Abstract

Geologic, geomorphic, and geophysical analyses of landforms, sediments, and structures in northern Cache Valley, USA, document a revised history of flooding and recession of Lake Bonneville, the world's premier pluvial lake. Crosscutting relationships suggest that the Riverdale fault produced a surface-rupturing earthquake ∼25 km southeast of Zenda shortly before, during, or after the Bonneville flood, as well as possible younger surface ruptures. Thus fluctuating stresses and pore pressure induced by changing lake levels may have triggered a large earthquake that, in turn, triggered the Bonneville flood. The flood scoured ∼25 km of Cache and Marsh Valleys and activated landslides during its ∼100 m incision to a new outlet near Swan Lake, Idaho. One to two thousand years of steady outflow produced the main ∼4775 ± 10 ft (1455 ± 3 m) Provo shoreline, ∼10 m above the commonly accepted altitude. Later Lake Bonneville oscillated below the main Provo shoreline, incised the Swan Lake outlet, rose and paused briefly at a new lower Provo sill (4745 ± 10 ft [1446 ± 3 m], P9?) 23 km south of Zenda, before reverting to a closed-basin condition. Correlation to the Blue Lake chronology of Benson et al. suggests that aridity during the Heinrich 1 event activated the lower Provo sill ∼15.9 ka. An abandoned, meandering riverbed, north of the lower Provo sill, records a large northward flowing river. The Great Basin's modern divide at Red Rock Pass formed in the Holocene when a small alluvial fan filled the dry bed of this paleoriver.

INTRODUCTION AND SETTING

General Statement

Singular high-energy events in geologic history often leave disproportionate geologic records because they represent unusual amounts and variations in energy (Chan and Archer, 2003). The great Bonneville flood in the northeast Great Basin, roughly 17,400 calendar years ago, is a well-known event of this type that affected one of the best known pluvial lakes of North America (Gilbert, 1880, 1890; Malde. 1968; Currey, 1990; Oviatt et al., 1992; O'Connor, 1993; calibration to calendar years in part from Guido et al. [2007] and Hart et al. [2004]). At the time of this flood, the shoreline in the Bonneville basin dropped 335 ft (∼102 m), and 380 mi3 (∼4750 km3) of water drained northward into the Snake River with a maximum discharge of about a million cubic meters per second (Figs. 1–4; Malde, 1968; O'Connor, 1993). Roughly half the water in the basin exited during the flood (O'Connor, 1993).

Surprisingly little is known about the geologic record of this flood and its aftermath near its threshold in northern Cache Valley (Figs. 1, 3, and 4). The subsequent events of Lake Bonneville at the Provo shoreline in this critical region are also poorly documented. The conflicting interpretations of these events are published mostly in abstracts, guidebook articles, geologic maps, and theses, and most in-depth analyses were based on examining landforms in the central part of the Bonneville basin, particularly in mid-lake locations (cf. Currey et al., 1984; Oviatt, 1987, 1997; Burr and Currey, 1988; Currey and Burr, 1988; Oviatt et al., 1992) and cores (Spencer et al., 1987; Oviatt et al., 1999; Benson et al., 2011; McGee et al., 2011); or focused primarily on flood deposits farther north (Malde, 1968; O'Connor, 1993). New mapping and analysis, and new shoreline-elevation data, in conjunction with previous and new geologic observations and inferences, support a new synthesis of the shifting outlets of Lake Bonneville, and clarify the likely sequence of events. Reservoir-induced seismicity and/or landsliding along the newly identified Riverdale structure are implicated as a possible cause of the Bonneville flood. This raises the possibility that tectonic processes were more important than climatic forcing in triggering the great Bonneville flood.

Isostatic rebound, after Lake Bonneville dropped to the two successive Provo shorelines and continuing for a time as it evaporated to its present level at the Great Salt Lake (∼1000 ft [305 m] below the highest Bonneville shoreline), tilted the highest shorelines of the Bonneville lake cycle gently northward in Cache Valley. The altitude drops from 5165 ± 10 ft (∼1575 m) at the south end of Cache Valley to 5107 ± 1 ft (∼1557 m) at Franklin, Idaho; 5110 ± 10 ft (∼1558 m) slightly north of Red Rock Pass; and 5090 ± 10 ft (∼1552 m) at the southern end of Oneida Narrows where the Bear River enters Cache Valley in the northeast (Gilbert, 1880, 1890; Williams, 1962a, 1962b; Crittenden, 1963; McCalpin, 1989, 1994). Locally this regional tilting was complicated by downward deflections near small Gilbert-type deltas and by Holocene faulting (McCalpin, 1994; Black et al., 1999, 2000). Shorelines of the Provo occupation of the lake are much less tilted than the Bonneville shoreline in Cache Valley (see below).

Previous Interpretations of Drainage Divides, Outlets, and the Bonneville Flood

Gilbert (1880, 1890) interpreted the geologic record near the current drainage divide between northern Cache Valley and southern Marsh Valley (Red Rock Pass) and provided the seminal interpretation of hydrology of the highest Bonneville lake level, the subsequent Bonneville flood, and the Provo shorelines (Figs. 1, 3, 4, 5, and 6). Gilbert (1880, 1890) interpreted both the Bonneville and Provo shorelines as resulting from an open, overflowing lake, and he identified the threshold at the time of the Bonneville flood at Zenda, Idaho (Figs. 1–5). He concluded that the outlet of Lake Bonneville then shifted south at least 12 km to a new northern lake margin, at the Provo shoreline “…between Swan Lake and the Round Valley Marsh” (Gilbert, 1890, p. 178), and that “…the point of outlet was at the edge of what is called Round Valley Marsh …, nine miles south of Red Rock (Pass)” (Gilbert, 1880, p. 348). He gave the name “Bonneville River” to the large river that flowed northward from Lake Bonneville.

Gilbert's argument for a large southward shift of the outlet was forgotten, and subsequent workers concluded instead that the outlet shifted only ∼2.5 km south from the hamlet of Zenda, Idaho, to Red Rock Pass, which implies that the modern drainage divide at Red Rock Pass formed in late Pleistocene time after recession from the Provo shoreline (Ives, 1948; Williams, 1962a; Williams and Milligan, 1968; Currey, 1982; Currey et al., 1984; Bright and Ore, 1987; Burr and Currey, 1988; Smith et al., 1989; Oviatt et al., 1992; O'Connor, 1993; Link et al., 1999). Most prior studies interpreted the highest position of Lake Bonneville as overflowing for a brief or fairly long period of time (Gilbert, 1880, 1890; Currey, 1982; Benson et al., 2011; McGee et al., 2010; and most others), but this conclusion is also debated (O'Connor, 1993). Smith et al. (1989) and O'Connor (1993) summarized key prior work in the Red Rock Pass and Zenda areas.

Megalandsliding near Red Rock Pass during Bonneville and Provo time is a common explanation for the multiple levels of shorelines developed near the Provo level of Lake Bonneville elsewhere in the Bonneville basin and for some of the events of the preceding Bonneville flood (Sewell, 1980; Sewell and Shroder, 1981; Currey and Burr, 1988; Burr and Currey, 1988; Shroder and Lowndes, 1989; Smith et al., 1989; Cornwell and Shroder, 1990; O'Connor, 1993). In one of these interpretations, a huge (> 17 km2) landslide was hypothesized to have collapsed the entire northeastern flank of the Bannock Range after the catastrophic Bonneville flood had undercut it at Red Rock Pass (Fig. 4; Shroder and Lowndes, 1989; Sewell, 1980; Sewell and Shroder, 1981; Sewell, cited in Smith et al., 1989; Link et al., 1999). Others suggested that subsequent smaller landslides in the Red Rock Pass area were responsible for the multiple shorelines associated with the Provo shoreline in the center of the lake basin (Burr and Currey, 1988; Smith et al., 1989; Godsey et al., 2005). This explanation of the multiple Provo shorelines may not be feasible if the outlet for the Provo shoreline was farther to the south, as Gilbert (1890) proposed.

The nature of the threshold near Zenda is also disputed, and therefore different hypotheses have been advanced for initiation of the Bonneville flood. One hypothesis concluded that two alluvial fans coalesced at their distal toes as a dam, or that an alluvial fan lapped against bedrock, constrained Lake Bonneville, and failed in the flood (Gilbert, 1880, 1890; Ives, 1948; O'Connor, 1993; Link et al., 1999). Others inferred that karst in the Cambrian carbonates beneath and flanking Red Rock Pass permitted throughflow that initially prevented overtopping but eventually failed due to ongoing subsurface dissolution (J. Stewart Williams, 1967, oral commun. to Oaks; Link et al., 1999), that resistant bedrock formed the Zenda sill before the failure of the threshold, that the flood was due to a drop from an altitude above the Bonneville shoreline to the Bonneville level (Malde, 1968), or that landsliding on a huge scale was a factor (Shroder and Lowndes, 1989).

Subsurface flow (throughflow in pores and/or piping along solution channels) of lake water through carbonate bedrock and/or overlying alluvial-fan material may have contributed to the failure (Cornwell and Shroder, 1990; O'Connor, 1993). No previous study thoroughly tested these numerous competing hypotheses nor has the outlet of Lake Bonneville been thoroughly mapped and analyzed since the work of Gilbert (1880, 1890) (Fig. S13 in the Supplemental File1).

Shoreline Altitudes in the Bonneville Basin

Shorelines in Cache Valley, Utah and Idaho, outline an elongate, complex graben that is up to 25 km wide and 110 km long and was produced by north- to north-northwest–striking Basin-and-Range normal faults (Evans and Oaks, 1996; Janecke and Evans, 1999) (Figs. 1 and 2). Its valley-floor altitudes range from ∼4315 ft (1315 m) where the Bear River exits westward at Cutler Narrows, upward to the highest late Pleistocene Lake Bonneville shoreline at 5165 ft (1574 m) in the south (McCalpin, 1989, 1994) (Figs. S1–S12 [see footnote 1]). We report altitudes in both feet and meters above sea level because the topographic maps on which we mapped the landforms are contoured in feet. In this paper we refer exclusively to Swan Lake, the shallow body of water, and not to the small community with the same name farther north (Figs. 1 and 2).

Interpretation of Lake Bonneville is complicated by uncertainty and disagreement about the altitudes of the successive Bonneville and Provo shorelines and about the positions of their northern threshold(s) (Fig. 3). Isostatic depression of the lithosphere by the load of the deep lake, and the subsequent rebound of the lithosphere when the lake water dropped produced vertical changes of up to 250 ft (∼75 m) in the altitudes of shorelines in the Bonneville basin (Gilbert, 1880, 1890; Crittenden, 1963; Currey et al., 1984; Bills and May, 1987; Bills et al., 1994). Northern Cache Valley was a shallow bay at the northeast margin of the lake, so deflections there were far smaller than in more interior parts of the Bonneville basin, and deflections of the Provo shorelines are thus negligible (Crittenden, 1963; Currey, 1982; McCalpin, 1994; Solomon, 1999; Biek et al., 2003; this study). As a result, several analyses of Lake Bonneville assumed little or no isostatic uplift of shorelines in the Red Rock Pass–Zenda area, and used features there as the reference altitude for deflected shorelines elsewhere in the Bonneville basin (Crittenden, 1963; Bills and May, 1987, 1994; Currey and Burr, 1988; Oviatt et al., 1992). The well-accepted reference altitudes are 5192 ft (1552 m) for the Bonneville shoreline and 4737 ft (1444 m) for the Provo shoreline in the Red Rock Pass–Zenda area (Currey, 1982).

Establishing the reference altitude in northernmost Cache Valley is not completely straightforward because the shorelines in the key area are weak. The highest Bonneville shoreline is weakly to moderately expressed in the Red Rock Pass–Zenda area (Bright, 1963), probably due to low wave energy in the shallow parts of the lake, and Provo shorelines are similarly poorly expressed on and north of the Provo delta of the Bear River (Figs. 5 and 7; Figs. S1–S13 [see footnote 1]) (Williams and Milligan, 1968). Furthermore, there is no Provo shoreline north of Swan Lake (Bright, 1963; Fig. 6; Figs. S1–S5 [see footnote 1]).

Altitudes of shorelines in the Bonneville basin that have been corrected for rebound should agree, and should therefore match altitudes in northern Cache Valley in the Red Rock Pass–Zenda area. However, the Supplemental Table2 shows that there has been considerable disagreement about the altitudes of the Bonneville and Provo shorelines in northern Cache Valley (Figs. S1–S11 [see footnote 1]). The highest level of Lake Bonneville there was thought to be ∼5135 ft (1565 m) (Figs. 1–5, 7, and 8; Figs. S1–S13 [see footnote 1]) (Hardy, 1957; Bright, 1966; Link, 1982a, 1982b) or 5090 ft (1551 m) (Burr and Currey, 1988; Currey and Burr, 1988; Smith et al., 1989; Godsey et al., 2005). Two different altitudes have also been identified for the Provo shoreline: a higher one between 4760 and 4780 ft (∼1454 m) and a lower level between 4737 and 4750 ft (∼1444–1448 m) (Supplemental Table [see footnote 2]). Comparison of the 1915 15-min topographic maps with 7.5-min topographic maps published in 1968 and 1969 (1927 North American datum) shows no differences in altitudes of basinal features. This rules out a datum change on topographic maps as the source of the different altitudes reported for the Bonneville and Provo shoreline.

The lower altitude of the Provo shoreline, which is more commonly reported since 1970, might have been influenced by a seismic-refraction study done by Williams and Milligan (1968) at Red Rock Pass. They interpreted an increase in velocity at depth to be a buried bedrock sill (rather than the water table) at an altitude of 4755 ft (∼1450 m) (Supplemental Table [see footnote 2]). The buried “sill” of Williams and Milligan (1968) is the only report of high-standing bedrock along the Swan Lake outflow channel in northern Cache Valley or southern Marsh Valley (Fig. 2), so it was natural to interpret its altitude at Red Rock Pass as the reference Provo shoreline.

All of these factors, plus the possibility that some shorelines assigned to the Provo level formed during pauses in the initial Bonneville transgression (Sack, 1999), make it critical to correctly identify shorelines and correlate them laterally (Fig. 3). We focused on and mapped the two best developed shorelines in northern Cache Valley—the highest Bonneville level and the main (higher) Provo shoreline of Lake Bonneville. In addition, we searched for, located, and mapped a more subtle lower Provo shoreline at the altitudinal range of an ancient north-flowing river that we discovered in Round Valley at ∼4745 ft (1446 m) (Fig. 9; see below).

We produced independent maps of shorelines instead of relying on those of prior reports (Figs. 1–8; Figs. S1–S13 [see footnote 1]) but compiled prior mapped shoreline locations, if those shoreline altitudes agreed with adjacent areas, could be traced laterally, and retained consistent and sensible geomorphic character along trend. Additional methodological details of our mapping criteria, landscape reconstruction, and analysis are in the Supplemental File (see footnote 1).

RESULTS

Shorelines

The highest Bonneville shoreline is very well expressed as a notch or beach ridge in the landscape (Gilbert, 1890; Figs. S8–S11 [see footnote 1]). It is easily identified around the basin, except north of Swan Lake (Bright, 1963). The singular Bonneville shoreline decreases in altitude northward across the area, except perhaps in its northernmost extent in Marsh Valley where it is weakly expressed and might be slightly higher (Bright, 1963; this study). The Bonneville shoreline reflects significant initial erosion, even in settings that were later shielded from waves by topsets of the deltas of the Bear River (Fig. 7). The erosional aspect of the shoreline is conspicuous around northern Cache Valley as extensive wave-cut and wave-built terraces and associated scarps (e.g., Gilbert, 1880, 1890). These formed at the bases of triangular facets, up to 50 m high, and commonly cut into the footwalls of basin-bounding normal faults (McCalpin, 1994; Solomon, 1999; this study). The Bonneville shoreline lies between 5090 and 5120 ft (1551–1561 m) in the northernmost locations of our study. The highest Bonneville shoreline is so well expressed (Fig. 8; Fig. S11 [see footnote 1]) that a fairly long and stable occupation likely produced it. In contrast, the Bonneville shoreline north of Swan Lake, Idaho, is weakly developed in northern Cache Valley (Fig. 5; Bright, 1963; this study) and in southernmost Marsh Valley, the Bonneville shoreline is barely discernible on the surface of the Marsh Creek pediment and alluvial fan, near Zenda, Idaho (Gilbert, 1880, 1890; Bright, 1963; this study).

Provo shorelines are the second most prominent shorelines of Lake Bonneville and lie ≥315 ft (≥96 m) below the Bonneville shoreline. Our analysis of the landscape and mapping shows that there are two important Provo shorelines in northern Cache Valley (Figs. 3, 8; Fig. S9 [see footnote 1]), and these coincide with the two altitudes reported previously in Cache Valley (Supplemental Table [see footnote 2]). The main Provo shoreline is the higher shoreline at ∼4775 ft (1455 m) in the northeast part of Cache Valley. The same shoreline is close to 4800 ft (1463 m) farther south and in the west, particularly on the west side of Cache Valley where the Dayton-Oxford and West Cache fault zones may have uplifted the shoreline in their footwalls (e.g., Fig. 8; Figs. S9 and S11 [see footnote 1]). Altogether this higher main Provo shoreline has altitudes between 4770 and 4810 ft (∼1454–1466 m) in northern Cache Valley (Figs. 1, 2, and 6; Figs. S1–S11 [see footnote 1]). We refer to this higher, main shoreline as the “4775 ft Provo shoreline (1455 m)” (Figs. 3 and 6) because it has this altitude in the northeasternmost part of Cache Valley where rebound and faulting are least likely to have modified it. The altitudinal variation of this shoreline is about ±10 ft (3 m).

The 4775 ft (1455 m) shoreline is distinctive because it typically coincides with a wide wave-cut cliff and wave-built bench, and separates highly eroded, gullied hillslopes above from smooth, much less eroded or gullied hillslopes below (Figs. 7 and 8; Figs. S8–S10 [see footnote 1]). Benches associated with this shoreline are among the most prominent ones in Cache Valley, and many exceed the Bonneville bench in width (e.g., McCalpin, 1994; Solomon, 1999; Biek et al., 2003; this study, Fig. 8; Figs. S7–S10 [see footnote 1]). Some of the best developed parts of this shoreline are nearly continuous wave-cut cliffs and associated basinward depositional benches on the western side of Cache Valley, from Weston Creek southward. Southwest of the Weston delta, the main higher Provo shoreline forms a prominent erosional scarp (notch) and wave-cut bench as high as 4800–4810 ft (1466 m) above and adjacent to the smooth topset of the Provo delta deposited by Weston Creek (Fig. 8; Figs. S8 and S9 [see footnote 1]). The 4775 ft (1455 m) main higher shoreline is absent north of Swan Lake, Idaho, where a flat-bottomed chute with an underfit stream has a floodplain at or slightly above this altitude (Figs. 2, 4, and 5; Figs. S1–S6 and S13 [see footnote 1]) (Gilbert, 1880, 1890; Bright, 1963; this study).

The main higher 4775 ft (1455 m) Provo shoreline corresponds to the singular Provo shoreline that Bright (1963, 1966) mapped in northern Cache Valley, in the only small-scale map of the shoreline near Lake Bonneville's outlet with a topographic base. Gilbert (1880, 1890) and Williams (1962a, 1962b) also identified this altitude as the main Provo shoreline. Gilbert's study and map of the shoreline (1880, 1890) does not report altitudes or have a topographic base, but a comparison between his map and modern topographic maps shows that he had identified this altitude as the Provo shoreline (compare Fig. S13 with Figs. S1–S6 [see footnote 1]). The Supplemental File (see footnote 1) provides additional evidence for the 4775 ft (1455 m) shoreline being the main Provo shoreline.

The lower Provo shoreline at ∼4745 ft (1446 m) is subtle, and well-developed stretches are uncommon. Its degree of development is similar to that of a myriad of subsidiary shorelines in the Bonneville basin. Thus, it would rarely be confused with the main, higher 4775 ft (1455 m) Provo shoreline. The lower Provo shoreline also varies somewhat in altitude, and it is highest in the footwall of the Dayton-Oxford and West Cache faults (Figs. 6, 8, and 10; Figs. S8–S10 [see footnote 1]). We searched for and located this shoreline, roughly between 4740 and 4750 ft (∼1446 m) near the Clifton sill because this is the altitude of an abandoned meander belt in Round Valley (see below) that was immediately downslope of Lake Bonneville when the meander belt was occupied by a large north-flowing river.

The lower Provo shoreline is especially well expressed on the Weston delta, where it forms a step down to the east of ∼30 ft (9 m) from the top of the smooth Provo delta that formed during the stable lake at the main Provo shoreline (Janecke and Oaks, 2011; Fig. 8). This shoreline is also expressed on deltas in central and southern Cache Valley (e.g., Solomon, 1999; Biek et al., 2003) and in some areas of strongly eroded hillslopes.

The 4745 ft (1446 m) lower Provo shoreline is present only south of the irregularly shaped Twin Lakes horst, which also marks the structural northern end of the Cache Valley basin. Scattered bedrock exposures, gravity data, and hilly topography show that the foothills of the Bannock Range merge with the foothills of the Portneuf Range from the Twin Lakes horst northward. The low point in the Twin Lakes horst is its western topographic saddle near Clifton, Idaho (Figs. 4 and 6; Figs. S2, S6, and S7 [see footnote 1]). The Salt Lake Formation is no deeper than ∼4645 ft (1416 m), and possibly as shallow as ∼4705 ft (1434 m) in this saddle, the Clifton sill, in two (of four) drillers’ logs of water wells that have reasonably detailed lithologic information. Roughly the altitude of the land surface in the saddle (∼4745 ft [1446 m]) was interpreted to be the main Provo shoreline in Cache Valley by Williams and Milligan (1968), Currey (1982), Currey et al. (1984), Burr and Currey (1988), Oviatt et al. (1992), and Link et al. (1999). All of these researchers also interpreted Red Rock Pass as the threshold of the Provo shoreline, yet none depicted the 23-km-long “finger” lake that is required by this interpretation to connect Red Rock Pass with the distant 4745 ft (1446 m) shoreline along the south edge of the Twin Lakes horst.

The pair of Provo shorelines that we identify in Cache Valley are quite different from another doublet of Provo shorelines in the Bonneville basin that typically lie a few meters apart and are variants of our upper Provo shoreline (Gilbert, 1880, 1890 called this “the underscore”; J. Oviatt, 2008, written commun.). Instead, the pair of shorelines reported here in Cache Valley is more analogous to the highest (P1 and P3) and lowest (P9) Provo shorelines in the central Bonneville basin, which are separated by 34–50 ft (∼13 m) in altitude (Gilbert, 1890; Currey and Burr, 1988; Burr and Currey, 1988; Godsey et al., 2005). The upper and lower Provo shorelines are separated by 30–50 ft (∼9–15 m) in most areas in northern Cache Valley, with the vertical distance between them increasing southward (e.g., Fig. 8; Fig. S9 [see footnote 1]).

Landforms Relevant to the Bonneville Flood and Its Aftermath

Shorelines in northern Cache Valley occupy different geographic settings in the landscape, and become progressively less numerous northward in Cache Valley as the bottom of the basin rises in altitude into the bedrock foothills between the Bannock and Portneuf Ranges. We identified five distinct geomorphic zones related to the Bonneville flood and its aftermath in northern Cache Valley. From north to south, they are: (1) southern Marsh Valley and the Red Rock Pass area; (2) Swan Lake terrain; (3) Round Valley terrain; (4) Twin Lakes horst; and (5) Lake Bonneville deltas of the Bear River (Figs. 1, 2, 4–6, 9, and 10; Figs. S1–S11 [see footnote 1]). These five terrains have markedly different geomorphic histories and deposits. For example, Lake Bonneville submerged parts of all five terrains when it was at its highest level. However, during occupation of the two Provo shorelines, Lake Bonneville was restricted first to south of the Swan Lake terrain and then to south of the Twin Lakes horst block. The deltas of the Bear River, the large southernmost area, are the topic of a separate paper (Janecke and Oaks, 2011).

Southern Marsh Valley and the Red Rock Pass Area

General statement. The relationships around Red Rock Pass and southern Marsh Valley are critical to understanding Lake Bonneville and the Bonneville flood because this terrain contained the outlet of the highest shoreline of Lake Bonneville and because Red Rock Pass has been interpreted as the threshold for Lake Bonneville after the flood. The southern end of Marsh Valley preserves high, abandoned pediment- and alluvial-fan complexes on the flanks of the Bannock and Portneuf Ranges, the underlying and adjacent Neogene Salt Lake Formation, and a high-standing, gently curving scour-and-discharge channel that angles northwest across Marsh Valley (Figs. 1, 2, 4, 5, and 10; DeVecchio et al., 2003; Thackray et al., 2011). Basinward-sloping Pleistocene pediment remnants, with caps of Pleistocene gravel, sand, and silt are plentiful in this region, and are concentrated in the northeastern part of the neotectonic parabola centered on the Yellowstone hotspot, where they record differential uplift, erosion, and subsidence as the hotspot passed nearby (Fig. 5; Figs. S3–S6 and S12 [see footnote 1]; Williams, 1948; Smith, 1997; Goessel, 1999; Goessel et al., 1999; Oaks et al., 1999; Oaks, 2000; Carney et al., 2003; DeVecchio et al., 2003; Janecke et al., 2003; Janecke, 2007; Thackray et al., 2011).

Pre-Tertiary bedrock (Late Proterozoic and early Paleozoic) in a major spur of the Bannock Range trends northeast, and drops in altitude toward Red Rock Pass and Zenda, beneath faulted Neogene Salt Lake Formation. The easily eroded, lacustrine tuffaceous Salt Lake Formation and some smaller fault blocks of pre-Tertiary rocks are truncated by a sloping erosional pediment surface, which is overlain by a thin pediment cap, overlying and inset alluvial-fan deposits in some areas, and later loess (DeVecchio et al., 2003; Janecke and Oaks, 2011; Thackray et al., 2011; this study). The thickness of the alluvial material on and adjacent to the pediment cap varies unpredictably due to cut-and-fill processes near paleochannels and due to aggradation near active streams, such as Marsh Creek, after formation of the pediment and pediment cap (e.g., DeVecchio et al., 2003).

The northeast-trending spur of the Bannock Range and its north- to east-sloping flanking pediments and Pleistocene alluvial fans lie directly opposite a radiating southwest- to northwest-sloping pediment cut across the Salt Lake Formation between upper Marsh Creek and Downey, Idaho (Figs. S4, S12, and S13 [see footnote 1]; DeVecchio et al., 2003; this study). Quaternary sediment is thickest in the east where cut-and-fill along upper Marsh Creek incised the Salt Lake Formation more deeply, but stream cuts at the downslope end show that the rest of this landform is underlain by Neogene Salt Lake Formation to within meters or tens of meters of the present land surface.

The Bonneville shoreline is lightly etched onto the southern part of the Marsh Creek pediment- and alluvial-fan complex at ∼5100 ft (∼1559 m), but the shoreline is nowhere developed on the northwest- to west-sloping parts (Figs. 5 and 10; Fig. S13 [see footnote 1]; Gilbert, 1880, 1890; Bright and Ore, 1987; this study). The Marsh Creek and Aspen Creek bench pediments, the deformed Salt Lake Formation beneath the pediments, and thin Quaternary sediment constrained Lake Bonneville at Zenda.

Groundwater flow through the Marsh Creek pediment. The surface of the northwest- to west-sloping remnant of the Marsh Creek pediment and alluvial fan, below ∼5120 ft (1561 m), is very different from the surface of the higher, smoother southern part (O'Connor, 1993; this study; Fig. 5; Fig. S2 [see footnote 1]). Above ∼5120 ft (1561 m), all gullies have V-shaped cross sections. On the northwest, lower part of the loess-covered pediment and alluvial fan, from ∼5120 ± 10 ft (1561 ± 3 m) to the floor of Marsh Valley near 4930 ft (∼1503 m), drainages have U-shaped cross sections; occupy much more surface area of the Marsh Creek pediment; and have steep-walled, theater-shaped scallops along their sides and at their heads. The entire surface of the northwest-sloping Marsh Creek pediment below ∼5120 feet (1561 m) has scalloped heads of streams and broad, shallow, dry stream beds (Fig. 5; Fig. S3 [see footnote 1]).

These landforms on the Marsh Creek pediment are diagnostic of sapping, undermining by groundwater, and collapse (O'Connor, 1993). All drainages on the Marsh Creek pediment surface are inactive, and flows have shifted to the deeply incised valley of upper Marsh Creek at the east edge of the pediment and alluvial fan, now ∼65 m below the pediment-alluvial fan surface (Fig. S4 [see footnote 1]).

Material forming the sill at Zenda, Idaho. The greater Red Rock Pass area exposes three main rock units beneath a variable thickness of Quaternary alluvial and loess cover (Fig. 5): (1) limestone, dolostone, and some interbedded shale of the lower Paleozoic miogeocline; (2) weak tuffaceous Neogene Salt Lake Formation with some conglomerate-bearing intervals; and (3) poorly consolidated latest Pleistocene to Holocene(?) landslide deposits. The Salt Lake Formation dominates in the north, and bedrock dominates south of two north-dipping normal faults at Red Rock Pass (Fig. 5; Mayer, 1979; Link, 1982a, 1982b; DeVecchio et al., 2003; Long and Link, 2007; this study).

The central 500 m of the Zenda sill area is flanked by gently northeast-tilted silty and conglomeratic Salt Lake Formation, a pediment, and a pediment cap composed of a few meters of loess. Beds of well-cemented pebble conglomerate of the Salt Lake Formation, exposed here and there in the threshold area, are between 0.3 and 10 m thick, and persist ∼6 km from Zenda, Idaho, northward to Downey, Idaho (Figs. S12 and S14 [see footnote 1]). An east-dipping section of thick upper Miocene ash beds underlies this conglomeratic interval farther south (north of Red Rock Pass). They were chemically correlated to ashes erupted from the Twin Falls center in the eastern Snake River Plain (DeVecchio et al., 2003).

There are several factors leading us to correlate this resistant conglomerate-bearing package with a conglomeratic member of the Salt Lake Formation instead of following DeVecchio et al.’s (2003) assignment of most of the sediment to a Quaternary fanglomerate deposit. The conglomerate in the Zenda area has: (1) fairly good cementation; (2) numerous conspicuous quartzite clasts (Fig. S14 [see footnote 1]; unlike the dark-chert- and carbonate-rich Quaternary fanglomerate nearby); (3) a high degree of sorting and rounding of these resistant clasts; (4) associated finer interbeds; (5) a strong resemblance to conglomerate beds in tuff-bearing Salt Lake Formation exposed a short distance to the west and in Cache Valley; and (6) conformity with an underlying dated ashy member of the Salt Lake Formation. Slightly east of the paleodivide near Zenda, conglomerate-bearing silty Salt Lake Formation persists upward to within a few meters of the original land surface, beneath several meters of loess in hillside and gully cuts. A valley-center conglomeratic knob at 4764 ft (1451 m) ∼0.8 km northwest of Zenda may be a flood-scupted remnant of the Zenda sill. Surprisingly, the outlet of Lake Bonneville at its highest shoreline at Zenda was not positioned over hard and durable bedrock, but instead was constrained by a conglomerate-bearing interval within the fairly weak Neogene Salt Lake Formation.

Material at Red Rock Pass. Fairly weak, erodable materials also form the current drainage divide between Cache and Marsh Valleys at Red Rock Pass. Although Red Rock Pass exposes resistant rock types immediately east and west of the Swan Lake scour and discharge channel (Fig. 5), the present divide is on the floor of the alluviated and smooth Swan Lake scour channel, roughly in the middle of its ∼25-km-long reach (Figs. 1, 2, 5, and 11; Figs. S1–S5, S12, and S13 [see footnote 1]; Gilbert, 1880, 1890). A post-Bonneville alluvial fan forms the divide.

From the northeast, Marsh Creek built a small alluvial fan into the abandoned scour and discharge channel. This must have occurred after recession of Lake Bonneville and after termination of large flows in the Bonneville River. It has produced enough relief to localize the drainage divide at the confluence of Marsh Creek and the Swan Lake scour and discharge channel (in Fig. S1 [see footnote 1], note the anomalously higher altitude of the present surface of the Swan Lake scour and discharge channel near Red Rock Pass where Marsh Creek empties into the channel) (Gilbert, 1880, 1890; Williams, 1890; Milligan, 1968; Bright and Ore, 1987).

Two scour and discharge channels at Red Rock Pass. There are two well-developed scour and discharge channels north of Red Rock Pass that diverge around Red Rock Butte (Sewell, in Smith et al., 1989; this study). The lesser known eastern channel is more curved than the western channel and has a concave-west geometry (Figs. 4, 5, and 10; Figs. S3–S5 [see footnote 1]). The eastern wall of this scour and discharge channel exposes uncemented Quaternary sediment, whereas the floor exposes more durable pre-Quaternary rocks of the tuffaceous member of the Salt Lake Formation (DeVecchio et al., 2003; this study). This relationship suggests that the flood cut easily through the Quaternary sand, silt, and minor gravel of Marsh Creek, but downcutting slowed when the more resistant tuffaceous Salt Lake Formation and underlying Cambrian dolostone of the St. Charles Formation were uncovered. See the Supplemental File (footnote 1] for additional data.

Landslides near Red Rock Pass. The volume of Quaternary landslides near Red Rock Pass is far less than the hypothesized 17 km2 megalandslide of Sewell, and landslides there are modest in volume (Fig. 5) (Sewell, 1980; Sewell and Shroder, 1981; Shroder and Lowndes, 1989; Smith et al., 1989; Cornwell and Shroder, 1990). Most of the proposed area of the megalandslide in the northern Bannock Range is in-place Neoproterozoic to Paleozoic bedrock overlain by in-place faulted and folded Salt Lake Formation, a pediment, pediment-capping deposits, and some modest to small landslides near the Swan Lake scour and discharge channel (Fig. 5; Raymond, 1971; Mayer, 1979; Link, 1982a, 1982b; Hennings, 2002; DeVecchio et al., 2003; Long and Link, 2007; this study). The mapped relationships reveal a fringe of landslides concentrated west of the scour and discharge channel in a belt 1–1.5 km wide. The toes of some of these complex slump-flow landslides were modified and sculpted by the floodwaters of Lake Bonneville, and some distal deposits form fluted to rounded knobs that project upward from the floor of the Swan Lake scour and discharge channel (Fig. 5; Figs. S3–S5 [see footnote 1]) (Gilbert, 1880, 1890, and many others). If all of these landslides were initially connected and completely filled the Swan Lake scour and discharge channel (Fig. 5), their maximum area would have been less than 7 km2. Even at their maximum possible extent, landslides occupied less than half of the hypothesized megalandslide area.

Original geometry of southern Marsh Valley and northern Cache Valley. We reconstructed the surfaces of pediments and alluvial fans near Zenda, Idaho, to pinpoint the original topographic saddle that constrained Lake Bonneville at its highstand (Fig. 10). The reconstruction of the upper surface of the pediments and thin alluvial deposits is robust because the original surfaces of the Marsh Creek and Aspen Creek pediments and overlying Marsh Creek and Chicken Creek alluvial fans are smooth, regular, and predictable in shape (Figs. S3–S5 [see footnote 1]), and all but the downslope ends of these large surfaces are preserved. Very little projection was required to reconstruct the landscape prior to the flood (Fig. 10).

The reconstructed landscape and reconstructed Zenda area show that the original divide and outlet at the north end of pre-flood Cache Valley was a saddle with essentially the same altitude as the highest Bonneville shoreline, within error limits (Fig. 10). Pediments and alluvial fans in the east merged laterally at ∼5100 ft (1555 m) with steeper piedmont slopes of the Bannock Range in the west, and encircled a small bedrock knob at Red Rock Pass. Cross-sectional analysis confirms that the pediment of Marsh Creek graded to slightly below the altitude of the highest Bonneville shoreline nearby at ∼5115 ft (1559 m; Fig. S12 [see footnote 1]).

The reconstructed landscape makes it possible to delineate the “Zenda sill,” a composite landform that persisted south far into Cache Valley and north into Marsh Valley before the flood (Fig. 10). The Zenda sill was >10 km wide (N-S) near 1510 m altitude due to its gentle slopes, and had a butterfly shape in map view (Fig. 10). En echelon strands of the north-striking Dayton-Oxford fault zone in the west, and a speculative northward continuation of the Riverdale fault zone (see below) cross at least the southern part of the Zenda sill. Offset across them might have caused failure of the sill. Pediments cut across the Neogene Salt Lake Formation and its pediment-capping sediment comprises most of the northern half of the Zenda sill in Marsh Valley (DeVecchio et al., 2003; this study), whereas middle to late Pleistocene alluvial material (e.g., Bright and Ore, 1987; this study) dominates the Zenda sill in Cache Valley.

Swan Lake Terrain, Northernmost Cache Valley

General statement. Cache Valley narrows northward to Red Rock Pass, and has a single highstand Bonneville shoreline along its margins north of Swan Lake (Figs. 2, 4, 6, and 10; Figs. S1–S5 [see footnote 1]; Gilbert, 1880, 1890; Bright, 1963). We name this part of Cache Valley, between Swan Lake (the body of water) and Red Rock Pass, the Swan Lake terrain. The highest Bonneville shoreline is at an altitude near 5100 ± 10 ft (1554 + 3 m) in this area (Fig. 5; Bright, 1963; Supplemental Table [see footnote 2]).

Bedrock ridges. The Swan Lake terrain is an area of low, bedrock-cored hills with a fairly thin carapace of Quaternary sediment. This terrain is dominated by two E- to ENE-trending bedrock ridges, at Swan Lake in the south and at Red Rock Pass in the north (Long and Link, 2007; Figs. 2 and 4; Figs. S4–S6 [see footnote 1]). The southern bedrock ridge (horst) forms an intravalley high of intermittent bedrock exposure that trends east-northeast through Swan Lake (Link, 1982a, 1982b; Long and Link, 2007; Figs. 1–5; Figs. S4–S6 [see footnote 1]). The 2- to 3-km-wide Swan Lake bedrock ridge (horst) is cored by poorly exposed faulted and brecciated Salt Lake Formation and Neoproterozoic quartzite, and is probably bounded by buried ENE-striking normal faults (Figs. 1, 2, 4, and 11) (Oriel and Platt, 1980; Link, 1982a, 1982b; Eversaul, 2004; Long and Link, 2007).

The poor exposures and hummocky topography in the eastern half of the Swan Lake horst below the Bonneville shoreline are suggestive of mass wasting (Figs. S5 and S6; see the Supplemental File [footnote 1] for more detail), and are expressed across an area as large as 15 km2. There are few obvious shorelines preserved on this landform, and even the highest Bonneville shoreline is obscure across this horst. Scouring produced by the Bonneville flood and later agricultural activity further modified the original, possibly mass-wasted, carapace of the Swan Lake horst (Fig. 4). The disorganized geomorphic expression of shorelines in the Swan Lake horst is in sharp contrast to that of the Twin Lakes fault block, only ∼10 km farther south (Figs. 2, 6, and 7; Fig. S2 [see footnote 1]), where dozens of continuous depositional and erosional shorelines and short, northerly to NNW-striking fault scarps adorn most of the hillslopes. Irregular topography in the Swan Lake terrain has up to 500 ft (∼150 m) of relief (Figs. 5 and 10; Figs. S1–S6 [see footnote 1]).

Scoured landscapes. The lower part of the Swan Lake terrrain is mantled by Bonneville sediment or has distinct fluting and scours with north- to NNW-trending long axes that converge northward toward Zenda (Figs. 1, 4, 5, and 7; Figs. S4–S6 and S13 [see footnote 1]; Gilbert, 1880, 1890). Scoured hillslopes of the Bonneville flood end abruptly along the SSE edge of the Swan Lake bedrock ridge, and scoured landscapes are completely missing short distances farther south at identical altitudes (Figs. 1–5). Some scouring may be localized in the hanging walls of normal faults. Scoured landscapes do not appear above 4920 ft (∼1500 m).

Swan Lake scour and discharge channel. The lowest part of the Swan Lake terrain is a nearly linear stream valley that extends NNW from Swan Lake through Red Rock Pass and 12 km into Marsh Valley (Figs. 1, 2, 5, and 6; Figs. S2–S6 [see footnote 1]). Its present expression is a flat-bottomed chute 1640–2460 ft (∼0.5–0.75 km) wide and at least 100 ft (30 m) deep that contains Swan Lake, marshes, an underfit, south-flowing stream, and irregular, flow-modified mounds of brecciated bedrock near Red Rock Pass and Zenda (Fig. 5; Gilbert, 1880, 1890). Both the northern and southern parts of this scour-and-discharge channel lie near or slightly below altitudes of both Provo shorelines. The lowest altitude of the ground or water is the surface of Swan Lake at 4756 ft (1450 m), at the southern end of this Swan Lake scour and discharge channel, and the highest is at Red Rock Pass (∼4780 ft [∼1457 m]) (Fig. 6; Figs. S1–S6 [see footnote 1]).

Most of the Swan Lake scour and discharge channel is partly infilled with post-Bonneville sediment, so that the current surface lies above the original base of the outflow channel of the Bonneville flood—and above the original bed of the Bonneville River (Fig. 5; Figs. S3, S5, and S14 [see footnote 1]; Gilbert, 1880, 1890; Bright, 1966; Williams and Milligan, 1968). Holocene alluvial fans of tributary streams and intervening marshy sediments comprise some of the infill (Fig. 5). Differential infilling produced the subtle saddle and modern drainage divide within this scour and discharge channel at Red Rock Pass (Gilbert, 1880, 1890; Williams and Milligan, 1968). Swan Lake is ponded between tributary fans that empty into the scour and discharge channel directly north and south of it, and the lake represents a shallow, less-filled relict low in the channel (Gilbert, 1890; this study).

Coring at Swan Lake showed that the top of post-flood gravel was ∼4730 ft (1442 m), below roughly 31.5 ft (9.6 m) of post-flood fine sediment (Bright, 1966). At Red Rock Pass, the original base of the outflow channel may lie near 4755 ft (∼1450 m), beneath ∼20 ft (6 m) of fill, if unreversed seismic-refraction data and assumed seismic velocities are correct (Williams and Milligan, 1968) (but see discussion below).

Drillers’ logs of water wells in the Swan Lake scour and discharge channel constrain its original depth before infilling roughly midway between Red Rock Pass and Swan Lake. Drillers’ logs of two water wells there reveal at least 100 ft (30 m) of horizontal gravel, clay, and sand, and the original base of the Swan Lake scour and discharge channel at 4680 ft (1427 m) or deeper (Fig. 11). The lateral continuity of nearly horizontal units in the channel fill rules out correlation of these sediments with the tilted Salt Lake Formation that is exposed east and west of the channel (Fig. S14 [see footnote 1]).

Round Valley Terrain

Geography. Round Valley is a nearly flat terrain immediately south of, and lower than, the irregular Swan Lake terrain (Fig. 9; Figs. S4–S8 [see footnote 1]). The valley bottom here is also known as Oxford Slough, and is named for the town on its northwest margin. The valley bottom, which lies between 4745 and 4750 ft (∼1446 m), has very little relief. The bounding highlands are the Swan Lake horst in the north, the Bannock Range in the west, the Twin Lakes horst in the south, and the eroded foresets of the Bonneville delta of the Bear River in the east.

Meander Belt

Round Valley preserves the channel, point bars, and flood plain of an old river system that formed below, and probably later than, the higher 4775 ft (1455 m) Provo shoreline (Fig. 9; see also Google Earth or other aerial imagery of the area southeast of Oxford, Idaho). The landform is 1.5–3 km wide and ∼6–8 km long. This topography is diagnostic of a meandering stream because it preserves fluvial channels, point-bar scrolls, an adjacent flood plain with marshes, a levee system, abandoned oxbow lakes, and numerous other diagnostic features. Alternate possible origins of this landform were considered and rejected (in the Supplemental File [see footnote 1]) because they cannot account for the features in Round Valley.

Clay and silt comprise the current surface materials of Round Valley (Web Soil Survey, accessed June 31, 2008; http://websoilsurvey.nrcs.usda.gov/app/WebSoilSurvey.aspx). Fine sediment also lined a pit dug to 34 inches (0.9 m) depth and persist to the depth of 1–2 m in drainage ditches (Robert Bundy, manager at Oxford Slough Waterfowl Production area, 2007, oral commun.). Either these fine materials mantle underlying coarser fluvial sediment, or the river system transported mostly fine sediment. The latter is expected along an outlet stream downflow from the large sand to mud-dominated deltas of the Bear River (Janecke and Oaks, 2011). Water wells through the deltas of the Bear River consistently encounter abundant silt and clay, as well as some sand and minor gravel (Janecke and Oaks, 2011).

Flow directions then and now. Relict channels and oxbows in Oxford Slough are now occupied by a marsh, wetlands, and standing water. There is no through-flowing drainage in the main channel or its disconnected oxbow remnants. In high-water years, some overland flows exit southward from Round Valley (Robert Bundy, manager of Oxford Slough Waterfowl Production area, 2007, oral commun.).

Every tributary to the meander belt in Round Valley joins the main channel at an acute angle and opens southward (Figs. 2, 6, and 9), which indicates northward flow at the time of their formation. This is opposite the sluggish southward flow during high-water years of the past century. Northward elongation of some point bars provides additional evidence of downstream meander migration during northward flow (Fig. 9). The meander belt merges seamlessly into the Swan Lake scour channel in the northeast, which requires that the entire floor of the Swan Lake scour and discharge channel was lower than 4745 ft (1446 m) when the meander belt in Round Valley was occupied by the Bonneville River.

Large paleodischarges. The average wavelength and width of the meander belt are many times larger than those of the modern Bear River (Fig. 2), but are similar in scale to incised meanders that were cut by the late Pleistocene Bear River and are preserved in strath terraces (Figs. 6, 7, and 9; Figs. S8 and S9 [see footnote 1]). Meander size is a function of discharge (Leopold et al., 1964), and the meanders of Round Valley are so large that the outflow from Lake Bonneville at the Provo shoreline is the only conceivable source of such a large volume of water. Our interpretation expands on Gilbert's concept of the Bonneville River by recognizing this youngest and most upstream part of the Bonneville River ∼20 km south of Red Rock Pass.

The wavelengths of the meanders in Round Valley are ∼1.4–2.8 km, and average 2.3 km (seven measurements from Google Earth, aerial photographs, and topographic maps), with an average radius of ∼0.6 km (Fig. 9). The part of the floodplain with relict lateral accretion structures is up to ∼2.75 km wide, and there are five to eight accretionary point-bar complexes visible in each meander loop despite the drape(?) of mud across the landforms (Fig. 9). The width of the relict channel of the meander belt in Round Valley is ∼43 m (average of 12 measurements from Google Earth across the channel courses). The average meander wavelength and average channel width permit other paleohydraulic parameters to be estimated: (1) channel depth ∼1–2 m; (2) channel slope ∼0.1 m/km; (3) mean annual discharge ∼7 m3/sec; and (4) mean annual flood ∼25 m3/sec (Dury, 1965; Schumm, 1972). These calculations lie within the data sets shown in Leopold et al. (1964), and the ratio of slope to discharge is within the field for meandering streams rather than braided to straight streams (Leopold et al., 1964).

Clifton sill at south end of meander belt. The southern extent of the relict meanders in Round Valley is less clear than the northern end at the Swan Lake scour and discharge channel. Meanders disappear in a 2-km-wide, flat-bottomed saddle or sill at 4745 ft (∼1446 m) in the Twin Lakes horst block near Clifton, Idaho. This complex bedrock horst is well expressed in the topography, and was further characterized by geologic mapping (Link, 1982a, 1982b; Link and LeFebre, 1983; Carney et al., 2003; Long and Link, 2007), and gravity data (Eversaul, 2004) (Figs. 1, 2, 6, 7, and 9). Exposures of Neoproterozoic to Cambrian bedrock and the Salt Lake Formation on either margin of the Clifton sill, and four drillers’ logs of water wells in the sill area show that partly consolidated Tertiary tuffaceous sedimentary rocks are close to the surface beneath the saddle, whereas hard Precambrian bedrock is at least 520 ft deep (Eversaul, 2004; Carney and Janecke, 2005; Long and Link, 2007). Two water wells there penetrated definitive Salt Lake Formation at 129 ft (49 m) and 136 ft (41 m) depth below the surface, respectively, but the contact may be as shallow as 35 ft (11 m).

This “Clifton sill” area has several contrasts across it. South of the sill, modern drainage is southward toward the Great Salt Lake in deeply incised streams, whereas north of the sill, marshy lowlands lack an organized modern stream. The southern ∼2 km of Round Valley has an intermediate morphology, and contains a small south-flowing stream but no incision. Tributaries to the axial stream channels change from opening southward to opening northward across the sill area (Figs. 2, 6, and 9; Fig. S2 [see footnote 1]). Undissected, abandoned fluvial landforms are all north of the sill, whereas deep, steep-sided gullies cut into a smooth valley bottom (topset of the Provo delta) and lie south of the sill (Fig. 6). Shallow Salt Lake Formation in the Clifton sill area may constrain a knickpoint of the modern south-draining Deep Creek in the same area where it once held back Lake Bonneville.

Riverdale Normal Fault (+Landslide?)

The northeast end of the meander belt of Round Valley also coincides grossly with the buried trace of the southwest-dipping Riverdale normal fault zone (+landslide?) (Figs. 1–3; Figs. S5–S7 [see footnote 1]; Oaks et al., 2005). Southeast of Preston, a few kilometers of the southwest-dipping Riverdale normal fault zone offset the Salt Lake Formation on both sides of Cub River (Fig. 6; Oriel and Platt, 1980; Willard, 1972; Scheu, 1985a, 1985b; Winter, 1985, 1989). There, and for ∼30 km along strike to the NNW, the Riverdale fault zone coincides with a prominent WSW-sloping gravity gradient at the ENE edge of the relatively flat, deep floor of northern Cache Valley basin (Figs. 1, 2, 6, 7, and 9; Eversaul, 2004; Oaks et al., 2005). The gravity gradient along the Riverdale fault zone shows that it is a major basin-bounding fault that separates the more open part of Cache Valley—including the relict meanders in Round Valley in its hanging wall—from uplifted Salt Lake Formation with a thinner Quaternary carapace in its footwall.

Analysis of aerial photographs and digital-elevation models of the delta of the Bear River revealed the location and multistranded geometry of the Riverdale fault zone, and provided the first evidence for Quaternary slip across this newly identified fault zone (Figs. 6, 7, and 9). Evidence for the Riverdale fault zone includes: (1) deep linear gullies eroded parallel to its trace north of the Bear River in foresets of the Bonneville delta of the Bear River; (2) misaligned gullies upslope and downslope of the trace that appear deflected near fault traces (Fig. 9); and (3) northwest-trending vegetation lineaments, groundwater barriers, and other subtle geomorphic lineaments in Round Valley farther to the NW. The Riverdale fault may cross the Swan Lake horst, and might persist as an en echelon strand within the Swan Lake terrain as far north as the Zenda sill (Figs. 5 and 6).

The Riverdale fault (+landslide?) zone is strongly expressed in the northern half of the Bonneville delta deposits of the Bear River, north of the Bear River, and is much less obvious in Provo-age sediment of Round Valley (Figs. 9 and 10). The fault zone does not appear to displace any Quaternary deposits south of the Bear River, although there are some slight topographic anomalies present in the Bonneville delta there that are suggestive of a left-stepping en echelon geometry (Fig. 7).

Between Round Valley and the Bear River, there are one to three gently curving gullies parallel to most of the Riverdale fault (+landslide?) zone, along 8 km of its trace. Gullies uphill from the Riverdale zone do not align with gullies downslope of it, and this suggests that gully erosion occurred during or immediately after the Riverdale fault ruptured to the surface in the Bonneville delta.

Where the fault zone continues into mostly Provo-related and post-Bonneville deposits to the NW, there are vegetation lineaments, groundwater barriers, and possible low fault scarps on the floor of Round Valley, with many possible traces parallel to one another (Fig. 9). The longest trace is 2.5 km long. North of Round Valley, the flood-scoured landforms contain west-facing escarpments that might be flood-modified scarps of the Riverdale fault zone (Fig. 10). Our preliminary reconnaissance map of the northernmost extent of the Riverdale fault zone shows its possible continuation into the southern half of the Zenda sill. Trenching and much more detailed mapping are needed to fully assess the Riverdale fault zone.

DISCUSSION, ANALYSIS, AND NEW FINDINGS

Could the Bonneville River Have Flowed North from Round Valley?

Round Valley is hemmed in on its northern and northwest edge by the Swan Lake horst, so that the Swan Lake scour and discharge channel is the only possible outlet for the meander belt (Figs. 1, 2, and 6; Figs. S1, S2, and S6 [see footnote 1]). The post-flood floor of the Swan Lake scour and discharge channel (∼4730 ft [1442 m]) was at least 6 m lower in altitude at Swan Lake (Bright, 1966) than the scroll bars of the meander belt. The channel's floor dropped an additional >50 ft (>15 m) in the next 3 km northward (cf. Fig. 11). This depth and slope are also consistent with a north-flowing river system, and together they rule out the conventional interpretation of the Provo threshold being beneath Red Rock Pass, at least during the stillstand at the later lower Provo shoreline.

The altitude of a buried Paleozoic bedrock sill (4755 ft [1449 m]) interpreted at Red Rock Pass by Williams and Milligan (1968) is too high to accommodate northward flow of a large river from Round Valley (starting at ∼4744 ft [∼1446 m]). However, the altitude of the buried bedrock there is poorly constrained, and we infer that it must be lower than 4755 ft (1449 m). Perhaps the increased velocity observed in the seismic-refraction data near 20 feet depth is the water table, which roughly doubles seismic velocities, or landslide materials composed of large intact blocks of pre-Tertiary rocks.

Was There Overland Flow Near Zenda Prior to the Bonneville Flood?

The reconstruction of the Zenda area prior to the Bonneville flood reveals a low-relief saddle between Marsh and Cache Valleys that dammed the lake near Zenda, Idaho. Once overtopped, this saddle probably allowed overland flow of excess water from the Bonneville shoreline during high-water years. The altitude of this saddle is well constrained, and coincides with the highest shoreline of Lake Bonneville, within error. There is little space within the narrow eroded landscape for a significantly higher barrier to have existed (Fig. 10). The coincidence of the altitude of the reconstructed saddle with the highest Bonneville shoreline strongly argues for at least episodic overland flow from the Bonneville level of the lake. To argue otherwise would require the unlikely coincidence of Lake Bonneville rising and falling repeatedly to within a few meters of an overflow point and stabilizing there as a closed basin for a protracted period of time.

The reconstructed surfaces in the greater Red Rock Pass–Zenda area reveal an outflow channel several hundred meters wide that had a point of origin ∼1 km west-northwest of Zenda, Idaho (Figs. 5 and 10), as inferred previously with less precision by Gilbert (1880, 1890), Currey (1982), and others. Most prior workers inferred overland flow during the Bonneville highstand, and our reconstruction of the landscape supports their interpretation (Currey, 1982, and others). The landscape in Marsh Valley may also indicate northward outflow before the Bonneville flood (Thackray et al., 2011).

Was There Subsurface Flow Near Zenda Prior to the Bonneville Flood?

In addition to the strong evidence for overland flow across the Zenda threshold area prior to the Bonneville flood, there is compelling evidence for sapping through the Zenda dam and/or sill (O'Connor, 1993). The widespread distribution of the sapping-related landforms in the Marsh Creek pediment cap and alluvium-loess cover requires subsurface flow across a wide area—from the altitude of the high Bonneville shoreline to the bottom of Marsh Valley. However, the duration of the sapping is not known, and it might have been short lived. Perhaps sapping was dominant during the transgressive phase of the lake, but was replaced as the primary outflow mechanism when overflow began across the Zenda sill.

Subsurface flow might have weakened the Zenda dam, and could have contributed to its failure during the Bonneville flood (O'Connor, 1993). However, field relationships suggest that groundwater precipitated considerable carbonate in the pore spaces of conglomerate beds (Fig. S14 [see footnote 1]). This raises the possibility that subsurface flow strengthened the Zenda sill instead of weakening it.

How Did Weak Materials Constrain Lake Bonneville?

The pediment-capping deposits and the underlying Salt Lake Formation are usually weak and easily eroded. However, the broad base and gentle curvature of the surface of the pedimented threshold (and some Quaternary alluvial and eolian sediment) at Zenda probably contributed to its being a fairly long-lived (∼1.2 k.y., Godsey et al., 2005) earthen dam for Lake Bonneville. Engineering principles show that earthen dams with broad bases and gentle slopes are very effective, despite being composed of weak materials, because the lateral pressure of the water is distributed across a wide area (Nelson, 1985; Stephens, 1991). The gentle northwestward slope of the seam between the Marsh Creek and Aspen Creek pediments may have been a slope approximately “graded” for the outflow volume and velocity of the Bonneville River, and thus may have resisted rapid erosion until some event triggered collapse or extremely rapid downcutting of the Zenda outlet.

Assisting in maintaining the Zenda outlet were fairly well-cemented conglomerate beds within the Salt Lake Formation (Fig. S14 [see footnote 1]). The conglomerate beds are interbedded with siltite and minor sand, and form ledges and overhangs above weaker sedimentary beds below. Individual clasts or boulders of conglomerate eroded from well-cemented pebble conglomerate beds near Zenda may have armored the broad crest of the Zenda sill and prevented downcutting by outflow during the occupation of the highest Bonneville shoreline.

The Role of Landslides in the Bonneville Flood

Enormous landslides, aggregating at least 17 km2 in area (Sewell, 1980; Sewell and Shroder, 1981; Shroder and Lowndes, 1989), did not collapse the north end of the Bannock Range during the flood. Geologic mapping by several workers, some of which dates back to the 1980s, refutes this interpretation, and documents fault blocks of well bedded, tilted Salt Lake Formation and lower Paleozoic units truncated by Pleistocene pediments, overlain by pediment caps, and younger alluvial deposits in the location of the hypothesized megalandslide (Mayer, 1979; Link, 1982a, 1982b; Hennings, 2002; DeVecchio et al., 2003; Kruger et al., 2003; Long and Link, 2007; this study). Our mapping delineates several smaller, more localized remnants of late Pleistocene landslides that were shed eastward into the Swan Lake scour and discharge channel (Fig. 5).

This modest amount of landsliding was initiated during downcutting of the Zenda threshold, and so, in combination with deepening of the scour and discharge channel, may have partly dammed and diverted Bonneville floodwaters at times. One larger landslide, preserved only in the east, seems to have filled and deactivated part of the eastern scour and discharge channel just north of Red Rock Pass, and may have diverted the floodwaters westward into the now-deeper western channel (Fig. 5). After the flood cut deeper in the pass, the base of another landslide in the same location is at the new lower, local, base level.

As the Zenda threshold continued to incise, the increasing flow rapidly removed most of the landslide material in the western scour and discharge channel near Red Rock Pass (Gilbert, 1880, 1890; Burr and Currey, 1988; Currey and Burr, 1988). The straighter and now lower western scour and discharge channel was more efficient at emptying the enormous Bonneville basin of half of its water than the higher, earlier, and more circuitous eastern scour and discharge channel.

The landslides adjacent to the scour and discharge channel in the Red Rock Pass area formed in response to a fairly modest (∼100 m) drop in the local base level. The parent material in the landslides includes faulted Paleozoic carbonate and sandstone, and Tertiary sedimentary rocks with Quaternary colluvial and alluvial cover (Mayer, 1979; DeVecchio et al., 2003; Long and Link, 2007).

Hillslopes of northern Utah and southern Idaho that are composed of similar materials with similar slopes rarely produce this much landslide material. We speculate that reservoir-induced seismicity during and slightly after the Bonneville flood may have been an important additional trigger for the failure of hillslopes adjacent to the scour and discharge channel. Reservoir-induced seismicity is documented to be particularly intense when water levels drop (Gupta, 2002; Telesca, 2010), but may occur whenever loads and pore pressures fluctuate. The great depth, rapid drop, many shoreline fluctuations, and large size of Lake Bonneville were ideal conditions to promote enhanced earthquake activity in this region. Hillslopes immediately west of Cache Valley released a large landslide after the Bonneville highstand and before the end of the Provo stillstand (Maw, 1968; Oviatt et al., 1986; Biek et al., 2003), possibly in response to this process.

Is the Riverdale Structure a Normal Fault and/or the Head Scarp of a Lateral Spread or Large Slump?

The Riverdale fault cuts obliquely across the foreset of the Bonneville delta of the Bear River, and has a position resembling delta-front slides that are induced by rapid drops in water level (Paola, 2000; Heller et al., 2001; Paola et al., 2001). It is therefore possible that the Riverdale fault is the headscarp of a large lateral spread, because it has a position midway up a hillslope rather than at a break in slope, it has some concave-southwest curvature where it is best expressed in the Bonneville delta, and its geomorphic features are unusual for a normal fault (Fig. 7). However, there are no identifiable lobate zones or compression ridges along the presumed toe of the landslide, where they should be expressed (Figs. 2, 6, and 7; Figs. S6 and S7 [see footnote 1]). Wave erosion might have redistributed the sediment from such lobes along the toe, but removal of a toe tends to trigger additional failures in landslides and multiple headscarps, and these are not apparent. There are no scours to indicate removal of the toe region by currents (Fig. 7).

Factors that are most consistent with a fault interpretation of the Riverdale structure include its position along strike of a mapped bedrock fault, its similar sense of displacement (SW down) as the mapped fault, its location and strike roughly coincident with a southwest-down gravity gradient (Fig. 2), its significant lateral persistence, and the presence of vegetation lineaments and low scarps along strike in the low-lying areas of Round Valley where a headscarp could not possibly have formed. The contrast between the fault's strong topographic expression in Bonneville deposits and its subtle traces in Provo deposits is probably the result of multiple tectonic slip events, likely with much less throw across the post-Provo surface ruptures.

Gully erosion destroyed much of the original displaced landscape in the Bonneville delta of the Bear River and makes it difficult to tell whether faulting alone produced the irregular topography in the foreset of the Bonneville delta along the Riverdale fault zone or if mass-wasting processes also contributed to it. We favor a fault model because it explains more relationships in northeast Cache Valley than a landslide model, but we cannot rule out enhanced escarpments from landsliding processes in the Bonneville delta. Because of this ambiguity, we call this structure the Riverdale fault (+landslide?) zone where it crosses the Bonneville delta.

Did an Earthquake, Overland Flow, or Sapping Trigger the Bonneville Flood?

It is challenging to pin down the main trigger for the Bonneville flood, and to develop a definitive test of competing models, because many processes could have triggered it. Sapping-related dam collapse provides one plausible trigger of the Bonneville flood (O'Connor, 1993). However, incision by overland flow could just as easily have triggered the Bonneville flood. The crosscutting relationships between the Riverdale normal fault (+landslide?) zone and Bonneville and Provo deposits of the Bear River raise another possibility, that a moderate to major earthquake on that fault (or emplacement of a lateral spread with a headscarp at the trace of the Riverdale fault) triggered the flood. After all, the Riverdale fault (+landslide?) cuts and deforms deltaic deposits and landforms that predate the flood (Bonneville deposits), yet is only weakly expressed in deposits and landforms that postdate the flood (Provo to Holocene deposits). An earthquake on the Riverdale fault (or on some other Cache Valley fault) could have produced seiche waves that overtopped the Zenda sill with high-velocity waters of sufficient energy to destabilize that dam, to breach part of the Zenda sill, to rapidly incise the length of the dam, or to cause other damage.

Theoretical considerations suggest that the rapid fluctuations of the level of Lake Bonneville could have been ideal conditions for triggering seismicity on tectonic faults beneath the lake. Numerous small and some large earthquakes develop beyond the background level as bodies of water fill and empty (Gupta, 2002; Telesca, 2010). It is possible that the rising lake, or climatically induced rapid fluctuations in lake levels, triggered an earthquake that then triggered the Bonneville flood. The crosscutting relationships are also consistent with the opposite sequence: that the abrupt drawdown of Lake Bonneville by the flood changed vertical stresses and pore pressures, and these in turn triggered an earthquake on the Riverdale fault. Trenching and dating of the Riverdale structure are needed to better assess the role of the Riverdale fault (+landslide?) zone in the Bonneville flood. Other possible triggers for failure of the Zenda threshold include a rapid release of an unusually large volume of water into the Bear River, large earthquakes on nearby faults, or some other mechanism.

On a regional scale, differential loading and unloading of Lake Bonneville and of Lake Lahonton in western Nevada are likely to have induced seismicity throughout the Great Basin. The data are incomplete, but the only two central segments of the Wasatch fault with paleoseismic records dating back to the Bonneville highstand both preserve evidence for major earthquakes around the time of the Bonneville flood (McCalpin, 2002; McCalpin and Forman, 2002). Liquefaction and faults within the highstand Bonneville delta of Green Canyon, in North Logan, Utah, within tens of meters of the East Cache fault, also reflect subaqueous earthquake-related shaking in Cache Valley shortly before or during the Bonneville flood. Sand volcanoes erupted along the East Cache fault near Logan Canyon and produced 17.4 ± 3.0 ka thermoluminescence dates. These relationships may record an earthquake coincident with the Bonneville flood, but the interpretation is complex (McCalpin, 1994).

REVISED HISTORY OF LAKE BONNEVILLE DURING AND AFTER THE BONNEVILLE FLOOD: LAKE POSITIONS AND OVERFLOW SCENARIOS

The landforms in northern Cache Valley were produced by three widely separated sills that successively controlled the shorelines of Lake Bonneville, the Bonneville flood, and the two younger Provo shorelines. First, a sill near the hamlet of Zenda, Idaho, restrained Lake Bonneville south of it (Gilbert, 1880, 1890, and many others). After the Bonneville flood, a new threshold was established as scouring modified the irregular bedrock surface of northern Cache Valley. This new outlet for Lake Bonneville formed along the Swan Lake scour and discharge channel at its southeast end where the flood incised a channel across the east-northeast–trending Swan Lake horst block. We infer that a now-eroded ∼4775 ft (1455 m) bedrock sill near Swan Lake constrained the 4775 ft (1455 m) higher Provo shoreline of Lake Bonneville for a fairly long period of time, because the shoreline at this altitude is very well developed and prominent in the landscape. Geochronology of Provo deposits indicates a stillstand at the Provo shorelines that lasted ∼1.5–2.5 ka, in agreement with our inference (Godsey et al., 2005; Benson et al., 2011). The altitude of the main 4775 ft (1455 m) Provo shoreline and the Swan Lake scour and discharge channel, the megaflood-style morphology (Baker, 2009) of the Swan Lake terrain, the northward slope of the floor of the Swan Lake scour and discharge channel (between Swan Lake and the wells in Fig. 11), and its position in the landscape are all consistent with this interpretation (Figs. S1–S11 [see footnote 1]). Gilbert (1880, 1890) came close to locating this bedrock sill when he observed that “…the outflowing (Bonneville) river headed … farther south, between Swan Lake and Round Valley Marsh.”

However, this cannot be the whole story, because Lake Bonneville must have been confined by a sill south of Round Valley when the Bonneville River meandered north through it. Later incision and removal of the Swan Lake sill briefly activated the Clifton sill as the new outlet of Lake Bonneville ∼10 km farther southwest, and established the 4745 ft (1446 m) lower Provo shoreline of Lake Bonneville. This younger sill was probably ephemeral based on the weak expression of the ∼4745 ft (1446 m) Provo shoreline. The landforms in northern Cache Valley suggest that the relict meander belt in Round Valley was active during a stillstand of Lake Bonneville at the subtle ∼4745 ft (∼1446 m) lower Provo shoreline south of the Clifton sill. Shallow Salt Lake Formation beneath the Clifton sill probably constrained this fleeting (?) stillstand of Lake Bonneville at the lower Provo shoreline. The Clifton sill is ∼23 km south of Zenda, Idaho, and coincides with the southernmost of the three prominent bedrock ridges that cross the floor of northern Cache Valley (Fig. 2) (Carney et al., 2003; Eversaul, 2004; Oaks et al., 2005).

The excellent preservation of the meander belt in Round Valley indicates that the lower 4745 ft (1446 m) Provo shoreline postdates the main upper Provo shoreline at ∼4775 ft (∼1455 m). The opposite sequence would have submerged the meander belt beneath Lake Bonneville for a considerable period of time, and would have covered the relict landforms of the Bonneville River in Round Valley with younger sediment.

Recent paleoclimate work in the Basin and Range indicate that there was either a persistent or brief interval of warmer and drier climate during the second half of the Provo occupation of Lake Bonneville, starting between 15.9 and 15 ka (Fig. 3; Wagner et al., 2010; Benson et al., 2011). This change may correspond with the Heinrich 1 event of the North Atlantic (Benson et al., 2011). During the transition to warmer conditions, melting of glaciers throughout the Bonneville basin might have increased runoff, which in turn probably increased the discharge across the outlet at Swan Lake for a time. A brief oscillation beneath the level of one or both Provo shorelines, documented by two independent studies, probably resulted from continuing warmer and drier climate. Godsey et al. (2005) suggested that the drop in water level occurred between roughly 15 and 16 ka (radiocarbon years corrected to calendar years), whereas Benson et al. (2011) dated the oscillation at 15.9 ka and described evidence for “…a brief oscillation of the lake beneath the Provo level, followed by a decreased rate of spill across the Provo threshold.” The latter agrees with our interpretation of two Provo sills in northern Cache Valley, and allows a drop in base level during the oscillation to produce some or all of the incision and fluvial erosion that eventually eroded the Swan Lake sill.

These temporal relationships between the sub-Provo oscillation and the change of sills suggest that a warmer and drier climate (during the Heinrich 1 event?) was the main cause of the change from the Swan Lake sill to the Clifton sill (Wagner et al., 2010; Benson et al., 2011). The data also suggest that the upper Provo shoreline started to form between 17.4 (calibrated age of Godsey et al., 2005) to 17.0 ka (age of Benson et al., 2011), and was abandoned around 15.9 ka during the Heinrich 1 event (Fig. 6). The lower Provo shoreline probably postdates 15.9 ka, and may have been completed by 15.2 ka, if the chronology of Benson et al. (2011) is correct. This correlation indicates a longer occupation of the higher Provo shoreline (1500–1100 yr) than the lower Provo shoreline (700 yr), as predicted by the relative intensity of shoreline development.

Shortly after the Clifton sill was the outlet of Lake Bonneville, the most recent glacial epoch drew to a close (Fig. 3; Wagner et al., 2010), and a forced regression of Lake Bonneville from the Clifton sill returned the basin to a closed condition. The regression deprived the meander belt of its main discharge, and changed Round Valley into the site of a sluggish marsh. Stream capture at its southern margin, by a tributary of the modern Bear River, reversed the flow.

OUTSTANDING QUESTIONS: TOPOGRAPHIC ANOMALIES AND PROBLEMATIC RELATIONSHIPS

There are relationships in northern Cache Valley that we cannot fully explain. We enumerate them here to encourage future study, and we describe some in more detail in the Supplemental File (see footnote 1).

(1) The location, slip history, and importance of the newly defined Riverdale fault are incompletely characterized. It is vital to determine whether the Riverdale structure ruptured entirely as a normal fault or also failed as a head scarp of a lateral spread or slump in the Bonneville delta of the Bear River. Perhaps tectonic faulting triggered a lateral spread with a headscarp at the fault scarp within the foresets of the Bonneville delta. Another central question is whether slip across the Riverdale fault triggered the Bonneville flood, either directly by rupturing part of the Zenda sill or indirectly by producing seiche waves, liquefaction, incision, or other secondary processes. Of less importance to Lake Bonneville, but critical to residents of Cache Valley, is understanding the fault's strong geomorphic expression in the Bonneville deposits and its weak expression in Provo deposits in Round Valley and south of the Bear River. It is possible that multiple, possibly Holocene, ruptures of the Riverdale fault zone produced these relationships. Our reconnaissance mapping shows that the Riverdale fault zone connects the East Cache fault to the Red Rock Pass area along a trace that is 30–45 km long. Such a long fault with a history of Quaternary slip poses an unquantified hazard in northern Cache Valley.

(2) The interpreted depth of bedrock in the Swan Lake scour and discharge channel at Red Rock Pass (Williams and Milligan, 1968) is too shallow (high) to allow northward flow from the lower 4745 ft (1446 m) Provo shoreline, and is inconsistent with drillers’ logs 4 km to the south (Fig. 11; Fig. S14 [see footnote 1]). According to Williams and Milligan (1968), the original base of the outflow channel at the pass may be near 4755 ft (∼1450 m), beneath ∼20 ft (6 m) of younger fill (Williams and Milligan, 1968). This altitude is also too low to have been the sill for the 4775 ft (1455 m) Provo shoreline.

Landslide materials composed of Paleozoic carbonate and Proterozoic quartzose rocks lie directly east and west of the seismic-refraction site at Red Rock Pass itself, and form knobs within the outflow channel (Fig. 5; Williams and Milligan, 1968). These materials may have seismic velocities that are similar to the bedrock beneath the pass, particularly because several landslide masses contain large intact blocks tens of meters in dimension. Remnants of the landslides are likely to be present in the subsurface beneath the refraction line because they surround the refraction site, and knobs of flood-sculpted landslide materials dot the landscape nearby (Fig. 5; Figs. S3–S5 [see footnote 1]). High-velocity pre-Tertiary bedrock blocks in landslides in the subsurface could have been mistaken for bedrock. In addition, the line spread was insufficient to determine the seismic velocities. Therefore we conclude that the depth to bedrock is poorly known at Red Rock Pass. Intact bedrock is likely to be lower in altitude than 4680 ft (1427 m) at Red Rock Pass, based on the altitude of channel-filling gravel in Figure 11, ∼4 km farther south. Drilling is recommended to resolve this question.

(3) The post-Provo reversal of drainage between Clifton and Red Rock Pass predicts temporary ponding along the path of the Bonneville River south of Red Rock Pass. The geologic relationships might also require some differential uplift along the path of the Bonneville River in Cache Valley, if the depth to bedrock at Red Rock Pass is above the altitude of the meanders in Round Valley, as Williams and Milligan (1968) interpreted.

All in all, we posit that the evolution of Lake Bonneville was very complex because both tectonic and climate-forcing mechanisms played a role in the levels and sills of the lake. Field studies and new geochronology can resolve these questions with future work.

CONCLUSIONS

Analysis of landforms on aerial photographs, digital-elevation models, and satellite imagery, consideration of gravity data, analysis of drillers’ logs of water wells, and geologic mapping show dynamic, rapidly changing relationships in northern Cache Valley and southern Marsh Valley and clarify the likely sequence of events at Lake Bonneville's outlet. The landscape there preserves a clear record of the Bonneville flood and the transition into the latest Pleistocene.

The Bonneville basin was hydrologically closed until occupation of the highest Bonneville shoreline. Then the Bonneville shoreline stabilized long enough to produce a strong wave-cut and wave-built shoreline throughout the basin. At that time lake waters started to exit the basin (intermittently?) northward, across a broad, gently sloping pediment cut across conglomeratic late Cenozoic Salt Lake Formation at the Zenda outlet.

The SW-dipping Riverdale normal fault (+landslide?) produced clear and anomalous landforms in the north half of the Bonneville delta of the Bear River, and it failed around the time of the Bonneville flood. Collapse of the Zenda sill during a surface-rupturing earthquake on the Riverdale fault (+landslide?) zone (or other nearby fault zone) is a possible trigger for the Bonneville flood that could have induced the catastrophic failure of the Zenda sill after a fairly long period of stability. Emplacement of a very large landslide with a headscarp at the Riverdale structure could have had similar effects on the outlet of Lake Bonneville, if it produced a major lacustrine tsunami (seiche wave). Alternatively, an earthquake or landslide on the Riverdale structure might have been the result of “lake-induced seismicity” during filling or emptying of the lake basin, after some other process triggered the Bonneville flood. Detailed geochronology is needed to test these competing hypotheses. There is no evidence for wholesale failure of the northern Bannock Range in a megalandslide, as had been inferred by others.

Flood water re-excavated channels filled by landslides, and created a fluted and scoured landscape across the northern 10 km of Cache Valley, north of the eventual sill that constrained the Provo shoreline near Swan Lake. Sediments of the Salt Lake Formation in the Swan Lake bedrock ridge prevented further rapid downcutting, and brought the Bonneville flood to a close.

Two significant shorelines of Lake Bonneville formed during Provo time in northern Cache Valley. Two separate bedrock sills controlled these shorelines, and sequentially marked the upstream ends of the Bonneville River and the outlets of Lake Bonneville. After the Bonneville flood, the first outlet formed along the Swan Lake scour and discharge channel near Swan Lake, where a bedrock sill controlled the level of Lake Bonneville when the more prominent older and higher 4775 ± 10 ft (1455 m) Provo shoreline stabilized after ∼17.4 ka. This shoreline is the main Provo shoreline, and is ∼30 ft (∼9 m) higher than has been reported in the literature since ∼1980.

A second important Provo shoreline at 4745 ± 10 ft (1446 m) is more subtle than the main Provo shoreline at 4775 ± 10 ft (1455 m). The lake that produced this lower Provo shoreline had a bedrock sill near Clifton, Idaho, ∼22 km south of the original Zenda threshold. We infer a brief occupation because a longer occupation would have produced a more definitive shoreline. The repositioning of the outlet of Lake Bonneville may have been a response to increased aridity during the Heinrich 1 event. The Bonneville River flowed north from this second, younger, and lower outlet near Clifton, Idaho, into Round Valley. There it meandered laterally and deposited point-bar scrolls within a large meander belt, and floodplain while flowing toward the Snake River Plain.

In the Holocene, a final drainage reversal affected northern Cache Valley. Twenty kilometers north of the Clifton sill, the incising Marsh Creek built an alluvial dam where it flowed onto the open floor of the Swan Lake scour and discharge channel at Red Rock Pass (Gilbert, 1880, 1890; Williams and Milligan, 1968; Bright and Ore, 1987). The slight high in the landscape at the crest of the alluvial fan established the modern drainage divide at Red Rock Pass, 2 km south of its original pre-flood position near Zenda, Idaho. Differential infilling of the once north-sloping dry bed of the Bonneville River by tributary streams explains the paradox of the former Swan Lake sill being the current site of a shallow lake, and the midpoint of the dry river bed being the current divide. Round Valley may have been occupied by a short-lived shallow lake before the very recent, incipient drainage capture by a tributary of the Bear River.

These events coincidentally repositioned the modern drainage divide close to its original one near Zenda, Idaho. This coincidence may explain why Gilbert's (1880, 1890) conclusion that the outlet of Lake Bonneville was just south of Swan Lake during occupation of the Provo shoreline has been overlooked for ∼120 yr.

We thank Victoria Bankey, Victoria E. Langenheim, Joseph M. Kruger, and Martin Eversaul for working on the gravity data with us and sharing gravity data. Images from Google Earth, U.S. Geological Survey Seamless server, FlashEarth.com, Microsoft Terraserver, GeoMapapp, and GoogleMaps permitted us to recognize and analyze the landscape. We thank these organizations for use of selected images. Conversations with Jim Evans, Paul Link, Robert Bundy, Jack Schmitt, Tammy Rittenour, Dave Mickelson, Bob Dott, and Glenn Thackray were helpful. Thad Erickson assisted in digging a pit in Round Valley, and various local people assisted us in locating water wells. The “Can humans cause earthquakes?” podcast of “Stuff you should know” led to our speculative correlation of the Bonneville flood and induced seismicity. We thank Jack Oviatt for a detailed review of an earlier manuscript that helped us to clarify, confirm, and strengthen our observations and interpretations. Reviewers Jim O'Connor and David Miller provided many helpful suggestions and urged brevity. Numerous others have assisted us through their observations and constructive comments after oral presentations and a field trip. We accept all responsibility for the correctness of the data and for our interpretations.

1Supplemental File. PDF file containing additional methodological information about the mapping criteria, landscape reconstruction, and analysis used in this paper, along with Supplemental Figures S1–S15. If you are viewing the PDF of this paper or reading it offline, please visit http://dx.doi.org/10.1130/GES00587.S1 or the full-text article on www.gsapubs.org to view the Supplemental File.
2Supplemental Table. PDF file of Lake Bonneville shoreline altitudes in feet (m) at various locations. If you are viewing the PDF of this paper or reading it offline, please visit http://dx.doi.org/10.1130/GES00587.S2 or the full-text article on www.gsapubs.org to view the Supplemental Table.