poole, 2002; Fig. 11), so the source of the Proterozoic grains is uncertain. Proterozoic plutonic rocks are common in southwestern North America outside of the Papago terrane, and zircons from them are known or likely in overlying Paleozoic strata (Stewart and Poole, 2002; González-León et al., 2005; Soreghan et al., 2007; Dickinson and Gehrels, 2008, 2009).

An interesting result of the proposed dispersal system is the requirement that two rivers with origins in the same part of the magmatic arc diverged, apparently immediately after leaving the highlands. This topographic setting is analogous to present-day rivers that head within a few tens of kilometers of each other in the Bolivian Andes, with the northern stream eventually joining the Amazon to reach the Atlantic Ocean in northern Brazil, and the southern stream a tributary of the Rio de la Plata, which flows into the Atlantic 4000 km to the south. The fate of the paleorivers that joined in southern Arizona remains uncertain.

CONCLUSIONS

Detrital zircons from three Upper Triassic sedimentary sections separated by several hundred kilometers are very similar in age, Th/U ratio, Hf content, and Yb/Gd ratio, leading to the hypothesis that the zircons were derived from magmas similar in time, space, and composition. Samples of the Chinle Formation on the Colorado Plateau are strongly dominated by grains ca. 205–230 Ma and Th/U ratios of 0.3–2. This pattern is mirrored in the lowest Vampire Formation sample in eastern California, although the youngest grains are older in this sample (ca. 215 Ma). The similarity of these grains in age and, in part, in Th/U to Triassic plutonic rocks in the Mojave Desert supports the interpretation that these rocks, or cogenetic volcanoes, were the source of some backarc material. More significantly, these patterns verify the correlation between the three units.

The Waterman formation, in contrast, has a tripartite division of grains. Group 1 is very similar to the other two samples, implying a cogenesis of the source rocks. Group 2 comprises several grains that are age equivalent to the Chinle and Vampire grains, but that have very different Th/U ratios, suggesting derivation from a different source. The third group consists of grains that are substantially older (245–280 Ma) than any grains in the other samples. These oldest grains are inferred to have derived from the early Mesozoic arc in Sonora, where plutonic remnants are older than those in the Mojave Desert. These data suggest that three major streams were sourced in the magmatic arc (Fig. 11): one flowed from plutonic sources through the Vampire depocenter and on to the present-day Colorado Plateau; one flowed from a source near the first, but southeastward to the Waterman depocenter; and the third flowed northeastward from the Sonoran arc. The added layer of distinguishing subgroups of Th/U ratios within the main sediment group, and comparing these subgroups to potential sources, suggests that multiple volcano-plutonic complexes were eroded and tapped by streams within the arc. Thus the added geochemical filter can provide important information to refine geologic interpretations.

Funding for this study was provided by the National Science Foundation through grants EAR 0711541 to Riggs and EAR 0711115 and EAR 0711119 to Barth. We are very grateful to Andrew Kylander-Clark, UCSB, for guidance through the LASS system and interpretation and presentation of results. Excellent reviews by Tim Lawton, Brendan Murphy, and Paul Stone helped us clarify our ideas and were extremely helpful. Ongoing and fruitful conversations with Ron Blakey and Zach Oberling are appreciated. Carmen Winn and Courtney Pulido did the mineral separations, and the warm hospitality of the Tiffney-Gowen family is much appreciated. Bill Dickinson first gave Riggs the invaluable advice to look at the Waterman Mountains rocks.

1Supplemental Table. Laser-ablation inductively coupled mass spectrometer results for sandstone samples from Petrified Forest National Park (Jim Camp Wash bed and Long Logs sandstone), Lowest Vampire Formation, and Waterman formation (Waterman Mountains, Mustang Mountains). Waterman Mountains sample divided into group 1 and group 2; see text for discussion. If you are viewing the PDF of this paper or reading it offline, please visit http://dx.doi.org/10.1130/GES00860.S1 or the full-text article on www.gsapubs.org to view the Supplemental Table.
2Supplemental File. PDF file containing a table of instrumental parameters of laser-ablation split-stream ICP-MS (inductively coupled plasma–mass spectrometry), information on the University of California Santa Barbara LASS (laser ablation split stream) procedure, and a comparison of LA-ICP-MS and SHRIMP (sensitive high-resolution ion microprobe) methods for zircon geochemistry. If you are viewing the PDF of this paper or reading it offline, please visit http://dx.doi.org/10.1130/GES00860.S2 or the full-text article on www.gsapubs.org to view the Supplemental File.