USGS Logo Geological Survey Professional Paper 644—F
History of Snake River Canyon Indicated by Revised Stratigraphy of Snake River Group Near Hagerman and King Hill, Idaho


The latest episodes of geologic history along the Snake River between Hagerman and King Hill, which are recorded by the various lava flows and river deposits of the Snake River Group, are portrayed on a series of paleogeologic sketch maps (fig. 11). The existing and inferred geologic features by which each of these maps has been constructed will be discussed in turn. The maps should be compared with the present geology shown in figure 3.

FIGURE 11.—Map of area between Hagerman and King Hill showing late Pleistocene drainage changed caused by local lava flows. (click on image for an enlargement in a new window)


The eruption of Sand Springs Basalt occurred after the Snake River had become entrenched to its present depth (fig. 11A). Before then, the entrenchment had been interrupted at various times by lava flows that had caused minor diversions of the river. The first of these older lavas, Madson Basalt, came from an eastern source and reached nearly as far west as King Hill. It occupied a shallow ancestral valley of the Snake River at about 3,000 feet in altitude at the approximate latitude of Gooding Butte. This valley, which is now mainly concealed north of the present mouth of the Big Wood River (Malad Canyon), passed just south of Bliss (where it is intercepted by the present canyon) and then turned west. Gravel under the Madson Basalt is exposed in Malad Canyon (fig. 3) and is a type that differs from the porphyry-rich gravel that is now carried to the Snake River Plain by the Big Wood River from central Idaho (Howard Powers, written commun., May 19, 1967). The lack of porphyry in this gravel suggests that the Wood River at that time was probably along the north edge of the Madson Basalt. The Sugar Bowl Gravel that overlaps the terminus of the Madson Basalt (locality 510 in fig. 3) resembles the modern gravel of the Big Wood River (locality 538) by being rich in porphyry and other rocks from central Idaho (table 2).

TABLE 2.—Selected pebble counts, in percent, of gravel pertaining to ancestral river history between Hagerman and King Hill

Locality No Sugar Bowl GravelGravel on Thousand Springs Basalt Crowsnet GravelCrowsnest Gravel Melon GravelModern Gravel of Big Wood River

Source chiefly in central Idaho:

   Vein quartz1

Source nearby:
Source indeterminate:
Number of pebbles counted279311313281272372

1This locality is 7 miles southwest of King Hill and 1 mile south of Glenns Ferry.
2Includes as much as 3 percent fine-grained dark rock, counted as argililte.

The Madson Basalt pushed the ancestral Snake River south to the latitude of Hagerman. Lava from Gooding Butte, one of the sources of the Thousand Springs Basalt, then covered the eastern part of the Madson, restricted the Wood River to a course north of Gooding Butte, occupied the site of the Snake River that was then east of Hagerman, and diverted the Snake into a course still farther south at the approximate latitude of 42°45'. Gravel southeast of Hagerman on this Thousand Springs Basalt (locally 531) reflects material that was transported by the ancestral Snake River and is more than 400 feet above the present Snake River. The gravel evidently closely marks the depth of canyon cutting at that time.

As entrenchment by the Snake River progressed, the base of the Thousand Springs Basalt from Gooding Butte in the former canyon east of Hagerman was exposed and began to discharge spring water. This discharge initiated the small spring-fed tributary (Billingsley Creek) that parallels the east wall of the canyon. The springs at the base of the basalt range in altitude from 3,100 to 3,150 feet.

A younger part of the Thousand Springs Basalt from vents about 20 miles east of Hagerman was then deposited on the floor of a shallow canyon 100 feet below the south edge of the basalt from Gooding Butte. This canyon was at the latitude of 42°45', as previously mentioned. Ground water carried by basalt in this former canyon is thought to account for the large discharge at Thousand Springs, 5 miles southeast of Hagerman (Nace and others, 1958, p. 45-48). Because of this younger Thousand Springs Basalt, the ancestral Snake River was pushed a little farther south to the position of the concealed canyon sketched as a dotted line in figure 11A.

The canyon was next partly filled with about 50 feet of Crowsnest Gravel, which overlaps the younger part of the Thousand Springs Basalt near Hagerman. Along Billingsley Creek, the base of this gravel is from 3,050 to 3,080 feet in altitude, which is 30-80 feet below the springs at the base of the basalt from Gooding Butte. If the Crowsnest Gravel extended eastward under the basalt, the springs would issue from a level equal to the base of the gravel. Their absence at this level confirms that the basalt from Gooding Butte is older than the Crowsnest Gravel, as the physiographic relations already described indicate.

The Crowsnest Gravel southeast of Hagerman (locality 532) closely resembles in composition the gravel on the basalt from Gooding Butte (locality 531), and this composition reflects nearby sources. The outcrops of Crowsnest Gravel downstream near King Hill, however, include a large component transported by the ancestral Wood River from central Idaho (locality 513). The Crowsnest Gravel near King Hill lies about 200 feet below the Sugar Bowl Gravel and thus indicates the depth of canyon entrenchment by this time.

Canyon cutting then resumed, eventually reaching a depth of more than 400 feet at Hagerman (250 ft. below Crowsnest Gravel) and probably a comparable depth near King Hill. In the lava plain southeast of Hagerman, the Snake River canyon was approximately in the position it had occupied since eruption of the last Thousand Springs Basalt (dotted line in fig. 11A). From Hagerman to Bliss, as indicated by confining outcrops of Madson Basalt and older rocks still preserved along the canyon rim, the ancestral canyon was approximately in its present position. At Bliss it was separated from the Wood River canyon on the north by a divide of old Quaternary and Tertiary deposits, as recorded in the log of well 6S-13E-6dc1 (table 3). West of Bliss the canyon must have been approximately in the position it occupied during the subsequent eruption of McKinney Basalt, as will be explained later.

TABLE 3.—Log of water well 6S-13E-6dc1 at Bliss

[From files of U.S. Geological Survey, Water Resources Division, Boise, Idaho. Alt. 3,620 ft. Assignment of lithologic units based on map by Malde, Powers, and Marshall (1963)]

Material Thickness
Glenns Ferry Formation (Pleistocene and Pliocene):
Clay, yellow8484
Lava, gray, hard; water at 188 feet34118
Clay, yellow92210
Clay, blue52262
Clay, yellow147409
Clay, blue21430
Clay and black sand22452
Banbury Basalt (Pliocene):
Lava, black40492
Sand, black3495
Clay, blue17512

When the Snake River canyon reached its present depth at Hagerman (2,800-ft alt), it was filled as far downstream as Hagerman by the Sand Springs Basalt, which erupted from a vent 50 miles east. The narrow ancestral canyon is now exposed in profile along the present canyon 7 miles southeast of Hagerman (fig. 12). In the wide part of the canyon at Hagerman—an area bounded on the west by weakly consolidated basin deposits (older Quaternary and Tertiary)—the Sand Springs Basalt spread as a thin terminal tongue. (For a map of outcrops of Sand Springs Basalt farther east, see Malde and others, 1963.)

FIGURE 12.—Canyon fill of Sand Springs Basalt about 7 miles southeast of Hagerman. The lava cliff rises 210 feet above the Snake River.

With emplacement of the Sand Springs Basalt, the geology resembled that shown in figure 11A.


The Wendell Grade Basalt (fig. 11B) was the immediate predecessor of the McKinney Basalt. Both display a rough surface of pressure ridges that project above a thin cover of surficial material. Glassy skins on ropy surfaces of these lavas are locally preserved, and the lavas are fresh. (The McKinney Basalt, however, at a single locality 7 miles east of King Hill, displays patterned ground, a curious landform characteristic of older weathered basalt and gravel in this region; see Malde, 1964, p. 204.)

Before the eruption of Wendell Grade Basalt, a thin mantle of windblown sand and silt had accumulated on the Sand Springs Basalt to form a smoothly rolling lava plain. The Snake River had been diverted at a place 30 miles upstream from Hagerman by the Sand Springs Basalt and had carved a new canyon southeast of Hagerman where the Thousand Springs Basalt had previously abutted an upland of basin deposits. This part of the Snake River canyon of that time was virtually in the position of the present canyon. (Small outcrops of Thousand Springs Basalt at the south rim of the present canyon are mapped 1-3 miles east of the area of fig. 11; see Malde and others, 1963.) Where the new canyon intercepted the former canyon 5-7 miles southeast of Hagerman, the Sand Springs Basalt had been removed. Near Hagerman the Snake River probably followed the west edge of the terminal tongue of Sand Springs Basalt, as it does now. Billingsley Creek probably was a little deeper than it had been during the eruption of Sand Springs Basalt and was flanked on the east by landslide and talus debris derived from the basalt rimrock. Downstream from Hagerman, the canyon probably had become deeper, as indicated by outcrops of pillow lava assigned to the McKinney Basalt that occur 100 feet lower than the terminus of the Sand Springs Basalt.

The Wendell Grade Basalt flowed west as several thin tongues on the old lava plain. One tongue crossed the future path of the Wood River southwest of Gooding Butte, and others opposite Hagerman reached the canyon edge. At least one of these ended as a lava cascade on landslide debris along the east wall of Billingsley Creek. The preservation of this cascade indicates that subsequent erosion from spring discharge along Billingsley Creek has been minor.


When the McKinney Basalt erupted from McKinney Butte (fig. 11C), the Snake River canyon from Hagerman to Bliss was as deep as it is now (about 500 ft) and occupied its approximate present position. At a place about 2 miles west of Bliss, where the only important gap in the continuity of the Madson Basalt occurs, I infer that the former canyon continued northwest a short distance to its confluence with the ancestral Wood River and then turned west. The canyon route to the west can be approximately located by the presence of confining upland areas of older rocks and by the lava-filled canyon intersected at Bancroft Springs (fig. 8). Similarly, the route of the ancestral canyon of the Wood River north of Bliss can be determined by its adjoining uplands (including the divide of older rocks under the site of Bliss); its path from the east must have been along the north flank of Gooding Butte.

The buried Wood River canyon is evidently encountered in a water well three-quarters of a mile east of Bliss, which penetrates a lava fill described in log 6S-13E-5dd1 (table 4). The bottom of this well is at an altitude of about 2,750 feet and probably marks the approximate floor of the filled canyon. Howard Powers (written commun,, May 19, 1967) infers that the confluence of this filled canyon with the ancestral Snake River was west of Bliss because no permeable connection exists between the well and pillow lava 100 feet lower at Bliss; this is discussed more fully on page F17.

TABLE 4.—Log of water well 6S-13E-5dd1, three-quarters of a mile east of Bliss

[From files of U.S. Geological Survey, Water Resources Division, Boise, Idaho. Alt. 3,270 ft. All beds penetrated are here assigned to McKinney Basalt]

Material Thickness
Clay (sloughs)229285
Clay, blue10295
Basalt, hard, black32440
Clay —16456
Sand rock, water-bearing6462
Clay —4466
Sand, water-bearing4470
Clay, blue 35505
Basalt, sand12517

It is unlikely that any lava in the filled canyon of the ancestral Wood River came from Gooding Butte. The Thousand Springs Basalt from this vent dates from a time before the ancestral drainage was deeply entrenched, as previously explained, and the filling of the canyon during more than a single eruptive episode would have caused diversions of drainage that are not recorded by known geologic features. I conclude that the ancestral Wood River canyon was filled solely by basalt from McKinney Butte, which simultaneously descended the Snake River canyon of that time nearly as far as King Hill. Most of the McKinney Basalt in the ancestral Wood River canyon, and farther west in the ancient Snake River canyon, probably is subaerial basalt; the flow of river water, which would have favored the formation of pillow lava, probably was interrupted early during the eruption of the basalt. A few feet of pillow lava locally occurs below subaerial basalt at The Pasture, near the terminus of the flow, and some "clay" and "sand" reported in the water well east of Bliss may represent basalt that fragmented in water; however, the massive lava fill of McKinney Basalt at Bancroft Springs displays only columnar layers that evidently cooled in air (fig. 9). This evidence indicates that Bancroft Springs lies downstream from the site where the McKinney first entered and dammed the ancestral Snake River.

As the McKinney Basalt filled the Snake River canyon west of Bliss, a lake formed in the canyon upstream. Eventually, the lava dam reached an altitude of 3,150 feet, thus impounding a lake 500 feet deep that extended upstream along the Snake River beyond Hagerman. Basalt that reached the lake formed pillow lava, the first of it being deposited on old talus that had been derived from Madson Basalt (Malde and Powers, 1962, p. 1216). Angular pieces of Madson Basalt (not necessarily talus) are also found under the McKinney Basalt at The Pasture near King Hill. Much of the basalt that became pillow lava probably arrived at Bliss by overflowing the ancestral Wood River canyon when the Snake River canyon downstream was already deeply blocked by lava. The copious amounts of McKinney Basalt that spilled into the temporary lake produced the massive subaqueous volcanic features described by Stearns (in Stearns and others, 1938, p. 78-80) as the "Bliss cone," "brecciated lava," and "dikes."

In the old lava plain southeast of Bliss, subaerial McKinney Basalt spread south and covered the Madson Basalt as well as some of the Thousand Springs Basalt from Gooding Butte and a small part of the Wendell Grade Basalt, Some McKinney Basalt on the lava plain undoubtedly spilled west into the lake held by the ancestral Snake River canyon, because outcrops of pillow lava in the canyon extend upstream along the east side as far as the McKinney rimrock (Stearns and others, 1938, p. 79). Discernible bedding in the pillow lava dips west.

The McKinney Basalt near the canyon rim southeast of Bliss has a relatively smooth surface and was formerly mapped separately from lava on the plain 1-3 miles east, which has a rough surface. (See Malde and others, 1963.) Detailed work might show that the smooth McKinney is a local physiographic effect produced by a mantle of loess, but the two contrasting areas may also represent different flows from McKinney Butte that were perhaps separated by a considerable span of time. Whatever the cause of the local differences, such a contrast cannot be recognized west of Bliss, where the McKinney surface is uniformly rough. All basalt from McKinney Butte is therefore classed as a unit even though its eruption may have been prolonged.

The McKinney Basalt west of Bliss, after filling the ancestral Snake River canyon, spread south as a subaerial flow over Madson Basalt to the edge of an upland formed by older basin deposits. In the gap between outcrops of the Madson about 2 miles west of Bliss, which marks the former canyon, the subaerial McKinney evidently rests on the pillow lava facies, which completely fills the former canyon to a depth of 500 feet. Although continuous exposures of pillow lava in the former canyon have been found no more than 250 feet above the canyon floor, the subaerial McKinney rimrock at the east edge of this ancestral canyon is separated from the Madson Basalt below by 25 feet of pillow lava, which reaches 500 feet above the canyon floor—that is, to an altitude of 3,150 feet. Identical pillow lava at a comparable height underlies McKinney rimrock 2 miles southeast of Bliss. The height of these outcrops implies that the dam of McKinney Basalt is concealed about 3 miles west of Bliss, near the buried confluence of the ancestral Wood and Snake Rivers, as shown in figure 11C. If the present canyon had then existed downstream from Bliss, it necessarily would have been filled with McKinney Basalt, thus providing a means for tile spread of the McKinney to its two outcrops now preserved on the south rim (fig. 11D). No sign of such a lava fill in the present canyon can be found; as explained in the discussion which follows, the present canyon west of Bliss surely formed after McKinney time.

Aspects of ground water support the inference that the ancestral Wood and Snake Rivers joined a short distance west of Bliss and indicate that the area between the Snake and the buried Wood River canyon is a drainage divide consisting of rather impermeable basin deposits (older Quaternary and Tertiary). These conclusions are deduced from a lack of springs in the McKinney pillow lava, whereas most other basalts of the Snake River Group in this area are notable aquifers. Small springs that appear to issue from pillow lava along the canyon wall southeast of Bliss, at altitudes between 2,825 and 2,875 feet, are possibly supplied by water that migrates along a concealed surface of Banbury Basalt. (See log of well 6S-13E-6dc1 and map by Malde and others, 1963.) Stearns also thought that this meager flow of spring water could not represent discharge from the pillow lava (Stearns and others, 1938, p. 164), but he attributed the discharge to water carried by the Madson Basalt. Because the former valley filled with Madson Basalt is intercepted upstream by the present Wood River in Malad Canyon, where it yields a phenomenal discharge of about 1,000 cubic feet per second at Malad Springs (Nace and others, 1958, p. 55-58), the amount of water carried by the Madson Basalt in the downstream area of pillow lava must be negligible. In summary, a lack of springs in the pillow lava indicates that this lava is isolated from ground water in the buried ancestral Wood River canyon by an impermeable barrier. Thus, the talus-covered canyon wall southeast of Bliss, which was previously mapped mostly as pillow lava (Malde and others, 1963), surely consists of older Quaternary and Tertiary basin deposits (fig. 3). The necessary corollary is that the dry pillow lava in the ancestral Snake River canyon connects with lava in the filled canyon of the Wood River at a somewhat lower altitude at a concealed junction west of Bliss.

A small amount of ground water, however, issues from the canyon-filling McKinney Basalt at Bancroft Springs, which is near the terminus of the flow. Although Bancroft Springs connects with the present Wood River via the filled ancestral Wood River canyon and also with the Snake River near Bliss via the canyon fill of McKinney pillow lava, its discharge is curiously small. Howard Powers (written commun., Sept. 25, 1967) suggested the following explanation. For the McKinney Basalt at Bancroft Springs to have solidified as columnar subaerial layers, its dam upstream must have been virtually watertight, a circumstance that may persist today. That is, under present conditions, the water carried by the Snake River flows readily down the existing channel, and the water pressure on the concealed lava dam is not great enough to cause substantial leaks. Also, contributions to Bancroft Springs from the Big Wood River must be small because this stream maintains its flow across the lava plain and seemingly loses little water by seepage.


After the eruption of McKinney Basalt, lake water impounded by the lava dam west of Bliss spilled along the south edge of the McKinney lava, where it abutted the upland of basin deposits, and thus established the Snake River in its present course (fig. 11D). As noted by Stearns (in Stearns and others, 1938, p. 77), this overflow descended 700 feet in 7 miles (actually more nearly 13 miles) and must have cut rapidly into the soft basin deposits. In the reach from 3 to 10 miles west of Bliss, a canyon about 1 mile wide exposed Madson Basalt (locally overlain by Sugar Bowl Gravel) below McKinney rimrock. When downcutting reached a depth 300 feet below the north rim, the river was confined to a narrow inner gorge in Tertiary basalt and eventually reached an additional depth of 200 feet (Malde and others, 1963). By the time of the Bonneville Flood of 30,000 years ago (Malde, 1968), the canyon had virtually attained its modern dimensions.

Two remnants of McKinney Basalt are preserved on the south rim of the new canyon. One of these, 11 miles west of Bliss, is subaerial basalt at the altitude of McKinney rimrock on the north side (fig. 10, section A—A'). The other remnant, 6 miles west of Bliss, consists only of basalt boulders 2-3 feet in diameter, which are slightly below the altitude of McKinney rimrock on the north, 1 mile distant. These boulders obviously match the distinctive McKinney lithology and must have been deposited near the edge of the McKinney Basalt during initial overflow of the temporary lake. Their size demonstrates the transport power of the turbulent water discharged from the lake.

The Wood River that had been blocked by the McKinney Basalt was diverted into its present route (the Malad River) along the southeast edge of the McKinney lava. Downcutting by the ancestral Wood River along this marginal path required removal of thick layers of basalt and need not have been especially rapid until erosion reached the base of the Madson Basalt, 200 feet below the lava plain. The rate of cutting must have then increased. Primarily because the soft basin deposits below the Madson were easily erodible, but partly because streamflow was thereby augmented by discharge from Malad Springs (p. F17). It is remarkable that the present Malad Canyon upstream from the base of the Madson Basalt (the site of Malad Springs) abruptly narrows and becomes shallow, whereas the canyon in basin deposits downstream to its mouth, 2 miles distant, rapidly widens and deepens. Even so, Malad Canyon is short and narrow when compared with the Snake River canyon of equivalent age west of Bliss. This difference is probably a function of contrasts in river gradients, discharge, and the resistance of rocks to erosion. As explained below, the comparatively large canyon west of Bliss was not a circumstance of the Bonneville Flood.

The Bonneville Flood deposited large amounts of basaltic debris, particularly in wide parts of the canyon (Melon Gravel of fig. 3). For a short distance downstream from the mouth of the Malad River, however, the flood debris contains gravel derived from the headwaters of the Big Wood River in central Idaho (locality 498). This fact alone, even if other geologic relations were lacking, would suggest that the Bonneville Flood occurred after diversion of the Wood River and was, therefore, younger than the McKinney Basalt. (This gravel was previously discounted as pertaining to the age of the McKinney because I assumed it was younger than the Bonneville Flood.) The basaltic flood debris elsewhere along the canyon, although mostly produced by spectacular erosion near Twin Falls 35 miles up stream from Hagerman, partly represents local smoothing of canyon walls by flood erosion. Such smoothing by the flood, however, had little effect on the previous size of the canyon, for calculations of flood discharge and canyon capacity indicate that the canyon had already attained its approximate present dimensions by the time of the flood. The movement of boulders 10 feet in diameter 9 miles west of Bliss, for instance, demonstrate a velocity of floodwater that required a canyon of the present dimensions to accommodate the Bonneville Flood at its known height and discharge (Malde, 1968, p. 34, 46).

Part of the erosional effect of the Bonneville Flood presumably was the removal of some McKinney pillow lava that had survived during the cutting of the canyon west of Bliss and also the removal of any lacustrine deposits that may have formed at the head of the temporary lake that was dammed by the McKinney Basalt. Such lake deposits, if they had been present, would have been mainly in a narrow part of the canyon upstream from Hagerman and therefore would have been in the direct path of the floodwater.

In descending the Snake River canyon, the Bonneville Flood overtopped rimrock of Sand Springs Basalt 250 feet above the canyon floor south of Hagerman (Malde, 1968, p. 32). Floodwater again overtopped the canyon rim at Bancroft Springs and partly covered the toe of the McKinney Basalt with Melon Gravel.

Since the Bonneville Flood, the Snake River canyon has changed little and still plainly bears marks of this catastrophic event.

<<< Previous <<< Contents >>> Next >>>

Last Updated: 28-Mar-2006