Interaction of Outburst Floods with Basaltic Aquifers on the Snake River Plain: Implications for Martian Canyons

Home , Big Lost River, Bonneville Flood, Island Park Caldera, Red Rock Pass (Idaho)

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/270216431

Interaction of outburst floods with basaltic aquifers on the Snake River Plain: Implications for Martian canyons

ARTICLE in GEOLOGICAL SOCIETY OF AMERICA BULLETIN · APRIL 2014 Impact Factor: 3.87 · DOI: 10.1130/B31141.1

2 AUTHORS, INCLUDING:

William Amidon Middlebury College

28 PUBLICATIONS 280 CITATIONS

SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate, Available from: William Amidon letting you access and read them immediately. Retrieved on: 28 January 2016 Geological Society of America Bulletin, published online on 6 November 2014 as doi:10.1130/B31141.1

Interaction of outburst ﬂ oods with basaltic aquifers on the Snake River Plain: Implications for Martian canyons

William H. Amidon† and Arthur C. Clark Geology Department, Middlebury College, Middlebury, Vermont 05753, USA

ABSTRACT lar, the incision of fl uvial canyons into a basal- does not easily explain the observed canyon tic substrate. Recent observations of basaltic locations or morphologies. Understanding the Idaho’s Snake River Plain is underlain by canyons on Mars have placed renewed empha- timing and routing of pre-Bonneville fl oods is a young sequence of basaltic lava fl ows that sis on linking basalt canyon morphology to thus essential to understanding the morphologi- house one of the most conductive aquifers mechanisms of formation such as groundwater cal evolution of amphitheater-headed canyons in the world and have been sculpted by at sapping or surfi cial fl ooding (Lamb et al., 2006, on the Snake River Plain and on Mars. least three megafl oods in the last ~100 k.y. 2008; Malin and Carr, 1999). Because canyons The question of whether Snake River Plain The timing and routing of these fl oods, and on the Snake River Plain are infl uenced by one canyons formed by spring sapping or megafl ood their interaction with the underlying aquifer, of the most conductive aquifers in the world and erosion has proven to be an excellent analog to have taken on renewed signifi cance because have also experienced multiple outburst fl oods, study the formation of Noachian-aged amphi- they have carved amphitheater-headed dry they provide a unique setting in which to tease theater-headed canyons on Mars. Canyons such canyons analogous to those found on Mars. apart the infl uences of these two processes on as Nanedi, Nirgal, Mamers, and Bahram Valles In this study, we use cosmogenic 3He and canyon morphology. Essential to this effort is meet the traditional morphological criteria for 21Ne dating of fl ood-deposited boulders to an understanding of when past outburst fl oods spring sapping, yet they have also been attrib- show that the Big Lost River and Bonne ville occurred on the Snake River Plain, how those uted to megafl ooding (Lamb et al., 2008; Malin fl oods were closely spaced in time at ca. 22.3 fl oodwaters were routed, and the nature of sur- and Carr, 1999). One argument for ground water and ca. 17.5 ka, respectively. Most of the dry face fl ood interaction with the basaltic aquifer. sapping is that weakly developed tributary val- canyons record signifi cant erosion during This study considers how fl oodwater routing ley networks suggest limited overland fl ow the Big Lost River fl ood, despite its much infl uenced development of four amphitheater- (Malin and Carr, 1999). A second argument is smaller magnitude than the later Bonne ville headed (or more precisely “theater”-headed) that canyon and valley networks are not evenly fl ood. We explain this puzzling observation canyons, which enter the Snake River from distributed across the landscape, as would by proposing a composite erosion model the north. Because these canyons have mor- be expected from precipitation-driven runoff in which erosion during the Big Lost River phological traits commonly associated with (Irwin et al., 2008; Whitehead, 1987). However, fl ood was partially accomplished by routing formation by spring sapping, previous studies a groundwater sapping mechanism has been of fl oodwaters through the Snake River Plain have focused on determining whether they were questioned by studies showing that sapping can- aquifer. Topographic analysis shows that Big formed by spring sapping or by overland fl ow. not form amphitheater-headed morphologies Lost River fl oodwaters ponded in the Terre- Large springs found at the head of each canyon in vertically jointed basalt due to its internal ton Basin, infi ltrated into the aquifer, and prompted early models to call upon ground- strength and resistance to chemical dissolution likely emerged as return fl ow in watersheds water sapping (Stearns et al., 1938), whereas (Lamb et al., 2006). upstream of the dry canyons. We propose plunge pools and fl ood-deposited boulders A clearer understanding of the formative that sustained and focused erosion associated caused later models to call upon erosion during mechanisms for Martian canyons will provide with return fl ow over months to years could the Bonneville fl ood (Malde, 1968). Neither of critical constraints on the Noachian hydrologic explain the unique morphology of some dry these models has proven entirely correct. The budget, which featured a large northern ocean canyons. Such a model also explains why original spring sapping model has been dis- fed by an interconnected network of ground- most dry canyons are coincident with springs carded because there is little evidence for seep- water and surface water (Andrews-Hanna and surface watersheds, and it may provide a age weathering, and it is widely recognized that and Lewis, 2011; Harrison and Grimm, 2005; model for the way in which morphologically typical spring discharges cannot erode consoli- Krause and Whitlock, 2013). The exact nature similar canyons evolved on Mars. dated basalt (Lamb et al., 2006). The Bonneville of hydrologic cycling through this system fl ood erosion hypothesis has been weakened remains unclear, including residence times of INTRODUCTION by pre-Bonneville cosmogenic ages on fl ood- surface and groundwater and the nature of sur- deposited boulders and spillways, which sug- face-atmosphere water vapor fl uxes. If canyons Idaho’s Snake River Plain is a recently resur- gest earlier fl oods carved the canyons (Lamb are interpreted to form from surface-water run- faced basaltic landscape that provides a fresh et al., 2008, 2014). However, the timing and off, they may directly constrain the amount and slate to record geomorphic processes, in particu- routing of pre-Bonneville fl ood events remain rate of precipitation (Matsubara et al., 2013). If poorly documented, and the most obvious route interpreted to form from groundwater sapping, †[email protected]. of fl oodwaters down the Snake River canyon they would suggest groundwater recharge in a

GSA Bulletin; Month/Month 2014; v. 1xx; no. X/X; p. 1–14; doi: 10.1130/B31141.1; 9 ﬁ gures; 1 table; Data Repository item 2015019.

Amidon and Clark restricted part of the planet, redistributed largely return fl ow in upstream watersheds combined was interspersed with periods of landscape by a globally interconnected aquifer (Harrison with direct discharge of groundwater from development in which well-developed rivers and Grimm, 2004). springs at canyon heads. If correct, this “com- traversed the Snake River Plain (Whitehead, To better understand the evolution of can- posite model” (e.g., Schumm and Phillips, 1984). For example, detrital zircons in buried yons on the Snake River Plain and on Mars, 1986) could reconcile some of the confl icting sediments near Wendell, Idaho, suggest that the this work establishes a chronology of three observations of the sapping and fl ood erosion Big Lost River once fl owed to the west, meet- recent outburst fl oods on the Snake River Plain. models, such as why most of the dry canyons ing the Snake River near Hagerman, Idaho We present cosmogenic 3He and 21Ne ages for are collocated with springs and are linked to (Hodges et al., 2009). The repeated inundation fl ood deposits that date the Bonneville and Big tributary watersheds yet also contain fl ood- of the landscape by basalt fl ows has fi lled these Lost River fl oods to ca. 17.5 and ca. 22.3 ka, deposited boulders and morphologies indicative ancient river valleys and pushed the modern-day respectively, and are consistent with a >100 ka of erosion during large fl ood events. Snake River to the southern edge of the Snake age for the Henry’s Fork fl ood. Dating of fl ood- River Plain, creating a network of lava-fi lled deposited boulders at four different dry canyons OVERVIEW paleocanyons (Malde and Powers, 1962). shows that only one canyon can be temporally Although the modern Snake River Plain is linked to the Bonneville fl ood at 17.5 ka and that Geologic Setting of the Snake River Plain largely devoid of rivers, numerous dry canyons four others record erosion during the older Big join the Snake River from the north; they lack Lost River fl ood. Because the larger Bonneville The fl at topography of the Snake River Plain modern rivers yet typically contain effusive fl ood did not heavily erode these canyons, it is belies a lava-covered tectono-fl uvial landscape springs. All dry canyons are carved in basalt, unlikely that the much smaller Big Lost River containing numerous conduits for ground- typically ~30–100 m deep and hundreds of fl ood could have eroded them by conventional water (Fig. 1). Basaltic lava fl ows younger meters wide. They have vertical walls and rela- overland fl ow down the Snake River Canyon. than ca. 6 Ma are underlain by silicic volcanics tively fl at fl oors, culminating in arc-shaped head- Instead, we propose that fl oodwaters may have emplaced between ca. 16 and 5 Ma (Armstrong walls, typically lacking well-developed fl uvial reached the western dry canyons by ponding in et al., 1975; Leeman and Bonnichsen, 2005). spillways along the rim. Most dry canyons sit the Terreton Basin and transmission through Recent drilling results reveal >2-km-thick at the mouths of weakly developed watersheds, the Snake River Plain aquifer. Canyon erosion basalt sequences with abundant intercalated which would have aggregated any upstream pre- would then be driven by the accumulation of fl uvial sediments, demonstrating that vol canism cipitation or groundwater discharge. The upper

Figure 1. Map of the Snake River Plain (SRP) showing the location of the dry canyons relative to the source regions of major fl oods. The Bonneville fl ood was sourced from Red Rock Pass, and the Henry’s Fork fl ood was sourced from either the Island Park or Falls River region. The Big Lost River fl ood was sourced from Glacial Lake East Fork, perhaps drawing additional water from the hypothetical Thou- sand Springs Lake. Water ponded in the Terreton Basin, overfl owing through pour points to the east, and transmitting through the Snake River Plain aquifer (white lines) toward the dry canyons of the Snake River (DC—Devil’s Corral, BL—Blue Lakes Canyon, B + B—Box and Blind Canyons, Bi—Billingsley Canyon, and S—Stubby Canyon).

2 Geological Society of America Bulletin, Month/Month 2014 Geological Society of America Bulletin, published online on 6 November 2014 as doi:10.1130/B31141.1

Interaction of outburst ﬂ oods with basaltic aquifers on the Snake River Plain

reaches of these watersheds are characterized by Snake River are probably also important con- that the Martian aquifer reached a saturated youthful chaotic lava fl ows abruptly giving way duits. As Stearns et al. (1938, p. 136) envisioned state during the Noachian in which ground- to moderately developed channels in the lower it, a network of highly conductive lava-fi lled water intersected the surface along topographic reaches of the watersheds. Low precipitation canyons “act as collecting channels and serve as escarpments (Andrews-Hanna et al., 2007). on a high-permeability basaltic substrate means a great underground drain for the Snake River Return fl ow to the surface is thought to have that these channels are typically devoid of natu- Plain aquifer.” fi lled many craters, leading to systems of inter- ral runoff, and the magnitude of historic storms The unique combination of active faulting and connected crater lakes, which ultimately drained is insuffi cient to induce signifi cant channel ero- basaltic eruptions makes the Snake River Plain into a large ocean basin (Clifford and Parker, sion (Lamb et al., 2008). Because the water- aquifer highly compartmentalized and unpre- 2001; Fassett and Head, 2008; Krause and Whit- sheds descend from the higher-elevation Snake dictable (Smith et al., 2004). Well observations lock, 2013). This model is supported by the River Plain interior, they have not been inun- show rapid variation in water table depth and fact that hydrologic models roughly predict the dated by the three most recent outburst fl oods hydraulic head across relatively short distances distribution of canyon networks and evaporite (the Henry’s Fork, Big Lost River, and Bonne- (Anderson and Liszewski, 1997; Bartholo may deposits (Andrews-Hanna et al., 2010; Nimmo ville). These fl oods emanated from the west, et al., 1997a). For example, in the Big Lost et al., 2002). Likewise, the density of Noachian north, and south, respectively, before joining the River area, modern isotopic tracers delineate a canyon networks and their elevation have been Snake River at or above Pocatello and follow- fast path fl owing into a region where the water shown to decrease with time, presumably due to ing the main canyon of the Snake River into the table drops to ~457 m (~1500 ft), probably the lowering of the global water table as Mars Columbia River. One exception to this is the Big fl owing into an ancient canyon of the Big Lost became increasingly cold and arid during the lat- Lost River, which fl owed directly into the Terre- River or into permeable deposits associated est Noachian (Harrison and Grimm, 2005). ton Basin, ponding ~3 × 108 m3 of water before with the axial rhyolitic vent system (Johnson Despite evidence for near-global saturation overfl owing from pour points at its western end et al., 2000). Shallow levels of the aquifer also of the Martian aquifer, regional geology must and continuing down the Snake River (Fig. 1). appear compartmentalized and in some cases have exerted a fi rst-order control on the rate and The basalts of the Snake River Plain create confi ned. For example, nearly half the springs pattern of groundwater fl ow. Many high-eleva- one of the most highly conductive aquifers in along the Snake River have discharges that are tion canyons seem to require local sources of the world, effectively an underground river sys- uncorrelated to water levels in nearby wells recharge, such as glaciation of the Tharsis high- tem that detours large mountain rivers through (Janczak, 2001). Likewise, tracer studies show lands (Harrison and Grimm, 2004) or localized porous basalts before they rejoin the Snake that groundwater can travel horizontally at least infi ltration through ponding in canyons (Per- River at springs between the towns of Twin several kilometers laterally within the unsatu- kins et al., 2000). It has also been noted that the Falls and Bliss. These springs include 11 of the rated zone, presumably through fl ow contacts location of outfl ow channels likely refl ects the largest springs in the United States, with a total and fractures (Nimmo et al., 2002). impact of aquifer heterogeneities (Harrison and annual discharge of ~5.4 × 109 m3 (Meinzner, The very high transmissivity of basaltic Grimm, 2009), and that some craters were likely 1927; Whitehead and Covington, 1987). aquifers makes them uniquely suited to rapidly fed by higher rates of groundwater recharge Because many of these springs are located at absorb ponded fl oodwaters. Aquifer charging than others (Fassett and Head, 2008). Strong the mouths of amphitheater-headed canyons, during the Missoula outburst fl ood is recorded heterogeneities in shallow groundwater fl ow they have long been thought to play a role in by depleted δ18O in deep groundwater within are likely on Mars given the role of impact frac- canyon formation (Stearns et al., 1938). Water the Columbia River Basalt (Brown et al., 2010). turing and repeated burial of the landscape by table depths in the western Snake River Plain Charging of the aquifer is thought to have basalt fl ows during and after the Noachian wet aquifer are typically <30.5 m (100 ft) from the occurred during ponding of ~1.2 × 1012 m3 of period (Goudge et al., 2012). Better understand- surface and average ~61 m (200 ft) across the water in the Pasco Basin, behind the water gap ing of what controls heterogeneous groundwater entire Snake River Plain (Whitehead, 1992). near Wallula, Washington. This example, com- fl ow in response to large-scale infi ltration on the Groundwater follows a westward hydraulic bined with new evidence presented here, sug- Snake River Plain would improve our under- gradient of ~5 ft/mile (~1 m/km) with regional gests that infi ltration of ponded fl oodwaters is standing of the Martian hydrologic system. conductivities of 0.1–104 m2/d, and average a fundamental aspect of basalt-megafl ood inter- linear transport rates of ~30 m/d, reaching hun- actions on both Earth and Mars. MEGAFLOODS ON THE dreds of meters per day along fast paths (Acker- SNAKE RIVER PLAIN man, 1991; Anderson et al., 1999). Martian Hydrology Flow in the Snake River Plain aquifer is Henry’s Fork Flood highly heterogeneous, and the exact plumb- The hydrologic characteristics of Mars are ing that feeds canyon-head springs is unclear. strikingly similar to those on the Snake River The earliest known fl ood on the Snake River Detailed studies show that hydraulic conductiv- Plain, setting the stage for a Noachian period Plain occurred along the Henry’s Fork River, ity is controlled by poorly mapped patterns of in which groundwater fl ux could have easily which drains west Yellowstone, the Tetons, and lithology, dikes, and fractures. At the local scale, exceeded surface-water fl ux. Like the Snake Island Park caldera, before joining the Snake the highest conductivities are observed in thin River Plain, Martian bedrock is composed pre- River near Idaho Falls (Fig. 1). Three lines of pahoehoe fl ows and their associated rubble, as dominantly of basalt and is an immature land- evidence suggest large fl ood event(s) were well as scoria and spatter deposits associated scape containing numerous closed basins that sourced from the upper Henry’s Fork during with volcanic vents. The lowest conductivities promote infi ltration (e.g., impact craters). The the late Pleistocene. First, deposits of imbri- are associated with thick tube-fed lavas and Martian subsurface is usually modeled as a cated megaboulders have been mapped by extremely rare dense dike swarms (Anderson single unconfi ned aquifer, assumed to have been Scott (1982) near the Highway 20 bridge east et al., 1999; Kuntz et al., 2002). On the regional recharged by rainfall that increased toward the of Ashton and ~4 km (~2.5 miles) downstream scale, numerous lava-fi lled paleocanyons of the equator (Andrews-Hanna et al., 2010). It appears from the Ashton reservoir (Fig. 2). Second,

Geological Society of America Bulletin, Month/Month 2014 3 Geological Society of America Bulletin, published online on 6 November 2014 as doi:10.1130/B31141.1

Amidon and Clark

turbance, we sampled only pristine boulders that were imbricated against neighboring boulders and displayed patterns of lichen growth consistent with long-term stability.

Big Lost River Flood

Another pre-Bonneville flood is recorded along the Big Lost River south of Arco, Idaho (Figs. 1 and 3). Evidence for this fl ood comes primarily from a basaltic scabland south of Arco, which includes eroded cataracts, streamlined topography, and elongate boulder bars (Rathburn, 1993). Most features are eroded into the <609-k.y.-old vent 5371 lava fl ow (Skipp and Kuntz, 2009). Peak discharge estimates range from ~60,000 to ~30,000 m3/s, based on step back-water modeling and critical shear stress calculations respectively (Bartholomay et al., 1997a; Brown et al., 2010). After entering the Snake River Plain from the north, the Big Lost River terminates in a basaltic basin currently managed as a settling zone for fl oodwaters by the Idaho National Laboratory (Ben- nett, 1990). Topographic analysis shows that fl oodwaters bypassed the modern settling area and fi lled the Terreton Basin with ~3 × 108 m3 of water, overfl owing to the Snake River at its Figure 2. Area near Ashton, Idaho, where evidence of the Henry’s western end (Fig. 1). This volume estimate is Fork fl ood is recorded in a system of cataract lakes (dashed lines) derived from fi lling of the 10 m National Eleva- and boulder deposits such as the one shown in the inset. Samples for tion Dataset (NED) digital elevation model to cosmo genic dating were collected from seven boulders on the outskirts the extent shown in Figure 1. of an excavated gravel quarry at the Vernon Bridge fi shing access. The source of the Big Lost River fl ood has stimulated much discussion, and the reader is referred to Knudsen et al. (2002) for a thor- several eroded cataracts (now lakes) are pres- evidence for fl ooding is abundant. No direct age ough summary. The most likely source is Gla- ent just north of Ashton, where fl oodwaters constraints exist on the timing of the Henry’s cial Lake East Fork, an ice-damned lake in eroded the surface of the 1.2-m.y.-old Huckle- Fork fl ood(s). Loess cover and other geomor- the Big Lost River headwaters (Barton, 2008; berry Ridge Tuff (Fig. 2). Third, a large volume phic relationships suggest that the major event Welhan and Reed, 1997; Whitehead, 1992). of obsidian-rich gravel was deposited in planar is at least Bull Lake in age (Scott, 1982). This However, Knudsen et al. (2002) questioned this sheets, which now underlie much of the Henry’s idea is supported by a 10-m-thick deposit of source due to a lack of fl ood deposits or scour- Fork valley. If interpreted as fl uvial in origin, a obsidian-rich gravels in Market Lake, which are ing immediately downstream, and the existence prominent dune near the town of Parker, Idaho, likely correlative to those found upstream in the of well-developed soils along the fl ood path. implies water at least 5 m deep across a 10-km- Henry’s Fork valley. These gravels at Market Moreover, the volume of Glacial Lake East Fork wide valley (Pierce and Scott, 1982). Lake are constrained to be ~100–140 k.y. old (~6.3 × 109 m3) may have been insuffi cient to The exact timing and source of the Henry’s based on bracketing optically stimulated lumi- generate the peak discharges inferred at the Big Fork fl ood(s) remain unknown. Scott (1982) nescence (OSL) ages (Allison, 2001). Lost River scabland (Bartholomay et al., 1997b; argued for two distinct events, one of which We collected seven samples from a boulder Barton, 2008). A solution proposed by Rathburn he thought emerged from Island Park cal- bar deposited at the confl uence of Henry’s Fork (1989) is the existence of a transient pluvial dera, where cataract lakes and boulder depos- and Black Spring Creek, ~4.8 km (~3 miles) lake in the Thousand Springs Valley, contained its may record fl ooding near Harriman State west of Ashton (Fig. 2). We infer that the bar behind large alluvial fans near Mackey Reser- Park (Fig. 1). Geomorphic evidence suggests was deposited in a zone of fl ow deceleration at voir (Fig. 1). The initial outburst fl ood would a fl ood may also have been sourced from the the confl uence of two channels. The rounded have caused this lake to overtop its alluvial dam, headwaters of the Falls River in SE Yellow- boulders are ~1.5–3 m in diameter and con- doubling or tripling the downstream discharge. stone (Fig. 1). The Falls River drains a region sist exclusively of the Huckleberry Ridge Tuff The timing of the Big Lost River fl ood is con- of streamlined topography containing numer- sourced from a canyon immediately upstream. strained by 3He and 21Ne dating of three fl ood- ous cataracts, headscarps, dry channels, and Because these boulders are heavily weathered deposited boulders collected from the scablands boulder deposits. Although some of these fea- and the boulder bar has been exhumed, the ages downstream of Arco (Cerling et al., 1994). These tures can be attributed to subglacial erosion presented here should be considered minimum boulders give ages of 19.1, 21.9, and 43 ka, with beneath the Pinedale and Bull Lake ice sheets, bounds on the fl ood event. To avoid human dis- a preferred fl ood age of ca. 20.5 ka. A sedimen-

4 Geological Society of America Bulletin, Month/Month 2014 Geological Society of America Bulletin, published online on 6 November 2014 as doi:10.1130/B31141.1

Interaction of outburst ﬂ oods with basaltic aquifers on the Snake River Plain

In 2011, six boulders were sampled from Bone Cataract, a fi eld of 1.5–3-m-diameter angular boulders deposited ~10–50 m downstream of an ~4-m-high headwall. No attention was paid to reconstructing fl ow surfaces at that time. In 2012, 10 boulders were sampled from Lost Moon Cataract, and three additional boulders were collected from Bone Cataract, with care taken to reconstruct (and avoid) the original surface of the lava fl ow prior to sampling. A shielded sample was collected from a nearby cave with ~3.5 m of vertical and ~2 m of lateral shielding (Fig. 3).

Bonneville Flood

The Bonneville fl ood has been studied in great detail, and the reader is referred to Malde (1968) and O’Conner (1993) for excellent reviews. The fl ood was caused by the catastrophic draining of Lake Bonneville when an alluvial dam failed at Red Rock Pass (Fig. 1), causing Lake Bonne- ville to drop ~100 m and release ~4750 km3 of water down the Portneuf River over a period of 6–12 mo (Godsey et al., 2005; Malde, 1960; Oviatt et al., 1992). Floodwaters joined the Snake River near Pocatello, Idaho, eroding large cataracts and depositing abundant gravel and boulders downstream. The timing of the Bonne- ville fl ood is well constrained to ~17.5 ± 0.5 ka by 14C dating of shorelines in the Bonneville basin (Godsey et al., 2005). Despite the fact that all fl ood-related features along the Snake River have been attributed to the Bonne ville fl ood, only a single fl ood-deposited boulder along the Portneuf River has actually been dated to Bonne ville age (Cerling and Craig, 1994). We dated the Bonneville fl ood at the Portneuf Narrows in Inkom, Idaho (Figs. 1 and 4). Because this site is ~50 km downstream from Red Rock Pass and well upstream of the Snake River Plain, the boulders were unambiguously depos- Figure 3. Map of the Big Lost River scabland where the Big Lost ited by the Bonneville fl ood. We sampled six River fl ood entered the Snake River Plain. Arrows denote the fl ow 1.5–3-m-diameter boulders from a well-imbri- direction of fl oodwaters, and streamlined basaltic topography is vis- cated deposit atop the 430 ± 70 ka Portneuf lava ible in the inset image. Bone cataract is shown in the photo at bot- fl ow, which once fi lled the entire valley (Rodgers tom, the location of which is shown by the camera icon in the inset. et al., 2006). Boulders were likely plucked from an eroded gap in the lava fl ow ~100 m upstream Bedrock geology and K-Ar ages are from Skipp et al. (2009): Qbd— (Fig. 4). All of the boulders showed well-devel- Blue Dragon fl ow, Qsb: Sixmile Butte fl ow, Qcb—Crater Butte fl ow, oped patinas and 1–2 mm hydration rinds sug- Qtb—Teakettle Butte fl ow, Qlr—Lost River fl ow. Numerical suffi xes denote the vent number of unnamed fl ows. gesting zero erosion of the boulder surfaces. A shielded sample was collected from a cave ~500 m upstream, with ~5 m of vertical over- tary record of the fl ood may be recorded by a a >12-k.y.-old basalt-clay diamict found in the burden and ~2 m of lateral shielding. highstand shoreline in Lake Terreton, which has stratigraphy at the North Lake embayment con- been dated to 21.2 k ka by 14C dating of gastro- tains basalt clasts that are likely fl ood-rafted DRY CANYONS OF THE SNAKE RIVER pods at Clay Butte (Anderson and Lewis, 1989). (Anderson et al., 1996). This shoreline would record a minimum age for To improve the age estimates of Cerling et al. To determine which fl oods eroded the dry the fl ood because it likely developed during a (1994), we collected fl ood-deposited boulders canyons, we dated fl ood deposits in Devil’s Cor- regressive phase of the fl ooded lake. Likewise, from two separate erosional cataracts (Fig. 3). ral, Blue Lakes, and Box Canyons (Fig. 5), and

Geological Society of America Bulletin, Month/Month 2014 5 Geological Society of America Bulletin, published online on 6 November 2014 as doi:10.1130/B31141.1

Amidon and Clark

The “lower canyon” site is a group of very large, >3-m-diameter boulders deposited near the base of a large cliff near the exit of canyon A. The “talus amphitheater” site is an apron of 1–3 m angular talus blocks ﬁ lling the large alcove between lobes A and B. A shielded sample was collected from a cave in canyon A, with ~4 m of lateral shielding and ~6 m of vertical shielding.

Blue Lakes Canyon

Blue Lakes Canyon is traditionally thought to have formed by fl oodwaters rejoining the Snake River Canyon from the Eden overfl ow channel (Figs. 1 and 7; Malde, 1968; O’Connor, 1993). It is characterized by a single, deep, and fl at-fl oored canyon lobe headed by a spectacular 95-m-high headwall directly above an active spring, which provides the municipal water supply for the city of Twin Falls. A well-developed spillway feeds the canyon head and a plunge pool below. Two bedrock strath terraces are preserved along the sides of the canyon, ~60 m above the canyon fl oor near the contact between the underlying Hansen Butte and Rocky Butte fl ows. Five samples were collected from 1.5- to 2-m-diameter fl ood-deposited boulders atop the eastern terrace, clearly deposited by a jet of water emerging from Figure 4. Shaded relief map of the Portneuf River Valley and the a plunge pool. Four additional samples were col- path that Bonneville fl oodwaters followed from the outlet at Red lected from 2 to 3 m angular blocks on the talus Rock Pass onto the Snake River Plain. Six samples were collected apron immediately to the east. A shielded sample from boulders atop the Portneuf lava fl ow, in the town of Inkom, was collected from a cave, with ~2 m of lateral Idaho. and ~25 m of vertical shielding.

Box Canyon we compared these ages with published ages which postdate erosion of the main canyon and from Stubby Canyon (Lamb et al., 2014). Here, record multiple episodes or stages of fl ooding. Box Canyon enters the Snake River ~32 km we describe the morphology and samples col- Finally, the uppermost strath terrace is littered (~20 miles) downstream from Blue Lakes Can- lected at each of the dry canyons. with angular boulders derived from the toppling yon, occurring partially within the mapped of basaltic columns during retreat of the basaltic extent of Bonneville fl oodwaters along the Devil’s Corral headwall (Fig. 6). contact between the ca. 395 ka Flat Top Butte The location and complex morphology of fl ow and the Bacon Butte Younger Flow (Giller- Devil’s Corral is situated where the Eden Devil’s Corral led to our initial hypothesis that it man et al., 2005). The canyon is ~3 km long, overfl ow channel rejoins the main Snake River, was formed from multiple fl ood events. To test single lobed, and fl at fl oored, terminating in an accommodating about two thirds of peak fl ood this hypothesis, we collected samples for cos- ~35-m-high headwall fed by a single ~3-m-deep discharge routed via upstream ponding and mogenic 3He dating from seven different sites spillway delivering water to two dry plunge overfl ow of the Burley Basin (Fig. 1; O’Connor, (Fig. 6). The “lobe A–upper” site contained pools (Fig. 8). A large spring discharges ~10 1993). The canyon rims are composed of the angular, 1–3 m boulders sourced directly from m3/s from the mouth of the canyon, giving rise Rocky Butte basalt dated at ~95 ka, which the headwall. The “lobe A–lower” contained to Box Canyon Creek. overlies older undifferentiated basalts (Kauff- similar boulders in a distinctly separate deposit man et al., 2005). The morphology of Devil’s ~200 m further down canyon A. The “high Corral differs considerably from the dry can- spire” site is a pile of 1–2 m boulders retaining TABLE 1. GEOMETRIC PROPERTIES yons found downstream to its west. First, it excellent columnar structure that can be restored OF DRY CANYONS Width Depth Length consists of two canyon lobes (A and B), with to an undercut basalt spire ~30 m away. The (m) (m) (m) L/W D/W higher width/length and width/depth ratios than “high plateau” site is a sparse fi eld of 1.5–2 m Devil’s Corral (A) 570 17 1050 1.8 0.03 the western canyons (Table 1; Fig. 6). Second, rounded boulders strewn across the lower Devil’s Corral (B) 633 25 800 1.3 0.04 Blue Lakes 370 95 1350 3.6 0.26 the canyon heads do not contain active springs. reaches of lobe A. The “plunge pool” site is a Box Canyon 150 65 2092 13.9 0.43 Third, canyon A contains stair-stepping strath pile of well-sorted and perfectly rounded basalt Stubby Canyon 125 70 2470 19.8 0.56 surfaces rather than a fl at canyon fl oor. The boulders, ~0.75–1 m in diameter, deposited in a Note: L/W—length/width ratio, D/W—depth/width straths are connected by incised inner spillways, plunge pool at the base of a canyon B headwall. ratio.

6 Geological Society of America Bulletin, Month/Month 2014 Geological Society of America Bulletin, published online on 6 November 2014 as doi:10.1130/B31141.1

Interaction of outburst ﬂ oods with basaltic aquifers on the Snake River Plain

1.6 A’ 25 km A’ ) Flat Top Surface 1.4 Butte Elevation Stubby Lake 1.2 Terreton A Hypothetical Water Table vation1.0 (km During BLR Flood

Ele Snake River 0.8 200 160 120 80 40 0 A Hager- man Billingsly

Sand Springs Flat Top Butte Box Jerome ID10-49 Blind

Blue Devil’s Lakes Corral

5 km Twin Falls

Figure 5. Six major dry canyons of the Snake River on a 10-m-resolution digital elevation model. Inset shows a surface-elevation profi le extracted along the red line, which approximates the direction of groundwater fl ow from Lake Terreton to Box Canyon during the Big Lost River (BLR) fl ood. Black arrows denote emergent channels possibly formed by return fl ow. White dot shows the location of sample ID10–49.

Geomorphic Map LAU LAL HS Lobe A A B Figure 6. The upper panel shows a map of the Devil’s LC 200 m Spring Corral canyon complex TA Panel B with lobes A and B la- beled. Seven sites were Snake River PP HP Lobe B sampled for cosmogenic 80 dating: LAU—lobe A up- Snake Riv 42 94 per, LAL—lobe A lower, er Talus 112 Boulder Bar Samples HS—high spire, LC— C Shoshone Shielded Main lower canyon, TA—talus 500 m Falls Canyon C.I. = 3 m Sample amphitheater, HP—high From Cave Boulders Spillways Edge of Strath plateau, and PP—plunge Talus Plunge Pools pool. The middle panel Scarps Photo shows a simplified geo- 250 m 1 2 morphic map of the site.

The photos at bottom Boulder Bar show examples of a typi- (2–4 m diam.) Plunge cal boulder-free inner Canyon Rim Pool canyon surface in lower Spring canyon A and an angular Spring Figure 7. (A) Shaded relief map of Blue Lakes Can- boulder lag at the head yon and the adjacent canyon of the Snake River. of canyon A. Photo loca- Samples along the Snake River are cosmogenic Photo 1: SW down Canyon A Photo 2: NW across Canyon A tions are shown by the 3 boulder-free surface angular boulders He ages (ka) from ﬂ uvially polished bedrock sur- camera icon in the mid- (2–4 m diam.) faces reported in Amidon et al. (2010). (B) Aerial dle panel. C.I.—contour photo showing the location of cosmogenic boulder interval. samples from the upper terrace, and a shielded way sample from a cave in the canyon wall. (C) Photo Edge of Spill taken from the canyon rim; the location is shown by the camera icon in panel B.

Geological Society of America Bulletin, Month/Month 2014 7 Geological Society of America Bulletin, published online on 6 November 2014 as doi:10.1130/B31141.1

Amidon and Clark

A fi ed over a liquid nitrogen cold fi nger, hot and cold SAES getters, and cryogenically focused at 12 K (32 K) prior to gas inlet at 32 K (75 K). Because the 40Ar2+ peak is pseudoresolved from the 20Ne+ peak, no correction for the 40Ar iso- 44 2+ bar was made. Corrections for the CO2 isobar B C were <2% of the mass 22 signal and were made 44 2+ 44 by determining a CO2 / CO2 ratio that was constant to ±10% for all samples, standards, and blanks (e.g., Amidon and Farley, 2012). Blank corrections were made using the average of re- extract values before and after each analysis, which were generally <1% of measured values. The 21Ne excesses were computed assuming all 20Ne was air derived and using a 21Ne/20Ne ratio of 0.002959 and a 22Ne/20Ne ratio of 0.102 Figure 8. (A) Aerial photo of Box Canyon with sample locations of (Nieder mann, 2002). All analyses were stan- boulders, eroded bedrock (S1-S4), and a shielded sample (ABC 5). dardized using the Caltech Air standard, for (B) Small straths beveled into the edge of the spillway. (C) Photo which the aliquot concentration is manometri- looking toward the headwall showing strath terraces beveled into cally determined to <2% accuracy. Precision on spillway. individual analyses was determined by linear regression of the signal to time zero. All cosmogenic exposure ages discussed in Existing geochronology suggests that fl ood cosmogenic 3He ages of spillways and scoured this paper were computed by adopting a sea- deposits in the canyon record pre-Bonneville surfaces along the rim of Stubby and Pointy level high-latitude (SLHL) 3He production rate events. Flood-deposited boulders yield rescaled Canyons record 18–24 ka ages, whereas the of 114 ± 4.5 at g–1 yr–1, and a 21Ne production cosmogenic 3He ages from 20 to 47 ka, and an youngest age in Pointy Canyon is ca. 11.5 ka. rate of 18.7 at g–1 yr–1, both using the Lal/Stone incised spillway notch at the canyon head gives Together, these ages suggest that Stubby Can- time variable scaling model (Balco et al., 2008; a rescaled age of ca. 42 ka (Lamb et al., 2008). yon was formed by pre-Bonneville fl ood events Kober et al., 2011). Because the 3He produc- To augment existing geochronology, we col- and then abandoned as the Wood River contin- tion rate was adopted from a compilation of lected samples from three large boulders in the ued to cut headward, forming Pointy Canyon production rate studies at broadly similar eleva- lower canyon, which were imbricated against (Lamb et al., 2014). tions in the western United States, we estimate one another on a terrace ~15 m above the river an uncertainty of ~5% (Goehring et al., 2010). level (Fig. 8). In addition, four bedrock samples FIELD AND LABORATORY METHODS No corrections were made for snow cover or were collected from small bedrock straths along erosion, and topographic shielding corrections a vertical transect from the deepest part of the Laboratory Methods were made using the Cronus calculator where spillway to the highest point on the immediately appropriate. adjacent lava fl ow (Fig. 8). An additional sample The 2–4-cm-thick samples were collected Because uncertainties in production rate scal- was collected from a fl uvially eroded basalt sur- from fl at, well-preserved boulder surfaces ing models are mostly systematic when applied face within a dry river channel ~5 km upstream with a hammer and chisel. Several kilograms to closely located sites at similar elevations, dif- from Box Canyon. A shielded sample was pre- of rock were crushed and sieved to the target ferences in 3He ages between sites are known viously analyzed from a road cut in the Flat Top grain size, which varied from 300–600 μm to more precisely than their absolute ages. Recog- Butte fl ow, with ~30 m of vertical shielding 75–200 μm depending upon the lithology. A nizing this, cosmogenic ages are reported with (Aciego et al., 2007). Frantz iso dynamic separator was used to iso- two sets of uncertainties: (1) uncertainties aris- late the least magnetic (and thus most Mg-rich) ing only from external reproducibility of sam- Stubby Canyon at Malad Gorge olivine or pyroxene fractions within the range ples at the same site, and (2) fully propagated of 0.3–0.6 A. Separates were further isolated uncertainties associated with the chosen scaling Stubby Canyon is a dry canyon that joins the to densities >3.3 g/cm3 using methylene iodide model (reported in parentheses). Snake River ~32 km (~20 miles) north of Box heavy liquid. The resultant separate was then

Canyon (Figs. 1 and 5). Recent work on Stubby sonicated in 5% HF/HNO3 until all adhering RESULTS Canyon by Lamb et al. (2014) is relevant to the groundmass or staining was removed, resieved, conclusions of this paper and is therefore worth and handpicked to remove any remaining con- Helium and neon data are reported in supple- summarizing. The morphology of Stubby Can- taminants. All samples were crushed under mentary Tables DR1–DR6,1 and ages are sum- yon is similar to Box and Blue Lakes Canyons, alcohol to a <26 μm powder and dried at 60 °C, marized in Figure 9. In general, 3He/4He ratios characterized by a single ﬂ at-ﬂ oored canyon and a 75–550 mg sample of powder was loaded are >20, suggesting that the 3He is cosmogenic lobe headed by an effusive perennial spring. into foil balls for analysis (Amidon et al., 2009). However, Stubby Canyon is unique from the Helium (neon) analyses were heated to 1GSA Data Repository item 2015019, Tables other canyons because it is closely associ- 1350 °C (1400 °C) in a double-walled furnace DR1–DR6 and Figure DR1, is available at http://www ated with a V-shaped canyon (Pointy Canyon), coupled to a MAP 215 (GV Helix-SFT) mag- .geosociety.org /pubs /ft2015 .htm or by request to containing the modern Wood River. Rescaled netic sector mass spectrometer. Gas was puri- editing@geosociety .org.

8 Geological Society of America Bulletin, Month/Month 2014 Geological Society of America Bulletin, published online on 6 November 2014 as doi:10.1130/B31141.1

Interaction of outburst ﬂ oods with basaltic aquifers on the Snake River Plain

to be associated with an outburst fl ood from that region (Licciardi and Pierce, 2008). Moreover, none of the downstream erosional surfaces dated in the Snake River Canyon yields ages less than ca. 17 ka (Amidon and Farley, 2011). Given these constraints, we interpret the boulder ages to rep- resent exhumation of the fl ood deposit by surface defl ation. However, apparent exposure ages are younger and more tightly clustered than would be expected from surface defl ation, suggesting that defl ation would need to have occurred very rapidly. This could occur in response to rapid fl uvial incision during the early Holocene, when regional aridity reduced sediment supply while sustaining high discharges from mountain snow- pack (Krause and Whitlock, 2013). We adopt a preferred age of ca. 22.3 ± 0.6 (1.7) ka for the Big Lost River fl ood based on the four youngest samples that overlap within 1σ analytical uncertainty (Fig. 9; Table DR2 [see footnote 1]). This interpretive model attributes the spread in boulder ages to prefl ood inheritance and is supported by occurrence of the youngest ages on boulders sampled from paleodepths of >3 m after surface reconstruction. A ca. 22.3 ka age is slightly older than the previous estimate of ca. 20.5 ka, but it is consistent with a fl ood from Glacial Lake East Fork because it over- laps the timing of maximum ice advance in the Sawtooth Mountains and the nearby Yellowstone Plateau (Knobel et al., 1997; Perkins et al., 2002; Thackray et al., 2008). It is also consistent with the observed highstand in Lake Terreton at ca. 21.3 ka (Anderson and Lewis, 1989). We estimate the age of the Bonneville fl ood as 17.5 ± 0.1 (0.9) ka based on the weighted mean of all boulders at the Inkom site (Fig. 9; Table DR3 [see footnote 1]). This age is in excellent agreement with the accepted age of the Bonneville fl ood (Godsey et al., 2005).

Ages of the Dry Canyons

Sites at Devil’s Corral yield evidence for two separate fl ood events that match the age of the Figure 9. Probability density functions showing individual boulder Bonneville fl ood at ca. 17.5 ka and the Big Lost ages at each of the sites discussed in the text. Dashed lines show River Flood at ca. 22.3 ka. Three sites yield individual boulders; black line shows the normalized probability mean boulder ages consistent with the Bonne- function for all samples combined. Vertical lines show the pre- ville fl ood: plunge pool at 17.6 ± 0.1 (1.2) ka, ferred age for the deposit. high spire at 18.0 ± 0.6 (1.6) ka, and upper lobe A at 18.2 ± 0.4 (1.7) ka (Fig. 9; Table DR4 [see footnote 1]). Two other sites yield mean boulder ages consistent with the Big Lost River fl ood: in origin. Shielded samples yield between 0.2 Timing of Floods on the Snake River Plain lower lobe A at 21.8 ± 1.1 (2.2) ka and lower and 2.4 Mat/g of 3He. Cosmogenic 3He concen- canyon at 21.2 ± 0.4 (1.5) ka. The high plateau trations were computed by direct subtraction of The mean 21Ne age of the fi ve youngest boul- site shows an intermediate age of 18.7 ± 1 (1.9) the 3He concentration measured in a shielded ders collected along the Henry’s Fork River ka. Boulder ages from the talus amphitheater sample from the same location. Neon isotope cluster at ca. 8.2 ± 0.9 (1.6) ka. Because most are widely scattered and range from 3.2 to14 ka, ratios are very close to air, with 21Ne excesses of glaciers in the Yellowstone region plateau began presumably recording sporadic block fall from 6%–18% of total measured 21Ne. retreat by ca. 14 ka, these ages appear too young the headwall.

Geological Society of America Bulletin, Month/Month 2014 9 Geological Society of America Bulletin, published online on 6 November 2014 as doi:10.1130/B31141.1

Amidon and Clark

Boulder ages from Blue Lakes Canyon are we propose that Devil’s Corral was most it explains why the western canyons (but not signifi cantly older than the Bonneville fl ood, recently eroded during the Bonneville fl ood Devil’s Corral) are all aligned at the mouths of giving a mean age of 21 ± 1 (2) ka. Ages from (ca. 17.5 ka), whereas canyons further to the watershed networks, despite supposedly being the talus pile generally postdate fl ooding and west (Blue Lakes, Box, and Stubby) were most formed by megafl oods that fl owed roughly cluster between 4.8 and 6 ka (Fig. 9; Table DR5 recently eroded during the Big Lost River fl ood perpendicular to those watersheds. Finally, it [see footnote 1]). (ca. 22.3 ka). This conclusion arises because explains how canyons could be cut during a At Box Canyon, we combined our results fl ood features at each of the western dry can- megafl ood despite the fact that the Snake River with the rescaled results of Aciego et al. (2009). yons are similar in age to the Big Lost River Canyon would be bankfull and potential knick- Because boulder ages show signifi cant scat- fl ood, but they are signifi cantly older than the points drowned during the peak fl ood hydro- ter, we adopted the mean concentration of the Bonneville fl ood. Given that both fl oods trav- graph. Lamb et al. (2008, 2014) demonstrated three youngest boulders that are within 1σ error eled the same path down the lower Snake River, that the canyons were formed primarily by sub- of each other, which give a mean age of 20.9 ± how is it possible that the Big Lost River fl ood aerial waterfalls and knickpoint retreat). Next, 1.3 (2.4) ka. The 3He concentrations in strath eroded the dry canyons, yet the ~750 times we explore the hydrologic conditions and mech- samples decrease with elevation from an appar- larger Bonneville fl ood did not? anisms of erosion that would be expected under ent age of ca. 227 ka at the top (S1) to an age One explanation is that the western dry the composite erosion scenario and speculate on of ca. 44 ka at the deepest point in the spillway canyons were eroded by Big Lost River fl ood- whether suffi cient discharge could be obtained (S4). The age of the lowest sample in the spill- waters that ponded in the Terreton Basin and by this mechanism. way is statistically indistinguishable from ages were transmitted through the Snake River Plain reported by Lamb et al. (2008, 2014) from the aquifer before resurfacing near the western Hydrologic Assumptions of the spillways at Box and Stubby Canyons, prompt- dry canyons. Infi ltration into the Snake River Composite Flow Scenario ing them to propose a fl ood event at ca. 46 ka. Plain aquifer would have raised the water table The composite fl ow scenario predicts that Calculations shown in supplementary Figure across much of the western Snake River Plain, downstream discharge would increase abruptly DR1 (see footnote 1) reveal that the age-elevation returning groundwater to the surface along a and discontinuously depending on the loca- relationship along the spillway cannot be easily topographic escarpment where the water table tions of return fl ow from groundwater fast explained by erosion during a single fl ood event intersected the surface of the Snake River Plain paths within the watershed. Rapid increases in at 46 ka, but requires multiple events (Fig. DR1 upstream of canyon heads (Fig. 5). This return discharge may preclude development of a tradi- [see footnote 1]). To constrain the timing of the fl ow then coalesced in surfi cial watersheds and tional dendritic watershed pattern, instead favor- fl oods, we performed a one-dimensional iterative was delivered as a focused fl ow through spilling rapid appearance of incised channels down- forward model to generate a fi t to the data. Our ways at the heads of the dry canyons. stream of the highest return fl ow zones. This model assumed an initially extensive S1 surface This scenario is supported by streamlined prediction is consistent with the morphology established at 227 ka, which was incised in two topography on the Big Lost River scabland that of fl uvial networks in the Blue Lakes and Box steps to create the strath terraces and spillway. record fl oodwaters fl owing into the Terreton Canyon watersheds, where channels are weakly The model thus superimposed inherited produc- Basin (Bartholomay et al., 1997a), and vol- developed in the headwaters and develop rapidly tion accumulated at depth with subsequent pro- ume estimates of Glacial Lake East Fork (6.3 × downstream of complex basaltic topography duction on the newly exposed strath surfaces. 109 m3) exceeding the volume of the Terreton along the topographic escarpment (white arrows We fi xed the timing of the youngest event at Basin (Barton, 2008). The idea that infi ltration in Fig. 5). Direct evidence for signifi cant fl uvial 22.3 ka and allowed timing of the earlier fl ood into the subsurface could outpace evaporation incision in one of these channels comes from a event to vary. A family of best fi ts was obtained is plausible given a cooler glacial climate and 70 ka exposure age on a 395-k.y.-old basalt in a for an additional fl ood event between ca. 76 and infi ltration rates up to 0.5 m3 s–1 km–1 observed dry bedrock channel ~5 km upstream from Box 86 ka, although none of the solutions provides a in the modern settling areas used by the Idaho Canyon (sample ID10–49, Fig. 5; Table DR6 match within the 1σ ana lyti cal uncertainties of National Laboratory to drain excess runoff from [see footnote 1]). Given the paucity of modern all data points. Importantly, fi tting the data with the Big Lost River (Bennett, 1990). The time discharge events and the limited stream power, a younger 17.5 ka event leads to very poor fi ts, scale of drainage was likely only a few months, as well as the similarity of this age to the S3 providing further evidence that the Bonneville given that the modern Snake River Plain aquifer strath in the Box Canyon spillway, this age is fl ood was not important in incising the spillway discharges ~5 × 109 m3 per year, and is presum- best explained by event-driven erosion. at Box Canyon. ably operating far below its saturated transport In addition to return fl ow, increased discharge Recalculated results from Stubby Canyon capacity. from springs at canyon heads would contribute (Lamb et al., 2014) show that the mean age of Recognizing that canyon erosion would pro- to erosion of the dry canyons. Discharge near the seven youngest erosional notches and aban- ceed by the combined discharge from return existing springs would increase for two rea- doned channels within 1σ error of each other is fl ow and direct groundwater discharge at canyon sons: (1) higher pore pressure due to increased 20.9 ± 2.2 (3.2) ka. heads, we subsequently refer to this scenario as hydraulic head, and (2) activation of additional erosion from “composite fl ow” (e.g., Schumm springs due to an elevated water table. If the DISCUSSION and Phillips, 1986). Such a model could explain basaltic aquifer is envisioned as a uniform and some of the seemingly contradictory observa- unconfi ned aquifer, then hydraulic head cannot Timing and Mechanisms of tions supporting either the groundwater sap- exceed the local relief of the canyon headwall. Dry Canyon Formation ping or fl ood erosion viewpoints of previous However, if groundwater fast paths are confi ned authors. First, it explains why canyons occur in by overlying basalt fl ows, the hydraulic head Our results suggest that the dry canyons of conjunction with springs, despite the fact that could reach hundreds of meters, resulting in the Snake River were not all formed simul- modern spring discharge is insuffi cient to physi- elevated pore pressures. Confi ned fl ow paths are taneously by the Bonneville fl ood. Instead, cally or chemically erode the canyons. Second, plausible given that only about half the springs

10 Geological Society of America Bulletin, Month/Month 2014 Geological Society of America Bulletin, published online on 6 November 2014 as doi:10.1130/B31141.1

Interaction of outburst ﬂ oods with basaltic aquifers on the Snake River Plain

along the Snake River have discharges that cor- of Lamb et al. (2008), who noted that a relatively of groundwater discharge during composite relate to water levels in nearby wells (Janczak, small spillway volume relative to the size of Box erosion may sustain a higher ratio of canyon- 2001). Likewise, recent drilling near Mountain Canyon implies a long-duration fl ood event. exiting to canyon-entering discharge during the Home, Idaho, encountered artesian fl ow after Hydrologic conditions of the composite sce- declining limb, which allows transport to keep penetrating a depth of 1745 m, demonstrating nario predict evolution of unique canyon mor- pace with toppling for a larger portion of the that an isolated portion of the aquifer had been phologies relative to those eroded dominantly declining hydrograph. recharged from elevations higher than the drill by overland fl ooding. A comparison of the west- Finally, it must be stressed that all of the dry site (Whitehead, 1986). Although the poten- ern canyons with Devil’s Corral reveals mor- canyons have likely been affected by multiple tial is clear, poorly understood aquifer plumb- phological differences that may refl ect different fl oods. The idea that Devil’s Corral experienced ing precludes quantitative estimates of direct fl ood-erosion scenarios. The western canyons erosion by surfi cial fl oodwaters in multiple groundwater discharge at canyon heads during are generally defi ned by a single fl at-bottomed fl oods is sensible given that it lies along the fl ow the Big Lost River fl ood. canyon lobe with a high length/width ratio. In path of two thirds of the peak Bonneville, and One simple approach to estimating discharge contrast, Devil’s Corral has: (1) multiple amphi- likely pre-Bonneville, fl oodwaters that were from return fl ow in upstream watersheds is to theater-headed canyon lobes with lower length/ routed through the Eden Channel (O’Connor, assume a short-lived steady state between infi l- width ratios, (2), a stair-step canyon fl oor with 1993). Passage of surfi cial Big Lost River fl ood- tration and return fl ow, such that the infi ltration incised inner spillways, and (3) an upper fl oor waters through Devil’s Corral is recorded by rate in the Big Lost River settling area (0.5 m3 in canyon A that is littered with angular basaltic ca. 21.8 ka boulders at lower lobe A. Preserva- s–1 km–1) is taken as the return fl ow rate over a columns preserving primary lava fl ow textures, tion of these boulders implies that later Bonne- large area (Bennett, 1990). With this assump- which must have been toppled but not trans- ville fl oodwaters may not have fully inundated tion, Box Canyon would yield a return fl ow dis- ported (Fig. 6). lobe A of Devil’s Corral, perhaps contributing charge of ~113 m3/s, Blue Lakes ~64 m3/s, and First, the higher length/width ratio in the to the high density of angular toppled blocks on Billingsley ~77 m3/s, with far higher, but more western canyons may result from unidirectional the canyon fl oor. uncertain estimates for the larger Blind and headward erosion in response to enhanced Importantly, our data do not refute the idea Stubby watersheds that exceed 600 m3/s. These groundwater discharge and focused delivery of that Bonneville floodwaters inundated the discharges fall short of the ~220 m3/s required to surface water to the canyon head. This contrasts western dry canyons, but they suggest that move basaltic boulders in Box Canyon (Lamb with a multidirectional cliff retreat induced by fl oodwaters accomplished relatively little ero- et al., 2008), suggesting that modern spring complete inundation of the canyon rim beneath sional modifi cation. This is sensible given that discharge at the canyon head (10 m3/s) would surficial floodwaters, which would tend to topographic considerations suggest the western need to increase by a factor of 10, or that return widen the canyon. canyons experienced relatively short-lived and fl ow would need to be roughly twice what we Second, fl at-bottomed canyons would be probably minor spillover during the Bonneville have estimated here. Given the known hetero- more likely to form during composite fl ow fl ood, apparently insuffi cient to leave a strong geneities in the Snake River Plain aquifer, it is because as headward erosion occurs, the loca- morphologic signature. A mild erosional over- conceivable that discharge within one watershed tion of maximum discharge is fi xed at the spring print of the western dry canyons by the Bonne- exceeds an average of 0.5 m3 s–1 km–1, but falls outlet. Maximum stream power thus occurs ville fl ood is consistent with the ~1 k.y. offset short of that in other watersheds, which may downstream of the spring outlet and migrates between boulder ages in those canyons and explain the stark differences in canyon develop- upstream with the spring during headward ero- those on the Big Lost River scabland. Because ment between adjacent watersheds, such as Box sion. Because toppled blocks must be removed the two fl oods are separated by ~5 k.y., a 22.3 and Sand Springs Canyons (Fig. 5). in order to induce further headward erosion, the ka boulder deposited during the Big Lost River pace and style of headward erosion are set by fl ood would need to be eroded by an average of Mechanisms of Erosion and the location of the spring outlet through time. ~10 cm during the Bonneville fl ood to give an Resultant Morphology In cases where springs emanate from fl at-lying apparent age of 21.3 ka. This amount of erosion Erosional conditions under a composite contacts (e.g., lava fl ows), the locus of maxi- could be accomplished by direct abrasion of model would differ from a conventional mega- mum erosion migrates headward at a constant boulder surfaces during fl ooding. fl ood in three ways: (1) Peak discharge would be elevation, creating a fl at-bottomed canyon. This lower, but surface discharge would be sustained idea is supported by observations of amphi- Implications for Martian Canyons for longer periods of time; (2) discharge would theater-headed canyons in Utah, where the slope be focused in a channel at the downstream end of the canyon fl oor is typically set by lithologic A composite discharge model is a plausible of the watershed; and (3) a larger fraction of contacts (Schumm and Phillips, 1986). It is also formation mechanism for amphitheater-headed discharge would be derived from groundwater consistent with numerical models suggesting canyons on Mars, many of which lack evidence discharge at the canyon headwall. This scenario that the rate of knickpoint retreat is set by the of large surfi cial fl oods. A growing body of evi- predicts a single large waterfall focused at the removal of bed-load debris from downstream dence suggests that many amphitheater-headed canyon headwall for months, instead of many reaches (Haviv et al., 2010). canyons are linked to weakly developed drain- diffuse waterfalls spilling over large portions of Third, high densities of basaltic blocks that age networks, which could have been fed by the canyon rim for a shorter period of days to have been toppled but not transported may be return fl ow. For example, Magnold et al. (2008) months. Enhanced groundwater discharge from less likely to form during composite fl ow. One showed that 14 out of 17 amphitheater-headed springs would also be focused at the canyon explanation is that narrower canyons formed by canyons were connected to tributary valley net- head. Prolonged surface discharge is supported focused fl ow obtain higher unit steam power works in the Valles Marineris region of Mars. by deeply incised spillways at Box and Blue under a given discharge, improving their abil- Although they suggested these networks were Lakes Canyon, which are not present at Devil’s ity to transport boulders during the declining fed by precipitation, the spatially discontinu- Corral. This is consistent with the observations limb of the hydrograph. Alternatively, addition ous nature of the valley networks is diffi cult to

Geological Society of America Bulletin, Month/Month 2014 11 Geological Society of America Bulletin, published online on 6 November 2014 as doi:10.1130/B31141.1

Amidon and Clark reconcile with distributed precipitation. Diffuse location of amphitheater-headed canyons may REFERENCES CITED discharge from return fl ow across a very large directly constrain the location of hydraulic fast Aciego, S.M., DePaolo, D.J., Kennedy, B.M., Lamb, M.P., area may produce tributary valley networks paths (e.g., Harrison and Grimm, 2009). Sims, K.W.W., and Dietrich, W.E., 2007, Combining while also explaining why such networks are He-3 cosmogenic dating with U-Th/He eruption ages discontinuous across the broader landscape. CONCLUSIONS using olivine in basalt. Earth and Planetary Science Letters, v. 254, p. 288–302, doi: 10 .1016 /j .epsl .2006 A composite discharge model is also con- .11 .039 . sistent with the occurrence of amphitheater- This paper provides new constraints on the Ackerman, D.J., 1991, Transmissivity of the Snake River headed canyons along regional-scale topo- timing and routing of major fl ood events on Plain Aquifer at the Idaho National Engineering Labora tory, Idaho. U.S. Geological Survey Water Re- graphic escarpments on Mars (Andrews-Hanna the Snake River Plain and the age of the dry sources Investigations Report 91-4058, 39 p. et al., 2007, 2010). For example, many of the canyons they helped to carve. By dating fl ood- Allison, R.R., 2001, Climatic, Volcanic, and Tectonic In- fl uences on Late Pleistocene Sedimentation along the best-developed amphitheater-headed canyon deposited boulders in their source regions, we Snake River and in Market Lake [M.S. thesis]. Poca- networks, such as Nanedi and Nirgal Valles, show that the Big Lost River fl ood occurred tello, Idaho, Geology Department, Idaho State Univer- are found along the broad topographic slopes at ca. 22.3 ± 1.7 ka, and the Bonneville fl ood sity, 128 p. Amidon, W.H., and Farley, K.A., 2011, Cosmogenic 3He descending to the Chryse Planitia (Harrison and occurred at ca. 17.5 ± 0.9 ka. Boulders depos- dating of apatite, zircon and pyroxene from Bonneville Grimm, 2005). Many others, such as Echus and ited in the headwaters of the Henry’s Fork River fl ood erosional surfaces. Quaternary Geochronology, Ganges Chasma, occur along regional topo- appear to have been rapidly exhumed at ca. v. 6, p. 10–21, doi: 10 .1016 /j .quageo .2010 .03 .005 . Amidon, W.H., and Farley, K.A., 2012, Cosmogenic 3He graphic slopes in the Valles Marineris region 8.2 ± 1.6 ka, leaving the exact timing of these and 21Ne dating of biotite and hornblende. Earth and (Ghatan et al., 2006). Global-scale numerical fl ood(s) unclear, with at least one of them prob- Planetary Science Letters, v. 313–314, p. 86–94, doi: 10 .1016 /j .epsl .2011 .11 .005 . models suggest that during a the wettest periods ably occurring >100 k.y. ago (Allison, 2001). Amidon, W.H., Rood, D.H., and Farley, K.A., 2009, Cosmo- in Martian history, regional water tables on Additional dating of fl ood-deposited boulders genic 3He and 21Ne production rates calibrated against Mars would have intersected these surfaces, near the dry canyons shows that Devil’s Cor- 10Be in minerals from the Coso volcanic fi eld. Earth and Planetary Science Letters, v. 280, p. 194–204, doi: releasing return fl ow across vast expanses of the ral was most recently eroded by the Bonneville 10 .1016 /j .epsl .2009 .01 .031 . landscape (Andrews-Hanna et al., 2007). fl ood, whereas four other canyons to the west Anderson, S.R., and Lewis, B.D., 1989, Stratigraphy of the Finally, the cratering history of the ancient record older ages consistent with the Big Lost Unsaturated Zone at the Radioactive Waste Manage- ment Complex, Idaho National Engineering Labora- Martian surface creates favorable hydrologic River fl ood at ca. 22 ka. tory, Idaho. U.S. Geological Survey Water-Resources conditions to promote infi ltration and transmis- Given that the Bonneville fl ood was ~750 Investigations Report 89-4065, 54 p. Anderson, S.R., and Liszewski, M.J., 1997, Stratigraphy of sion of vast amounts of groundwater. Many times larger than the Big Lost River fl ood, it is the Unsaturated Zone and the Snake River Plain Aqui- impact craters were once part of “fi ll-and-spill” diffi cult to understand how Big Lost River fl ood- fer At and Near the Idaho National Engineering Labo- river systems in which they fi lled with surface waters could have heavily eroded the western ratory, Idaho. U.S. Geological Survey Water-Resources Investigations Report 97-4183, 70 p. water before overfl owing into downstream cra- dry canyons, but Bonneville fl oodwaters did not. Anderson, S.R., Liszewski, M.J., and Ackerman, D.J., 1996, ters. This situation is quite analogous to pond- We thus propose a composite erosion model in Thickness of Surfi cial Sediment At and Near the Idaho ing of the Big Lost River fl ood in the Terreton which waters of the Big Lost River fl ood were National Engineering Laboratory, Idaho. U.S. Geologi- cal Survey Open-File Report 96-330, 16 p. Basin, and it implies huge volumes of ground- ponded in the Terreton Basin and partially routed Anderson, S.R., Kuntz, M.A., and Davis, L.C., 1999, Geo- water must have infi ltrated into the fractured through the highly conductive Snake River Plain logic Controls of Hydraulic Conductivity in the Snake River Plain Aquifer At and Near the Idaho National En- crater fl oor (Krause and Whitlock, 2013). More- aquifer. These waters would have returned to gineering and Environmental Laboratory, Idaho. U.S. over, fractures radiating from impact craters the surface along a topographic escarpment just Geological Survey Water-Resources Investigations would provide a network of hydraulic fast paths upstream of the western dry canyons and com- Report 99-4033, 44 p. Andrews-Hanna, J.C., and Lewis, K.W., 2011, Early Mars of potentially greater magnitude than those bined with elevated groundwater discharge at hydrology: 2. Hydrological evolution in the Noa- observed on the Snake River Plain. canyon heads to drive canyon erosion. chian and Hesperian Epochs. Journal of Geophysi- If some amphitheater-headed canyons were This composite discharge scenario predicts a cal Research–Planets, v. 116, E02007, doi:10 .1029 /2010JE003709 . in fact formed by composite discharge, it has unique fl ood hydrograph and routing in which Andrews-Hanna, J.C., Phillips, R.J., and Zuber, M.T., 2007, several implications for the hydrologic cycle fl oodwaters are focused at the canyon head for Meridiani Planum and the global hydrology of Mars. Nature, v. 446, p. 163–166, doi: 10 .1038 /nature05594 . on Mars. First, it suggests that canyons could months to years. A sustained hydrograph of this Andrews-Hanna, J.C., Zuber, M.T., Arvidson, R.E., and be fundamentally attributed to groundwater nature is broadly consistent with observed mor- Wiseman, S.M., 2010, Early Mars hydrology: Meridi- discharge, but on a more rapid time scale than phological differences between Devil’s Corral ani playa deposits and the sedimentary record of Ara- bia Terra. Journal of Geophysical Research–Planets, required by a seepage weathering mechanism and the younger dry canyons to its west, which v. 115, E06002, doi: 10 .1029 /2009JE003485 . (e.g., Lamb et al., 2006). Second, it implies that have higher length/width ratios, fl atter canyon Armstrong, R.L., Leeman, W.P., and Malde, H.E., 1975, large quantities of local precipitation may not fl oors, and a lower density of immature toppled K-Ar dating Quaternary and Neogene volcanic rocks of the Snake River Plain, Idaho. American Journal of be required to erode canyons found at higher boulders. The composite discharge model may Science, v. 275, p. 225–251, doi: 10 .2475 /ajs .275 .3 .225 . latitudes, and that canyons could potentially provide an explanation for some amphitheater- Balco, G., Stone, J.O., Lifton, N.A., and Dunai, T.J., 2008, A simple, internally consistent, and easily accessible be attributed to groundwater recharge at lower headed canyons on Mars, where cratered and means of calculating surface exposure ages or erosion latitudes, or from nearby craters. Finally, our fractured basalt creates numerous opportuni- rates from 10Be and 26Al measurements. Quaternary results from the Snake River Plain suggest that ties for infi ltration of surface waters into a vast Geochronology, v. 3, p. 174–195, doi: 10 .1016 /j .quageo .2007 .12 .001 . the location of amphitheater-headed canyons is basaltic aquifer. Bartholomay, R.C., Tucker, B.J., Ackerman, D.J., and Liszewski, M.J., 1997a, Hydrologic Conditions and controlled by three key variables: (1) a supply ACKNOWLEDGMENTS of groundwater, (2) the location of topographic Distribution of Selected Radiochemical and Chemi- Thanks go to Ken Farley for use of his laboratory, cal Constituents in Water, Snake River Plain Aquifer, escarpments, and (3) the location of ground- Idaho National Engineering Laboratory, Idaho, 1992 to Michael Lamb for many helpful conversations, and water fast paths. Because large-scale topo- through 1995. U.S. Geological Survey Water-Re- to Louise McCarren for help with fi eld logistics. The sources Investigations Report 97-4086, 57 p. graphic escarpments of Mars are well mapped manuscript was improved by reviews from Vic Baker Bartholomay, R.C., Williams, L.M., and Campbell, L.J., and groundwater was once ubiquitous, the and Sarah Rathburn. 1997b, Radiochemical and Chemical Constituents

12 Geological Society of America Bulletin, Month/Month 2014 Geological Society of America Bulletin, published online on 6 November 2014 as doi:10.1130/B31141.1

Interaction of outburst ﬂ oods with basaltic aquifers on the Snake River Plain

in Water from Selected Wells and Springs from the Hodges, M.K.V., Link, P.K., and Fanning, C.M., 2009, The ciations in greater Yellowstone and the Teton Range, Southern Boundary of the Idaho National Engineer- Pliocene Lost River found to west: Detrital zircon USA. Quaternary Science Reviews, v. 27, p. 814–831, ing Laboratory to the Hagerman Area, Idaho, 1996. evidence of drainage disruption along a subsiding doi: 10 .1016 /j .quascirev .2007 .12 .005 . U.S. Geological Survey Open-File Report 03-168, hotspot track. Journal of Volcanology and Geothermal Malde, H.E., 1960, Evidence in the Snake River Plain, 37 p. Research, v. 188, p. 237–249, doi: 10 .1016 /j .jvolgeores Idaho, of a catastrophic fl ood from Pleistocene Lake Barton, I., 2008, Calculating the Peak Discharge of the Gla- .2009 .08 .019 . Bonneville. U.S. Geological Survey Professional Paper cial Lake East Fork Outburst Flood, Big Lost River, Irwin, R.P., Howard, A.D., and Craddock, R.A., 2008, Flu- 400b, p. 295–297. Idaho [M.S. thesis]. Bethlehem, Pennsylvania, Earth vial Valley Networks on Mars, in Rice, S.P., Roy, A.G., Malde, H.E., 1968, The Catastrophic Late Pleistocene Bonne- and Environmental Sciences, Lehigh University, 53 p. and Rhoads, B.L., eds., River Confl uences, Tributaries ville Flood in the Snake River Plain, Idaho. U.S. Geo- Bennett, C.M., 1990, Streamfl ow Losses and Ground-Water and the Fluvial Network. Chichester, UK, John Wiley logical Survey Professional Paper 596, 52 p. Level Changes along the Big Lost River at the Idaho & Sons, p. 419–451. Malde, H.E., and Powers, H.A., 1962, Upper Cenozoic stra- National Engineering Laboratory, Idaho. U.S. Geologi- Janczak, L., 2001, Relationships between Spring Discharge tigraphy of western Snake River Plain, Idaho. Geologi- cal Survey Water-Resources Investigations Report 90- and Aquifer Water Levels in the Thousand Springs Re- cal Society of America Bulletin, v. 73, p. 1197–1219, 4067, 49 p. gion, Idaho [M.S. thesis]. Moscow, Idaho, University doi: 10 .1130 /0016 -7606 (1962)73 [1197: UCSOWS]2 .0 Brown, K.B., McIntosh, J.C., Baker, V.R., and Gosch, D., of Idaho, 139 p. .CO;2 . 2010, Isotopically-depleted late Pleistocene ground- Johnson, T.M., Roback, R.C., McLing, T.L., Bullen, T.D., Malin, M.C., and Carr, M.H., 1999, Groundwater formation water in Columbia River Basalt aquifers; evidence for DePaolo, D.J., Doughty, C., Hunt, R.J., Smith, R.W., of Martian valleys. Nature, v. 397, p. 589–591, doi: 10 recharge of glacial Lake Missoula fl oodwaters? Geo- Cecil, L.D., and Murrell, M.T., 2000, Groundwater .1038 /17551 . physical Research Letters, v. 37, L21402, doi: 10 .1029 “fast paths” in the Snake River Plain Aquifer; radio- Matsubara, Y., Howard, A.D., and Gochenour, J.P., 2013, /2010GL044992 . genic isotope ratios as natural groundwater tracers. Hydrology of early Mars: Valley network incision. Cerling, T., and Craig, H., 1994, Cosmogenic 3He produc- Geology, v. 28, p. 871–874, doi:10 .1130 /0091 -7613 Journal of Geophysical Research–Planets, v. 118, tion rates from 39N to 46N latitude, western USA and (2000)28 <871: GFPITS>2 .0 .CO;2 . p. 1365–1387, doi: 10 .1002 /jgre .20081 . France. Geochimica et Cosmochimica Acta, v. 58, Kauffman, J.D., Othberg, K.L., Gillerman, V.S., and Gar- Meinzner, O.E., 1927, Large Springs in the United States. p. 249–255, doi: 10 .1016 /0016 -7037 (94)90462 -6 . wood, D.L., 2005, Geologic Map of the Twin Falls U.S. Geological Survey Water-Supply Paper 557, 94 p. Cerling, T.E., Poreda, R.J., and Rathburn, S.L., 1994, Cos- 30 × 60 Minute Quadrangle, Idaho. Moscow, Idaho, Niedermann, S., 2002, Cosmic-ray-produced noble gases mogenic 3He and 21Ne age of the Big Lost River fl ood, Idaho Geological Survey, scale 1:100,000. in terrestrial rocks: Dating tools for surface processes, Snake River Plain, Idaho. Geology, v. 22, p. 227–230, Knobel, L.L., Bartholomay, R.C., and Orr, B.R., 1997, in Porcelli, D., Ballentine, C.J., and Wieler, R., eds., doi: 10 .1130 /0091 -7613 (1994)022 <0227: CHANAO>2 Preliminary Delineation of Natural Geochemical Re- Noble Gases in Geochemistry and Cosmochemis- .3 .CO;2 . actions, Snake River Plain Aquifer System, Idaho try. Reviews in Mineralogy and Geochemistry, v. 47, Clifford, S.M., and Parker, T.J., 2001, The evolution of the National Engineering Laboratory and Vicinity, Idaho. p. 731–784. Martian hydrosphere: Implications for the fate of a U.S. Geological Survey Water-Resources Investiga- Nimmo, J.R., Perkins, K.S., Rose, P.E., Rousseau, J.P., primordial ocean and the current state of the Northern tions Report 97-4093, 52 p. Orr, B.R., Twining, B.V., and Anderson, S.R., 2002, plains. Icarus, v. 154, p. 40–79, doi:10 .1006 /icar .2001 Knudsen, K.L., Sowers, J.M., Ostenaa, D.A., and Levish, Kilometer-scale rapid transport of naphthalene sulfo- .6671 . D.R., 2002, Evaluation of glacial outburst fl ood hy- nate tracer in the unsaturated zone at the Idaho National Fassett, C.I., and Head, J.W., 2008, Valley network-fed, pothesis for the Big Lost River, Idaho, in House, P.K., Engineering and Environmental Laboratory. Vadose open-basin lakes on Mars: Distribution and implica- Webb, R.H., Baker, V.R., and Levish, D., eds., Ancient Zone Journal, v. 1, p. 89–101, doi: 10 .2113 /1 .1 .89 . tions for Noachian surface and subsurface hydrology. Floods, Modern Hazards: Principles and Applications O’Connor, J.E., 1993, Hydrology, Hydraulics, and Geomor- Icarus, v. 198, p. 37–56, doi: 10 .1016 /j .icarus .2008 .06 of Paleofl ood Hydrology. Washington, DC, American phology of the Bonneville Flood. Geological Society .016 . Geophysical Union, p. 217–235. of America Special Paper 274, 84 p. Ghatan, G.J., Head Iii, J.W., and Wilson, L., 2006, Mangala Kober, F., Afflinov, V., Ivy-Ochs, S., Kubik, P.W., and Oviatt, C.G., Currey, D.R., and Sack, D., 1992, Radiocarbon Valles, Mars: Assessment of early stages of flood- Wieler, R., 2011, The cosmogenic 21Ne production rate chronology of Lake Bonneville, eastern Great-Basin, ing and downstream fl ood evolution. Earth, Moon, in quartz evaluated on a large set of existing 21Ne-10Be USA. Palaeogeography, Palaeoclimatology, Palaeo- and Planets, v. 96, p. 1–57, doi:10 .1007 /s11038 -005 data. Earth and Planetary Science Letters, v. 302, ecology, v. 99, p. 225–241, doi: 10 .1016 /0031 -0182 -9009 -y . p. 163–171, doi: 10 .1016 /j .epsl .2010 .12 .008 . (92)90017 -Y . Gillerman, V.S., Kauffman, J.D., and Othberg, K.L., 2005, Krause, T.R., and Whitlock, C., 2013, Climate and vege- Perkins, K.S., Nimmo, J.R., Rose, P.A., Rousseau, J.P., Orr, Geologic Map of the Thousand Springs Quadrangle, tation change during the late-glacial/early-Holocene B.R., Twining, B.V., and Anderson, S.R., 2002, Flow Gooding and Twin Falls Counties, Idaho. Moscow, transition inferred from multiple proxy records from System Analysis Using a Surface-Applied Tracer at Idaho, Idaho Geological Survey, scale 1:24,000. Blacktail Pond, Yellowstone National Park, USA. the Idaho National Engineering and Environmental Godsey, H.S., Currey, D.R., and Chan, M.A., 2005, New Quaternary Research, v. 79, p. 391–402, doi: 10 .1016 Laboratory, Idaho. U.S. Geological Survey Open-File evidence for an extended occupation of the Provo /j .yqres .2013 .01 .005 . Report 02-89, p. 82–83. shoreline and implications for regional climate change, Kuntz, M.A., Anderson, S.R., Champion, D.E., Lanphere, Perkins, K.S., Nimmo, J.R., Rose, P.E., Rousseau, J.P., Orr, Pleistocene Lake Bonneville, Utah, USA. Quaternary M.A., and Grunwald, D.J., 2002, Tension cracks, erup- B.R., Twining, B.V., Anderson, S.R., and Anonymous, Research, v. 63, p. 212–223, doi: 10 .1016 /j .yqres .2005 tive fi ssures, dikes, and faults related to late Pleisto- 2000, Contaminant-transport implications of large- .01 .002 . cene–Holocene basaltic volcanism and implications scale, rapid, horizontal fl ow in the unsaturated zone Goehring, B.M., Kurz, M.D., Balco, G., Schaefer, J.M., for the distribution of hydraulic conductivity in the at the Idaho National Engineering and Environmental Licciardi, J., and Lifton, N., 2010, A reevaluation of eastern Snake River Plain, Idaho, in Link, P.K., and Laboratory. Eos, Transactions, American Geophysical in situ cosmogenic He-3 production rates. Quaternary Mink, L.L., eds., Geology, Hydrogeology, and Envi- Union, v. 81, p. 414. Geochronology, v. 5, p. 410–418, doi: 10 .1016 /j .quageo ronmental Remediation: Idaho National Engineering Pierce, K.L., and Scott, W.E., 1982, Pleistocene episodes of .2010 .03 .001 . and Environmental Laboratory, Eastern Snake River alluvial-gravel deposition, southeastern Idaho, in Bon- Goudge, T.A., Mustard, J.F., Head, J.W., and Fassett, C.I., Plain, Idaho. Geological Society of America Special nichsen, B., and Breckenridge, R.M., eds., Cenozoic 2012, Constraints on the history of open-basin lakes Paper 353, p. 111–133. Geology of Idaho. Idaho Bureau of Mines and Geology on Mars from the composition and timing of volcanic Lamb, M.P., Howard, A.D., Johnson, J., Whipple, K.X., Bulletin 26, p. 685–702. resurfacing. Journal of Geophysical Research–Planets, Dietrich, W.E., and Perron, J.T., 2006, Can springs Rathburn, S.L., 1993, Pleistocene cataclysmic fl ooding v. 117, E00J21, doi: 10 .1029 /2012JE004115 . cut canyons into rock?: Journal of Geophysical along the Big Lost River, east central Idaho. Geomor- Harrison, K.P., and Grimm, R.E., 2004, Tharsis recharge: Research–Planets, v. 111, E07002, doi:10 .1029 phology, v. 8, p. 305–319, doi: 10 .1016 /0169 -555X A source of groundwater for Martian outfl ow chan- /2005JE002663 . (93)90026 -X . nels. Geophysical Research Letters, v. 31, L14703, Lamb, M.P., Dietrich, W.E., Aciego, S.M., DePaolo, D.J., Rodgers, D.W., Long, S.P., McQuarrie, N., Burgel, W.D., p. 14,701–14,704. and Manga, M., 2008, Formation of Box Canyon, and Hersley, C., 2006, Geologic Map of the Inkom Harrison, K.P., and Grimm, R.E., 2005, Groundwater- Idaho, by megafl ood: Implications for seepage erosion Quadrangle, Bannock County, Idaho. Moscow, Idaho, controlled valley networks and the decline of surface on Earth and Mars. Science, v. 320, p. 1067–1070, doi: Idaho Geological Survey, scale 1:24,000. runoff on early Mars. Journal of Geophysical Re- 10 .1126 /science .1156630 . Schumm, S.A., and Phillips, L., 1986, Composite chan- search–Planets, v. 110, p. 1–17. Lamb, M.P., Mackey, B.H., and Farley, K.A., 2014, Amphi- nels of the Canterbury Plain, New Zealand; a Martian Harrison, K.P., and Grimm, R.E., 2009, Regionally com- theater-headed canyons formed by megafl ooding at analog? Geology, v. 14, p. 326–329, doi: 10 .1130 /0091 partmented groundwater fl ow on Mars. Journal of Malad Gorge, Idaho. Proceedings of the National -7613 (1986)14 <326: CCOTCP>2 .0 .CO;2 . Geophysical Research–Planets, v. 114, E04004, doi:10 Academy of Sciences of the United States of America, Scott, W.E., 1982, Surfi cial Geologic Map of the Eastern .1029 /2008JE003300 . v. 111, p. 57–62, doi: 10 .1073 /pnas .1312251111 . Snake River Plain and Adjacent Areas, 111 Degrees Haviv, I., Enzel, Y., Whipple, K.X., Zilberman, E., Matmon, Leeman, W.P., and Bonnichsen, B., 2005, Overview of si- to 115 Degrees W., Idaho and Wyoming. U.S. Geo- A., Stone, J., and Fifi eld, K.L., 2010, Evolution of ver- licic volcanism of the Snake River Plain–Yellowstone logical Survey Miscellaneous Field Investigations Map tical knickpoints (waterfalls) with resistant caprock: (SRPY) province. Geochimica et Cosmochimica Acta, I-1372, scale 1:250,000. Insights from numerical modeling. Journal of Geo- v. 69, p. A237. Skipp, B., and Kuntz, M.A., 2009, Geologic Map of the Arco physical Research–Earth Surface, v. 115, F03028, doi: Licciardi, J.M., and Pierce, K.L., 2008, Cosmogenic expo- 30 × 60 Minute Quadrangle, South-Central Idaho. Mos- 10 .1029 /2008JF001187 . sure-age chronologies of Pinedale and Bull Lake gla- cow, Idaho, Idaho Geological Survey, scale 1:100,000.

Geological Society of America Bulletin, Month/Month 2014 13 Geological Society of America Bulletin, published online on 6 November 2014 as doi:10.1130/B31141.1

Amidon and Clark

Smith, R.P., Lenhard, R.J., Yonk, A.K., Wright, P.M., and 10 .1130 /0016 -7606 (1997)109 <0855: GAORHC>2 .3 eastern Oregon. U.S. Geological Survey Professional Herzog, J.D., 2004, Geologic setting of the Snake .CO;2 . Paper 1408-B, 39 p. River Plain aquifer and vadose zone. Vadose Zone Whitehead, R.L., 1984, Geohydrologic Framework of the Whitehead, R.L., and Covington, H.R., 1987, Thousand Journal, v. 3, p. 47–58, doi: 10 .2136 /vzj2004 .4700 . Snake River Plain, Idaho and Eastern Oregon. U.S. Springs Area Near Hagerman, Idaho, in Beus, S.S., Stearns, H., Crandall, L., and Steward, W.G., 1938, Geology Geological Survey Professional Paper, Open-File Re- ed., Boulder, Colorado, Geological Society of Amer- and groundwater resources of the Snake River Plain in port 1408-B, 3 sheets. ica, Centennial Field Guide, Rocky Mountain Section, southeastern Idaho. Washington D.C., U.S. Govern- Whitehead, R.L., 1986, Geohydrologic Framework of the v. 2, p. 224–228. ment Printing Ofﬁ ce, 263 p. Snake River Plain, Idaho and Eastern Oregon. U.S. Thackray, G.D., Owen, L.A., and Yi, C.L., 2008, Timing and Geological Survey Hydrologic Atlas 681, 6 maps on SCIENCE EDITOR: CHRISTIAN KOEBERL nature of late Quaternary mountain glaciation. Journal 3 sheets. ASSOCIATE EDITOR: NADINE BARLOW of Quaternary Science, v. 23, p. 503–508, doi: 10 .1002 Whitehead, R.L., 1987, Geohydrologic Framework of the /jqs .1225 . Snake River Plain Regional Aquifer System, Idaho and MANUSCRIPT RECEIVED 27 MAY 2014 Welhan, J.A., and Reed, M.F., 1997, Geostatistical analysis Eastern Oregon. U.S. Geological Survey Open-File REVISED MANUSCRIPT RECEIVED 1 OCTOBER 2014 of regional hydraulic conductivity variations in the Report 87-107, 72 p. MANUSCRIPT ACCEPTED 23 OCTOBER 2014 Snake River Plain Aquifer, eastern Idaho. Geological Whitehead, R.L., 1992, Geohydrologic Framework of the Society of America Bulletin, v. 109, p. 855–868, doi: Snake River Plain Regional Aquifer System, Idaho and Printed in the USA

14 Geological Society of America Bulletin, Month/Month 2014