Oligocene onset of uplift and inversion of the Cascadia forearc basin, southern Oregon Coast Range, USA

The Cascadia margin of western North America is an archetypal active subduction zone where Eocene to Miocene pre-arc and forearc sedimentary and volcanic rocks are exposed in the slowly uplifting (~0.1 mm/yr; Kelsey et al., 1996) and eroding Oregon Coast Range (OCR) (Fig. 1). Although uplift of the OCR is recognized, the timing of initial uplift, spatial patterns of topographic growth, and geodynamic mechanisms responsible for them are not well understood. The onset of forearc uplift is difficult to address in the OCR because erosion has removed most upper Eocene to Miocene strata from the axis of the range (Fig. 1B). A useful existing constraint is provided by Columbia River Basalt that erupted from vents in eastern Oregon at ca. 16 Ma and flowed south to Salem and Newport (Fig. 1B; Beeson et al., 1989) and farther west based on offshore seismic and well data (McNeill et al., 2000). Substantial differences in the elevations of inboard, coastal, and offshore Columbia River Basalt (McNeill et al., 2000; Scanlon et al., 2021) record post–16 Ma warping of the pre–Columbia River Basalt surface, subsidence offshore, and uplift in the OCR. However, the scarcity and isolated occurrences of the southernmost Columbia River Basalt hint at an emergent, and likely dissected, topographic barrier in the OCR prior to ca. 16 Ma (Beeson et al., 1989), but the timing of earliest uplift and basin inversion is unknown.

To address this problem, we present a large (n = 6324) detrital zircon (DZ) U-Pb geochronology data set from southwestern Oregon (Fig. 1). Coeval production and delivery of primary magmatic zircon from the Cascades arc to the adjacent forearc make this an ideal setting for DZ studies, which are widely used for evaluating depositional ages, provenance, and patterns of orogenic exhumation (e.g., Dickinson and Gehrels, 2009; Surpless, 2015; Shekut and Licht, 2020). First, we established the age of the Ancestral Cascades arc (ACA) in southern Oregon using DZ samples from modern rivers that drain the Western Cascades range. We then integrated these with new bedrock DZ data and existing paleomagnetic and biostratigraphic data to establish depositional ages of Cascadia forearc strata near Coos Bay, and to evaluate their provenance. These data record Oligocene uplift, forearc basin inversion, and emergence of the OCR and provide insight into mechanisms that govern the structural and topographic evolution of forearc regions.

Prior to ACA magmatism, the Siletzia oceanic plateau formed offshore 56–49 Ma and accreted to the North American margin, causing a westward jump in subduction to the modern Cascadia trench by ca. 49–46 Ma (Wells et al., 2014). Docking of Siletzia was accompanied by a regional magmatic flareup ca. 52–40 Ma that produced the Clarno Formation in central and eastern Oregon (Bestland et al., 1999) and the Challis volcanic field in western Idaho (Gaschnig et al., 2010) (Fig. 1A). Siletzia accretion drove exhumation in the Klamath Mountains orogen and deposition of the lower Eocene Umpqua Group in a syn-collisional foredeep, followed by middle Eocene north-northwest progradation (modern coordinates) of the Tyee Formation delta system in western Oregon (Dott, 1966; Heller and Ryberg, 1983; Santra et al., 2013). The onset of arc volcanism in the ACA (Western Cascades) occurred after deposition of the Tyee Formation and is loosely constrained to ca. 45–35 Ma in Oregon (du Bray and John, 2011).

Strata in the Coos Bay area comprise an ~6.5-km-thick succession of lower Eocene to upper Miocene marine and marginal marine strata with extensive angular unconformities at the base of the Tyee Formation, Tarheel formation (informal name; Armentrout, 1967, 1981), and Empire Formation (Madin et al., 1995; McNeill et al., 2000). The composite stratigraphic section (Fig. 2) is well exposed in coastal cliffs along the steeply dipping western limb of the South Slough syncline (Fig. 1B; Walker and MacLeod, 1991; Madin et al., 1995; Wiley et al., 2015).

Detrital zircon were separated from 22 bedrock samples and three samples of modern river sand using standard procedures. Between 50 and 525 (n_mean = 283) unpicked (random) zircon grains from each sample were analyzed for U-Pb geochronology via laser ablation–inductively-coupled–plasma mass spectrometry (LA-ICP-MS) at the University of Arizona LaserChron Center (Tucson, Arizona, USA). DZ age spectra are displayed as kernel density estimates using a fixed, optimized bandwidth for each sample group. All errors are reported at 2σ. For bedrock samples, only concordant analyses that overlap concordia within error are shown (Fig. 2; Fig S1 in the Supplemental Material¹). Depositional age estimates exclude dates that do not overlap concordia or have unusually high effective uranium, eU (>1000 ppm). Additional details and raw data are provided in the Supplemental Material.

Many different procedures have recently been proposed for calculating maximum depositional ages (MDAs) from DZ data (Coutts et al., 2019; Copeland, 2020; Vermeesch, 2021). This study used three methods for calculating MDAs (see Figs. S2–S4): (1) the “youngest single concordant grain” (YSCG; Dickinson and Gehrels, 2009; Copeland, 2020); (2) the “maximum likelihood age” (MLA), a robust statistical estimate that converges to the actual or “true depositional age” (tDA) with increasing sample size (Vermeesch, 2021); and (3) the youngest cluster of three or more analyses whose dates overlap within 2σ error (YC2σ), identical to the “YC2σ(3+)” method of Dickinson and Gehrels (2009).

In forearc basins with a relatively continuous supply of zircon sourced from an adjacent volcanic arc, the youngest zircon in a sample may approximate the tDA of the host deposit (Dickinson and Gehrels, 2009; Cawood et al., 2012; Coutts et al., 2019; Copeland, 2020). Using synthetic DZ simulations, Coutts et al. (2019) showed that YSCG overlaps tDA within 2σ uncertainty in 97.4% of large-n (n >300) samples that contain abundant (>9%) near-depositional-age (nDA) dates. Following this precedent, we define nDA dates as those whose central age falls within 10% of the central age of the YSCG of the sample. Hence, for samples with relatively few nDA dates (<10%), the MLA is interpreted as the MDA; for samples with abundant nDA dates (≥10%), the YSCG may be interpreted as within error of the tDA (Table 1).

We dated DZ from modern point-bar deposits in the Middle Fork Willamette River, North Umpqua River, and upper Rogue River at locations where upstream bedrock sources consist exclusively of volcanic rocks of the ACA and transitional High Cascades (Fig. 1B). These data therefore directly constrain the DZ age signature of the ACA source area and the onset and timing of arc magmatism in southern Oregon.

Of the combined 457 DZ dates from these samples, virtually all (95.2%) are 40–12 Ma (Fig. 2). Sparse early to middle Eocene (0.7%), Cretaceous (0.9%), Jurassic (2.0%), and Precambrian (1.3%) zircon, mostly found in the southernmost river sample (20COA-2), are likely xenocrystic and may reflect inheritance from sub-arc basement. Other than these trace inherited dates, the oldest ages in this data set are ca. 40 Ma, consistent with the oldest published volcanic rock age in the southern half of the arc in Oregon (39.1 ± 0.7 Ma; du Bray and John, 2011). Together, these data sets show that ACA magmatism initiated ca. 40 Ma in southern Oregon.

The composite DZ age spectrum shows a bimodal distribution with discrete age modes at 34 and 25 Ma (Fig. 2; Fig. S5). The absence of DZ dates <12 Ma reflects localization of late Miocene to recent magmatism in the High Cascades (e.g., du Bray and John, 2011), which lie east of and mostly outside the watersheds upstream of our river samples (Fig. 1B).

For all bedrock samples, the youngest zircon dates (YSCGs) and MDAs are progressively younger upsection (Table 1; Fig.2), as expected in forearc settings (Cawood et al., 2012). We determine an MDA of 52.5 ± 0.9 Ma for the Fivemile Point Sandstone (Fig. 2) from DZ data reported by Wiley et al. (2015). Abundant nDA dates (22.3%) in the upper Tyee Formation (sample 20COA-14; Fig.1) indicate a tDA of 45.4 ± 1.6 Ma in the Coos Bay area, slightly younger than prior estimates of ca. 49–47 Ma (Wells et al., 2014). The age of the beds of Sacchi Beach was previously unconstrained (Bird, 1967). We calculate a DZ-based MDA of 45.0 ± 0.4 Ma for the oldest exposed beds of Sacchi Beach near Coos Bay, supporting a lithostratigraphic correlation to similar slope facies of the middle Eocene Elkton Formation (Bird, 1967).

Abundant nDA dates in the Coaledo Formation and Bastendorff Shale provide robust constraints on the tDA for these units (Table 1). A tDA of 40.9 ± 1.1 Ma for the top of the lower member of the Coaledo Formation suggests correlation to chron C19n or C20n, and shifts the normal-polarity base of the Coaledo Formation to chron C20n at 43.4–42.3 Ma (Prothero and Donohoo, 2001a; Ogg, 2012). These data suggest maximum age ranges of ca. 43.4–38.8 Ma for the Coaledo Formation and 38.8–33.5 Ma for the Bastendorff Shale (Table 1), consistent with their Narizian and Refugian benthic faunas, respectively (Rooth, 1974; McDougall, 2008). Samples from the base and top of the Tunnel Point Sandstone have identical MDAs of ca. 33.5 Ma, consistent with paleomagnetic correlation to chron C13n or C12n (Prothero and Donohoo, 2001b) and sparse molluscan fauna suggesting correlation to the 34–30 Ma Eugene Formation (Retallack et al., 2004). These data suggest a maximum age span of ca. 33.5–30 Ma for the Tunnel Point Sandstone.

The middle and top of the Tarheel formation yield DZ MDAs of 24.5 ± 0.6 and 18.1 ± 0.4 Ma, respectively, consistent with correlation of marine fauna to the Newportian molluscan stage (22–15 Ma; Moore and Addicott, 1987) and the Saucesian foraminiferal stage (22–17 Ma; McDougall, 2008). We infer that the Tarheel formation was deposited at ca. 18–15 Ma, although a maximum age of ca. 22 Ma is permissible based on available constraints. We calculate a DZ MDA of 15.9 ± 0.5 Ma for the base of the Empire Formation and an eruption age of 8.2 ± 0.1 Ma for a prominent tuff in the middle Empire Formation (Fig. S2). These data are consistent with its assignment to the Wishkahan molluscan stage (11.5–8.5 Ma) and North Pacific diatom zones XI–X (8.5–5.0 Ma) (Armentrout, 1981; McDougall, 2008) and suggest deposition of the Empire Formation sometime between 11.5 and 5 Ma.

All Coos Bay strata contain abundant late Jurassic to Cretaceous zircon (33.5%–72.9%; Fig.2) that correspond to igneous and sedimentary source-rock ages in the Klamath Mountains (175–130 Ma) and northern Sierra Nevada arc (130–85 Ma) (Surpless, 2015, and references therein), demonstrating those sources were major primary or secondary source areas. Middle to early late Eocene units (Tyee, Sacchi Beach, Coaledo) contain substantial populations (17.8%–27.3%) of coeval 52–40 Ma zircon (Fig. 2) that correspond to Clarno-Challis volcanic centers east of the Cascades (Figs. 1A, 3A, and 3B). The late Eocene Bastendorff Shale shows a decrease in Clarno-Challis zircon (8.9%) at the expense of new input of ca. 40–34 Ma zircon from the ACA (9.5%), followed by a notable reduction in Clarno-Challis (3.5%) and ACA zircon (2.0%) in the Oligocene Tunnel Point Sandstone (Figs. 2 and 3B).

The late Oligocene–early Miocene unconformity below the Tarheel formation marks a distinct change in provenance characterized by a scarcity of 40–12 Ma zircon from the ACA (0.6%–2.9%) and by the reappearance of a significant proportion of ca. 52–40 Ma Clarno-Challis zircon (8.8%–8.9%; Figs.2 and 3C). The relative abundance of northern Sierra Nevada versus Klamath Mountains ages changes upsection from relatively more Klamath Mountains ages at the base to relatively more Sierra Nevada arc ages in the Tunnel Point Sandstone, and back to relatively more Klamath Mountains ages in the Tarheel and Empire formations (Fig. 2; Fig. S6).

Data presented here constrain the timing of uplift in the southern OCR. The angular unconformity at the base of the middle Miocene Tarheel formation and correlative Astoria Formation along the central Oregon coast (Madin et al., 1995; McNeill et al., 2000) records pre–middle Miocene deformation and erosion (Beeson et al., 1989). The scarcity of ACA (40–12 Ma) zircon in the Tunnel Point, Tarheel, and Empire Formations (Fig. 2) suggests that the Coos Bay area has been isolated from rivers that drain the ACA since early Oligocene time (Figs. 3C and 3D). This is also evident in the absence of nDA dates in those units and their contrasting abundance in most underlying units (Table 1), which implies a fluvial connection between the ACA and Coos Bay only during late Eocene time (Fig. 3B). The decrease of Clarno-Challis (ca. 52–40 Ma) zircon in the Bastendorff Shale and Tunnel Point Sandstone suggests that the early ACA formed a topographic barrier that restricted sediment flux from eastern Clarno-Challis sources by latest Eocene to Oligocene time (Figs. 2, 3B, and 3C).

We interpret the decrease of ACA zircon in the Tunnel Point Sandstone to signal incipient isolation of the Coos Bay area from the ACA due to the onset of forearc basin inversion and uplift at ca. 33–30 Ma (Figs. 2 and 3C). The subsequent increase of Clarno-Challis detritus and shift to more Klamath Mountains relative to northern Sierra Nevada arc ages in the overlying Tarheel and Empire formations is best explained by erosion and recycling of underlying Eocene strata and transport of recycled sediment into the Coos Bay area by the ancestral Coos and Coquille Rivers (Figs. 1B and 3D; Fig. S6). Therefore, we attribute the disappearance of ACA zircon and the simultaneous reappearance of Clarno-Challis zircon in the Tarheel formation to uplift and exhumation of the previously subsiding forearc basin and emergence of the OCR between early Oligocene and early Miocene time (30–18 Ma).

In Cascadia, other evidence for the onset of forearc uplift and emergence above sea level is sparse. DZ data suggest that crustal thickening and uplift started as early as late Oligocene time, but with delayed emergence of the Olympic Peninsula (Washington) until after ca. 16 Ma (Shekut and Licht, 2020). In contrast, Columbia River Basalt flows that unconformably overlie folded Paleogene rocks in the Willapa Hills (Washington) and Nehalem, Portland, and Tualatin Basins (Oregon and Washington) imply uplift and erosion before ca. 16.5 Ma in southwestern Washington and northwestern Oregon (Wells et al., 2020; Scanlon et al., 2021).

Thermomechanical models identify several processes that drive rock uplift in forearcs over geologic time scales. Forearc highs typically are produced by crustal thickening due to tectonic underplating or subcretion at the base of the accretionary wedge (Menant et al., 2020) (Fig. 3D). Uplift can also be driven by the insulating effect of crustal thickening and forearc sedimentation, which promote thermally activated viscous flow in the lower crust (Pavlis and Bruhn, 1983; Fernández-Blanco et al., 2021). These mechanisms can generate significant forearc topography (>1 km elevation) after only ~15–20 m.y. of subduction (Menant et al., 2020; Fernández-Blanco et al., 2021), similar to the time span between Cascadia subduction initiation (49–46 Ma) and the onset of forearc uplift at ca. 33–30 Ma proposed herein. Sediment supply thus exerts a first-order influence on the structural and topographic evolution of forearcs both by providing material to the trench for subcretion and by insulating the crust and promoting ductile flow. Along-strike variability in the volume of subcreted material can explain patterns of uplift and forearc topography along parts of the modern Cascadia (Delph et al., 2021) and southern Anatolia margins (Fernández-Blanco et al., 2021). We postulate that space-time variability in forearc uplift at Cascadia and other subduction margins may be governed by along-strike variations in sediment supply and related positive feedbacks between sedimentation, deformation, and crustal recycling.

Field work for this study was conducted on the traditional homelands of the Coos, Coquille, Kalapuya, Modoc, Molalla, Siltez, Takelma, Umpqua, and Yoncalla peoples. We thank the Geology Department Fund of the University of Oregon Foundation and the University of Nevada, Reno, College of Science for generous funding support. We greatly appreciate field assistance from David Blackwell, Noel Blackwell, Lucy Walsh, and Charlie Ogle, and analytical assistance from Mark Pecha, Martin Pepper, and Sarah Evans at the University of Arizona LaserChron Center (Tucson, Arizona, USA). Editorial handling by Jerry Dickens and constructive reviews from Mark Brandon, Kathleen Surpless, and Rebecca VanderLeest improved the quality of the manuscript.