USGS Logo Geological Survey Circular 838
Guides to Some Volcanic Terrances in Washington, Idaho, Oregon, and Northern California



Flows within the Grande Ronde, Wanapum, and Saddle Mountains Basalts range from a few tens of centimeters to more than 100 m thick, averaging 30-40 m. The thick flows generally record ponding in pre-basalt valleys, in structurally controlled basins that developed during volcanism, or in narrow canyons eroded into older flows; such intracanyon flows are common only in the Saddle Mountains Basalt. Even the thinner flows generally show evidence of having ponded. This evidence consists of the columnar-jointed nature of the basalt (fig. 6). Such columns can apparently form only under static cooling conditions; their development therefore implies that the lava had ponded. What impounded the lava can rarely be determined. Natural levees several meters high have been observed in places and probably account for most of the ponding. Elsewhere, flows could have pinched out against opposed topographic slopes.

Figure 6. Cross section of flow in Yakima Basalt Subgroup, showing, in idealized form, jointing patterns and other structures. PPC, pillow palagonite (hyaloclastite) complex, present at base or throughout flows that entered water. (click on image for an enlargement in a new window)

Flows that cooled under stagnant conditions contracted and developed a characteristic jointing habit, shown in idealized form in figure 6. The terms colonnade and entablature were borrowed from classical architectural usage by Tomkeieff (1940). Columns in the colonnade are from 10 cm to 5 m in diameter, averaging about 1 m, and can be as long as 50-75 m although generally 5-10 m. Most are straight, but curved columns are rather common and generally unexplainable in terms of simple cooling models. Columns in the colonnade are commonly subdivided into prismatic blocks by cross joints; the coarsest-grained part of a flow may be platy owing to the close spacing of cross joints.

The colonnade-entablature contact is relatively sharp, the change commonly taking place within 1-2 cm. The contact is traceable in many places for several kilometers before other complexities obscure it. The glass content of the groundmass increases abruptly from the colonnade to the entablature for an unknown reason (Swanson, 1967; Long, 1978). The entablature consists of columns of smaller diameter, generally less than 25 cm, and less consistent orientation than those in the colonnade. Columns in many entablatures are bundled into fans, synforms, tents, or other unusually shaped arrangements. Most columns in an entablature are highly segmented by irregular cross joints, so that the columns can be readily broken into fist-size pieces. The entablature generally comprises about 70 percent of the thickness of a flow but can make up 100 percent (one example that we know of) to zero percent. The upper part of the entablature is scoriaceous and commonly merges into a zone of short, wide, generally poorly defined columns that some workers call the upper colonnade. A rubbly, clinkery zone occurs above the entablature of some flows. Such a zone is absent from the base of flows. The origin of this rubbly zone is in question. The rubble is, in our experience, more common near vent areas than elsewhere; this observation suggests that the rubble may represent material thrown out of the vent near the end of eruption and modified during movement of the flow.

Idealized jointing patterns can be satisfactorily explained by existing theory for the cooling of bodies of igneous rock (Jaeger, 1961), but such patterns are seldom found in nature. Acceptable thermo-mechanical explanations for the typically complex jointing patterns, particularly in the entablature, are not available despite considerable descriptive information (Tomkeieff, 1940; Waters, 1960; Mackin, 1961; Spry, 1962; Swanson, 1967; Schmincke, 1967b; Long, 1978; Ryan and Sammis, 1978). Complications related to the mutual interference between columns growing inward from irregular contacts, ponding of water on a flow surface and percolation down joint planes during solidification, the influence of chemical composition on tensile strengths and heat conduction, and inadequate knowledge of rock mechanics under high temperature-low pressure conditions are some of the difficulties that plague attempts at analysis of natural jointing habits.

Some flows have a tiered appearance defined principally by alternating layers of vesicular and relatively nonvesicular rock rather than by joints. These layers may record separate gushes or thin flows that piled up and solidified as a single compound cooling unit. Tiers much more commonly occur in entablatures than colonnades.

The upper surface of a flow is rarely exposed in plan view. Where seen, the surface is rather flat, smooth, filamented, and locally ropy—surface features characteristic of lava ponds at Kilauea. The surface of a flow with a rubbly upper zone is rough, has a relief of as much as 6 m, and otherwise appears unlike typical surfaces of ponded flows.

Many flows entered water and formed pillows. Recent studies (Jones, 1968; Moore, 1975) have demonstrated conclusively that pillows are nothing more than the subaqueous equivalent of pahoehoe toes. Many of the pillowed flows occur near the margin of the plateau as it existed at the time of eruption, apparently because lakes were formed as flows ponded rivers draining from marginal highlands. Other pillowed flows are much more extensive, perhaps signifying entry into shallow lakes standing on the plateau surface. One such extensive flow, in the Priest Rapids Member, is pillowed throughout an area of tens of square kilometers in the Cheney-Palouse scabland southwest of Spokane.

In places, lava deltas (Fuller, 1931; Moore and others, 1973) formed as lava poured into shallow lakes and ponded streams. The direction of dip of foreset "bedding" defined by elongate pillows and thin sheet flows in the lava deltas indicates the local flow direction of the lava. Particularly good examples of lava deltas can be seen near Malden south of Spokane (Griggs, 1976), near the mouth of Moses Coulee (Fuller, 1931), and at the mouth of Sand Hollow south of Vantage.

Other criteria for defining flow directions include inclined pipe vesicles, which plunge upcurrent, and bent spiracles (fig. 6), formed by steam blasts beneath a flow, which tail out down-current. Flow directional data for basalts must be treated in the same way as those for current-produced structures in sedimentary rocks—carefully. A few data in a small area show the local direction but say little about regional patterns. Nonetheless, careful studies by Schmincke (1967a) and on-going work by others are succeeding in defining patterns of lava advance within the plateau.


One of the major results of recent mapping has been identification of sources for most stratigraphic units and even single flows. Feeder dikes have been found for flows in all formations and members of the Columbia River Basalt Group except the Wilbur Creek, Asotin, Esquatzel, Buford, and Lower Monumental Members of the Saddle Mountains Basalt. The feeder dikes average about 8 m wide but vary from a few centimeters to more than 60 m. They may tend to thin upward, but this is far from certain. The dikes generally trend north to north-northwest. They cannot be traced far along strike, in part because of exposure problems. Obviously related dike segments, offset a few meters to form an en echelon pattern, form systems extending tens of kilometers. Compound or multiple dikes, consisting of two or more pulses of magma related to the same intrusive event, are common, but composite dikes, containing two or more phases of contrasting compositions, have not been reported. In other words, each fissure was used just once, not repeatedly.

The chance of finding a dike connecting with a flow it fed is small, owing to problems of exposure. Nonetheless, several dikes have been found displaying such a connection. The top of most such dikes is rubbly, apparently consisting of slabs of crust once floating on a flow before it poured back into the fissure (for example, Plate 1a in Swanson and others, 1975). In one dike lacking such rubble, the dike merges imperceptibly with its flow of the Frenchman Springs Member (fig. 1, number 6, Swanson and others, 1975).

Many other dikes can be inferred to correlate with particular flows, or at least sequences of flows, on the basis of chemical composition and magnetic polarity. In this, way, feeders have been identified for most of the named stratigraphic units.

The shape and extent of vent systems for specific flows or related sequences of flows can be reconstructed from the locations of feeder dikes, thick piles of degassed flows (presumably near their vent), abundant collapsed pahoehoe (Swanson, 1973), and accumulations of basaltic tephra (in places occurring in still recognizable spatter cones and ramparts). Such reconstructions show that eruptions of single flows or related flows took place from fissures concentrated in long, narrow vent systems on the order of tens of kilometers long and several kilometers wide (Swanson and others, 1975, and later work).

Vent systems for the Grande Ronde Basalt are distributed across the eastern half to two-thirds of the Columbia Plateau, but those for other units are more restricted (Swanson and Wright, 1979). For example, feeder dikes for the Picture Gorge Basalt are confined within the John Day Basin and neighboring areas, those for the Frenchman Springs Member within a zone 60 km wide and probably more than 200 km long, and those for the Ice Harbor Member within a zone 15 km wide and about 90 km long. On a still finer scale, vent systems for specific flows are nearly linear, as they occur along single dikes or closely spaced related dikes. Examples are the vent system for the Roza Member (probably less than 5 km wide and now known to be more than 165 km long Swanson and others 1975; P.R. Hooper, unpub. data, 1978 ), the basalt of Robinette Mountain in the Eckler Mountain Member (a single dike extending at least 25 km across the Blue Mountains), the basalt of Basin City in the Ice Harbor Member (possibly a single dike at least 50 km long), and several chemically distinct flows in the Grande Ronde Basalt (T.L. Wright and D.A. Swanson, unpub. data, 1978).

Knowledge of the geometry of the vent systems not only allows prediction of where vents for specific flows should occur once one such vent is found but also further provides important constraints regarding magma generation, storage, and eruption mechanics. Using such knowledge, attempts have been made to estimate the rates of eruption and advance for single flows. The estimates take into account the observation that flows, even those that advanced tens to hundreds of kilometers from their sources, quenched to a crystal-poor sideromelane glass when they entered water; this indicates little cooling during transport and hence rapid advance, since the lava apparently moved as sheet floods rather than through insulating tube systems. Application of rheologic models, developed in part from this observation by Shaw and Swanson (1970), to vent systems of known dimensions suggests eruption rates of about 1 km3/day per linear kilometer of active fissure for the largest flows, such as those in the Roza Member, and about 10-4km3/day/km for the smaller flows (Swanson and others, 1975). For flows of "average" volume, probably several tens of cubic kilometers, rates of 10-1 to 10-2km3/day/km may be inferred. By comparison, sustained rates of eruption at Kilauea and Mauna Loa are 10-3 to 10-4km3/day/km. Using observed dike widths, theoretical modeling suggests that the high eruption rates could indeed have been sustained by supply from depth (Shaw and Swanson, 1970). Such eruptions probably lasted a few days. Flow rates of 5 to 15 km/hr down slopes of 1:1000 are calculated from the model, adequate to allow thick flows to move far with little cooling.

Rapid eruption rates do not necessarily imply rapid melting rates in the mantle. Flows were erupted only once every ten thousand years or so in any one area on the plateau during even the peak of volcanic activity (Grande Ronde time), as estimated by counting the number of flows in a magnetostratigraphic unit of assumed duration based on comparison with seafloor magnetic anomalies of roughly comparable age. Calculations show that continuous melting at the present Hawaiian rate, 10-1km3/yr (Swanson, 19721, could account for the volume of the Columbia River Basalt Group in the alloted time. Episodic rapid melting events ("flash melting") cannot be excluded but are not required.

If melting progressed at the Hawaiian rate, then large, deep storage reservoirs are required in order to account for the large volume of single flows. This contrasts with the Hawaiian situation, where eruptions are much more frequent and lava "leaks" to the surface more or less continuously. The presence of large, deep storage reservoirs, possibly in the upper mantle, may be a principal and distinguishing characteristic of flood-basalt provinces in general.


Two major geochemical breaks occur within the Yakima Basalt Subgroup. The older break separates a high-SiO2, low-FeO and TiO2 sequence below (Grande Ronde Basalt) from a relatively low-SiO2, high-FeO and TiO2 sequence (Wanapum Basalt). This chemical change took place over a short time, as magnetic stratigraphy is continuous across the break, K-Ar ages above and below agree within analytical error for these rocks (+ 1 m.y.), and flows of different chemical types are locally interlayered (fig. 4). Trace-element levels are similar and Sr isotopic ratios are similar and relatively low (0.704-0.706) in both sequences. Major- and trace-element chemistry of the Grande Ronde Basalt, which makes up 75 percent of the volcanic pile, shows cyclic variation rather than an evolutionary trend.

The younger geochemical break, separating the Wanapum and Saddle Mountains Basalts, is marked by 1) increase in Sr isotopic ratios to 0.708-0.715, and 2) greater abundance of most incompatible elements and steeper chrondrite-normalized REE petterns in those flows of the Saddle Mountains Basalt whose major oxide chemistry is similar to older flows. These changes took place over a relatively long time (.5-1 m.y.?), as there is evidence of erosion and deposition of sediments between the Wanapum and Saddle Mountains Basalts.

The Columbia River Basalt Group shows far less coherent chemical variation than is typical for oceanic tholeiites. Ratios of incompatible elements vary widely among the various formations as well as among flows belonging to the same formation. Enrichment factors differ for different incompatible elements in units which show smooth chemical variation trends. For example, the least magnesian flows of the Grande Ronde Basalt (MgO/3 percent), have contents of P2O5. Hf, Ta, and LREE that are 2-3 times higher than in the most magnesian flows of the Grande Ronde (MgO/6 percent), whereas the enrichment of Th and K in these same flows is greater, about 3-4.

We make the following inferences as a start toward a model to explain the generation of magmas in the subgroup.

1. We infer that magma compositions are controlled principally by partial melting. Enrichment factors in the Grande Ronde Basalt are not consistent with a fractionation model in which crystallization of 55-70 percent of stored liquid would have to take place repeatedly. Mixing calculations fail to balance both major oxides and incompatible trace elements using any reasonable fractionating mineral assemblage.

2. The scarcity of phenocrysts, absence of high-pressure "megacrysts", and absence of fractionation or accumulation trends suggest that magmas accumulated near the site of melting and may have been superheated during transport to within a few kilometers of the surface.

3. We infer little or no high-level crustal storage for most of the magma, as evidenced by 1) absence of grabens or calderas associated with single flows of large volume and 2) absence of phenocryst-related chemical fractionation or accumulation trends even between phyric and aphyric flows of the same stratigraphic unit. We feel that plagioclase phenocrysts present in some flows may have formed during ascent of magma just prior to eruption and perhaps even after eruption.

4. If a partial melting model is accepted, melting must be relatively "wet" to yield quartz-normative magmas. Some metasomatic enrichment of trace elements is required to explain the incoherent ratios of incompatible elements and possibly the highest 87Sr/86Sr ratios.

5. Bulk-lava chemistry suggests that the source for most of the magma was relatively iron-rich and olivine-poor clinopyroxenite. The major chemical and stratigraphic units and many of the individual flows require chemically distinct source rocks. The overall source for magmas of the Columbia River Basalt Group must be heterogeneous both in space and through time.


Weakly consolidated sedimentary rocks, generally medium-grained sandstone to siltstone, occur between many flows near the margin of the plateau and between some of the younger flows in local structural basins on the plateau. The sedimentary rocks rest depositionally on the underlying flow in some places, but in many other places the contact relations show that subjacent basalt intrudes or invades the sediment. Schmincke (1967c) was one of the first to recognize this, and recent work has demonstrated how common such invasive relations are. We estimate that more han half of the observed contacts between basalt and sedimentary rocks on the Columbia Plateau are invasive.

How are such contacts interpreted? Do they signify "normal" intrusive relations in which basaltic magma never reached the surface before solidifying as in classic dikes and sills, or are they formed as lava flows burrow into unconsolidated sediments accumulating on the ground surface (invasive flows of Byerly and Swanson, 1978)? Both processes produce similar results.

The key to proper interpretation lies in the stratigraphy. If the basalt is at its proper stratigraphic position relative to overlying flows, it almost certainly was a flow that invaded sediments at the ground surface. This is because thin sedimentary deposits, generally less than 10 m thick on the plateau, are light and hence exert little confining pressure; vesiculating magma rising and encountering such sediments would certainly blast through rather than spread laterally into them.

Mapping and chemical studies have shown that, in every example so far found on the plateau, the invasive basalt is at its expected stratigraphic position relative to overlying flows and hence is in invasive flow (Schmincke, 1967c; Camp, 1976; Byerly and Swanson, 1978; D.A. Swanson, G.R. Byerly and T.L. Wright, unpub. data, 1978). We include in this interpretation two thick sills previously interpreted conventionally, the Hammond sill of Hoyt (1961) near Wenatchee, Washington and the "Whiskey Creek sills" of Bond (1963). Work by Byerly and Swanson (1978 and unpub. data) and V.E. Camp (unpub. data, 1978) shows that even these thick sills, more than 120 m thick locally, are in proper stratigraphic position relative to overlying flows.

These conclusions are significant, because they show that invasive contacts provide insufficient, in fact totally misleading, evidence for the former presence of magma beneath the area. For example, invasive relations are particularly well displayed along the northwest margin of the plateau, where all other evidence negates the former presence of magma bodies; no intrusive contacts are found here in any situation other than one involving sediments, the invading flows are in their proper stratigraphic position relative to other flows, and flow directions show that surface flows moved toward, not away from, correlative sills. Invasive relations also demonstrate the unreliability of using sedimentary interbeds as stratigraphic guides on the Columbia Plateau. The basalt flows are always in their proper stratigraphic position, but the interbeds, at least fine-grained ones, are commonly not, owing to rafting by invasive flows.

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

Last Updated: 28-Mar-2006