As both a driver and a response to landscape evolution, river networks evolve in conjunction with the dynamic interplay of regional tectonic and climatic forces, commonly resulting in cascading effects to sedimentary and biologic systems. For example, steep topographic gradients and relatively rapid uplift or incision in mountainous terrain correlate with increased species richness through time (Badgely et al., 2017; Smiley et al., 2024). The rates and underlying mechanisms responsible for topographic uplift and river incision can vary temporally, such as with changes in fault slip rates, climate, sediment load, or base level (Kirby and Whipple, 2001; Sklar and Dietrich, 2004; Wobus et al., 2010). Steep and deeply incised fluvial corridors and canyon systems commonly signal recent landscape evolution, as the juxtaposition of subdued relief on elevated plateaus with steep marginal topographic gradients are a hallmark of transient topography. Although landscapes are generated by a complex interplay of various mechanisms, the spatial and temporal patterns of river incision can help decipher the dominant processes in landscape evolution (e.g., Clark et al., 2006; Cook et al., 2009; Darling and Whipple, 2015; Crow et al., 2014).
The topography along major river corridors throughout the northern U.S. Cordillera and Pacific Northwest shows evidence for recent incision, such as along the Columbia, Snake, Clearwater, and Salmon rivers. Here, diverse datasets generated from incision markers, cosmogenic basin-wide erosion rates, landscape evolution modeling, thermochonology, and cave-trapped fluvial sediments have been used to estimate the onset and rate of uplift and incision through time (Meyer and Leidecker, 1999; Ely et al., 2012; Vogl et al., 2014; Larimer et al., 2019; Mitchell and Yanites, 2019; O’Connor et al., 2021; Mitchell et al., 2023; Pivarunas et al. 2024, Morriss et al., 2025). Although these rates provide critical controls on landscape evolution, they do not provide information on the routing, provenance, and reorganization of rivers through time. Yet, when paired with sedimentary provenance analysis (e.g., Beranek et al., 2006; Staisch et al., 2021), these complementary datasets can help resolve long-standing controversies on middle Miocene to Quaternary incision and drainage patterns of these fluvial systems and provide insights into possible drivers, such as the Yellowstone hotspot migration or regional and global climate change (Zachos et al., 2001, 2008; Kasbohm and Schoene, 2018; Love et al., 2023a).
In this work, we focus on the history of the rivers that incised Hells Canyon, the deepest river gorge in North America. The modern Snake River flows through this canyon, linking the low gradient Snake River Plain in Idaho with the Columbia basin in Washington and Oregon. Several hypotheses have been posed on how and when the Snake River incised through Hells Canyon, and where it flowed before (Lindgren, 1898; Livingston, 1928; Wheeler and Cook, 1954; Taylor, 1960; 1985; Repenning et al., 1995; Smith et al., 2000; Reidel and Tolan, 2013; Stearley and Smith, 2016; Staisch et al., 2021); nearly all assert that a throughgoing Snake River in Hells Canyon is a Pliocene or Quaternary development which led to enhanced Quaternary incision. To assess this concept, we analyze the stratigraphy, fluvial clast compositions, and detrital zircon provenance of the middle to late Miocene Clarkston Heights gravel, deposited downstream of Hells Canyon in the Lewiston basin. We also model detrital zircon provenance of coeval sediments upstream of Hells Canyon, within the western Snake River Plain, for comparative analysis of zircon sources. Our analysis shows that Hells Canyon was an active fluvial corridor linking the western Snake River Plain and Columbia basin since at least ~6 Ma, and maybe as far back as ~10 Ma. These results directly oppose hypotheses that the Miocene-Pliocene Lake Idaho, occupying the western Snake River Plain, had outflow into southern Oregon or northern California prior to a drainage capture event (Livingston, 1928; Wheeler and Cooke, 1954; Taylor, 1960; 1985; Repenning et al., 1995; Smith et al., 2000; Reidel and Tolan, 2013). They also complicate understanding of the rapid recent incision of Hells Canyon documented by Morriss et al. (2025), similarly attributed to drainage capture. We reconcile our new evidence for long-term connection with regional stratigraphy, incision, and biologic evidence to suggest an intermittent but sustained fluvial corridor between the western Snake River Plain and Columbia basin via Hells Canyon and an associated two-phased incision history driven by a downstream cascade of drainage basin changes, sedimentation, and outlet conditions of lacustrine systems in the western Snake River Plain.
Along its pathway between the Rocky Mountains and the Pacific Ocean, the Snake River flows through Hells Canyon, a rugged and deeply incised landscape that connects the Snake River Plain to the Columbia basin (Fig. 1). The canyon system is located near several major structural discontinuities. Just east of Hells Canyon, the Western Idaho Shear Zone (WISZ) forms the boundary between the North American craton (Laurentia) and several accreted terranes, defined by strong north-south oriented shear structures and the 87Sr/86Sr 0.704 and 0.706 isopleth (Fig. 1). The Wallowa Mountains, Blue Mountains, Olds Ferry, and Baker terranes were a series of island arcs that accreted in the Mesozoic as the Farallon plate subducted beneath the western margin of Laurentia (Avé Lallemant, 1995; Vallier, 1995; Schwartz et al., 2011, 2014; Kurz et al., 2012). Within the Hells Canyon region, exposed basement rocks are primarily the Permian to Triassic Seven Devils Group (part of the Wallowa terrane), composed of intermediate to felsic volcanic rocks, clastic and carbonate units, and intermediate-composition intrusive rocks (Vallier, 1977). These units underwent greenschist to amphibolite facies metamorphism ca. 244 Ma prior to the amalgamation of island arcs, displaying retrograde chlorite, and were later intruded at 227 and 145 Ma (Snee et al., 1995; Vallier, 1998). Suturing between the accreted terranes and Laurentia occurred by Cretaceous time, as suggested by a series of metamorphic events of increasing pressure between 130 and 101 Ma, culminating in suture-zone metamorphism and pluton emplacement between 93 and 90 Ma (Snee et al., 1995). Unconformably capping the accreted metamorphic rocks in Hells Canyon are lava flows of the Columbia River Basalt Group, a Miocene large igneous province that predominantly erupted from dikes near the Wallowa Mountains (Fig. 1). Given the exposed geology, the fluvial deposits within and downstream of Hells Canyon contain abundant metamorphic and mafic volcanic clasts (Webster et al., 1982).
East of the accreted terranes, the Bitterroot and Atlanta lobes of the Idaho Batholith were formed by arc magmatism from the Farallon subduction zone and were largely emplaced into the Precambrian basement rocks along the western margin of Laurentia (Fig. 1). The main phase of batholith intrusion occurred between 98 and 67 Ma to create the Atlanta lobe (Gaschnig et al., 2010) and continued between 66 and 53 Ma north to create the Bitterroot lobe (Foster and Fanning, 1997; Foster et al., 2001, 2007; Gaschnig et al., 2010; 2017). The younger Eocene Challis volcanic and intrusive suite (53-43 Ma) are of a more diverse geochemical assemblage and represent a later phase of magmatism (Moye et al., 1988). These plutonic units are exposed predominantly east of the WISZ and north of the western Snake River Plain (wSRP), a landscape that is currently incised by the Clearwater and Salmon rivers, as well as tributaries that enter the Snake River upstream of Hells Canyon, such as the Payette and Boise rivers (Fig. 1). Precambrian basement exposures in the Salmon and Clearwater rivers are common and thus the contributing geologic landscape is diverse in age and composition including sandstone, siltstones, and carbonates that have been metamorphosed to amphibolite facies to produce augengneiss, orthogneiss, pelitic schists, and amphibolites (Webster et al., 1982). The bedrock exposed in tributaries located north of the western Snake River Plain is dominated by the Idaho batholith.
Middle Miocene extension across southwestern Idaho was accommodated along north-south oriented normal faults related to the Oregon-Idaho graben (OIG; Fig.1; Cummings et al., 2000). Extensional basin-fill strata include the ~15.5 Ma Succor Creek Formation (sometimes published as Sucker Creek formation) and are composed of tuffaceous sandstone, siltstone, and arkosic channel sands (Kimmel, 1982; Lawrence, 1988; Ferns et al., 1993). Deposition within the OIG continued until ~10.5 Ma (Cummings et al., 2000), a time when the Yellowstone hotspot was erupting the proximal Owyhee-Humboldt and Bruneau-Jarbidge calderas (Fig. 1). The mapped extent of the Succor Creek Formation is generally confined to the OIG, suggesting that middle Miocene subsidence of the graben facilitated clastic deposition (Fig. 2)
The NW-trending extensional basin associated with the present topography of the western Snake River Plain cross-cuts the Oregon-Idaho graben. The strata associated with the OIG are not exposed within the NW-trending Snake River Plain, suggesting that the western Snake River Plain developed after ~10.5 Ma (Kimmel, 1982; Cummings et al., 2000). Offset volcanic units in the Bennett Mountains and the Owyhee Mountains, on the northern and south of the western Snake River Plain, respectively, suggest that most fault-accommodated extension occurred by 9.5–9 Ma followed by much lower rates of displacement (Fig. 1; Wood and Clemens, 2002). Across the western Snake River Plain, there is a >100 mGal gravity anomaly (Mabey, 1982) as well as a high velocity layer at ~8 km depth that contrasts strongly with the slower crustal velocities to the north and south (Wood and Clemens, 2002). Structural extension across the graben bounding faults does not entirely account for the variable geophysical properties, suggesting the mid to lower crust of the western Snake River Plain was invaded by mafic intrusions (Wood and Clemens, 2002). The increased density of the western Snake River Plain mid to lower crust may have contributed to subsidence and the establishment of Lake Idaho, the lacustrine system that formed within the basin throughout Miocene and Pliocene time.
Lake Idaho, poetically described by Clarence King (1903, p. 229) as a “wide and beautiful expanse of water” but more formally and geologically described by Malde and Powers (1962), was a large, deep, lacustrine system between middle Miocene and earliest Pleistocene time. Deep phases of the lake existed over two main time periods, recorded by the Chalk Hills and Glenns Ferry formations. The Poison Creek Formation is an early fluvial facies of the Chalk Hills Formation and represents the onset of deposition within the Miocene-Pliocene Lake Idaho system (Fig. 2; Wood and Clemens, 2002). The maximum depositional age of this 250-500 m thick Gilbert-style deltaic strata is constrained by the ~11.4 Ma Idavada Volcanics (Smith et al., 1982; Wood, 1994). U-Pb zircon ages from interbedded volcanic units are between 10.7–9.9 Ma (Love et al., 2023a). Chalk Hills Formation deposition continued until about ~6 Ma and includes fluvial, floodplain, and lacustrine facies (Kimmel, 1979). The maximum lake extent during Chalk Hills time is estimated based on local stratigraphy and fossil fish localities (Fig. 2). Near the western Snake River Plain basin margins, the Chalk Hills strata are faulted and tilted towards the basin center, indicating that local tectonic subsidence continued after ~6 Ma. The strata are dissected by an erosional surface, indicating substantial lake regression into deeper basin remnants (Wood and Clemens, 2002). Detrital zircon provenance analysis of several samples of Chalk Hills Formation indicates that it was sourced from local drainages of the western Snake River Plain, but lacks the older zircons typically sourced from the eastern Snake River Plain and western Wyoming (Staisch et al., 2021).
Unconformably overlying the Chalk Hills Formation is the Glenns Ferry Formation (Malde and Powers, 1962). The Glenns Ferry Formation is composed of deep-water lacustrine and low-energy fluvial sediment deposited mainly between ~4.3–2.5 Ma but continuing on locally as late as 2.18 Ma (Malde and Powers, 1962; Malde, 1991; Repenning et al., 1995). Deposit ages indicate a >1 Myr depositional hiatus between the Chalk Hills and Glenns Ferry formations, yet fish-fossil similarities among the formations suggest that the lacustrine system did not fully disappear from the western Snake River Plain (Smith et al., 2000). Local to the Boise foothills, the Terteling Springs formation is a ~120 m thick set of oolitic sand lenses and lacustrine muds between the Chalk Hills and Glenns Ferry formations, interpreted as a transgressive lake sequence (Swirydczuk et al., 1980a, 1980b; Wood and Clemens, 2002). Though locally differentiated, we consider the oolitic Terteling Springs and broader Glenns Ferry formations as evidence for a renewed lake filling of the western Snake River Plain. Compared to the underlying Chalk Hills strata, the Glenns Ferry lacustrine sediments are more widespread and lie at higher elevations, up to 1,100 m elevation, than the Chalk Hills lake strata. However, the basin geometry apparently was little changed among both phases of Lake Idaho (Fig. 2; Kimmel, 1979). In contrast with the earlier Chalk Hills lacustrine phase, the detrital zircon U-Pb age spectra from the Glenns Ferry Formation indicate that it contains sediment derived from the eastern Snake River Plain and upper portion of the modern Snake River drainage (Staisch et al., 2021). The more extensive lacustrine system of the Glenns Ferry Formation is likely the consequence of Pliocene eastward growth and expansion of the Snake River drainage contributing to Lake Idaho. The mechanism for ~6–4.3 Ma lake regression and later transgression separating the deeper lake phases is not clear (Wood and Clemens, 2002) but proposals include vertical tectonics involving the outlet, climate change, outlet erosion, and episodic volcanism and blockage near the outlet (Kimmel, 1982).
The lacustrine and low-energy fluvial environmental setting of Lake Idaho transitioned to a throughgoing gravel-transporting fluvial system in the late Pliocene – early Pleistocene. At this time, major rivers draining into western Snake River Plain began to prograde coarse sand and gravel deposits across the broad and relatively flat basin. Mapped as the Tuana and Tenmile gravels, these post-Glenns Ferry Formation units are the first major evidence for a Snake-River-style fluvial system traversing the western Snake River Plain (Malde, 1991). The upstream exposures of the Tuana Gravel likely post-date a 2.18 ± 0.18 Ma basalt flow conformably atop of Glenns Ferry Formation and may be as young as 1.92 Ma (Amini et al., 1984; Malde, 1991). Tuana Gravel remnants are exposed ~245 m above the Snake River (~991–1,058 m asl) near Mountain Home and Hagerman, Idaho (Fig. 2; Amini et al., 1984; Malde, 1991), The Tuana Gravel grades downstream to similar elevations as the younger Tenmile Gravel, which lies ~120 m above the modern river (~761–816 m asl) near Nampa and Ontario (Fig. 2; Malde, 1991). Incision into the Tenmile Gravel preceded emplacement of the 1.58 Ma Pickles Butte rim basalt (Othberg, 1994). Together, these data indicate that some combination of lake-level lowering and sediment infilling triggered a change from low-energy fluvial and lacustrine conditions of the Glenns Ferry Formation time to a through-going gravel-transporting river at about 2.2 Ma, likely as a consequence of progradation of the Snake River across the basin (Malde, 1991; Othberg, 1994; Wood and Clemens, 2002). Prior to incision, the broad braided floodplain was initially as high as 245 m above the modern Snake River. After 1.6 Ma, the system incised a further ~120 m towards the modern river level.
As the Snake River flows northward out of Hells Canyon, it enters the Lewiston basin where it joins the Clearwater River and turns west toward its confluence with the Columbia River. The Lewiston basin, bound to the north by the Clearwater escarpment, has been a persistent topographic and structural low since at least Miocene time (Fig. 3; Kehew, 1979). Between 16 and 6 Ma, lavas of the Columbia River Basalt Group (CRBG), erupting from the northwest trending Chief Joseph dike swarm, entered the Lewiston basin from the southwest, south, and east (Figs. 1, 3). The distribution and thickness patterns of these flows and associated fluvial deposits provide evidence of the late Miocene structural and topographic evolution of the Lewiston basin.
The ~15.9 Ma Priest Rapids Member of the Wanapum Basalt of the CRBG (Fig. 3; Mv(wpr); Kasbohm et al., 2023) erupted southeast of the Lewiston basin, producing extensive sheet flows that moved westward across the Snake and Clearwater River systems. The flows thicken slightly in the Lewiston basin region (Hooper et al., 1985), suggesting the region was a slight topographic low, but that relief was relatively subdued across a broad region. Overlying the Priest Rapids Member, the Miocene Latah Formation (Fig. 3; Mc(l)) is predominantly lacustrine sediments, possibly indicative of impounded drainage within the Lewiston basin (Camp, 1976). Latah strata are thickest near Tammany Creek (Reidel and Fecht, 1987; Fig. 3). Similarly, the overlying ~13 Ma Asotin and Wilbur Creek members of the Saddle Mountains Basalt (Fig. 3; Mv(sl); Tolan et al., 1989) are thick in this region and apparently followed a developing east-west oriented synclinal axis in the Lewiston basin now occupied by Tammany Creek valley just south of Lewiston, Idaho.
The emplacement pattern of the overlying 11.21 ± 0.21 Ma Pomona Member (Mv(sp)) and 10.7-10.2 Ma Elephant Mountain Member (Mv(sem)) flows (Figs. 3 & 4A; ages from Barry et al., 2013) differ substantially from the underlying basalt. The Pomona Member is the first major canyon-filling member in this region, possibly sourced just east of Tammany Creek (Kauffman et al., 2009). Intracanyon flows along a northwest striking paleocanyon immediately southwest of Clarkston, Washington, show that a deep river canyon was established before the 11.2 Ma Pomona Member flowed into the canyon system near the modern Snake River pathway. Here, the basalt filled the canyon to depths up to 125 m in the Clarkston and Lewiston areas (Swanson et al., 1980). The overlying Elephant Mountain Member erupted from the Chief Joseph dike swarm southwest of the Lewiston basin (Reidel et al., 2013). Map patterns show that it flowed over a broad swath of landscape from the highlands southwest of the Lewiston basin and became a canyon-filling flow as it entered the ancestral Clearwater and Snake River systems (Figs. 3 & 4B; Kauffman et al., 2009) and then, similar to the Pomona Member, flowed far down the Snake River valley and into the Columbia basin. The Pomona Member, Elephant Mountain Member, and Clarkston Heights gravels are not mapped along the modern Clearwater River corridor, suggesting that it was not at its current location in the middle to late Miocene. Though it is possible that these rocks were eroded away from the modern Clearwater River valley, the greater thickness of the Latah Formation in Tammany Creek and the distribution of the Pomona Member exposed in the Lapwai Creek drainage instead indicate the Clearwater River was about 8 km south of its current path between 11.2 and 14 Ma, following the synclinal trough now occupied by Tammany Creek (Fig. 3; Kehew, 1979; Kauffman et al., 2009).
The youngest local flow of the CRBG followed river courses similar to the modern river configuration. The ~6 Ma Lower Monumental Member basalt flow is typically exposed near river level along the Snake and Clearwater rivers, as well as along Asotin Creek (Figs. 3 & 4D; Hooper et al., 1985; Soderberg et al., 2024). Several dike sources of the Lower Monumental Member are mapped in the Snake River Canyon, upstream of the Grande Ronde River confluence, suggesting it flowed into the Lewiston basin from the south (Camp, 1976, Kauffman et al., 2009). Faulting along the Clearwater fault system may have shifted the Clearwater northward, closer to its current position (Kehew, 1979).
Along the Clearwater River and Asotin Creek, two gravel units are deposited along similar pathways marked by the Lower Monumental Member. The Clearwater gravels (Fig. 3; Mc(cl)) are deposited on top of the Lower Monumental Member and are therefore younger than ~6 Ma. They are not exposed downstream of the Snake-Clearwater confluence, suggesting that downstream Snake River canyon filled with the Lower Monumental Member causing aggradation in the lower Clearwater River valley. This gravel unit marks a similar trend of the Miocene Clearwater as the Lower Monumental Member which, according to the mapped extent, back-flooded 4 km up the Clearwater River valley from the Snake River confluence (Fig. 4). Deposited below and above the Lower Monumental Member, the Asotin Creek gravels (Fig. 3; Mc(ag)) are dominantly composed of basalt and mark a paleo-channel derived from similar basaltic terrain as the modern Asotin Creek (Fig. 4).
Overlying the Pomona and Elephant Mountain members, the Clarkston Heights gravels are deposited along the northwest oriented paleo-canyon trending along the modern Snake River pathway and partly underlie the Clarkston Heights suburb of Clarkston, Washington (Figs. 3 & 4B). The gravel is coarse, well-rounded, clast supported, and composed of diverse lithologies, together indicative of a high-energy big-river fluvial environment. Previous work on clast lithology of the Clarkston Heights gravels identifies a large component of metamorphosed volcaniclastic rocks similar to the Seven Devils Group, indicating a component of the sediment is sourced from the Salmon and Snake River canyons (Waggoner, 1981). The gravel lithology is distinct from that of the Clearwater gravels, which are composed of light-colored quartzites and granodiorites and lack the abundance of darker toned metavolcanics and metasedimentary rocks (Lupher, 1945).
A key inference that we develop from the stratigraphy and map patterns in the Lewiston basin is that the Clarkston Heights gravel post-dates the 10.2 Ma Elephant Mountain Member and pre-dates the ca. 6 Ma (Soderberg et al., 2024) Lower Monumental Member flow. This interpretation derives from (1) the overlying relation of Clarkston Heights gravel to the Elephant Mountain Member and (2) the geomorphic position of the Clarkston Heights gravel following a paleo-river course that pre-dates the incised valley filled by the Lower Monumental Member. This interpretation seems secure given the geographic and stratigraphic positions yet is in opposition to previous interpretations by Kehew (1979), Webster e al. (1982) and Othberg et al. (2003) inferring that the Clarkston Heights gravel was contemporaneous with the post-6 Ma Clearwater gravel overlying the Lower Monumental Member. Their correlation is premised on the similar aggradation levels reached by both gravel units; however this assessment does not account for the stratigraphic and geomorphic relations that we consider stronger evidence for depositional age. The 10.2–6.0 Ma age inference for the Clarkston Heights gravel is critical for our provenance assessment because it allows us to compare with the similar-age Chalk Hills Formation in the western Snake River Plain.
To establish baseline measurements of clast compositions derived from regional fluvial networks, we collected gravels from the Snake, Salmon, and Clearwater River systems. For each sample, 30-40 kg of material was collected and sieved to separate out clasts >22 mm in size. These clasts were sorted by lithologic categories defined from broad survey of fluvial gravel throughout the Pacific Northwest major river systems (O’Connor et al., 2021), including: Felsic, intermediate, mafic, or glassy volcanic, felsic or intermediate phanerite, welded tuff, metamorphic, quartzite, rose quartzite, greenish metabasalt, greenish metasediment, greenish phanerite, clastic, carbonate, chert, silicified wood, autochthonous. Not all categories were present in the ancestral and modern gravel samples considered here.
The late Pleistocene–Holocene river clast compositions are shown in Figure 5 and are reported in Table S1. From these data, we identify that the Snake River at Pittsburg landing contains a large fraction of metamorphosed sediments and mafic volcanics, likely eroded from the accreted terranes and CRBG strata, respectively, exposed along Hells Canyon. The Salmon River sampled near Copperville also contains a substantial fraction of greenish metasediment and mafic volcanics but has an overall more diverse range of gravel and cobbles including the largest fractions of felsic volcanics, intermediate volcanics, greenish phanerites, and quartzites of the sampled rivers. The modern Clearwater was sampled at two locations, both of which show a majority of clasts are mafic volcanics. At the confluence between the Snake and Clearwater Rivers, the fraction of CRBG clasts is mildly diluted. This trend carries further downstream at the Lower Granite Dam site, where mafic volcanic clasts compose <50% of the gravels and there is an increase in the relative abundance of greenish metamorphic compositions.
In this survey, we included a single large-N sample (for which many clasts were analyzed) of the Clarkston Heights gravels collected from the upper exposed stratigraphy (Fig. 3). Here, 342 clasts were identified and show strong compositional similarity to the modern Snake-Clearwater confluence sample (Fig. 5). Although this uppermost sample suggests that the Clarkston Heights gravels have a similar source compared to the modern Snake-Clearwater confluence, we develop more detailed assessment of stratigraphy, compositional change, and provenance below.
The Clarkston Heights gravels are well exposed in quarries along a paleochannel southwest of Clarkston, Washington (Fig. 3). We examined several quarry exposures to log the stratigraphy and clast compositions, measure flow direction, and sample for age control and provenance. Most of the detailed stratigraphy was logged from a large quarry exposure near Dry Creek, however lower strata are exposed in a smaller quarry to the northeast (yellow stars on Fig. 3; Table 1). For both quarries, we measured stratigraphy using a Jacobs staff with decimeter scale markings and a laser range finder for leveling and thickness measurements along steep and unconsolidated hillslopes. To measure flow direction, we focused on layers with strong imbrication fabric with multidimensional exposure where we could confidently identify the 3D geometry of sedimentary forms. The long axis of each clast was identified, and the direction of maximum inclination was measured with a Brunton compass. Twenty-five to thirty clasts were measured for each discrete layer. In addition to flow direction, we also categorized clast composition for each cobble to aid in identifying any compositional changes over the stratigraphic unit. Typically, 25-30 clasts were measured and identified per layer and are shown in Fig. 6 as rose diagrams colored by clast composition.
The basal contact between the Clarkston Heights gravels and underlying CRBG strata is at 231 m elevation near the mouth of the paleo-channel. This strath terrace is only 4 meters higher than the nearby modern Snake River. Above this contact and exposed well in the northwest quarry, the lowermost Clarkston Heights strata are dominantly conglomerate with a fine to medium grained sandy matrix. The typical clast size is between gravel and cobble (Fig. 6A). Individual gravel packages ranged from ~2 to 5 meters thick, commonly with ~1 m tall foresets and imbricated rounded cobbles (Fig. 6B). Here, we sampled sand within the lower exposures for U-Pb zircon dating (19LMS0730-01) and logged clast compositions and flow direction ~30 meters above the strath contact. Imbrication shows north-directed flow and clasts were mainly mafic volcanic (basalt) and greenish metabasalts and metasediments (Fig. 6) typical of Mesozoic accreted terranes exposed within Hells Canyon.
The upper stratigraphy was measured and sampled at the southwestern quarry (Fig. 3). Here, the lower exposed deposits are similar to the strata exposed in the northeastern quarry, chiefly conglomeratic strata deposited in tabular beds that typically fine upward. Maximum clast size is approximately 20 cm. Two meters above the quarry floor (81 m in the stratigraphic section), we sampled fine-medium grained micaceous sand for U-Pb zircon dating (21LMS0511-01) and measured north-directed paleocurrent indicators (Fig. 6A). The gravel here contained more quartzite and clastic cobbles than the sample from the lower quarry, but measurements higher in the section (meter 86) were similar to that in the northeastern quarry (meter 30). Above the ~91 m level (Fig. 6A and 6C), clast size decreased and interbedded sand lenses become more common and thicker within a sequence of repeated fining up packages. Cobble imbrication and cross-bedded sand continues to indicate northward flow direction in what we infer as a thick aggradational sequence of a large-river channel-facies deposit. Given the lack of evidence for depositional hiatus, we infer these strata were rapidly deposited until an abrupt change in depositional environment marked by stratigraphic changes at meter 103 (Fig. 6A and 6D).
Within the upper 20 meters of the measured section, exposed in the southwestern quarry, there are three stacked sedimentary layers that are each indicative of a change in depositional environment. First, at 103 m, a 2–4 m thick silty sand conformably overlies the fluvial gravel. It is poorly exposed but shows no primary structures. The upper 40 cm contains thin seams of carbonate but no features indicating surface stabilization such as a pedogenic O horizon. We interpret this deposit as actively accumulating loess. The loess sequence is conformably overlain by 3–5 m of horizontally laminated silt and clay between 105 and 110 m. The horizontal laminae consist of ~1–2 cm thick rhythmic silt and clay beds bounded by ~2 mm thick clay lamina. Some of the silt laminae are interbedded with light-colored carbonate-rich horizons (Fig. 6A, D). These beds are inferred to be lacustrine strata, including varves that may represent annual deposition. Both the loess and lacustrine deposits are abruptly overlain by >3 m of round-cobble conglomerate. Though the bedding style is similar to the underlying conglomeratic strata, the flow directions measured from imbricated clasts and observed from foreset orientations suggest a westward paleocurrent. In addition, the overlying gravel clasts are predominantly mafic volcanics and quartzites, conspicuously lacking the green metamorphic clast types so common lower in the section. We interpret this stratigraphic succession to indicate that there was a hiatus in fluvial aggradation during which loess accumulated on the upper surface of the Clarkston Heights gravels, possibly due to moderate base level lowering and incision of the river systems. The onset fine laminated lacustrine deposition signals a period of lake inundation, possibly from a temporary landslide dam or volcanic blockage downstream. The abrupt return to coarse fluvial deposition with westerly flow direction indicates that when local base level rose within the ephemeral lake, the original northwest oriented paleochannel was abandoned. From the upper fluvial strata, we collected a sand sample (21LMS0511-04) for U-Pb zircon dating (Fig. 6A) to identify whether the change in flow direction was also accompanied by a change in the upstream drainage of the river system.
To better understand the provenance of the Clarkston Heights gravel, we analyze U-Pb zircon ages of ancestral fluvial sands of the measured strata. We present the U-Pb zircon age analyses of three new ancestral river samples and include a previously analyzed sample 19LMS0730-01 from Staisch et al. (2021) (Figs. 3 & 6). We also supplement our ancestral detrital zircon spectra with existing datasets collected elsewhere in Idaho and Washington (Beranek et al., 2006; Staisch et al., 2021) and compare the resulting ages for all ancestral samples with previously published modern river U-Pb zircon age data (Beranek et al., 2006; Staisch et al., 2021; Link et al., 2023).
All new samples were sieved and subjected to density separation in heavy liquids to isolate the zircon mineral fraction. We dated ~120 zircons from each sample using U-Pb LA-ICPMS (Laser Ablation Inductively Coupled Mass Spectrometry) at the U.S. Geological Survey in Denver. Zircon was ablated with a Photon Machines Excite™ 193 nm ArF excimer laser coupled to a Nu Instruments AttoM high-resolution magnetic-sector inductively coupled plasma mass spectrometer. The laser spot sizes for zircon were ~25 µm. Raw data were reduced using the Iolite™ 2.5 program (Paton et al., 2011). Ages were corrected by standard sample bracketing with the primary zircon reference material Temora2 (ca. 417 Ma)(Black et al., 2004) and/or secondary reference materials FC-1 (ca. 1099 Ma; Paces and Miller Jr, 1993), Plešovice (ca. 337 Ma; Sláma et al., 2008), and Fish Canyon Tuff (ca. 28 Ma; Schmitz and Bowring, 2001). Reduced data were compiled using Isoplot 4.15 (Ludwig, 2012). 206Pb/238U ages are reported for igneous zircon samples less than ~1300 Ma and 207Pb/206Pb ages are used for older ages following the recommendations of Gehrels (2014).
U-Pb zircon ages for the Clarkston Heights gravel samples range from 41 Ma to 2813 Ma (Fig. 7, Table S3). All samples show strong age peaks at ~50 Ma and between 70–100 Ma (Fig. 7). All samples also show a moderate to strong Grenville age peak (1000–1200 Ma) and age groups similar to the “Lemhi doublet” (~1380 Ma and 1650–1800 Ma) likely derived from Proterozoic units in central Idaho (Dumitru et al., 2016). The absence of Neogene zircon ages in these datasets prevents useful estimations of the maximum depositional age of the Clarkston Heights gravel from the U-Pb ages, which is not uncommon given that modern rivers sampled in the Pacific Northwest and northern Rocky Mountains sometimes also lack young zircon ages (e.g., Staisch et al., 2021). Therefore, we rely on the stratigraphic relationships described earlier indicating that the gravel was deposited between 10 and 6 Ma.
We use a series of detrital zircon unmixing models to estimate the relative abundance of zircons in Miocene fluvial strata derived from potential upstream sources. The Clarkston Heights and western Snake River Plain strata are “sinks” whereas the modern river samples we use are “sources”. For modern river sands, we remove any zircons dated younger than 5 Ma, as these ages are younger than the strata we are comparing to the modern data. We use the source code from Sundell and Saylor (2017)’s detrital zircon unmixing program (DZMix) for each model and ran the models for 105 iterations. The goodness of fit between synthetic zircon mixtures and the reference Miocene sampled zircon spectra is determined using three different statistical metrics: D-stat, V-stat, and R2. Model fit ranges from 0 to 1, with lower values indicating better fit for the D-stat and V-stat but higher values indicating better fit for the R2 statistic. We retain the results for the top 1% of the results (top 100 models) and the best fit model for further investigation. For the subsequent interpretations, we use a weighted mean of the top 100 model results (Figs. 8, S2, S4, S5) and plot model spread to estimate how uncertain the results are for any particular “source” (Figs. S2, S4, S5).
The modern river networks that we use as “sources” are the configuration of today’s landscape which does not likely reflect the landscape of the middle-late Miocene. Furthermore, Miocene strata collected from basins, such as the western Snake River Plain and Lewiston basin, are likely derived from multiple sources, rather than from an individual river system. We therefore set up DZMix models to test several different possible river network configurations. In the models below, we include many possible sources and lean towards using models with more small tributaries than large rivers because they do a better job at partitioning the landscape and geology.
Our motivating question is whether or not the detrital zircons in the Clarkston Heights gravel show an upstream Snake River connection. Consequently, we model multiple possibilities of potential contributing area. For all Clarkston Heights models, we use existing detrital zircon datasets from the modern Salmon, Clearwater, Grande Ronde, Powder, and Imnaha Rivers. These rivers all have moderate to large watersheds and flow into the Snake River downstream of the Snake River Plain and were likely tributary to the river system depositing the Clarkston Heights gravel even if the western Snake River Plain did not yet drain through Hells Canyon. We also include detrital zircon analyses from Sheep, Granite, and Deep Creeks, which are small modern creeks that flow into the Snake River within Hells Canyon. The models vary in three ways in how the terrain is considered upstream of Hells Canyon: (1) The SnakeFB source model uses a Snake River sample collected at Farewell Bend, just upstream of Hells Canyon (Fig. S6A). At Farewell Bend, the modern Snake River has accumulated sediment from all the main tributaries draining into eastern and western Snake River Plain. This model tests whether the entire modern Snake River drainage basin can explain the past Clarkston Heights gravel zircon provenance. (2) The NoSnake source model only uses tributaries to the Snake River within and downstream of Hells Canyon—that is the tributaries listed above that are common to all models (Fig. S6B). This model is set up to test whether the Clarkston Heights zircon results are explicable without any contribution from the Snake River Plain or upper Snake River, thereby not requiring a fluvial connection through Hells Canyon. (3) The HC-wSRP source model does not use a mainstream Snake River sample source, but rather uses tributaries flowing into the Snake River Plain downstream of the 10.4–6.6 Ma Picabo volcanic center of the Yellowstone hotspot (Fig. 1; Drew et al., 2013), such as the Weiser River, Payette River, Boise River, Wood Rivers, Big Lost River, Raft River, Portneuf River, Salmon Falls Creek, Bruneau River, Reynolds Creek, Owyhee River, and Malheur River (Fig. S6C). Considering just these tributaries enables isolating the Snake River Plain sediment sources from those contributing to the upper Snake River drainage basin, consistent with a hypothesized drainage divide separating the western and eastern Snake River Plains near the 10.4–6.6 Ma Picabo volcanic center of the Yellowstone hotspot (Staisch et al., 2021).
The best fit and top 100 detrital zircon mixing results for the three source models show that the Clarkston Heights gravel provenance is best explained by substantial contributions from western Snake River Plain drainages such as the Payette and Boise rivers. The NoSnake and SnakeFB model results give much poorer fits to the Clarkston Heights gravel detrital zircon age spectra (Fig. 8). The two samples from the middle portion of the Clarkston Heights gravel stratigraphy are in particular poorly fit by these models. Only the HC-wSRP model performs well for all four samples, reflected in the improved statistical fit to the analytical data (Table 2).
The stratigraphically highest and lowest samples we collected from the Clarkston Heights gravels have similar zircon spectra and DZmix results suggesting ~50% of the zircons are sourced from the Clearwater and the Salmon Rivers (Fig. 8). These model results also suggest that the samples contain zircons from the Boise River, Payette River, Reynolds Creek, and Wood River system, all tributaries to western Snake River Plain. DZmix model results for samples 21LMS0511-01 and 21LMS0512-05, sampled from the middle of the sedimentary section highlight the same major contributors, but with different relative proportions. For these two samples, the Payette and Boise rivers are more substantial contributors, accounting for nearly half of the estimated zircon sources.
The detrital zircon age spectra for the Clarkston Heights gravels have a few age peaks that similarly suggest that the landscape currently drained by the Payette and Boise rivers was an active source to strata downstream of Hells Canyon. Specifically, the abundance of Idaho Batholith ages as well as the presence of Neoproterozoic ages in the Clarkston Heights gravels are well represented in the modern Payette and Boise river zircon age spectra. The substantial ~1380 Ma age peak in the Clarkston Heights gravel samples is a strong support for Salmon River contribution (Fig. 7).
We use three published detrital zircon datasets from the Miocene strata deposited in southwestern Idaho to test the sensitivity of Miocene strata provenance to the inclusion or exclusion of the eastern Snake River Plain (see white circles in Fig. 1; Beraneck et al., 2006). The Terteling Springs formation unconformably overlies the Chalk Hills Formation and may be a transgressive phase of the Glenns Ferry Formation, whereas the Poison Creek Formation is part of the lower Chalk Hills Formation. The Succor Creek Formation is from a middle Miocene stratigraphic unit deposited in the Oregon-Idaho graben. We designed two DZMix model setups to identify the sources for Miocene strata in southwestern Idaho using detrital zircon data from sand collected from modern river systems. (1) The allSRP model uses the compiled detrital zircon datasets from tributaries in the eastern and western Snake River Plain as potential sources (Fig. S6D). (2) The wSRP model uses only the compiled detrital zircon datasets from Snake River tributaries in the central and western Snake River Plain as possible sources, west of the proposed Miocene drainage divide (Fig. S6E; Beraneck et al., 2006, Staisch et al., 2021).
The best fit model and top 100 detrital zircon mixing results for the Terteling Springs (TSFm on Fig. 1) and Succor Creek (SCFm on Fig. 1) samples indicate a strong contribution (>60%) from the Payette drainage network (Figs. S4-S5). The Poison Creek sample (PCFm on Fig. 1) has lower (~20%) contribution from the Payette River, balanced by an increase to ~45-50% from the Boise River. Terteling Springs DZmix models also suggest moderate contribution from Boise River, Wood River system and Big Lost River system (Figs. S4-S5). Poison Creek model results show moderate contributions from the Big Lost River system. The Succor Creek DZ model results show moderate contribution from the Big Lost and Boise rivers, as well as an increased contribution from the Owyhee and Malheur rivers compared to the other analyzed strata (Figs. S4-S5).
Between the wSRP and allSRP models, the best-fit and top 100 models for the Poison Creek strata have better statistical fit to the wSRP model (Table 2), suggesting that the eastern Snake River Plain was not a major contributor during Chalk Hills time. The Succor Creek and Terteling Springs strata are only slightly better fit by the wSRP model, suggesting the strata may be indifferent to the inclusion or exclusion of eastern Snake River Plain drainages. In the allSRP models, the cumulative contribution from drainages east of the proposed divide is between 2-5% for best fit models and between 8-13% on average for the top 100 models (Fig. S5).
A critical result of our detrital zircon provenance analysis is that deposits in the western Snake River Plain, upstream of Hells Canyon, and the Clarkston Heights region, downstream of Hells Canyon, share common sediment source terranes during the 10–6 Ma periods represented by the Clarkston Heights gravel and the Chalk Hills Formation. Specifically, the source terrain for both depocenters included areas now drained by Payette, Boise, Wood River, and Big Lost rivers, all entering the western Snake River Plain. These findings are strong evidence of at least an intermittent fluvial connection through the Hells Canyon corridor since at least 6 Ma and perhaps as long ago as 10 Ma.
This finding starkly contrasts to the long-standing ‘capture’ hypothesis for the formation of Hells Canyon (Wheeler and Cook, 1954; Reidel and Tolan, 2013) and may seem to directly contradict the recent evidence that Hells Canyon began rapidly incising at ~2.1 Ma (Morriss et al., 2025). However, there is strong evidence for recent rapid incision that can be framed within the context of a persistent connection of the evolving greater Lake Idaho system by way of Hells Canyon and provides important clues on the diverse mechanisms involved in river incision and drainage reorganization, such as Yellowstone hotspot migration and climate change, over Miocene-Pliocene time.
The ~11.4 – 6.0 Ma Chalk Hills Formation and ~10 – 6 Ma Clarkston Heights gravel deposition occurred during a time of regional uplift throughout the northern Cordillera and localized basin formation. Stream power models and 10Be cosmogenic dating show that canyon incision along the modern Snake, Salmon, and Clearwater rivers initiated before ~10 Ma and that the amplitude of transient uplift diminished northward (Mitchell et al., 2023). Overall, uplift and incision across the northern Cordillera were relatively modest according to regional low-temperature thermochronologic data and geomorphic analyses (Mitchell and Yanites, 2019; Kahn et al., 2020). In some regions, however, there may have been localized uplift, such as the Pioneer-Boulder Mountains north of the Picabo volcanic center where thermochronologic data indicate 2–3 km of exhumation between 11 – 8 Ma (Vogl et al., 2014).
The mechanisms suggested for middle-late Miocene northern Cordillera uplift include faulting along the Western Idaho Shear Zone (Tikoff et al., 2001; Kahn et al., 2020), damming during emplacement of the CRBG (Reidel and Tolan, 2013), and Yellowstone hotspot driven mantle lithospheric instabilities and root loss (Vogl et al., 2014; Larimer et al., 2019). The consequences of volcanic damming would include a temporary decrease in incision during emplacement and temporary base level rise. This effect would likely affect the paleo-rivers draining central Idaho similarly because the basalt was emplaced largely downstream of this network. However, the evidence of late Miocene incision in the Lewiston basin as shown by the Pomona, Elephant Mountain, and Lower Monumental members (Fig. 9F) and pronounced northward decay of incision in the northern Cordillera suggests that this was not the main mechanism for middle-late Miocene incision (Larimer et al., 2019; Mitchell et al., 2023). Faulting along the WISZ may have occurred over this time, as suggested by a single young apatite (U-Th)/He age within Hells Canyon (Kahn et al., 2020), but shearing along the WISZ would likely induce relatively localized uplift as a dominantly vertical structure (Richardson et al., 2024) and its north-south orientation is inconsistent with the northward decay of uplift and incision as well.
During the period of middle-late Miocene regional uplift, the Yellowstone hotspot was in the central Snake River Plain and south of pronounced uplift in the Pioneer-Boulder Mountains (Vogl et al., 2014). Although uplift solely due to increased thermal buoyancy does not explain the wavelength of uplift or its duration, other related instabilities in the mantle lithosphere may. For example, Larimer et al. (2019) modeled the crustal flexural response driven by a buoyant load north of the Twin Falls and Picabo volcanic centers of the YHS, showing that this mechanism fits the profiles of regional uplift and incision. In their hypothesis, the buoyant load may have been a YHS-driven instability, either from removal of mantle lithospheric material or as a thermal process, focused along the intracontinental boundary between oceanic island arc and continental lithospheric material. The hypothesis that YHS-driven mantle instabilities affected the northern Cordillera is not unique to this region or time-period. Mantle root loss beneath the Wallowa Mountains, west of the SRP, has also been used to explain a pronounced high-velocity anomaly beneath the range and the timing of voluminous CRBG eruption at ~16 Ma (Hales et al., 2005; Darold and Humphreys, 2013).
To evaluate how regional uplift influenced incision rates in the Lewiston basin, we extract and compare topographic data from various units along the paleo-canyon systems (Figs. 3, 9F). The basal elevations for each unit roughly mark river position prior to basalt emplacement or gravel deposition. Although the ~11.2 Ma intracanyon Pomona Member captures the onset of uplift and incision in this region, the subsequent Elephant Mountain Member is found at higher elevations than the Pomona flows, suggesting the emplacement of the canyon-filling Pomona basalt temporarily arrested vertical incision (Fig. 9F). To evaluate incision rate, we use the Elephant Mountain Member and younger units. We also include the base of the ~18 Ka Bonneville flood gravel (O’Connor et al., 2021) to avoid biasing our analysis with the modern river channel (Gallen et al., 2015). The resulting 0.009 mm/yr incision rate is similar to the pre-2.1 Ma incision measured in Hells Canyon (Morriss et al., 2025; Fig. 9E-F), indicating that rapid incision after 2.1 Ma was focused upstream of Lewiston basin.
Coincident with the onset of regional uplift, faulting along graben bounding structures to the north and south of the western Snake River Plain initiated after 10.5 Ma and deformation was largely accommodated by 9 Ma (Kimmel, 1982; Cummings et al., 2000; Wood and Clemens, 2002), creating accommodation space for Lake Idaho. Low diversity of fish fauna in lower Chalk Hills strata indicates a smaller lake extent in western Snake River Plain or that it was blocked from fish migration into the area at the onset of Lake Idaho deposition, but fish fauna also display an increasing dominance of cool water species through time, possibly caused by a deepening lake or global cooling event (Kimmel, 1979). Markers of Lake Idaho shoreline elevation from the stratigraphy generally suggest early rapid infilling and subsequent lake-level fluctuation during the time period that the Chalk Hills Formation was deposited (Fig. 9D).
The stratigraphy and provenance of the 11.4–6.0 Ma Chalk Hills and Poison Creek formations suggest that early Lake Idaho was a low energy depositional system contained within the newly formed trough of the western Snake River Plain and that a large portion of sediment (or at least zircons) was derived from watersheds entering from the north. Low fish diversity in the early record of the Chalk Hills Formation also implies that the lake was small or had limited migration from other basins (Kimmel, 1979). Likewise, ostracod (Sheppard, 1991) and fish (Smith, 1982) fauna indicate water chemistry conditions similar to modern fish-populated lakes of the Great Basin (Wood and Clemens, 1982). Taken together, evidence indicates that the western Snake River Plain did not receive input from the east- the terrain between western Wyoming and the eastern Snake River Plain now traversed by the modern Snake River. Instead, integration of stratigraphic, provenance, and fossil data between the Snake River Plain and Columbia basin suggests that the paleo-Snake River, originating in today’s upper modern Snake River drainage, circumnavigated the uplifting northern Cordillera (Staisch et al., 2021; Fig. 9A). The northward diminishment in transient middle-late Miocene uplift likely facilitated a paleo-Snake River path north through the northern Rocky Mountains and then into the Columbia basin from the northeast.
Although the upper Snake River may not have flowed into the western Snake River Plain and then through Hells Canyon, our provenance analysis of Clarkston Heights gravels in the Lewiston Basin gives strong evidence that the western Snake River Plain did drain through Hells Canyon for periods during the middle-late Miocene time. Furthermore, fish faunal evidence from the Chalk Hills Formation supports an intermittent connection to the Pacific Ocean, with several species of Pacific salmon identified (Smith et al., 2000). Thus, early Lake Idaho was at least occasionally an open basin system (Fig. 9A). Without more precise age information for the Clarkston Heights gravels, we cannot assess the duration of this fluvial connection, though it was likely intermittent given the variable stages of Lake Idaho and at least occasionally alkaline water chemistry during Chalk Hills time and leading up to the Pliocene Glenns Ferry Formation deposits (Wood and Clemens, 2002).
Incision rates calculated in the Lewiston basin and Hells Canyon remained at a steady and slow rate throughout the middle-late Miocene (Fig. 9E-F), but the elevation of Lake Idaho dropped late during the time of Chalks Hills Formation deposition (Fig. 9D). Furthermore, the stratigraphic record of Lake Idaho between 6 and 4.3 Ma is scant, marked as a basin-wide disconformity between the Chalk Hills and Glenns Ferry formations (Wood et al., 2002). Comparisons of fish fauna from the Chalk Hills and Glenns Ferry strata suggest that there was at least a 1 Ma lapse in time between these units (Smith et al., 2000). Many of the lake species present in Lake Idaho during the deposition of the Chalk Hills Formation were also present, though modified, during deposition of the Glenns Ferry Formation. These data show that despite substantial lake-level lowering, lacustrine habitats did not fully disappear in the western Snake River Plain between 6 and 4.3 Ma.
The cause of ~6 Ma regression of the Chalk Hills lacustrine system is unknown (Wood and Clemens, 2002). One hypothesis is the onset of a drier climate, similar to the Quaternary history of closed lake basins of the Great Basin (e.g. Currey, 1990; Reheis et al., 2014). Global climate proxies from deep sea sediments record a reduction in carbon isotopic ratios over several periods of the Miocene associated with global aridification and the expansion of C4 grasses (Zachos et al., 2001). The timing of C4 grassland expansion was not coincident across the globe given variable aridity, temperature, and fire regime (Strömberg, 2005; Edwards et al., 2010; Strömberg, 2011; Strömberg and McInerney, 2011). Regional climate proxies from the proximal 8–6 Ma Beaverhead site in southwestern Montana and other sites in the North American Great Plains suggests that C4 grasses peaked in North America slightly after their global expansion initiated and during regional aridification (Hyland et al., 2019; Strömberg and McInerney, 2011; Fox et al., 2012). This climatic trend may explain the reduction in the elevation of Lake Idaho after ~6.5 Ma and its subsequent absence from the stratigraphic record (Fig. 9I). Conversely, Kimmel (1979) suggested Lake Idaho drained externally toward the southwest at this time based on variations in the elevation of stratigraphic units near Adrian, Oregon (Fig. 2), with outlet conditions possibly affected by faulting. But we have found little stratigraphic or geomorphic evidence in the literature or the field that is indicative of an extensive lake system draining through this hypothesized outlet.
Recorded by the Terteling Springs ooilitic lenses and Glenns Ferry lacustrine strata, lake conditions rapidly returned to the western Snake River Plain by 4.3 Ma and lasted until at least 2.18 Ma (Amini et al., 1984; Malde, 1991). Stable isotopic measurements from the Terteling Springs strata suggest an early closed-basin system (Fig. 9D; Wood, 2004; Talbot, 1990). Lake Idaho reached its maximum extent and highest stages (~1,100 m) during deposition of the Glenns Ferry Formation (Figs. 2, 9D). Glenns Ferry Formation detrital zircon data show that distally sourced zircons became increasingly abundant in Lake Idaho, specifically from the eastern Snake River Plain where the Proterozoic and Archean zircons are recycled from strata in the upper Snake River drainage (Beranek et al., 2006; Staisch et al., 2021). The early Pliocene Glenns Ferry strata at Hagerman, Idaho, similarly suggest possible capture of the upper Missouri or eastern Snake Rivers to explain migration of several northern fish species into the Snake River Plain (Smith et al., 1984, 2000; Smith, 1987). Othberg (1994) reports quartzite gravel clasts within the Glenns Ferry Formation, indicative of upper Snake River sources. Together, these observations indicate a substantial eastward expansion of the watershed initiated by 4.3 Ma. Such an expansion would increase hydrologic input, thus enabling the return of Lake Idaho (Fig. 9B; Wood and Clemens, 2002; Staisch et al., 2021).
Eastward expansion of the Lake Idaho watershed was likely due to Miocene–Pliocene migration of the Yellowstone hotspot. During deposition of the Chalk Hills Formation, the active Yellowstone hotspot Picabo volcanic center was in the central Snake River Plain likely dividing the basin into separate eastern and western drainages (Fig. 9A). In late Miocene and Pliocene, the eastward migration of the Yellowstone hotspot toward the Heise volcanic center and beyond allowed for thermal relaxation of the eastern Snake River Plain lithosphere and that localized densification of the crust. This resulted in flexural subsidence and eventual westward diversion of the paleo-Snake River corridor across southwestern Montana to the western Snake River Plain, thus increasing flow into Lake Idaho (Fig. 9B; McQuarrie and Rodgers, 1998; Staisch et al., 2021). Throughout Glenns Ferry Formation stratigraphy, the depositional environment remained fairly low-energy, dominated by lacustrine and fluvial floodplain strata (Malde and Powers, 1962; Kimmel, 1979; Malde, 1991). However, based on the size, duration, and the freshwater fauna identified in Glenns Ferry strata by Repenning et al. (1995), the lake was an open fresh-water system connected to the Columbia River by way of Hells Canyon.
The uppermost Glenns Ferry Formation is unconformably cut by the coarse fluvial deposits of the Tuana gravel and the Tenmile gravel (Malde, 1991; Othberg, 1994). These gravels suggest a through-going Snake River by 1.92 Ma (Malde, 1991; Wood and Clemens, 2002) and certainly by the time of the 1.58 Ma Pickles Butte basalt (Othberg, 1994). It is unlikely that the lacustrine conditions of the Glenns Ferry Formation coexisted with through-going gravel transport. The elevation of these gravels are ~40–340 m lower than the ~1,100 m maximum lake level corresponding with the highest elevation of lacustrine shoreline deposits of the Glenns Ferry Formation (Othberg, 1994; Wood, 1994, 2004; Table S6), thus we infer substantial erosion of the basin outlet by spillover, consistent with the stratigraphic evidence of substantial shoreline regression and a lake-level drop late in the Glenns Ferry Formation (Kimmel, 1982; Smith et al., 1982; Malde, 1991).
After the initial <2.2 Ma establishment of a through-going Snake River marked by the coarse gravel deposits, the river has incised another 120-150 m to its modern elevation in the western Snake River Plain (Malde, 1991; Othberg, 1994, Wood and Clemens, 2002). Malde (1991) suggested ~240 m of incision from the level of the Tuana gravel to modern river level near Bruneau, Idaho by 1.8–1.7 Ma in conjunction with at least four basalt-filling and canyon incision episodes recorded by lava flows, palagonite, and lake beds forming the 1.9–1.0 Ma Bruneau Formation (Malde and Powers, 1972; Malde, 1985, Malde, 1991). This finding implies an incision approaching 1 mm/yr during early Quaternary fluvial incision of Glenns Ferry Formation deposits.
Along the Boise River, which joins the Snake River in the far western Snake River Plain, the ~2 Ma Tenmile gravel sits 137–146 m above river level, whereas terrace gravels beneath the 0.572 ± 0.210 basalt of Gowan Terrace are only 43–47 m above the Boise River and the basalt of Mores Creek was emplaced within 2.5 m of the Boise River (Fig. 9; Othberg et al., 1995). The elevations and ages give a net incision rate between the Tenmile gravel and younger terrace and basalt markers of 0.05–0.19 mm/yr. These rates are similar to the 0.06–0.16 mm/yr of post 1.7 Ma incision of the Owyhee River (Ely et al., 2012), which also drains into the western Snake River Plain.
What were the regional geologic and physiographic conditions leading to the formation and demise of the Miocene-Pliocene lake systems in the western Snake River Plain? This question has been much investigated, summarized by Malde (1991); Othberg (1994), Repenning et al. (1995), Wood and Clemens (2002), and Staisch et al. (2021). Lake Idaho occupied the western Snake River Plain, its location governed by the structural basin formed mainly between 10.5 and 9 Ma (Kimmel, 1982; Cummings et al., 2000; Wood and Clemens, 2002). But a key component controlling the lake history is the location and evolution of the lake outlet, which we consider in light of our results showing a connection between the western Snake River Plain and the Clarkston Heights gravel at some time between 10 and 6 Ma.
Our findings indicate a fluvial connection between the western Snake River Plain and the Lewiston basin since at least 6 Ma. This contrasts with the early Pleistocene Hells Canyon capture hypothesis of Wheeler and Cook (1954), suggested by Malde (1991) to have been at about 2 Ma. Wheeler and Cook (1954) postulated that a new lake outlet through Hells Canyon and integration of the Snake River into the Columbia basin resulted from breaching of a drainage divide separating the Columbia basin from drainages flowing south into the Snake River Plain near Oxbow, Oregon, (river mile 270; Fig. 2).
This capture hypothesis, in conjunction with the primarily freshwater fauna associated with Lake Idaho deposits (e.g. King, 1903; Smith et al., 1982; Middleton et al., 1985; Repenning et al., 1995; Wood and Clemens, 2002), spawned several possible alternative outlets to address the question of where Lake Idaho drained if not through Hells Canyon. Proposed routes include (1) west of Hells Canyon but preserving a connection with the Columbia River in Pasco basin or downstream (Livingston, 1928; Smith et al., 2000; Van Tassell et al., 2001; Reidel and Tolan, 2013); (2) through southeastern Oregon to northern California (Wheeler and Cook, 1954; Cook and Larrison, 1954; Taylor, 1985; Malde, 1991); or (3) through northern Nevada by way of the Humboldt River drainage and into the Sacramento River basin in northern California (Repenning et al., 1995). Most of these proposed routes are based on biogeographic evidence, as summarized by Malde (1991), Repenning et al. (1995), Smith et al. (2000), Van Tassell et al. (2001), and Hovingh et al. (2008). Geologic mapping, reports, and our regional reconnaissance bare no evidence for late Miocene or Pliocene fluvial gravels indicative of an alternative drainage west of Hells Canyon. Furthermore, detrital zircon analyses of <4.5 Ma Baker Valley strata, along the proposed route west of Hells Canyon, do not support a connection with the western Snake River Plain (Staisch et al., 2021). The Clarkston Heights gravel detrital zircon provenance data strongly conflict with the latter two hypotheses.
We conclude that these alternate routes are unlikely and that a persistent yet evolving lake outlet near the present position of Farewell Bend in the western Snake River Plain—at the head of Hells Canyon—was the singular point of exit for the lacustrine systems represented by both the Chalk Hills and Glenns Ferry formations. At this location, near the northern limit of the extent of the Lake Idaho strata (fig. 2), the Snake River has incised a narrow canyon between the Slaughterhouse Range (Oregon) and its extension to the southeast across the Snake River in Idaho (Fig. 2), thus providing the physiographic setting to impound a lake to 1,100 m (Wood and Clemens, 2002). Furthermore, it seems most plausible that Lake Idaho during deposition of the Chalk Hills and Glenns Ferry formations, when full and spilling, had similar outlets based on the similarity in extent and geometry of mapped geologic units (Fig. 2; Kimmel, 1979).
Recently dated fluvial deposits trapped in caves provide strong evidence for a 2.1 +/- 1.0 Ma onset of rapid incision in Hells Canyon, increasing from ~0.01 mm/yr to ~0.16 mm/yr (Morriss et al., 2025). The coincidence of rapid incision in Hells Canyon with the drop in base level throughout the western Snake River Plain (Fig. 9D-E) suggests a causal link. Morriss et al. (2025) attribute and model rapid incision to the capture and resulting integration of the upstream drainage basin of the Snake River, and they propose several possible mechanisms on how capture occurred, including: (1) “fill and spill” event in which Lake Idaho overtopped the drainage divide with the Clearwater-Salmon River system, (2) headward erosion of a north-draining stream, and (3) groundwater sapping through permeable units. The finding of a >6 Ma fluvial connection between the western Snake River Plain and the Lewiston basin through Hells Canyon challenges the concept that rapid incision was caused by southward expansion of the drainage network but instead modifies the “fill and spill” hypothesis to promote the importance of bedload material.
Although the drainage area contributing to Lake Idaho started to increase by 4.3 Ma (Fig. 9B; Staisch et al., 2021), rapid incision within Hells Canyon initiated about 2 Myr later at about the same time as the onset of gravel transport across the western Snake River Plain. We hypothesize that the coincidence of continuous fluvial gravel transport and rapid incision in Hells Canyon are connected: Fluvially transported gravel enhances incision (Sklar and Dietrich, 2001; Stock et al., 2005, Pratt-Sitaula et al., 2007; Cowie et al., 2008), and this ‘tools effect’ has been inferred to be a controlling factor for rapid incision of Quaternary lava-flow dams on the Owyhee River, southeastern Oregon (Ely et al., 2012). Based on our synthesis of provenance results with regional geologic evidence for incision and depositional settings, we propose that the through-going fluvial transport of durable upper-basin Snake River clasts across the western Snake River Plain promoted the rapid incision commencing at ~2 Ma of the divide near Farewell Bend, not just overtopping of Lake Idaho waters, and consequent enhanced incision in Hells Canyon downstream.
An important implication of our model is that the recent incision of the Snake River in Hells Canyon should be unique relative to other rivers in the region, a factor which should manifest in valley morphology. This indeed seems to be the case, as comparison of canyon morphology of the adjacent Snake, Clearwater, and Salmon rivers in the vicinity of Hells Canyon shows that the Salmon and Clearwater River valleys system are much wider than that for the Snake River Hells Canyon, especially for the lowermost ~300-400 m, which would be the approximate incision depth associated with the enhanced erosion rate of the last ~2 My (Fig. S7). Furthermore, downstream of Hells Canyon, there is no evidence for increased incision rates during this time (Fig. 9F), nor do the proximal Clearwater or Salmon River drainages show an increase in fluvial downcutting (Mitchell et al., 2023). Therefore, the mechanisms driving rapid incision at ~2 Ma and the draining of Lake Idaho appear to be localized to Hells Canyon and the western Snake River Plain.
The top-down spill-over scenario and persistent Hells Canyon connection since the late Miocene proposed here contrasts with the long-standing hypothesis of headward erosion capturing the Snake River in the Pliocene or early Pleistocene, thereby initiating rapid Quaternary incision. Though a Snake River diversion into Hells Canyon may have been first proposed by Livingston (1928), the “headward erosion” capture hypothesis apparently took hold with Wheeler and Cook’s (1954) paper focused on the physiographic setting near the sharp “Oxbow” bend within Hells Canyon, near Snake River rivermile 270. On the basis of differences in canyon morphology upstream and downstream of the Oxbow, and what they described as barbed tributaries that are only present upstream of the Oxbow, Wheeler and Cook (1954) proposed that incision and headward erosion of a north-flowing tributary of the Salmon River extended its length southward, it intercepted a south-flowing (into Lake Idaho) tributary, triggering lake spillover and rapid incision of a through-going drainage to the north, leading to the now-narrow and steep Hells Canyon. This idea of a Salmon River tributary capturing the Snake River has been repeated by many considering the evolution of the Snake River Plain, including Kimmel (1979, 1982), Smith et al. (1982), Malde (1991), Smith et al. (2000), Van Tassell et al. (2001), and Reidel (2013).
As for Hells Canyon, such headward-erosion capture scenarios have been a prevalent explanation for river basin integration for more than a century (e.g. Davis, 1898; Lane, 1899; Lucchita et al., 2001; Enzel et al. 2003; Dickinson, 2013, 2015). But in their review of basin-integration mechanisms, Hilgendorf et al. (2020, p. 3) assert that this mechanism “may be overprescribed,” chiefly because of the general inefficacy of headward erosion to extend drainages (Douglass and Schmeeckle, 2007; Douglass et al., 2009). Hunt (1969, p. 118), recognizing the implausibility of piracy by headward erosion in his refutation of the headward-erosion capture hypothesis for the incision of Grand Canyon, asserted that such a process would require action of an especially “precocious” drainage.
The spillover mechanism for top-down fluvial integration and drainage basin expansion is gaining traction to explain geomorphic and geologic observations, especially in extensional environments (Guerts et al., 2018; Hilgendorf et al., 2020). Spillover has long been recognized as a plausible mechanism for drainage basin integration, exemplified by J.S. Newberry’s (1861, 1862) inferred mechanism for incising the Grand Canyon, but has consistently lived in the shadow of alternative mechanisms for establishing transverse drainage paths. A recent description of such a top-down, fill-and-spill mechanism is for the early Pliocene integration of the lower Colorado River (House et al., 2008; Pearthree and House, 2014; Crow et al., 2021). As described above, this is essentially the mechanism we advocate for triggering the Hells Canyon incision—a cascade of drainage-basin changes and shifting depositional environments working their way downstream within context of a long-established but intermittent fluvial connection between the Snake River Plain and the Columbia basin. Perhaps a locally relevant recent analog or continuation of this process is the ~18–15 ka overflow of pluvial Lake Bonneville (Oviatt, 2015) into the Snake River Plain, thereby temporarily connecting the Bonneville basin to the Snake and Columbia River drainage basins. This connection was subsequently severed by the warmer and drier post-pluvial climate, lowering Lake Bonneville below the outlet threshold and returning it to closed-basin conditions of Great Salt Lake. But it also seems plausible that repetition of this event sequence could lead to further lowering of the divide, ultimately resulting in incorporating parts of the Bonneville basin more permanently into the Snake River drainage. This late Quaternary example is also possibly analogous to the rise and fall of the lacustrine system depositing the Chalk Hills Formation, and the temporary overflow transporting sediment through Hells Canyon to the Clarkston Heights gravel between 10 and 6 Ma.
Collection of paleoclimatic and paleoenvironmental proxies may lead towards an improved understanding of the varied biologic and geologic implications of an intermittent but persistent Hells Canyon fluvial connection, as well as the role of climate change in the evolution of Lake Idaho’s connection downstream. However, based on this work, future analyses may consider alternative scenarios instead of a singular ‘capture’ and entry of the Snake River into Hells Canyon in the late Pliocene or Quaternary.
Detrital zircon analyses of Clarkston Heights gravel and the coeval Chalk Hills Formation show that the Lewiston Basin, downstream of Hells Canyon, and the western Snake River Plain, upstream of Hells Canyon, shared a common sediment source sometime between 10 and 6 Ma. We integrate these data with the regional history of Lake Idaho and the Columbia basin to conclude that Hells Canyon is a longer-lived fluvial corridor between the western Snake River Plain and Columbia basin than previously understood, but whether this fluvial connection was continual or intermittent remains unclear. This contrasts with the long-standing capture hypothesis of a new Pliocene or Quaternary pathway for the Snake River through Hells Canyon leading to regional incision.
Instead, our results are consistent with a two-phased incision history through Hells Canyon. The initial phase started around ~12–10 Ma, in which Hells Canyon and the Salmon and Clearwater rivers were incised at a rate of ~0.01 mm/yr in response to broad uplift across the northern Cordillera (Larimer et al., 2019; Mitchell and Yanites, 2019; Mitchell et al., 2023). During this phase, the 11–6 Ma Chalk Hills lake system at least occasionally drained through Hells Canyon but did not result in rapid incision. Rerouting of the upper Snake River and expansion of the Lake Idaho watershed instigated lake-level rise and basin infilling within the western Snake River Plain starting by 4.3 Ma. The second phase of Hells Canyon incision was a top-down spill over scenario, where progressive basin infilling eventually allowed gravel transport across the western Snake River Plain as early as 2.2 Ma, subsequent incision through the lake outlet at Farewell Bend, and the demise of Lake Idaho. Additionally, the through-going downstream transport of durable clasts provided the incision tools enabling the rapid downcutting of Hells Canyon at ~0.16 mm/yr starting at ~2.1 Ma (Morriss et al., 2025). Thus, recent rapid incision of Hells Canyon was localized and dominantly caused by broadscale changes in watershed dynamics and the evolution of the Snake River sediment regime.
The two-phased incision history along the deepest canyon in North America was effectively controlled by geodynamics of the Yellowstone hotspot; initially from mantle instability driven broad uplift (Vogl et al., 2014; Larimer et al., 2019), and later by drainage reorganization along the Yellowstone hotspot track (Beraneck et al., 2006; Staisch et al., 2021). Though some specifics remain unknown, it is an example of the interplay between the complicated geodynamic environment of western North America and the fluvial systems that drain and incise the landscape. Moreover, it shows how the character of specific places, such as Hells Canyon, are produced by processes and events occurring both locally and at continental scales. Some outstanding questions include the precise timing and duration of late Miocene connection between Lake Idaho and the Clarkston Heights gravels and the role of climate change in lacustrine systems within the western Snake River Plain.
| Sample | Latitude [°N] | Longitude [°E] | Elevation [m] |
| Lower strath contact with basalt | 46.4168 | -117.1142 | 229.72 |
| Base of lower quarry exposure | 46.4104 | -117.1088 | 252.30 |
| 19LMS0730-01 | 46.4104 | -117.1087 | 253.17 |
| Base of upper quarry exposure | 46.4051 | -117.1164 | 307.85 |
| 21LMS0511-01 | 46.4051 | -117.1164 | 309.62 |
| 21LMS0512-05 | 46.4039 | -117.1174 | 326.92 |
| 21LMS0511-04 | 46.4045 | -117.1170 | 341.89 |
| Sample | Model | D-statistic | V-statistic | R2-statistic | |||
| 19LMS0730-01 | NoSnake | 0.128 | 0.156 | 0.149 | 0.196 | 0.881 | 0.914 |
| SnakeFB | 0.113 | 0.126 | 0.136 | 0.166 | 0.902 | 0.931 | |
| WSRP | 0.099 | 0.170 | 0.892 | ||||
| 21LMS0511-01 | NoSnake | 0.214 | 0.245 | 0.340 | 0.345 | 0.811 | 0.837 |
| SnakeFB | 0.193 | 0.217 | 0.251 | 0.281 | 0.891 | 0.904 | |
| WSRP | 0.106 | 0.171 | 0.965 | ||||
| 21LMS0511-04 | NoSnake | 0.116 | 0.128 | 0.174 | 0.191 | 0.862 | 0.871 |
| SnakeFB | 0.100 | 0.115 | 0.137 | 0.163 | 0.893 | 0.918 | |
| WSRP | 0.094 | 0.159 | 0.929 | ||||
| 21LMS0512-05 | NoSnake | 0.197 | 0.208 | 0.231 | 0.240 | 0.844 | 0.856 |
| SnakeFB | 0.171 | 0.184 | 0.227 | 0.246 | 0.909 | 0.922 | |
| WSRP | 0.104 | 0.170 | 0.958 | ||||
| Terteling Springs | allSRP | 0.124 | 0.155 | 0.206 | 0.969 | 0.990 | |
| WSRP | 0.109 | 0.183 | 0.985 | ||||
| Poison Creek | allSRP | 0.140 | 0.201 | 0.215 | 0.293 | 0.948 | 0.978 |
| WSRP | 0.172 | 0.259 | 0.957 | ||||
| Succor Creek | allSRP | 0.164 | 0.221 | 0.214 | 0.275 | 0.815 | 0.959 |
| WSRP | 0.107 | 0.173 | 0.965 | ||||
Regional map of river systems across the Snake, Salmon, and Clearwater drainages. Colored polygons are watersheds above sampled modern detrital zircon samples (yellow diamonds) from Beranek et al. (2006), Staisch et al. (2021), and Link et al. (2023). Site labeled FB is Farewell Bend along the modern Snake River. TSfm: Terteling Springs formation. PCfm: Poison Creek Formation. SCFm: Succor Creek Formation. The Yellowstone hotspot (YHS) pathway through North America is denoted by warm-toned zones, with estimates ages and volcanic center names. Inset map shows the geologic units across the region. WISZ: Western Idaho Shear Zone. OIG: Oregon-Idaho Graben. wSRP: Western Snake River Plain. LGG: La Grande Graben. LFZ: Limekiln fault zone. WF: Wallowa fault. WMtns: Wallowa Mountains. OMtns: Owyhee Mountains; BMtns: Bennett Mountains. P-BMtns: Pioneer-Boulder Mountains. 87Sr/86Sr 0.704 and 0.706 isopleths are plotted to denote the edge of Laurentia (Pierce and Morgan, 2009). Accreted terranes are denoted by blue dotted lines and labeled accordingly. Lower graph shows the elevation profile of the Snake River. Blue elevation envelopes show the magnitude of relief along the Snake River corridor taken from 0.5-, 1.0-, 2.5-, 5.0-, and 10.0-kilometer distances from the river. Eastern (eSRP) and western (wSRP) are labeled, as is the Hells Canyon. Blue and black squares show the extent of the geologic map in Figure 3. Basemap hillshade was generated from 90-meter digital elevation data (NASA SRTM, 2013).
Mapped extent of Miocene-Pliocene fluvial and lacustrine strata in the western SRP (Brook, 1979; Mitchell and Bennett, 1979a-b; Rember and Bennett, 1979; Worl et al., 1991; Ferns et al., 1993a-b; Jenks et al., 1993; Madin and Ferns, 1997; Othberg et al., 2005a,b; Bond et al., 2011; Lewis et al., 2012; Feeney, 1993; Feeney et al., 2023a-e; Feeney et al., 2025a-b; Feeney, 2025; Franczyk et al., 2023; Lewis et al., 2023; Love et al., 2023a,b). We note that the Idaho Group, undivided, is not partitioned out into the Poison Creek, Chalk Hills, or Glenns Ferry formations and may therefore represent any of these units. Pleistocene fluvial terrace gravels, mapped in blue, include the Tuana Gravel, Tenmile Gravel, and other concordant terrace gravels mapped at similar elevations. Thin black lines are mapped faults that bound the Oregon Idaho Graben (OIG) and wSRP (Lewis et al., 2012; Franczyk et al., 2023). The 1,100-m elevation isopleth is mapped in blue and shadowed in transparent white to represent the highest elevation that Lake Idaho strata are identified. The estimated extent of Lake Idaho during Chalk Hills (11.4 – 6 Ma) and Glenn’s Ferry (4.3 – 2.2 Ma) time periods is shown by pink and orange lines, respectively, and based on stratigraphic observations and fossil fish localities (Bjork, 1971; Malde, 1972; Kimmel, 1979). Basemap hillshade was generated from 90-meter digital elevation data (NASA SRTM, 2013).
Geologic map along the Snake-Clearwater River confluence (Hooper et al., 1985; Garwood and Bush, 2005; Kauffman et al., 2009). The middle to late Miocene Snake River paleochannel through Clarkston Heights is labeled. Cyan dots are sampled Clarkston Heights gravel and Clearwater gravels for zircon provenance analysis. Yellow stars are the base of measured stratigraphic sections. Basemap is a slope derivative from local lidar datasets (Washington Department of Natural Resources, 2016; 2023; 2024).
Panels of the geologic map (Figure 3) highlighting specific basalt and gravel units that identify past fluvial corridors in the Lewiston-Clarkston region. The geology in each panel is subdued in pastel colors except for key units. The mapped extent of key geologic units for each panel is emphasized with hachured fill and in opaque tones, with colors consistent with the legend in Fig. 3.
Topographic map showing the locations of modern river clast composition measurements, and one upper Clarkston Heights gravel clast count (in lavender). Inset pie charts show the relative proportion of different clast compositions measured at each site (Table S1). Basemap topography and hillshade were generated from 90-meter digital elevation data (NASA SRTM, 2013).
(A) Stratigraphic log of the Clarkston Heights gravels, compiled from two quarry exposures and the lowermost strath contact. Cyan dots show the stratigraphic levels sampled for detrital zircon provenance analysis. Grain size acronyms include: s=silt; vf=very fine sand; f=fine sand; m=medium sand; c=coarse sand; vc=very coarse sand; g=granule; p=pebble; cb=cobble-boulder conglomerate. Rose diagrams show the flow direction measured from imbricated clasts, with north at the top. Rose diagrams are also colored by the measured clast compositions, color coded by general rock type. Small cameras indicate where photographs of strata shown in Fig. 6B and 6D were taken. (B) Field photo showing strong imbricated gravel fabric common to the Clarkston Heights gravels. (C) Field photo of the main quarry measured and sampled, with Jim O’Connor for scale. (D) Rhythmically bedded lacustrine stratigraphy near the top of the Clarkston Heights gravels, with cheese knife for scale. All field photos taken by L.M. Staisch (USGS).
Detrital zircon age spectra for sampled Chalk Hills and Clarkston Heights gravels and published major modern rivers that are identified as important contributors in DZMix (Sundell and Saylor, 2017) model results. Colored vertical bars represent known zircon age ranges from source terrane. Red-yellow vertical bars are from central Idaho sources. Pink-purple vertical bars are from recycled Proterozoic and Archean sources in the upper Snake River terrane. Acronyms are as follows: Yellowstone hotspot (YHS); Challis volcanics (CV); Absaroka volcanics (AV); Atlanta lobe (AL); Contact pluton (CP); Ghost Neoproterozoic (GN); Grenville (GV); Lemhi Doublet (LD); Yavapai (Y); Mazatzal (M); Peace River Arch (PR); North Belt supergroup (NB); Wyoming craton (WC).
Maps of modern sampled drainages used for various detrital zircon unmixing models and accompanying detrital zircon provenance results for the Clarkston Heights gravels. Cumulative density function (CDF) plots are organized by sample stratigraphic elevation, with higher samples at the top and lower samples at the bottom of the figure. Graphs show the cumulative probability density functions for analyzed data in red, good fit in grey, and best fit in black. The D-statistic for the best-fit model are noted on each graph. Pie charts to the right of the HC-wSRP CDF graph show the relative proportion of modeled sources based on the top 100 models fit to the analytical data. Watersheds are color coded the same as the maps and in Figure 1.
(A) Drainage network during Chalk Hills and Clarkston Heights depositional timeframes (11.4 – 6 Ma). Lake Idaho drained through Hells Canyon (yellow line, abbreviated as HC), at least intermittently, between 10.2 – 6 Ma. A drainage divide between the eastern and western SRP was possibly due to the Yellowstone hotspot (YHS) track near the Picabo volcanic center. The upper Snake River drained into the Columbia basin via the route proposed by Staisch et al. (2021; red line). Both rivers met in the lower Columbia basin and drained towards the Pacific Ocean through the lower Columbia River Gorge. (B) After a depositional hiatus between the time that the Chalk Hills and Glenns Ferry formations were deposited, the SRP expanded eastward due to the YHS eastward motion and the onset of subsidence in the eastern SRP. Lake Idaho at this time grew to its largest extent. It is unclear whether Lake Idaho drained through Hells Canyon during this time. Downstream in the Columbia basin, the Ringold Formation’s lacustrine phase catalogues continued deposition of upper-Snake River derived provenance, suggesting the paleo-Snake River continued to circumnavigate the northern Cordillera for some time. The Columbia River incised through Wallula Gap and takes it course through the Dalles basin within this timeframe, likely draining “Lake Ringold”. (C) After 2.2 Ma, lacustrine deposition has largely ceased in the wSRP and fluvial gravels transgress over the basin surface, reaching Hells Canyon. This is the first sign of large-scale fluvial Snake River deposits, similar to bedload today, in the SRP. Major incision through bedrock in Hells Canyon initiates ca. 2.1 Ma. (D) Elevation of dated markers in Lake Idaho (Table S6) relative to the nearby Snake River (small colorful markers), relative to the Boise River (black markers), and relative to the Snake River at Farewell Bend (white open markers). Error bars indicate elevation ranges (horizontal) and age uncertainty (vertical). Blue dotted line signifies lake level change relative to its current outlet. Unit ages are shown on the far left. Acronyms are as follows: CCA: Cove. Rd - Crane Rd. intersection ash, SCA: Sunnyside Canal ash; SCRA: South Crane Road ash; APB: Aldape Park basalt; HCP: Haw Creek pumice; SGP: Sulfur Gulch pumice; CHT: Chalk Hills tephra; TSO: Terteling Springs ooids; SB: Shoestring basalt; TG-H: Tuana gravel – Hagerman; TG-MH: Tuana gravel – Mountain Home; WBB: Walters Butte basalt; JBB: Jackass Butte basalt; TmG-N: Tenmile gravel – Nampa; TmG-O: Tenmile gravel – Ontario; PBB: Pickles Butte basalt; BGT: basalt of Gowan Terrace; BMC: Basalt of Mores Creek; TmG: Tenmile gravel; TG: Tuana gravel; GF Fm: Glenns Ferry Formation; TS Fm: Terteling Springs Formation; CH Fm: Chalk Hills Formation; PC Fm: Poison Creek Formation. (E) Incision rate through Hells Canyon from Morriss et al. (2025). (F) Height of unit bases in Lewiston basin above the modern Snake River based on geologic mapping and lidar data (Fig 3). Because the Clearwater River likely shifted northward sometime between 10 – 6 Ma, we extract basal elevations near the modern Snake River – Asotin Creek confluence, upstream of the paleo-Clearwater River. Incision rate estimated using a linear least squares regression. (H) Global deep sea oxygen isotopic data plotted in blue and red (Zachos et al., 2001) compared to mean annual temperature (MAT) estimates in black squares from nearby sites in southwest Montana (Hyland et al., 2019). Global isotopic data are plotted such that temperature increases to the right. (H) Global deep sea carbon isotopic data plotted in blue and red (Zachos et al., 2001) compared to relative abundance of C4 vegetation from regional climate proxy data in southwest Montana in black dots (Hyland et al., 2019). (I) General timing of global and regional climatic changes (Zachos et al., 2001; Hyland et al., 2019).