Sheeted Clastic Dikes in the Megaflood Region

Sheeted Clastic Dikes in the Megaflood Region, Washington, Oregon and Idaho

Skye W. Cooley

Cooley Geoscience | Mission Valley, MT

Clastic dikes in the Channeled Scablands of Washington, Oregon, and Idaho are vertically-sheeted wedges of silt, sand, and gravel filled from above. They are most numerous and best developed in slackwater deposits (Touchet Beds), but also found in coarser grained flood deposits that contain significant silt. For this study, I collected width measurements and sheet counts from >3000 dikes at >300 sites along the floodway between Priest River, ID and The Dalles, OR. All dikes exhibit identical characteristics and are sourced in flood-laid sediments. I interpret the dikes as sediment-filled hydraulic fractures (flood injectites) that grew incrementally in length and width during floodwater-loading events during the Pleistocene. This interpretation is supported by numerous lines of evidence including the source of fill material, taper direction, and crosscutting relationships with dated tephras, paleosols, and older bedrock. The dikes are not liquefaction features, feeder conduits to sand blows, or other soft sediment deformation features triggered by seismic shaking (i.e., seismites). They are neither larger nor more numerous near active Quaternary faults. In fact, evidence of liquefaction in the Channeled Scablands is rare, despite its transection by more than a dozen faults of the Yakima Fold Belt. Instead, dike size, abundance, and complexity appear primarily controlled by grain size, sediment thickness, and location within the floodway. The largest dikes occur in thick, silty slackwater sequences near the deepest flood channels. Analogous wedge-shaped dikes with sheeted fills are found in marine turbidites, subglacial tills, and sediments overridden by lahars - substrates subjected to rapid sedimentation and repeated loading.

Clastic dikes are threshold features that, if interpreted one way, may brand a landscape hazardous and unfit for occupation or development. Interpreted another way, the same dikes become Ice Age relicts of little importance to anyone other than academics and megaflood enthusiasts.

Definition

The term "clastic dike" is a general term that describes thin, sheet-like intrusive bodies of sediment that crosscut layering, whether in unconsolidated sediments or bedrock.

Types of Clastic Dikes

More than 400 articles on clastic dikes have been published since the early 1800s. The literature describes dikes formed by liquefaction (Fuller, 1912; Obermeier, 1998), injection (Jenkins, 1925; Hardie, 1999; LeHeron and Etienne, 2005; Hurst and other, 2011; Ravier and others, 2015), release of volcanic fluids and gasses (Diller, 1890; Gonzales and Koch, 2017), meteorite impact (Kriens and others, 1999; Huntoon and Shoemaker, 1995; Buchner and others, 2022), tectonic extension (Haluszczak, 2007), mass wasting (Winterer and others, 1991), diagenesis (Maher and Persinger, 2023), syneresis (McMahon and others, 2020), desiccation (Lawler, 1923), and freeze-thaw action (Pewe, 1959; Demoulin, 1996).

Classifying Soft Sediment Deformation

A clastic dike is a type of soft sediment deformation feature (SSD). Various classification systems for SSD devised to help aid in interpretation of features in the field have been developed (Potter and Pettijohn, 1963; Dzulynski and Walton, 1965; Nagtegaal, 1965; Hyashi, 1966; Dionne, 1974; Lowe, 1975; Allen, 1982; Mills, 1983; Owen, 1987; Guirard and Plaziat, 1993; Collinson, 1994; Greb and others, 2002; Jolly and Lonergan, 2002; Wheeler, 2002; Obermeier and others, 2005; Castilla and Audemard, 2001; Montenat and others, 2007; Van Loon, 2009; Hurst and others, 2011; Owen and others, 2011; Waldron and Gagnon, 2011; He and Qiao, 2015; Shanmugam, 2017; Alsop and others, 2019; Zhong and others, 2022. However, all of the systems present inconsistencies, which frustrates the user and thwarts widespread adoption. The terminology surrounding clastic dikes can be particularly confusing, as most articles describe a handful of features exposed in some local outcrop without much perspective on dikes described elsewhere. To date, no system captures the full menu of sedimentary dikes in the geologic record and few provide clear rules of thumb for determining the origin of a dike if more than one potential trigger exists. What's needed is an illustrated compendium of clastic dikes that standardizes terminology, addresses common questions, offers guidance to those working with ambiguous features, and provides numerous examples from dozens of locations around the world. I'd be happy to author one for pay.

Avoid Splitting

The main problem with classification systems is they unnecessarily encourage us to become splitters. The temptation to begin delineating one type of dike from another based on perceived differences - especially when your project is in its infancy - must be resisted. It is unlikely that multiple mechanisms conspired to produce different types of dikes of the same age in your field area. Don't be like Karl (Fecht and others, 1999). Pick one mechanism and stick with it for as long as possible.

Six Main Types of Clastic Dikes

Liquefaction Conduits - Dikes and pipes formed by the upward escape of pore water and sediment toward the ground surface. Seismic shaking is often the trigger. Shaking reduces grain-to-grain contact, induces consolidation, and increases pore fluid pressure that leads to their escape as dikes. Water and sand ventdd at the surface produces cone-shaped edifices called sand blows or sand volcanos. Example: Sand blow region near New Madrid, MO.

Hydraulic Fracture Fills - Loading-triggered fractures that open, propagate, and fill with sediment in a pressurized environment. Fracture propagation initiates when fluid pressure exceeds the resistance of the confining material (fluid pressure > fracture toughness). Hydraulic fractures are often triggered by rapid overloading by glaciers, debris flows, lahars, turbidites, or megafloods. Example: Sand injectites in the Panoche Hills, CA.

Passive Infill Structures - Subaerial fractures formed in brittle bedrock or frozen soils that fill with loose material that is washed in by water, blown in by wind, or collapsed in by gravity. Surface fractures may be the result of mass wasting (lateral spreads, slides), thermal contraction (freeze-thaw action), desiccation (drying), or extension (joints, faults, crests of folds). Examples: Permafrost of Arctic Alaska and coal mines near Kleszczow, Poland.

Volcanic-Phreatic Dikes - "Pebble dikes" are breccia-filled dikes that form by the rapid escape of volatile gasses and fluids through fractured rock. Pebble dikes are commonly ore-bearing and associated with hydrothermal systems (i.e., contact aureoles near plutons, plumbing systems of ancient hotsprings, and phreatic vent complexes). Example: Mining districts of the High Rockies.

Syneresis Cracks - Subaqueous shrinkage cracks that fill with sediment, authigenic minerals, or both. Common in shallow, clay-rich marine deposits and certain lacustrine sequences. Syneresis dikes are often long, thin, and filled by materials available nearby. An excellent overview is provided by McMahon and others (2020). Example: White River Badlands, SD.

Impact Injections - Shock waves and intense heat generated during meteorite impacts can trigger the temporary fluidization and injection of sand into the surrounding bedrock. Impact dikes, most often of sandstone, commonly radiate outward from the impact location and contain shocked grains. Example: Upheaval Dome, UT.

Previous Work

The earliest reports on clastic dikes date to the 1800s (Strangways, 1821; Murchison, 1827; Darwin, 1834; Dana, 1849; Gilbert, 1880; Diller, 1890; Cross, 1894). Olaf P. Jenkins (1925) authored the first report on clastic dikes in the Channeled Scablands, specifically those in the Touchet Beds near Walla Walla, WA. A number of similar reports on dikes in slackwater deposits exposed elsewhere in the region were published soon after (Lupher, 1940, 1944; Black, 1979; Woodward-Clyde Associates, 1981). Reports containing measurements, dike distribution maps, and examples of dikes in formations other than the Touchet Beds came next (Alwin and Scott, 1970; Cooley and others, 1996; Neill and others, 1997; Fecht and others, 1999; Ward and others, 2006). While numerous authors have speculated on the origin of clastic dikes in the Touchet Beds (Flint, 1938; Newcomb, 1962; Bingham and Grolier, 1966; Jones and Deacon, 1966; Beaulieu, 1974; Carson and others, 1978; Shaw and others, 1999; Pritchard and Cebula, 2016; Reidel and others, 2021), few support their assertions with measurements, maps, or models. Fresh ideas have been slow to emerge. Articles by Smith (1993), Pogue (1998), and Howard and Pritchard (2020) offer a refreshing departure from this trend.

Omitted or Ignored

Despite their ubiquity in Missoula flood deposits, clastic dikes are not shown in several important descriptions of classic Scabland outcrops (Waitt, 1985; Smith, 1988a,b; Lindsey and others, 1996; Benito and O'Connor, 2003; Sweeney and others, 2017). Some authors have told me, off the record, that they avoided them on purpose because the dikes complicated their work, focused primarily on the stratigraphy.

Proposed Origins of Dikes in the Channeled Scablands

Six origins for clastic dikes in the Channeled Scablands have been proposed: Earthquakes (Jenkins, 1925), ground ice (Lupher, 1944; Alwin and Scott, 1970), desiccation (Grolier and Bingham, 1978), mass wasting (Brown and Brown, 1962; Baker, 1973; Cooley and others, 1996), dewatering (Newcomb, 1962), and hydraulic fracture (Pogue, 1998). A dubious seventh, “multigenetic” (Black, 1979; Fecht and others, 1999), suggests the dikes formed by a combination of processes. Cooley (2015) provides a concise summary of the arguments for and against each hypothesis.

Literature Review

The PDF linked below contains a list of all (or nearly all) articles that mention clastic dikes in the Channeled Scabland. Commentary on and quotes from each is also provided.

Cooley_Clastic Dike Article Review_v9.pdf

Download PDF • 741KB

World's first publication on clastic dikes. Strangways (1821) is the first article in the geological literature to describe clastic dikes. According to his sketch of a shoreline exposure near St. Petersburg, Russia, yellow clay veins descend from a gravel-capped bed into blue clay below. The dikes criss-cross the gently sloping beach, forming a polygonal pattern.

First investigation in the Channeled Scablands. Olaf P. Jenkins examines a large clastic dike exposed in a gravel pit near Lowden, WA in 1923. The caption of his photo reads, "Clastic dikes in Touchet Beds and dust dune between Touchet and Walla Walla". The dike is sourced in light-colored slackwater sediments that overlie dark, plane bedded sands. Holocene loess caps the section. Jenkins, who would later become the Chief Geologist for the California State Division of Mines, recognized the the dikes as fluid-driven injection features similar to those he had seen in the San Joaquin Valley, CA (Jenkins, 1930). Though Jenkins' article is very early, his work was informed by an even earlier report on the giant sand injectite complex in the Panoche Hills by Anderson and Pack (1915). Photo: Washington Geological Survey Archives (No. 00604).

This Study

I searched for clastic dikes in unconsolidated sediments, partially-lithified sediments, and flood-scoured bedrock exposed along roads, streams, rail lines, and escarpments between Priest River, ID and The Dalles, OR. I also surveyed dozens of excavated pits, trenches, and quarries. I traversed the valleys of the Columbia, Snake, Yakima, Spokane, Walla Walla, Sanpoil, Touchet, Tucannon, Umatilla Rivers, and those of numerous tributaries. Thick sections of non-flood sediments at Saddle Mountains, Smyrna Bench, Frenchman Hills, White Bluffs, and Palouse Hills were carefully examined along with dozens of cuts north of the Channeled Scablands. Many roadcuts in Palouse Loess both inside and outside the margins of the floodway were inspected as well as dozens of floodplain exposures of Holocene alluvial fills. Pliocene loess sections at Selah and White Bluffs were checked for dikes. I measured the widths of >3000 dikes at >300 exposures and made vertical sheet counts for >1000 dikes. Field work was conducted between 1995 and 2025 using the same field and data collection procedures. The vertically-sheeted sand, silt, and gravel-filled dikes vary in width and length from place to place, but otherwise exhibit identical characteristics wherever found. The dikes occur only within Ice Age floodways and at elevations no higher than Missoula flood trimlines. All appear to have formed by the same mechanism during the Pleistocene, not before or since.

Study sites. Locations where sheeted dikes were measured are black dots (n = 326). Each location generally contains multiple dikes, typically more than 10. White dots are locations that were searched, but where no dikes were found (n = 318). Many white dots lie just outside the floodway, shown in gray, or where flood deposits are gravelly. Though a few black dots appear to fall outside the floodway, this is an artifact of the geologic mapping. Flood-laid sediments actually extend a bit beyond the areas mapped in gray, but are hidden beneath younger deposits such as loess and alluvium. The Cordilleran Ice Sheet terminus is shown in light blue. Glacial Lake Columbia in blue is shown filled to its 600m-elevation shoreline. Exposures containing dikes are more numerous in the southern part of the study area, where many streams, roads, and railways create more outcrops. Walla Walla County's road maintenance crews are particularly good at keeping their roadcuts fresh. Dikes are sparse between Moses Coulee and Cheney, where loose, patchy gravels lacking in silt rest atop scoured bedrock. The few dikes observed north of the Channeled Scablands occur in sandy outwash and till. Bar gravels and Palouse Loess generally contain few dikes unless interfingered with or overlain by slackwater rhythmites. No dikes cut loess anywhere above Missoula flood trimlines, the indicators of maximum flood stage. No dikes were found east of Priest River, ID (Glacial Lake Missoula basin), east of Lewiston, ID (Hells Canyon-Snake River Valley), west of White Swan, WA (Yakima River Valley), south of Cecil, OR (Willow Creek Valley), north of Hunters, WA (Columbia River Valley), north of Bridge Creek, WA (Sanpoil River Valley), or north of Tonasket, WA (Okanogan River Valley). The Willamette, Wenatchee, and Methow Valleys were not part of this study.

A typical clastic dike in Eastern Washington. A vertically-sheeted clastic dike intrudes silty-sandy late Pleistocene slackwater rhythmites (Touchet Beds) at Smith Hollow Rd in the Tucannon Valley, WA. Wall-parallel sheeting is the product of hydraulic crack-and-fill cycles. The wedge-shaped dikes were filled from the top. S.W. Cooley photo.

The largest dikes occur in the thickest stacks. A very large composite dike at Lowden, WA. Similarly large dikes are found in the thickest sections of Touchet Beds, typically near basin centers. S.W. Cooley photo.

Horizontal outcrops. Benches in quarries and surfaces bladed flat by a dozer often provide a unique view of dikes in cross section. Weathered surfaces often show better than fresh cuts. Ash Hollow Quarry, WA. S.W. Cooley photos.

Sheeted sills. Sheeted sills are less numerous than sheeted dikes, but are otherwise identical. Horizontal sills branch from and are fed by vertical dikes. Sills generally follow sandy bedding, which provide efficient injection pathways. Intrusion of both dikes and sills occurred together and at shallow depths. Because dikes greatly outnumber sills, it would appear fracturing in a vertical plane required less force than lifting the overburden against gravity. Most sills that I have observed were high in the section, suggesting sill formation may be depth-dependent. Hwy 240 at the Hanford Site. S.W. Cooley photo.

Dike-sill-dike geometry. Clastic dikes follow least-resistance pathways. In the Touchet Beds, most are vertical and crosscut bedding at high angles. The dike shown here cuts vertically across low-permeability silts (tan) and horizontally within higher-permeability sands (gray). Sheeting flips between vertical and horizontal, which is challenging to visualize in three dimensions. Vertical paths represents flow in fractures, while horizontal paths represent flow through the matrix. What causes a dike to switch modes is often unseen, but has to do with micro-scale differences in efficiency that the fluid-driven fractures are able to exploit. Hellsgate Recreation Area at Lewiston, ID. S.W. Cooley photo.

Vertical sheets with stratified fills. Sediments that fill the dikes is commonly stratified with fine-scale planar to concave-up bedding consistent with fractures that opened to the surface and filled from the top. Repose angles of laminated fills approach 55 degrees, indicating the sediments entered the fractures as turbulent slurries and accumulated rapidly. Bedding also indicates deposition was aided by a component of lateral flow (flow along strike) during the brief periods when fractures were widening and lengthening. Structureless fills are also found with some regularity and are typically composed of silt and very fine sand. Silt is notoriously poor at preserving bedding, however, some structureless fills appear to signal the abrupt closure of fractures and 'freezing' of unsettled sediment. S.W. Cooley photo.

Sheets, fill bands, or dikelets - all three terms work. More than 40 sheets comprise this dike. Both stratified and structureless fills occur together, indicating the sediment source was generally fine-grained, but continually changing. A few of the more distinct fills suggest formerly wider sheets were split by new injections as the dike widened. It is possible to reconstruct the injection history of some dikes. S.W. Cooley photo.

More than a Touchet Bed story. In addition to flood deposits, the dikes intrude older sediments and bedrock. Here, a sheeted dike sourced in silty slackwater deposits cuts what appears to be a crudely-bedded fan gravel. The coarse, angular basaltic material was originally deposited as scree at the base of cliffs along the Columbia Gorge, but was picked up by floodwaters and swept a short distance downstream. Alder Ridge, WA. S.W. Cooley photo.

Huge dikes formed where slackwater lakes were deepest. Very large dikes with widths exceeding a meter often contain >100 fill bands. They are most common in the southern part of the Lake Lewis basin, where rhythmite stacks are thickest and slackwater lakes were deepest. The dike shown here strikes obliquely to the bladed cutface. Foster Wells Rd at Hwy 395, some 30 km north of Wallula Gap. S.W. Cooley photo.

Burlingame Canyon. About 40 slackwater beds are exposed at this classic locality near Gardena, WA (Waitt, 1980, 1985; Moody, 1987). Clague and others (2003) counted 53 beds in south-central Washington, specifically 22 beds below the Mount St. Helens Set S tephra and 31 above. The size and completeness of the Burlingame Canyon section is unique in the Scablands region, though access to the land is complicated. Nearby outcrops along Touchet River Road and other drainages are easier to access and, frankly, provide superior information, especially if your group is large. Though a visually stunning record of repeated backflooding, the Burlingame section actually isn't all that illuminating when it comes to clastic dikes. Also, remember that Touchet Beds are not exclusive to the Lake Lewis basin. Touchet-equivalent rhythmites are also found in the Upper Columbia Valley, Latah Creek, Lacrosse, Cecil, Portland, and Salem. Photo: Washington Geological Survey Archives (1978, No. 3455).

Dike Distribution

Sheeted dikes are not unusual features in Missoula flood deposits, nor are they confined to fault zones. The dikes are widespread and number in the hundreds of thousands (if not millions), distributed throughout an area exceeding 30,000 km2. Identical dikes are found near Kettle Falls, WA, Salem, OR, and Lewiston, ID, sites separated by hundreds of kilometers. The dikes are not present above the flood elevation maximums (~366m in south-central Washington and higher to the north), nor in sandy, unconsolidated, non-flood sediments beyond the margins of the floodway. No dikes are known to intrude Palouse Loess outside of flood-cut coulees. Dikes are present where loess and flood deposits interfinger (i.e., northern Walla Walla Valley and eastern Pasco Basin). Identical dikes occur in close proximity to mapped Quaternary faults (i.e., Wallula fault zone) and at distances >150 km from them. The dikes are largest, most abundant, and best-developed in the thickest sections of silty slackwater rhythmites (i.e., near Wallula Gap through which all megafloods drained).

Size of Dikes

Dike width ranges from 1 cm to 3 m and length regularly exceeds 10 m. Most dikes measure <15 cm wide, contain fewer than a dozen vertical sheets, and have length-to-width ratios of approximately 40:1. The largest dikes contain >100 sheets and penetrate to depths greater than 50 m. The average width of an individual sheet is around 1 cm with the widest examples around 30 cm. Sills rarely widen beyond 30 cm and tend to be short, often pinching out within a few meters.

Shape of Dikes

Dikes in silt-sand rhythmites tend to be slender (H >> W), while those in coarse sand and gravel are stubby, few in number, and crudely sheeted. Dikes that penetrate bedrock are slender and follow obvious joints. The three-dimensional shape of a dike resembles an axe blade with a curved arris and some volume (i.e., PKN fracture model of Perkins and Kern, 1961; Nordgren, 1972; Belin and Carey, 1997; Rahman and Rahman, 2010). Fracture aperture predictably scales with volume. Dikes taper in the direction fractures propagated, in most cases downward. The 2D shape of a dike's cross section changes depending on where the line of section is drawn (i.e., how it intersects the outcrop).

Strike, Dip, and Orientation of Stresses

When viewed from above, the dikes form polygonal networks, so for the most part, dike strike is random. Dips are also quite consistent, typically just shy of vertical. Most dikes crosscut bedding at angles >70 degrees. Sills are less abundant and are always fed by larger dikes. Most sills form short spurs. Overall, dike and sill orientations are consistent with vertical loading. That is, the maximum principle stress (O1) is vertical. The minimum and intermediate principle stresses are horizontal and nearly equal (O2 = O3). Nearly-equal horizontal stresses explain why the dikes twist about their vertical axes similar to how hydraulic fractures propagate in dyed gelatin (i.e, Ham and Kwon, 2019; Nicolas Espinoza on YouTube, youtube.com/watch?v=3pJgZnPUIYE).

Bedding-plane Slip

Shear offsets across the width of a dike are uncommon. Most fractures opened against O3 or O2 without shear. Stair-stepping and en echelon forms are not common, but where observed, involve many small bedding-parallel slips. Centimeter-scale slips are most common, but larger slips exceeding 10 m are found in places. Low-angle slips appear to be either a mass wasting phenomena or perhaps related to drag imposed by fast-moving currents moving over the stack of rhythmites. I've been able to trace one conspicuous low-angle slip surface for several kilometers through a number of outcrops in the northern Walla Walla Valley.

Polygonal Networks

Burned fields, bladed cutslopes, and dry creek beds expose polygonal dike networks in plan view. Horizontal exposures often confirm crosscutting relationships seen in vertical cuts, but uniquely reveal delicate intertwining that occurs where dikes intersect. Intertwined dikes share fill bands and appear to have grown simultaneously despite having different strikes.

Bedrock Influence

There is no indication that the orientation, abundance, or density of dikes are influenced by joint patterns in the underlying bedrock. Dike injection appears entirely governed by stresses created by megafloods, not from bedrock structure. Columnar joints in the Columbia River Basalt are not mimicked by dikes contained in the overlying flood sediments. Identical dike networks develop atop columnar basalt and unjointed bedrock alike. Where dikes do penetrate the bedrock, however, they follow existing weaknesses - joints, faults, and margins of pillows. Similar behavior is seen outside the floodway, in sand dikes at Fort Rock Lake, OR (Allison, 1953; 1979), for example.

Source of Fill

Sediments that fill the dikes is sourced from Missoula flood deposits. The character of fills is governed by flood energy, therefore resembles local flood facies (slackwater rhythmites, coarse laminated sands, bar gravels). I concur with Lupher (1944), who visualized "currents above the fissures" sweeping sediments into fractures, creating dikes with stratified fills in a range of grainsizes. He also found that "many dikes are traceable to overlying current-bedded sand." While most dikes formed during vigorous flooding, some are sourced in soupy lake-bottom sediments deposited during slackwater stillstands. Dikes originate from the base, middle, and top of floodbeds, positions which correspond to the stages of rhythmite deposition described by Baker (1973) and Smith (1993).

Fills Reflect the Local Geology

The number of clastic dikes in the Touchet Beds far outnumbers those in any other formation. However, a significant number of dikes penetrate sediments of the Miocene Ellensburg/Latah Formation, Miocene-Pliocene Ringold Formation, Miocene-Pliocene Dalles Group (Selah, Chenoweth, etc.), early-middle Pleistocene Palouse Loess (L2, L3), pre-Late Wisconsin gravels, and a Pliocene-Pleistocene fanglomerate-calcrete-loess complex known as the 'Cold Creek unit'. Dike fills contain chips of the local bedrock, grains from sedimentary cover units, and chunks of cemented paleosols in addition to substantial suspended-load material carried from areas upstream. Grains of quartz, plagioclase, and muscovite in sand and silt size fractions are most common. Courser fills are typically dominated by basaltic sand and clasts of mixed lithology. Near the western margin of the floodway, where the quartzite-bearing conglomerates crop out, dike fills contain quartzite clasts. Similarly, along the floodway's northern margin, the dikes contain gruss shed from granitic batholiths in the Okanogan Highlands. Pebbles of felsic volcanics, weathered mafics, various granite rocks, and schist (so-called 'exotic' lithologies) are regularly found in gravelly deposits and dikes. Chunks of cemented loess, calcrete, and cemented cicada burrow 'clasts' are common along the eastern margin of the floodway, where thick deposits of old loess reside. In protected slackwater valleys filled with stacks of slackwater rhythmites, Touchet Bed sediment and loess comprise dike fills.

Do Dikes Parallel Drainages?

Orientation data collected in southeastern Washington reveal some parallelism between dikes and modern stream courses exists (Silver and Pogue, 2002). While sets of stream-parallel dikes can occur locally, other sets in the same outcrops crosscut the trend and cast doubt on the notion that dike orientation is controlled by the orientation of the local stream or the valley it occupies. Plots of strike for large numbers of dikes will always approach random (i.e., polygons).

Sheeting and Growth

Vertical sheeting developed as dikes widened and lengthened with advance of the crack tip. Sheeting defines three types of dikes: Single-fill, compound, and composite (Hayashi, 1966). Recognition of the three different types is important only because it helps establish whether a dike formed in a single event or grew over time through multiple events. Single-fill dikes contain a single wedge of sediment between two skin walls; a fracture that opened and filled once. Compound dikes contain two or more sheets with skin walls between. In compound dikes, multiple fractures opened and filled during a single diking event. Composite dikes contain multiple sheets injected during multiple diking events (i.e., reinjection over time). In composite dikes, new sheets of sediment intrude older dikes during successive events separated by hiatuses. Each new set of sheets is sourced from a different source with its own grainsize, composition, or sorting characteristics. Each successive crack-and-fill episode has the potential to create sheeted fills that contrast with those that came before.

Stacks Inside Sheets

The filling of vertical sheets occurred in pulses, resulting in the accumulation of coherent 'stacks' of sediment separated by thin silt breaks. These sub-horizontal to dipping skins represent depositional breaks between pulses of infilling. Grainsize and repose angle often change abruptly at silt breaks between stacks. The height of a stack may range from a few centimeters to a few meters.

Age

Field relationships constrain the age of all clastic dikes in the study area to the Pleistocene, between ~1.8 Ma to ~11 ka, the period coinciding with the growth of the Cordilleran Ice Sheet, filling of proglacial lakes, and outburst flooding across Eastern Washington and parts of Idaho, Oregon, Montana, and British Columbia. While a few cemented and truncated outliers exist, the vast majority formed during the last glacial, between 18–14 ka. Many dikes cut the Mount St. Helens Set S tephra (16 ka), but none the Mazama Ash (6.8 ka). No sheeted dikes are known to intrude Holocene alluvium in the study area. Except for crystalline bedrock of the Okanogan Highlands and Belt quartzite in steptoes of the eastern Palouse, the dikes penetrate all formations scoured by megafloods.

Growth of composite dikes. Diking triggered by fracture, injection, and filling during three successive flood events. Since each individual sheet tapers downward, the composite dike is wedge-shaped. Figure modified from LeHeron and Etienne (2005). Similar sheeted dikes described in subglacial setting by Ravier and others (2015).

Compound, Composite & Single-fill Dikes. Single-fill dikes involve the one-time filling of a fracture. Compound dikes contain two or more sheets that were filled during a single diking event. Composite dikes contain multiple sheets injected during multiple diking events separated by hiatuses. For the purposes of this article, diking events = megaflood loading events. Composite dikes form by reinjection of new sheets into older dikes and by the merging of subparallel dikes to form one. For example, if you were only to observe the upper half of the two dikes on the left (green and yellow), you might identify a single dike. In reality there are two dikes that have merged over a portion of their length. Numbers indicate the relative age of injection. Crosscutting relationships reveal which dike is the oldest ("1", cyan) and which is the youngest ("3", yellow), but the three dikes labeled "2" (green, red, purple) do not touch each other, so their relative ages are assumed to be the same until additional information from outside the frame of the photo can be acquired. Ash Hollow Quarry, WA. S.W. Cooley photo and figure.

Stacks are a high resolution record of dike infilling. I measured the height of 20 coherent "stacks" in dikes exposed near Touchet, WA. Each stack constitutes a portion of a sheet, therefore each sheet is composed of many stacks. Planar or cup-shaped silt breaks separate each stack from those above and below. Abrupt changes in grainsize and bedding repose angle commonly occur at the silt breaks. Stacks are interpreted as discrete pulses of sediment entering a fracture and represent increments of filling as new space is created with advance of the crack tip. The height of most stacks measured less than a meter, but one example exceeded 3 m. Stack height appears randomly distributed.

Descend and branch. Sheeted dikes cut downward through sandy Missoula Flood deposits at Latah Creek west of Spokane, WA. Small downward-terminating spurs mimic the form of the larger trunk dike. Downward pinchouts and the lack of a connection to a liquefied source bed at depth - at Latah Creek or anywhere else in the Scablands - help to rule out an origin involving upward fluid escape (i.e., sand blows). The geometry of the sheeted dikes is consistent with pressurized injection, not brittle fracture or open-standing cracks formed by lateral spreading or desiccation. S.W. Cooley photo.

Arris and aperture. A clastic dike is a sediment-filled fracture with a 3D shape loosely resembling an axe blade. Hydraulic fractures have a width (aperture), a volume, and a curved, irregular leading edge (arris). They thin in the directions they propagate (downward and outward). A dike's cross section changes depending on where the section line is drawn, that is where a dike intersects the plane of an outcrop. The figure above is a simplified version of a hydraulic fracture that tapers both vertically and horizontally, as the clastic dikes in this study do. Its 2D cross section appears to taper downward if sliced at Plane C, upward if sliced at at Plane B, and both upward and downward if sliced at at Plane A. The apparent taper direction (interpreted propagation direction) can vary depending on where a dike intersects an arbitrary plane of section (A,B,C) or a bladed roadcut.

Tapered at both ends. The injection direction of dikes that taper in two directions can be confusing. Upward or downward? Recall that in cross section a dike's shape is determined by the intersection of the axe blade-like arris and the plane of the outcrop. What we see here is the curved leading edge poking through the cutface. If you were to excavate back into the outcrop, the dike would grow in length and likely width, becoming wedge-shaped and connecting to a larger trunk. S.W. Cooley photo.

Pleistocene dikes in older sandstone. Energetic flooding down the Columbia Gorge incised deep, v-shaped gullies into partially-lithified sandstones of the Chenoweth Fm (Dalles Group) and filled them with gravel. Floodwaters at The Dalles filled the lower 7 km of Chenoweth Creek valley to an elevation of 340 m (O'Connor and others, 2020). The photo shows a gravel-filled, parallel-sided dike exiting the bottom of a flood-cut gully and descending below the road grade. Hundreds of flood-deposited boulders mantle the bench-like surface above the roadcut. Quartzite and other non-basaltic clasts are abundant in the fills and present in the Chenoweth exposed at nearby Dry Hollow and Signal Hill (Tolan and others, 1996). Like many locations along the floodway, dike fills at Chenoweth Creek contain both material quarried from the local bedrock and material transported from areas upstream. S.W. Cooley photo.

Depth of Intrusion

Numerous examples of dikes intruded to depths exceeding 10 m are found in the study area. Borehole logs in the Hanford Plain, where the thickest flood deposits in the Channeled Scabland are found, intersect dikes at depths greater than 50 m. At most sites, dikes pinch out 1-10 m below the surface, which scales with the thickness of flood deposits in most places.

Rate of Injection

Both field evidence and theoretical studies suggest individual clastic dikes form very rapidly, on the order of seconds. Fracture dynamics controls the rate, not the infilling. Fractures open, grow, and fill quickly and do so in jumps. 1-4 m/sec seems a reasonable estimate for the rate for advance of the fracture tip through unconsolidated sediment. Similar rates for the propagation of hydraulic fractures through weak sediments were found by Levi and others (2011) and Davies and others (2012). Diking events appear to last sufficiently long to allow compound dikes to propagate and crosscut one another. A reasonable estimate for the length of a single diking event - the period when flood-load forces exceed substrate resistance - is a few days to a couple weeks.

Skin Walls

Thin silt partitions (skin walls) form the outer boundaries of the dikes and separate internal sheets from one another. Skin walls typically achieve thicknesses of 1–10 mm, sufficient to form an effective seal. A silt skin builds rapidly, progressively thickening into a continuous layer, as pore water migrates out of the fill, through the fracture wall, and into the comparatively drier surrounding material. Leakoff to the formation appears to be the primary dewatering mechanism of Touchet-type dikes. Dikes that penetrate impermeable bedrock, most often basalt, lack outer skin walls but contain internal ones, indicating leakage diverts laterally into adjacent sheets, where it resides temporarily before draining to the formation via fractures. Mostly silt and variable amounts of clay comprise the skins. Muddy sediment that forms dike walls is not injected first, as suspected in some till dikes (Phillips and others, 2013), rather the fines are the finest constituent of the entering slurry that is quickly screened against the walls. Skin building and fracture sealing begin immediately. An analogous process occurs during the formation of concrete slurry walls used in heavy construction, such as trench-type building foundations.

Flutes on Skin Walls

Flutes are created by erosive scour. Their asymmetric shape precisely records the direction of flow. Upward-pointing flute casts ornament the interior faces of skin walls. The flutes adorn the inside surfaces of the outer walls and all faces of interior walls. Fluting unambiguously indicates sediment entered the fractures from the top. The preserved fluting is not subtle or sparse; the forms are obvious and present in nearly all dikes.

Rip-up Fragments in Fills

Fragments of older fills, chips of skin walls, and broken bits of the host material are a significant component of many dike fills. In places where dikes cut white tephras such as the Mount St. Helens Set S tephra, traces of ash are sometimes visible in the fill.

Slickensides

Slickensides, while often observed in dikes from other regions, are generally not present in Touchet-type dikes. Preservation of slicks in unconsolidated sediment is uncommon. In the few cases where I have seen slickensides, the dikes were cemented with CaCO3.

Stratified Fills

Cross-lamination is a conspicuous characteristic of most dike fills, but especially coarse sandy fills. As a turbulent sediment slurry enters an open fracture, it settles and stratifies almost immediately. Stratification develops under Newtonian flow during the brief period when the fracture is open and hydraulically connected to the surface (voluminous water-sediment reservoir). Laminated fills accumulate in open or widening fractures, whereas structureless fills likely record the instantaneous closure of the fracture and the resulting “freezing” of unsettled sediment. Abrupt termination of intrusion following a drop in fluid pressure has also been documented in sand injectites associated with deep-sea fans (Jonk et al., 2010; Dodd et al., 2020). Sheets filled with silt commonly lack clear stratification. But what can appear structureless at first glance often turns out to be subtly laminated after some light work with a soft brush. I disagree with Lupher's interpretation of structureless fills were deposited by wind into open-standing cracks without the aid of water (Lupher 1940, 1944). The field evidence simply does not support the filling of open surface cracks by dry sediment.

Leakoff Halos

Leakoff halos are firm and discolored zones that extend a few centimeters beyond the outer walls of some dikes. Leakoff halos are a product of dewatering consisting of fines that were not screened at the dike wall.

Lumpkins

Lumpkins are bulbous forms on the exterior walls of some dikes formed by leakoff. Leakoff indicates the host sediment during diking was ice-free and drier than fracture-filling slurries (i.e., vadose zone). The Lumpkins were a terrible band from Bellingham.

What causes a dike to terminate?

There are several ways to halt the growth of hydraulic fractures, thus the growth of clastic dikes. Lost of fluid pressure via leakoff to the formation will do it (i.e., fluid pressure, Pf < confining pressure, O3). Barriers and strongly contrasting layers with a high Young’s modulus or strength greater than a fracture's ability to break it can too (i.e., cemented sediment, frozen sediment, hard bedrock). Weak or permeable bedding planes can cause a fracture to branch or become a sill. Increased depth, a proxy for confining pressure, can terminate vertical advance. Reduced injection rate at the top of the fracture can also cause a fracture to close. Rate reduction may result from shallower water (less head pressure) or completed filling of the fracture with sediment.

Skin wall with flutes. Thin, continuous silt partitions form both the outer walls of dikes and inner partitions between infills. S.W. Cooley photo.

Fluted walls of mostly silt. Flute casts decorate remnants of skin walls. Touchet Valley, WA. S.W. Cooley photo.

Flute casts. Flute casts with upward-pointing noses decorate the interior surfaces of skin walls. The flutes are an unambiguous directional indicator: Sediment entered from the top. Identical fluting is present in nearly all dikes in the study area. Quarter for scale. S.W. Cooley photo.

Injection and leakoff. Initiation of a hydraulic fracture establishes a temporary pressure gradient between a voluminous source of sediment-laden water and the crack tip. The combination of tip advance and infilling is how dikes grow. Its a jumpy, incremental process. As pressure rises inside a fracture, it opens, its volume increases, and the new space is rapidly filled by a slurry of sediment (natural proppant). Leakoff to the formation beings immediately, plastering silt and clay against the fracture walls. Repeated pressure cycling, fracturing, filling, and dewatering creates the conspicuous vertically-sheeted fills. Figure modified from Phillips and others (2013).

Leakoff halo. A firm, slightly discolored halo is the result of water and fines transported through the dike wall as the dike dewaters. Slight cementation is present in the halos of some dikes. While the migration of water and fines occurs rapidly during diking, cementation is a diagenetic process and takes longer. Hwy 24 near crest of Yakima Ridge. S.W. Cooley photo.

Leakoff lumpkins. The outer surface of this dike's skin wall is adorned with bulbous structures formed by leakoff from the sandy fill into the drier surrounding sediment. These lumpkins look very much like tiny load casts, but occur here on vertical skin walls rather than horizontal bedding contacts, which is more common. Walla Walla Valley, WA. S.W. Cooley photo.

Party like a lumpkin. Bulbous forms on delicate skin walls, here larger in size than in the previous photo. They are easily damaged during excavation, even with a soft brush. White Bluffs, WA. S.W. Cooley photo.

Truncated tops. Dikes in the Channeled Scablands intrude more than a dozen geologic units. Dike tops are commonly truncated by bedding contacts, low-angle slide planes, and local unconformities. Dikes that penetrate non-flood sediments or bedrock are invariably sourced in overlying flood sediments. The figure above highlights relationships between dikes and the stratigraphy in the study area. (A) Five Walla Walla Valley sites from Spencer and Jaffee (2002), (B) Lind Coulee site from Daugherty (1956), (C) Moxee Mammoth site from Lillquist and others (2005), (D) Hanford's FMEF site from Bjornstad and others (1990), (E) Rulo site from Bader and others (2016). A = Alluvium, C = Colluvium, CRB = Columbia River Basalt, DIA = Silt diamict, EG = Exotic-clast bearing gravel, FG = Fanglomerate/Alluvial fan gravel, L = Loess, P = Paleosol, S = Sandy, SCR = Silt-clay rhythmites, TB = Touchet Beds/Hanford Fm.

Dikes in Faulted Sediments

In general, the dikes are not clustered along prominent joint systems or otherwise controlled by bedrock faults. Dikes are found in major fault zones, but in fewer numbers than in areas well away from faults. More often, dikes and local faults interact at the meter scale. Small normal faults that cut unconsolidated flood deposits likewise do not act as a primary control on diking. The dikes far outnumber faults and dip more steeply. Single-fill and thin compound dikes most often cut cleanly through bedded sediments, pioneering new pathways. Where dikes and faults interact, dikes cut faults, faults cut dikes, and dikes follow portions of faults. It is not unusual to find all three relationships in a single outcrop. Rotational slumps in slackwater sediments and in bedrock do seem to influence dike location, size, orientation, and number, at least locally. Bedding-parallel slip is common in slackwater sections, although it is often quite subtle. Small slips (<1 m offsets) that repeat in successive beds produce a stairstepping offset pattern in vertical dikes. Larger slips (>10m offsets) may shift entire packages of strata laterally and truncate the dikes. It is usually possibly to locate and match up the offset portions. Thrust faults are uncommon in scabland deposits and their influence on diking is minor.

Dikes Near Mapped Bedrock Faults

Eastern Washington is criss-crossed by more than a dozen mapped thrust faults associated with the Yakima Fold Belt and other systems such as the Hite Fault (Schuster and others, 1997). Sheeted clastic dikes sourced in glacial-age material intrude folded and faulted basalt flows and interbeds at Umapine, Touchet, Rattlesnake Hills, Horse Heaven Hills, Alder Ridge, Gable Mountain, Cecil, and elsewhere. I disagree with studies, notably Camp and others (2017, Fig. 50) and Reidel and others (2021, Fig 8.), which suggest a causal link between faults and dikes exists. I have traversed most of the sedimentary sections preserved atop the fault-bounded ridges, finding fewer and smaller dikes directly atop Quaternary faults than in sections located many kilometers from them. For example, I observed no sheeted dikes in the 3-25 m-thick Plio-Pleistocene section preserved at the crest of the Saddle Mountains anticline. Similarly, the gullied and fault-bounded Smyrna Bench section contains only a few small dikes, all below the elevation of Missoula flooding. I surveyed 20 km of the 150 m-thick section of Ringold Fm at White Bluffs, finding a few clusters of dikes in sandy Missoula flood deposits and a few isolated dikes in the Ringold. The White Bluffs lie directly along the strike of the Gable Mountain fault. I found very few dikes in scattered exposures at Frenchman Hills and neighboring Royal Slope. Plio-Pleistocene sediments exposed in lower Lind Coulee and in fault trenches near O'Sullivan Dam (West and Shaffer, 1988) are devoid of dikes, though a set of thin dikes cuts a hard, white bed in the Ringold at Othello Canal. No dikes were observed in a 10 m-thick Miocene diatomite exposed near the Frenchman Hills fault and in an active surface mine off the Beverley-Burke Road. No dikes were observed in a 4 m-thick interbed (Miocene overbank deposits) cut by the Arlington-Shutler Butte Fault exposed west of Arlington, OR (more about the ASBF HERE).

Water Table Position During Inter-flood Periods

The presence of silt skins, leakoff halos, rodent burrows, numerous brittle fractures, and minimal wetland soil indicators in the Touchet Bed suggest that a thick, well-drained, and ice-free vadose zone was reestablished each time floodwaters drained from the landscape. The field evidence hints at a wet-over-dry-over-wet condition that promoted brittle fracture in the subsurface. Brittle fractures initiated between the base of the flood and top of the water table served as entry points for hydraulic injection through the vadose zone. I believe the Pleistocene water table fell during inter-flood periods, returning to a position approximating the modern water table. Streambanks then and now reside well below the flat tops of benches composed in modern valleys.

Pristine Preservation

While the Touchet Beds contain clear evidence of recolonization by plants, rodents, insects, and mammoth during the decades-long periods between megafloods (Waitt, 1985; Spencer, 1989; O'Geen and Busacca, 2001; McEachern and others, 2013; Last and Rittenour, 2021), bioturbation was modest overall. Original bedding, often quite fragile, remains intact and pristine nearly everywhere, a strong argument against wholesale reworking. Fine laminae and thin silt walls are nearly always perfectly preserved in dike fills. In contrast to the Palouse loess and most Holocene alluvium, thorough mixing by organisms is rarely found in Scabland deposits.

Subsurface Imaging of Dikes

A few attempts to image the subsurface geometry of individual dikes and large networks using ground penetrating radar (GPR) have been made at the Hanford Site (Murray et al., 2001; Williams and others, 2002; Clement and Murray, 2003; Ward and Gee, 2003; Ward et al., 2006). Radar images prove that the general shape of dikes >1 m wide can be resolved to a depth of about five meters. GPR images show no widening at depth and no connection to buried feeder beds (i.e., dikes are not fed from below).

Dikes as Fast Fluid Conduits?

Sandstone dikes have long been recognized as fast migration conduits for oil, natural gas, and ore-bearing fluids (Murchison, 1827; Rickard, 1903; Anderson and Pack, 1915; Jenkins, 1930; Braccini et al., 2008). In fact, sand injectites are targeted by drillers in offshore oil fields specifically for their high permeability. Likewise, in hard rock mining districts (i.e., Colorado Mineral Belt), sedimentary dikes that often parallel bedrock fracture systems are known to concentrate economic minerals such as Au, Ag, Pb, Cu, Zn. Despite the well-established conduit behavior of sand dikes in shallow terrigenous sediments, offshore fan-turbidites complexes, and bedrock mining districts worldwide, Hanford scientists think that the thousands of sand, silt, and gravel-filled dikes beneath the Hanford Nuclear Site function differently. According to Murray and others (2007), the dikes do not quicken vertical transport of chemical waste fluids leaked from storage tanks into the deeper subsurface and actually retard lateral flow in the vadose zone. Unsurprisingly, the findings of the Murray study contrast with reports by scientists at other U.S. Department of Energy facilities, the rest of the geologic literature, physics, and common sense (Reichert and Fenimore, 1964; Pollard and Aydin, 1988; Finfrock, 1994; Caggiano, 1996; USDOE, 1996; Faybishenko et al., 2000; Serne et al., 2002, 2004, 2020; Bjornstad and Lanigan, 2007; Fang and Mayes, 2007; Gee et al., 2007; Reidel and Chamness, 2007; Rockhold et al., 2015; Springer et al., 2017). The lightning just ain't distributed right.

Touchet-type dikes are slender, sheeted, and wedge-shaped. Each Missoula flood rhythmite, labeled R1 through R7, corresponds to a distinct flood event. A clastic dike that extends downward through the section originates at the base of the youngest bed, R7. Nearby dikes also descend through the sequence in a similar manner. Notably, the dike cuts a clean path through the host sediment without following preexisting fracture sets or rubbly zones between laterally displaced blocks. Bedding contacts are neither offset nor tilted into the dike, and there is no evidence of a low-angle sliding surface in the outcrop. The stack of Touchet Beds remains in its original position atop the basalt bedrock, which is exposed at the base of the section. The sediment filling the dike was supplied from above, rather than from a liquefied layer below R1. The dike does not feed a sand blow (i.e., Obermeier, 1998). Both the top and bottom of the dike are clearly visible. The dike starts at the base of R7, widening gradually from a small sag. Bedding in R7 smoothly grades upward from the sag, suggesting that diking occurred early in the deposition of this bed. Both branches of the dike taper to a point. This dike is representative of thousands of others found throughout the megaflood region. Burlingame Canyon, Walla Walla Valley near Gardena, WA.

Small dikes reveal early stages of injection. Diking is sometimes interrupted early, resulting in thin, short, single-fill dikes. These modest features provide a clearer picture of the initial stages of the fracture-and-fill process than do larger, more impressive examples. Here, two small dikes descend from the coarse-grained base of a Missoula flood rhythmite. Both dikes propagated downward and were filled from above. No visible crack, root cast, or other flaw appears to control where either initiated. No matter what range of clasts are present in the bed above, clasts in the fill are limited in size by the width of the dike's opening (fracture aperture). Since both dikes originate at the base of the bed, we can assume injection was triggered by the initial surge of water up the valley. The dike on the left strikes obliquely to the outcrop face and appears wider than the one on the right, which strikes perpendicular to the face. Both dikes are approximately 3 cm wide. Tucannon River Valley near Starbuck, WA. S.W. Cooley photo.

Take a second look. When a dike is oriented in a plane similar to that of an outcrop, things can look weird. This dike is a good example. Its not actually folded, but straight as an arrow. The strike and dip of the dike and the cutface are almost the same, but the face of the outcrop is irregular. So when it intersects the planar dike, an illusion of folding is created. A trick of 3D geometry that your eye will begin to identify with time in the field. Its the second look, the considered observation that will more often yield a correct interpretation. Hwy 397 at Finley, WA.

Warden Canal. Not all dikes cut from top to bottom. Several beds overlie the top of this truncated dike near Warden, WA. A marked change in the character of the sediments above the erosion surface reflect a change in the water table. Sediments below the truncation surface show evidence of complete drainage between floods. Sediments above indicate a wetland was abruptly established and persisted until the Holocene. Appears to be a local phenomenon perhaps involving changes in Lind Coulee. S.W. Cooley photo.

Warden Canal stratigraphy. About 10 sandy rhythmites are exposed along a canal near Warden, WA. The exposure contains many important structures and relationships. At nearly every bedding contact, sags, load casts, contorted bedding, and sets of small wedge-shaped dikes are present. The repetitive features suggest they formed during the deposition of successive beds. A large clastic dike descends through the outcrop is truncated at its top by a prominent erosional surface, above which lies a wetland soil. Apparently, a high water table was established after deposition of bed R8. Above the truncation surface is a conspicuous 5 cm-thick gray layer, likely reworked volcanic ash, which is also present in other outcrops nearby. Near the top of the exposure, at least two wetland units are deformed. The overlying gravel, likely deposited by a late flood, appears to have swept across the wetland, causing the saturated sediment to liquefy. If the t-shaped mudsquirts and dish structures in the upper beds are interpreted as seismites, one must explain what appear to be structures formed by rapid loading and sedimentation lower down. The two different styles of soft sediment deformation are best explained by the same trigger, repeated overriding floods, rather than by two different triggers in upper and lower beds.

Injection during flooding. My conceptual model for sheeted clastic dikes in the megaflood region is consistent with relationships observed in the field. Downward injection occurred mostly during overland flood events and occasionally as slackwater lakes drained. Flood loads fractured the substrate allowing sediment circulating in currents at the base of the flood (or soupy lake bottom) to fill the fractures, forming dikes.

Flood injectites vs. sand blows. (A) The sketch illustrates differences between clastic dikes formed by liquefaction (sand blows and fluid escape structures) and those formed by floodwater loading and hydrofracture (flood injectites). Liquefaction dikes propagate upward and are sourced in wet, sandy beds deposited sometime in the past and remobilized by strong shaking. Liquefaction often produces feeder dikes that vent to the surface as sand blows (volcanic edifices of sand). Flood injectites are sediment-filled filled hydrofractures that propagate downward from the surface. The fractures are immediately filled with sediment sourced in circulating bottom currents of glacial floods. Liquefaction dikes in A cut younger strata and are filled with older sediment. Injection dikes in B cut older strata and are filled with younger sediment. (B) My dike-fill generations concept sketch explains the formation of sheeted clastic dikes in aggrading flood sediments (each bed = one flood). The four geometries represent the range of forms found in the study area: a). Unsheeted - Single-fill, b). Sheeted - Multi-fill Compound, c). Sheeted - Single-fill Composite, d). Sheeted - Multi-fill Composite. Compound = Multiple fill bands injected during a single event. Composite = Multiple fill bands injected during two or more events separated in time.

Dike abundance, shape, and grainsize. Sediment porosity (%) and permeability (md), predict dike shape. The two end members are silt and gravel. (A) The tightness of the formation, largely a function of grainsize in young Scabland deposits, determines whether pore fluid pressures will build or disperse and whether slender or stubby dikes will form. Silty mixtures are tight and tend to fracture when stressed (i.e., pore fluids escape in fractures). Gravelly mixtures accommodate the same stress via matrix flow (i.e., pore fluids flush through the interconnected pore space). (B) Diking can occur in all phases of flooding, but dike fills are nearly always sourced at the flooded surface. Coarser fills correspond with coarser source material typically at the base of beds deposited during the initial flood rush. Finer fills correspond with lower energy deposits that came slightly later, including sands and slackwater silts. If a flood carries only silty material, then dikes will contain silty fills (i.e., northern Walla Valley and Skyrocket Hills). Dikes formed earlier will generally contain coarser fills. (C) Slender, sheeted dikes correspond with higher silt content beds, slackwater deposition, and protected valley settings. Stubby, crudely-sheeted, stubby dikes form in coarse sand, laminated sand, and gravelly bar deposits sourced at the base of beds and inside coulees.

Dike descends from base of bed. This dike originates within the coarse-grained base of a Touchet Bed. It cuts downward through the silty top of the bed below, which was deposited by an earlier flood or possibly a pulse within the same flood. The dike formed early in the flood event and is associated with flame structures of silt. Note how the coarse sand of the new flood invades the top of the underlying bed, pulling thin layers of tan silt away and deforming them slightly into flames. Both the flames and the dike are products of water loading and rapid deposition during a flood. Tucannon Valley, WA. S.W. Cooley photo.

Dike descends from top of bed. This dike originates within the silty, fine-grained top of a Missoula flood rhythmite and cuts downward through sandy material below. The dike formed late in the flood event. Tucannon Valley, WA. S.W. Cooley photo.

Stubby dikes. As grainsize increases, length-to-width ratio decreases. Dikes become shorter and fatter in coarse, laminated sands and gravelly deposits. Smith Coulee, WA. S.W. Cooley photo.

Detail of stubby dike. Base of dike terminates abruptly. There is no bedding-parallel offset as is seen in other locations. Smith Coulee, WA. S.W. Cooley photo.

Stubby wubby. Short, fat dikes in gravelly flood beds at Starbuck, WA. The presence of silt (tan bed) allows fractures forming in the permeable material to seal, but the injection momentum fades quickly as the silt runs out and the fluid is lost to the formation. S.W. Cooley photo.

Did a dry vadose zone play a role? A dry vadose zone sandwiched between the base of an overland flood and the water table may have created a wet-over-dry-over-wet situation that facilitated hydraulic fracture. Fractures initiated at the ground surface (top of the dry interval) propagated in dry sediment before reaching the water table. The water table must have risen during each flood, temporarily eliminating the dry gap before returning to a lower level as floodwaters drained.

Clean breaks and discrete deformation. Deformation associated with diking does not extend beyond the dike wall. Material surrounding the dikes has been laterally extended slightly to accommodate the fill, but otherwise appears undeformed. Clear bedding contacts and bedforms can be traced undisturbed across each dike. Here, dikes filled with gray Touchet Bed sediment intrude an oxidized fluvial sandstone of the Miocene Ellensburg Fm. Snipes Mountain near Granger, WA. S.W. Cooley photo.

Clean cuts. A sand-filled dike cuts cleanly across several silt-sand rhythmites at Starbuck, WA. Most dikes pioneer new pathways through the host sediment; they rarely follow faults or existing flaws (frost cracks, plant roots, etc.). Also, compaction is minimal in most Touchet Bed sections. Dike walls remain straight and are not crenulated. Circular back-filled rodent burrows show only a minor degree of flattening in most places. S.W. Cooley photo.

Touchet dikes intrude an older fanglomerate-calcrete-loess complex. A sheeted dike sourced in late WisconsinTouchet Beds cuts the 'Cold Creek unit', a Pliocene-Pleistocene complex of cemented loess, calcrete, and weathered fanglomerate. Hwy 397 near Finley, WA. S.W. Cooley photo.

Pleistocene dikes in Miocene basalt. Sheeted sand- and silt-filled dikes intrude Columbia River Basalt in various locations. The dikes are sourced in overlying flood deposits and exploit joints and other weaknesses in the bedrock. A.) Weaver Pit near Gardena, Walla Walla Valley, B.) Hwy 12 at Alpowa Creek, Lewiston Basin, C.) Hwy 14 near Alderdale, Umatilla Basin. S.W. Cooley photos.

Flood gravel fills dike in Tertiary sandstone. Gravelly material sourced from above fills a clastic dike in tuffaceous sandstone of the Miocene-Pliocene Chenoweth Fm at The Dalles, OR. S.W. Cooley photo.

Pleistocene dikes in Ellensburg sandstone. A silt-sand dike cuts crossbedded fluvial sandstone of the Ellensburg Fm (Latah Fm) at West Foster Creek near Bridgeport, WA. The location was overridden by outburst floods and glacial ice of the Okanogan Lobe. S.W. Cooley photo.

Touchet dikes intrude fractured basalt. A sheeted sand dike fed from above cuts Columbia River Basalt at Prosser, WA. Sediment filled a widening a joint in the unstable cliff (incipient block topple) during one or more floods. S.W. Cooley photo.

Early to Middle Pleistocene dikes cut Pliocene Ringold Fm. Cemented, downward-tapering dikes are sourced in gravelly, deformed, ancient flood deposits (pre-Missoula) at Ringold Road, WA. S.W. Cooley photo.

Touchet dike cuts Pliocene fan gravel. Gray dike sourced in unconsolidated flood-laid sediment cuts older, reddened fanglomerate shed from the north flank of the Saddle Mountains anticline. Smyrna Bench, WA. S.W. Cooley photo.

Liquefaction in Eastern Washington?

To date, no significant liquefaction has been identified in more than a dozen trenches excavated across young fault scarps in the Channeled Scabland. Trenching reports are listed here and briefly reviewed below.

• Ahtanum Ridge-Burbank trench near Yakima, WA (Bennett and others, 2016).

• Horned Lizard trench in the western Boylston Mountains, WA (Barnett and others, 2013).

• Toppenish Ridge trenches above Pumphouse Rd, WA (Campbell & Bentley, 1981; Campbell & Repasky, 1995; Repasky and others, 1998).

• Ahtanum Ridge (Repasky and others, 1998)

• Wenas Valley trench, WA (Sherrod and others, 2013).

• Saddle Mountains trenches at Smyrna Bench, WA (Bingham and others, 1970, Plates 4,5,6).

• Buroker roadcut southeast of Walla Walla, WA (Farooqui and Thoms, 1980; Foundation Sciences, 1980).

• Lower Lind Coulee trenches east of O'Sullivan Dam, WA (GEI/West & Shaffer, 1988).

• Gable Mountain trenches at the Hanford Site, WA (Bingham and others, 1970; Golder Associates/PSPL, 1982).

• Spencer Canyon trench near Entiat, WA (Sherrod and others, 2015).

• Finley Quarry west of Wallula Gap, WA (Sherrod and others, 2016; Coppersmith and others, 2014).

• Starthistle trench east of Wallula Gap, WA (Angster and others, 2020, 2023; Mahan and others, 2022).

• Kittitas Valley trench, WA (Huddleston, 2022; Dr. Walter Szeliga, personal and written communications, 2023).

• Gate Creek trench near The Dalles, OR (Bennett and others, 2021; Madin and others, 2021).

• Gales Creek trenches near Portland, OR (Horst and others, 2021; Redwine and others, 2017).

Trenched faults in the Channeled Scabland. Trenches excavated through young fault scarps east of the Cascades have revealed no record of liquefaction and have established no connection between the thousands of Pleistocene clastic dikes and movements of Quaternary faults. The map shows locations of paleoseismic trenches opened by USGS, WGS, DOGAMI, and USBOR. In all but one (Spencer Canyon), the trenched scarps were formed by Yakima Fold Belt deformation of Miocene basalts and cover sediments. The Spencer Canyon trench lies outside the fold belt in older crystalline rocks of the Cascade Range. Deformation there occurs along a different fault system perhaps responding to different tectonic stresses. The Gales Creek trench near Portland is not shown on the map. Basemap by Czajkowski and Bowman (2014).

Liquefaction rarely reported in the Columbia Basin. Just 2 of 107 studies that examined clastic dikes found evidence of liquefaction. The two studies, both on the Wallula Fault Zone, were authored by employees from the same USGS office in Seattle (Sherrod and others, 2016; Angster and others, 2023).

Finley Quarry - Two investigations claim to have found liquefaction features in sediments near Wallula, WA (Wallula Fault Zone). The first is Finley Quarry located west of Wallula Gap (Sherrod and others, 2016). Several conspicuous vertical, rubble-filled fractures were initially investigated by Kienle (Foundation Sciences, 1980). In an accompanying report, Farooquoi and Thoms (1980) observed "thin clastic dikes of sand" and "clastic dikes of very light terracotta-colored silt" intruding zones of fault breccia in the older basalt. Rockwell Hanford Operations also opened a 310 m-long trench in 1977, finding sheared and brecciated basalt, but no conclusive evidence of deformed Quaternary deposits (Jones and Fecht, 1977; Gardner, 1977). Reinvestigation of faulting at the quarry by USGS (Sherrod and others, 2016) lead to discovery of an old clastic dike cutting two beds of silt-pebble diamict. Sherrod misinterpreted it as a liquefaction feature formed in response to strong shaking. The beds hosting the dike are silty flood deposits identical to those exposed nearby in large roadcuts along Hwy 397 (i.e., reworked loess not wind-deposited silt). A team of co-investigators disputed the interpretation in a separate report (Coppersmith and others, 2014).

Same same. When a clastic dike is oriented similar to the plane of the outcrop, it can appear disaggregated. This is an artifact of geometry and quite common. Dikes are not perfectly planar; they are a bit wrinkled. The top sketch is of a dike observed at Finley Quarry by Brian Sherrod. The dike crosscuts a few layers and has two branches. The bottom photograph shows a similarly-branched dike from a different outcrop. Sherrod attributes his Finley Quarry feature to liquefaction. His sketch implies upward fluid escape where the two branches are becoming sills. Sherrod's interpretation is incorrect. This is a typical Touchet-type clastic dike identical to thousands of others in the region formed by hydraulic fracture and downward injection during megaflooding. The two branches are not sills, but lower-angle portions of the same crosscutting dike. Sherrod's error can be chalked up to unfamiliarity with these features and the sediments that host them.

Starthistle Trench - Angster and others (2020, 2023) reported finding liquefaction features in Holocene loess east of Wallula Gap. The set of small, blobby structures that occur within and below the modern soil profile post-date the 13 ka Glacier Peak G tephra. I note several problems with their trench log interpretations. a.) The surface lineament they targeted for trenching is actually an old ranch road, not a fault scarp. Remnants of old pavement are clear in the trench wall and the old road network appears in historic aerial photographs (see THIS POST). b.) Touchet Beds that underlie the "liquefied" loess remain undeformed, a logical flaw that invalidates their argument. c.) A windblown origin for beds at Starthistle interpreted as "loess" may be incorrect; the bed might be a water-laid silt (loess reworked by water). d.) The blobby features themselves are anomalous, yet they bear no resemblance to the "liquefaction dike" just across the river at Finley Quarry (Sherrod and others, 2016) or features associated with thousands of sand blows described near New Madrid, MO (i.e., Obermeier, 1990, Fig. 4 and 22). Though numerous nearby roadcuts and streambanks expose the same sediments, none contain blobs. I believe the blobs are local, minor anomalies related to soil forming processes that have been misinterpreted by Angster as earthquake-induced liquefaction. They might krotovina, or features produced during dewatering after a heavy rainfall event, or any number of other things. Wild claims of "widespread liquefaction" caused by shaking along the Wallula Fault Zone (i.e., Mahan and others, 2022) are entirely unsupported by observations made at the Starthistle trench or Finley Quarry. Strange that the USGS team did not visit the nearby Wallula Bluffs (<3 km away), where repeated fluidization in successive Touchet Beds is well displayed. There, deformed and undeformed bedding (silty upper portions, sandy lower portions) alternate through a stack of about a dozen "late" rhythmites. Deformation is tied to grainsize and deposition during a certain flood stage, not seismic shaking (see Wallula Bluffs stratigraphy figure below).

Liquefaction in the Starthistle trench? Sketch of the Starthistle trench wall by Angster. Blue unit is undeformed Touchet Beds. Tan unit is a young loess. Orange blobby features were interpreted by the study authors as post-depositional liquefaction features formed by seismic shaking along the Wallula Fault Zone. I disagree. Liquefaction has not previously been reported in either unit anywhere in the region. Abundant outcrops nearby expose the same section, but contain no similar features.

Wallula Bluffs stratigraphy. Flat-lying and deformed strata alternate within each bed in a stack of about dozen "late" rhythmites at Wallula, WA. The deformation is partitioned within rhythmites by grainsize (sandy portions) and formed synchronous with flooding and sediment deposition. It was not caused by an earthquake at some later date. The entire stack is not deformed, as would be the case if shaking were the trigger. The sediments here are giddily susceptible to deformation. It is possible that these rhythmites, deposited low in the Columbia Valley, remained soupy and undrained during periods between floods. The firm, flat-lying, laminated silts (SFF) may represent long-term ponding rather than temporary slackwater deposition (bases not tops), but opinions differ on where to place the breaks between rhythmites in this section. Note that these are the smaller "late" rhythmites deposited by flows confined to the Columbia Valley first recognized in the 1980s by Richard Waitt and Brian Atwater and discussed in detail in a recent GSA Field Guide 62 (Waitt and others, 2021). A few sheeted clastic dikes cut the entire section and two small dikes descend from the middle of rhythmites R5 and R6. Section is 4 m tall. Cutbank located along the Columbia River near its confluence with the Walla Walla River (46.0681, -118.9098). Figure by S.W. Cooley, 2026.

Gable Mountain Trenches - Several trenches were opened across two thrust faults at Gable Mountain on the Hanford Site by Golder Associates (Bingham and others, 1970; Golder Associates/Puget Sound Power and Light, 1982). The South Fault displaces Miocene Pomona and Elephant Mountain basalts and the Rattlesnake Ridge interbed by about 15 m. Unfaulted Missoula flood deposits (Hanford Fm) overlie the fault. The Central Fault displaces the Rattlesnake Ridge by 55 m, but the Hanford Fm by only 6 cm (Reidel and others, 1992, p. 43-44). No liquefaction was found in trenches at Gable Mountain. A wedge-shaped clastic dike filled with flood-laid sediment exploits a weak zone of brecciated basalt (Trench log GT-2 in Reidel and others, 1992, Figure 39, p. 45). The Late Pleistocene dike post-dates the fault breccia.

Lind Coulee Trenches - Trenches were opened across the Lind Coulee Fault in the 1980s by consultant Michael West on behalf of dam managers at the U.S. Bureau of Reclamation (West and Shaffer, 1988; Shaffer and West, 1989; Geomatrix Consultants Inc., 1990). The fault, an eastern extension of the Frenchman Hills thrust, places Miocene Roza over Pleistocene Palouse loess. Trenches near the shoreline of Lower Crab Creek/O'Sullivan Reservoir revealed relationships similar to those at Gable Mountain, namely that clastic dike injection post-dated faulting. Initially, crews believed they had uncovered a colluvial wedge and Missoula flood sands "injected into the fault zone", but their interpretation changed as trenching proceeded. Ultimately, the GEI team found,

...no evidence of shearing, tectonic displacement or colluviation characteristic of surface fault rupture. The [flood-deposited] sands along the shear plane appear to have been injected hydraulically along the plane rather than dragged along it...the sand was injected hydraulically from the top...The last surface fault displacement, therefore, occurred before 40 to 50 Ka...Similar injection of flood sands along shear planes was noted in fault trenches excavated on Gable Mountain (DOE/Westinghouse, 1987b).

Toppenish Ridge Trenches - Four trenches were opened across splays of the Toppenish Ridge Fault by Ted Repasky and Newell Campbell in the 1990s (Campbell and others, 1995; Repasky and others, 1998). Trenching, GPR surveys, hammer seismic surveys, and age dating at the site were supported by USGS in partnership with the Yakama Indian Tribe. No liquefaction features or clastic dikes were observed in the trench walls. The latest movement on the fault was estimated at 500-700 years BP, consistent with previous reports (Campbell and Bentley, 1981). Seventeen kilometers along the ridge to the east, two gravel pits straddle the same fault, exposing both hanging wall and footwall strata. The upper pit (hanging wall) is located <200m from the fault at an elevation of 265-295 m. There, a thick, partially-indurated, quartzite-bearing Miocene conglomerate dips steeply south into the fault. No liquefaction features or clastic dikes were observed in the tilted beds. At the lower pit, located to the south and some 40 m lower, the quartzite-bearing Snipes Mountain Conglomerate is nearly flat-lying and capped by light gray Touchet Beds. Several light gray clastic dikes crosscut the darker conglomerate. Touchet Beds are the unambiguous source for all dikes exposed in pits off Tule Road. They are filled from above, pinch downward, and show no connection to a liquefied source below. The dikes post-date deposition of the conglomerate and most, if not all, of the tilting.

Sand dikes cut the underlying Snipes Mountain Conglomerate at Toppenish Ridge. Pleistocene Touchet Beds are the unambiguous source for clastic dikes that descend into Miocene gravels below. Diking appears related to flooding, not recent faulting or seismic shaking. Pit is located near Tule Rd on the Yakama Indian Reservation. S.W. Cooley photo.

Smyrna Bench Trenches - Trenches excavated through loess, basaltic fanglomerate, calcic paleosols, and sheared basalt along the north flank of the Saddle Mountains revealed no liquefaction features or clastic dikes (Bingham et al., 1970). A few conspicuous vertical features noted in field sketches are loess-filled tension cracks, not injection structures. The loess-filled openings are gaps between broken blocks of bedrock. According to project geologist John Bingham, "Some of these are filled with loess; others contain fragments derived from the walls of the crack. Several of the cracks show some stratigraphic offset, but no gouge zones or slickensides". The geology of Smyrna Bench is described in reports on the Saddle Mountains (Twiss, 1933; Jahns, 1966; Grolier and Bingham, 1978; Campbell, 1979; Reidel, 1978, 1984, 1988; ARCO, 1982; West and Shaffer, 1988; West and others, 1989, 1996; West, 1997, 1998; Reidel and Fecht, 2002; Tincher and Reidel, 2009; Hart and others, 2012; Staisch and others, 2016, 2018).

Wenas Creek Trenches - Scarps in the Wenas Creek Valley, identified from lidar imagery, were trenched by USGS in 2009. Two trenches located 5 km apart were opened in alluvial fan deposits and revealed several small faults offsetting the bedrock, cover sediments, and soils. No liquefaction features or clastic dikes were observed.

Buroker Roadcut - A small reverse fault exposed along Russell Creek Road six miles east of Walla Walla was investigated in the late 1970s by Rockwell (1979), Foundation Sciences (1980), and Farooqui and Thoms/Shannon & Wilson Consultants (1980) on behalf of Washington Public Power Supply System. A sketch by Foundation Sciences shows the fault offsetting Miocene basalt of the "Dodge" flow (lower Wanapum), post-basalt stream gravel, and oxidized Palouse loess with a caliche stringers. Younger gray and dark brown loess units above are not cut by the fault. Clastic dikes are indicated in the Palouse loess, but they do not appear in the sketch. According to Swanson in Rockwell (1979), the fault cuts "fluvial gravels and an older loess, but does not deform overlying young loess." A different interpretation of the same roadcut is provided by Farooquoi and Thoms (1980). Their Figure 11 shows the fault offsetting weathered Miocene basalt and reddened Palouse loess. Throw on the fault is 56 cm. Holocene loess is not cut by the fault. The report identified no clastic dikes, but notes some "fractures are lined with caliche". I was unable to locate clastic dikes in the roadcut during a visit in 2026.

Boylston Mountains Trenches - Scarps identified from lidar imagery along Johnson Creek and Park Creek, were trenched by USGS in 2010 (Barnett and others, 2013). The Horned Lizard trench exposed two colluvial wedges atop a prominent fracture. A buried soil consisting of dark brown, silty-clay with prismatic structure separates the two bodies of colluvium. Two episodes of fault movement were interpreted, separated by a length of time sufficient for a soil to accumulate. No liquefaction features or clastic dikes were observed.

Arlington-Shutler Butte Roadcut - The Arlington-Shutler Butte Fault is well exposed along I-84 west of Arlington, OR. A wide road shoulder for the east-bound lanes provides access to the large, clean, vertical roadcut. The oblique-normal fault cuts Miocene basalt and a sedimentary interbed. The fault is truncated by a boulder gravel deposited by the Missoula floods. The unfaulted flood gravel fills a swale cut right across the fault. The fault is believed to have last moved during early to middle Pleistocene time (<780,000 years). Its length indicates it is capable of producing <4.0 magnitude quakes. The fault strikes northwest across the Columbia River, connecting Jones Canyon with Old Lady Canyon. No clastic dikes or liquefaction features were observed during a visit in 2024.

Ahtanum Ridge Trenches - A Ted Repasky and Newell Campbell study just south of the Yakima Valley. No liquefaction features or clastic dikes were observed. Limited information exists on this investigation.

Gate Creek Trench - A study by Ashley Streig (PSU) and Scott Bennett (USGS) on a portion of the Mt. Hood Fault Zone in Oregon. An offset of 1.5 m was observed and a last-movement age of ~1000 years was determined from carbon dates on charcoal. No liquefaction features or clastic dikes were observed. The site lies well outside the Pleistocene floodway.

Kittitas Valley Trenches - The Dead Coyote and Bitterbrush trenches were excavated by Dr. Walter Szeliga and student Craig Huddleston across a strand of the Dead Coyote Fault northeast of Ellensburg (Huddleston, 2023). The project built on a previously unsuccessful attempt to assess the history of the north-dipping reverse fault by USGS/Brian Sherrod (c. 2022). The scarp, visible in lidar imagery, crosses Miocene basalt, old fan remnants, and Quaternary gravels of the Reecer and Naneum fans. Trenching exposed small shear zones that offset fan conglomerates, colluvium, and a buried soil. Two rupture events occurred since about 470 ka. An undeformed colluvial unit and the modern soil overlie the fault. No liquefaction features or clastic dikes were observed.

Spencer Canyon Trench - The Spencer Canyon trench was opened across a prominent scarp observed in lidar imagery near Entiat, WA (Sherrod and others, 2015; Brocher and others, 2017; Brocher and others, 2018; Sherrod and others, 2021). The scarp is believed to have formed during the 1872 North Cascades/Chelan quake (>6.5 M). Bedrock at Spencer Canyon is not Columbia River Basalt, but older crystalline rock of the North Cascades. It is unclear if the fault is part of the Yakima Fold Belt or an older set of faults in the older bedrock. Vertical structures shown in sketches of the trench wall are sediment-filled root casts. No liquefaction features or clastic dikes were observed.

Holocene Dikes West of the Cascade Divide

Floodwater spilling out of the Columbia River Gorge ponded in the Willamette Valley, blanketing its floor with sand and silt and from Portland to Eugene. Ice-rafted boulders of "granite and schist" were first noted by Diller (1896) and later more lithologies were mapped throughout the basin (Bretz, 1919; Allison, 1935; Minervini and others, 2003).

While the Willamette Silt does contain clastic dikes, they are far fewer than in Columbia Basin some 350 km upstream. PhD student Jerry Glenn documented a few sheeted dikes at his River Bend and Irish Bend sites near Corvallis, OR (Glenn, 1965). Dr. Ira Allison (Allison, 1978, Figure 14) photographed a clastic dike cutting slackwater rhythmites near St. Paul. Dikes exposed in the basement of the Oregon State Capital Building at Salem were sent to me by Ray Wells in 2012. Photos of dikes exposed in highway excavations near Portland were taken by Ian Madin and also shared with me in 2014.

Willamette Valley rhythmites. Graded beds nearly identical to the Touchet Beds of south-central Washington are present in Willamette Valley, OR. Dark layers are the coarse-grained lower portions of each bed. Clastic dikes are relatively sparse in these slackwater deposits. A clastic dike is seen cutting steeply through the stack at the center of the photo. To the left is a line of steps cut into the slope by the geologist to allow closer inspection of the exposure, a common practice. Photo by Allison (1978, Fig. 14) taken about five miles southeast of St. Paul, OR.

Thurber and Obermeier (1996) reported finding 16 clastic dikes at 7 sites along the lower Calapooia River, a tributary to the Willamette River. The largest features measured 10 cm wide x 5 m long. They attributed the dikes to liquefaction triggered by a Holocene earthquake. No specific fault or event was identified. The copy of the Calapooia report that I obtained contained no photos of dike fills or sketches detailing field relationships between the dikes and the sediments they intrude. Consultant John Sims (2002) reviewed the report, concluding Thurber and Obermeier's data set was too small to support their interpretation,

The limited area surveyed by [Thurber and Obermeier] in the Willamette Valley does not allow for a high level of confidence in determining if the features result from large subduction events or local intracrustal events. The age of the structures is somewhat in doubt as few radiocarbon dates are available for the host deposits and Thurber and Obermeier (1996) do not report any radiocarbon dates as part of their study. They also do not mention any evidence for liquefaction in post Pleistocene deposits of which there are many in the banks of the Willamette River and its tributaries. Thus, with incomplete coverage and lack of dating of paleoliquefaction features, the question of source zones is moot. Earthquake source determination can only be addressed with broader coverage of liquefaction features and better age data to constrain timing of events and to allow regional correlations of liquefaction features. In addition, we need a more complete picture of the size distribution of similar-aged features for the purposes of evaluating the magnitudes of prehistoric earthquakes.

River Bend section. Glenn (1965, Figures 3 and 15) found a few clastic dikes in largely undeformed Touchet-equivalent flood rhythmites in the Willamette River Valley. Outcrop photos of various other sites around the valley taken by Glenn record information largely lost to development. Few mentions of soft sediment deformation or clastic dikes are found in the numerous descriptions of the Willamette Silt by Bretz (1925, 1928), Allison (1932, 1933, 1936, 1953, 1978), Piper (1942), Treasher (1942), Lowry and Baldwin (1952), Baldwin and others (1955), Allison and Felts (1956), Wells and Peck (1961), Trimble (1957, 1963), Balster and Parsons (1969), Hampton, (1972), Robert (1984), McDowell (1991), Yeats and others (1996), and McDowell and Roberts (1987).

Obermeier and Dickenson (2000), working in the nearby Columbia River Valley west of the Cascade divide, found "relict liquefaction features" in low shoreline bluffs of sandy islands between Astoria, OR (Marsh Island) and Kalama, WA (Bonneville Dam) and in cutbanks of 10 tributary streams in the Hood River area. The thickest dikes they measured were 30 cm wide, on par with dikes in Missoula flood rhythmite sections I have documented in side canyons east of the divide (i.e., Sixprong, Glade, Rock, Old Lady, Arlington, Chenoweth, Willow, etc.). The authors attributed the dikes to lateral spreading, hydraulic fracturing, ground shattering, and warping triggered by earthquakes. Similar investigations by USGS and DOGAMI were conducted in the Columbia gorge (Obermeier, 1993; Peterson and Madin, 1997; Atwater, 1994) and contain some of the same information, which has created a record that is a bit confusing.

Atwater (1994) and Takada and Atwater (2004 + Appendix A Supplement) describe sandy riverbank sediments in the lower Columbia River gorge deformed by the 1700 AD Cascadia earthquake. They note Holocene-age dikes filled with sand and sills that,

...mostly follow and locally invade the undersides of mud beds. The mud beds probably impeded diffuse upward flow of water expelled from liquefied sand. Trapped beneath mud beds, this water flowed laterally, destroyed bedding by entraining (fluidizing) sand, and locally scoured the overlying mud.

Peterson and Madin (1997) and Peterson and others (2014) describe unsheeted sand dikes and sills in Holocene overbank muds at sites near the mouth of the Willamette River and in bluffs along Pacific beaches near the mouth of the Columbia. They also interpret the dikes as features triggered by the 1700 AD event. A field guide was prepared for a Friends of the Pleistocene outing (Peterson and others, 1993).

All of the dikes and sills described by Atwater (1994), Thurber and Obermeier (1996), Obermeier and Dickenson (2000), Sims (2002), Takada and Atwater (2004), and Peterson and others (2014) are fluid escape structures that formed in wet floodplain deposits west of the Cascade divide. They do not resemble the sheeted, wedge-shaped injection dikes found in scabland deposits. While a seismite interpretation is reasonable, none of the reports clearly document a source bed for the dikes and I wonder if some sections they called Holocene are actually Pleistocene. Perhaps a deeper dive into their unpublished field notes would clarify.

Holocene liquefaction dikes in the lower Columbia gorge. Caption for Figure 13b in Atwater (1994) reads, "Dikes with raised edges at upper Wallace Island [near Longview, WA]...The dikes transect mud beds that extend parallel to shoreline." This is the same dike pictured in Peterson and Madin (1997, Fig. 11b) and probably the largest example seen by all parties involved. Guessing that's Atwater's shovel in the photo.

Liquefaction model for Pacific beaches and western Washington floodplains. The caption for this figure from Peterson and Madin (1997, Fig. 2) reads, "Drawing of subsurface fluidization features including clastic dikes and sills and flames. Internal structures include intruded contacts with host deposit and disoriented mud blocks in sandy matrix. Fluidization features such as clastic sills are often enhanced under thin capping deposits of mud overlying thick source beds of sand." The cartoon, originally conceived by Fiegel and Kutter (1994), is not a sketch from the field or meant to depict features observed at a specific study site. Rather, it is a conceptual model of liquefaction features and relationships that may or may not be present in any one outcrop. The figure has been reproduced in several articles (i.e., Obermeier, 2005, Fig. 2).

Clastic Dikes and Seismic Hazard Maps

Clastic dikes are commonly observed in earthquake-prone regions of the world and are often highlighted in post-quake damage assessments (i.e., Walsh and others, 1995). Systems for describing sediments deformed by strong shaking have matured over the past century thanks to dedicated staff at USGS, state geological surveys, and consulting firms (McCulloch and Bonilla, 1970; Gohn and others, 1984; Atwater, 1994; Obermeier, 1990, 1992, 1996, 2009; Peterson and Madin, 1998; McCalpin, 2009; Holtzer and others, 2011). Maps of liquefaction features often help geologists delineate the extent of deformation following a large quake. The value of such maps largely depends on the intensity of the field effort (size of the dataset) and experience of the geologist. A small number of measurements or measurements collected within a small area (i.e., one valley or one trench) have low value because they lack statistical power and may not reflect the actual pattern of damage. Misinterpretation of features and field relationships can also be a problem, especially for inexperienced staff or where exposure is poor. The assumption that all clastic dikes form by liquefaction triggered by earthquakes has led many to incorrectly identify features formed by aseismic processes seismites. Investigators should be especially aware of knowledge gaps, their own biases, and those of their managers and editors. Caution and restraint are helpful when interpreting paleoseismic information in the field (Borradaile, 1984; Bonilla and Lienkaemper, 1990; Holtzer and Clark, 1993; Moretti and Van Loon, 2014).

Seismic hazard in the Columbia Basin vs. New Madrid. Earthquake hazard probability map generated by the 2023 USGS model (fault-slip rates, frequency, magnitude). The map shows modeled results, specifically the 2% in 50-year probability of exceedance for fixed VS30 760 m/s. Red-orange indicates a high probability for damaging quakes. Green-blue indicates a low probability. Note the stark contrast between the Columbia Basin (green-yellow) the New Madrid Fault Zone (red-orange). Dikes in the Columbia Basin are wedge-shaped injectites filled from above. Dikes in the New Madrid Seismic Zone are feeder conduits to sand blows. Columbia Basin dikes are Pleistocene age and occur entirely within the Ice Age floodway. New Madrid dikes are Holocene features that occur in floodplain alluvium of the Mississippi River and several large tributaries.

Paleoseismology Community Audits Itself

In 2017, an international conference was convened to review reporting on seismites in sedimentary sequences. Participant emphasized the need for caution (Feng, 2017). It seems “seismite” (Seilacher, 1969; Montenat and others, 2007; Van Loon, 2014) has for some time been assigned too liberally to features of nonseismic or ambiguous origin, making reexamination of "classic" seismite localities necessary. Clear-eyed geoscientists who participated reattributed many features formerly identified as seismites to nonseismic processes, most commonly to rapid sedimentation and loading (Moretti and Van Loon, 2014; Shanmugam, 2016 and references therein). The following quotes capture the feelings of some participants:

Nonseismic events can create structures that are virtually indistinguishable from seismically-deformed sediments, or seismites. Therefore, paleoseismologists must correlate candidate seismites over regions and rule out nontectonic origins before concluding that an earthquake occurred.

– L.B. Grant

A great progress has been made in researches [sic] of soft-sediment deformation structures (SSDs) and seismites in China. However, the research thought was not open-minded. About the origin of SSDs, it was almost with one viewpoint, i.e., almost all papers published in journals of China considered the beds with SSDs as seismites. It is not a good phenomenon.

– Z-Z. Feng

At present, there are no criteria to distinguish...soft-sediment deformation structures formed by earthquakes from SSDs formed by the other 20 triggering mechanisms...the current practice of interpreting all SSDs as “seismites” is a sign of intellectual indolence.

– G. Shanmugam

Obermeier's Maximum Width Method is Inappropriate for Sheeted Dikes

The relationship between liquefaction and shaking intensity is well established (Ambraseys, 1991; Galli, 2000; McCalpin, 2009; Zhong and others, 2022).

A field-based method relating sand blow feeder dikes to shaking intensity was developed in the 1990s by Steve Obermemer of the USGS (Obermemeier, 1998). The "maximum width method" involves measuring the width of the widest dikes at a number of locations and contouring the values to produce a map that (hopefully) reveals the epicenter. Since seismic shaking is often most intense near the epicenter, the largest dikes should occur there.

The method was applied the New Madrid Seismic Zone in the southeastern United States, a region struck by magnitude 7.2–8.2 quakes with Modified Mercalli Intensities >VIII in 1811-1812 prior to the invention seismographs (Obermeier and others, 2005). Sand that was vented to the surface during shaking is distributed over hundreds of square kilometers and remains clear in aerial photos. Obermeier's investigations identified not one, but several potential epicenters for the quakes, an improvement over earlier feature-mapping efforts (Fuller, 1912; Russ and others, 1978; Munson and others, 1992; Boyd and Schumm, 1995).

Obermeier's "maximum width method", while appropriate for sand blows in active floodplains, is in appropriate for sheeted injectites in the Touchet Beds. The method assumes most dikes are single-fill structures that rose from a liquified source bed at depth during a shaking event. It presumes dike width (fracture aperture) corresponds with shaking intensity. However, these assumptions don't apply to dikes in the Channeled Scablands. The widths of Touchet-type dikes grows incrementally, flood by flood, by the addition of new fillings with apertures of varying widths. The dikes widen sheet by sheet during the course of a single flood event and many continue to expand during additional flood events. The widths of single-fill feeder-type dikes and sheeted injection-type dikes are simply not comparable. One involves the involves the upward escape and venting of fluidized sand at the ground surface, while the other involves pressurized injection of sediment into hydraulic fractures propagated downward. For an apples-to-apples comparison, match widths for the widest liquefaction dike to the widest sheet in any dike at each site.

Clusters of sand blow feeder dikes at New Madrid. Obermeier (1998) used the widths of sand blow feeder dikes to delineate the extent of liquefaction and identify potential paleoepicenters. Obermeier collected width measurements only in floodplains where the depth of the alluvium regularly exceeds 30m (Saucier, 1964). Black dots are measurement locations. Dot size reflects a width category (<15 cm, 15-50 cm, >50 cm). Six dashed ovals are the interpreted damage halos associated with historic quakes. I find it odd that Obermeier identifies six separate clusters of large dikes in his study area. It is unclear whether clustering of dikes is related to alluvium thickness or proximity to an epicenter. Likewise, dike width variation within individual stream valleys is equal to the variation seen across the entire map area; the largest and smallest classes of dikes are mapped at the westernmost and easternmost edges of the study area. The fact that more dikes were found near Vincennes, Indiana could be due to more outcrops, a function of stream network density, not necessarily to stronger shaking. Also, the map would benefit from clear delineation of valley bottoms (where sand blows are common) from dry uplands between (where no sand blows formed). A map in an earlier report does just this (Obermeier, 1992, Fig. 2).

Columbia Basin Crust vs. New Madrid Crust The crust beneath the Columbia Bains and the New Madrid region are fundamentally different. By almost any measure, stark contrasts are found - thickness, composition, tectonic setting, stress regime, or structural grain. The New Madrid is an ancient failed rift in crystalline basement. Seismicity >M 7.0 is generated by deep, steeply-dipping faults in strong crust. The Columbia Basin, by contrast, is a young back-arc flood basalt province resting atop extended Tertiary crust capable of ~M 7.0 quakes (Madin and others, 2021). Clastic dikes at New Madrid occur in Holocene floodplain deposits that liquefied during shaking and vented sand upward. In Columbia Basin, dikes were injected downward into sedimentary and bedrock substrates during Ice Age megaflood events. New Madrid sand blow region is one of the world's classic earthquake liquefaction localities. Holocene sediments in the Columbia Basin contain no record of liquefaction.

Megaflooding, not bedrock, controls where dikes formed. Sheeted clastic dikes are abundant where Ice Age floods swept through basins floored mostly by Miocene basalt. No dikes are found where basalts are overlain by non-flood sediments (Blue Mountains or Idaho-Nevada Graben). Cataclysmic flooding and the response of unconsolidated surficial deposits to flood loads, not bedrock lithology or structure, appears to control where dikes formed. Map modified from Tolan and others (2009, Figure 1).

Missing Record of Holocene Liquefaction in Eastern Washington

Sections of thick, unconsolidated alluvium preserved in many valleys of creeks of Eastern Washington lack clastic dikes, liquefaction features, and other soft sediment deformation structures commonly found in regions subjected to strong shaking. The absence of such features in young, wet, fine-grained alluvium is difficult to explain if the Yakima Fold Belt has been generating earthquakes with magnitudes >6 M (Intensities >VII) every 500-1000 years since the Miocene. It is difficult to believe modern floodplain sediments would deform dramatically differently than Missoula flood sediments, or that earthquakes of exceeding 6 M have not occurred in the past 10,000 years, or that deformation features in floodplain sediments have been erased by channel processes or vigorous bioturbation while Pleistocene sediments remain pristine.

Alluvium of Dry Creek. Thick Holocene alluvium (>4m) like this section along Dry Creek near Walla Walla, WA shows no evidence of strong shaking, pre- or post-Mazama ash. Widespread deformation would have long ago been identified by local farmers, geologists, and soil scientists, especially given the strong visual contrast between the light-colored ash and dark overbank alluvium. Mapped Quaternary faults in the vicinity include the Wallula Fault Zone (21 km away), Hite Fault (33 km away), Kooskooskie Fault (23 km away), and Promontory Point Fault (6 km away). Intersection of Harvey Shaw Rd and Dague Rd ~8 km north of Walla Walla. S.W. Cooley photo June 2021.

Alluvium of Union Flat Creek. No soft sediment deformation has been found in the floodplain of Union Flat Creek near Dusty, WA. S.W. Cooley photo.

Alluvium of Touchet River. Thick deposits of alluvium along the modern Touchet River contain no evidence of deformation consistent with strong seismic shaking. S.W. Cooley photo 2026.

Alluvium of Willow Creek. Thick alluvial fills along Willow Creek near LaCrosse, WA remain undeformed. S.W. Cooley photo 2026.

Alluvium of Latah Creek. A mix of sandy Ice Age flood deposits, reworked colluvium, and varved lake beds capped by Holocene alluvium and Mazama Ash is exposed along Latah Creek west of Spokane, WA. While the Pleistocene lake beds near the Qualchan Golf Course are prone to landsliding and folds in the overlying Touchet Beds appear to be flood-formed, the Holocene sediments that overlie oxidized gravel remain undeformed. I've seen nothing upstream of the Hatch Rd bridge resembling liquefaction. Cutbank near Hangman Valley Rd northwest of Hangman Valley Golf Course. S.W. Cooley photo.

Shaking Intensity-Liquefaction Distance Relationships Fail

Shallow, intraplate faults in the study area are believed capable of producing magnitude 6.5 earthquakes and MMI VII–VIII intensities (Lidke and others, 2003). However, intensity-liquefaction curves (Ambraseys, 1988; Galli, 2000; Qiao and others, 2017; Zhong and others, 2022) do a poor job of predicting where dikes in the Channeled Scabland occur. Large dikes are routinely found at distances twice those predicted. For example, many large dikes are separated from mapped faults by distances well in excess of the ~150 km liquefaction limit. In their review of sand blows produced by 28 large earthquakes, Castilla and Audemard (2007) determined only subduction zones capable of producing liquefaction a distances greater than 150 km. An epicenter placed at...

• Wallula Gap, WA (Wallula Fault Zone) is >285 km from large dikes near Kettle Falls, WA.

• Burbank, WA (Umtanum–Gable Mountain Fault) is >260 km from large dikes in Lewiston Basin, ID.

• Arlington, OR (Arlington–Shutler Butte Fault Zone) is >230 km from the central Willamette Valley, OR.

• Entiat, WA (1872 North Cascades Earthquake) is >210 km from Touchet, WA.

• Smyrna, WA (Saddle Mountains Fault) is 225 km from Kettle Falls, 205 km from Lewiston, 135 km from Cecil, OR, and 120 km from Bridgeport, WA.

Magnitude–distance curves reveal weaknesses. Distances between earthquake epicenters and liquefaction features were compiled from studies on several continents (Ambraseys, 1988; Galli, 2000; Qiao and others, 2017; Zhong and others, 2022). A robust relationship exists between earthquake magnitude and the radial distance away from an epicenter liquefaction features will form. Liquefaction produced by a M 6.5 quake is predicted to occur out to ~75 km. For a M 7.5 quake, the limit approaches 150 km. My field work throughout the Scablands region shows that many large dikes are located at distances exceeding 150 km from mapped Quaternary faults and potential epicenters.

Distances of dike-bearing outcrops from an assumed epicenter at Wallula Gap. Distances measured from an assumed epicenter at Wallula Gap (Wallula Fault Zone) to outcrops containing clastic dikes are shown (circles). Most dikes occur within 150 km of the assumed epicenter, but many are found at distances far beyond liquefaction limits established by Galli (2000), which suggests the dikes are not seismites. Black bars represent the boundaries of subbasins along the Ice Age floodway. Outcrops containing dikes are abundant in the 10 subbasins located within 150 km of Wallula Gap, but become less abundant beyond 150 km due to a reduced subbasin count. Subbasin count, at bottom of figure, is a proxy for exposure. Outcrops are more numerous in valleys near Wallula Gap, where several rivers converge and the road network is extensive. The dikes themselves are largest and most abundant in exposures immediately upstream and downstream of Wallula Gap, though very large dikes are found in a number of distant exposures. Subbasins: CC = Crab Creek Valley, GT = Gorge Tributary valleys downstream of Wallula Gap to The Dalles, LB = Lewiston Basin, OK = Okanogan Valley, PB = Pasco Basin, RP = Rathdrum Prairie, QB = Quincy Basin, SR = Snake River Valley, TV = Tucannon River Valley, UB = Umatilla Basin, UC = Upper Columbia River Valley, WC = Willow Creek Valley, WW = Walla Walla Valley, WV = Willamette Valley, YV = Yakima Valley.

Large dikes located hundreds of kilometers from Quaternary faults. Very large clastic dikes in the Upper Columbia River gorge are too distant from Quaternary faults to have been created by shaking and liquefaction. Colville River mouth south of Kettle Falls, WA. S.W. Cooley photo.

Strong Shaking East of the Cascades?

If the dikes in the Channeled Scablands are the products of seismic shaking, then one or more of the Yakima Fold Belt structures would be the likely trigger. However, the dikes are distributed over too large an area for a single fault to be the culprit. If movement on the Saddle Mountains Fault, for example, triggered diking, then we should see evidence of repeated dike injection dating to the Miocene consistent with its fault recurrence interval. Since the Saddle Mountains (and other ridges) have been rising for at least the past 15 million years, dikes and other evidence should be present in nearby exposures of Miocene, Pliocene, Pleistocene, and Holocene strata. The Ellensburg, Latah, and Ringold Formations should be riddled with soft sediment deformation features if strong shaking were as common occurrence in the Yakima Fold Belt as some would have you believe. To date, no such pattern has been recognized.

Geologic evidence of strong shaking is absent in fine-grained Neogene sediments of the region. Despite more than a century of investigation, no local examples or regional pattern have been recognized in the following:

• Hundreds of borehole cores logged in supra-basalt sediments at the Hanford Site.

• Dozens of measured sections through the Pliocene Ringold Fm at White Bluffs and equivalent sediments in the Dalles-Umatilla syncline.

• Cores from alpine lakes in the Cascades and Okanogan Highlands.

• Dozens of large exposures of Ellensburg/Thorp/Latah Fm fills in Kittitas, Yakima, and Naches Valleys.

• Dozens of large exposures of sedimentary interbeds in the Columbia River Basalts.

• Holocene floodplains east of the Cascade divide.

• Soft sediment deformation features east of the Cascade divide attributable to large historic quakes (1918 Vancouver Island M 7.2, 1946 Vancouver Island M 7.5, 1949 Olympia M 6.7, or 2001 Nisqually M 6.8).

Seismic station data for the Yakima Fold Belt have likewise failed to reveal a pattern of strong shaking. The often-referenced Milton-Freewater/Stateline earthquake of 1936 that struck the Walla Walla Valley was a sub-magnitude 6.0 event that formed no clastic dikes and caused no damage to speak of beyond the immediate epicenter, the tiny outpost of Umapine, OR. Was the Umapine School, built in 1911 and still standing today, damaged by the quake? The equally beloved Hite Fault, located in the Blue Mountains southeast of Walla Walla, shows no indication of Quaternary movement and appears to have last been active in the Miocene (Foundation Sciences, 1980; Brocher and others, 2018). I am aware of no young scarps, liquefaction features, or other evidence of strong shaking associated with the Hite Fault. Its an old structure that is conspicuously straight, thus prominent in aerial photos, nothing more.

Modest shaking east of the Cascades. Map of earthquake epicenters in Washington recorded between 1970-2015 (Brocher and others, 2017, Fig. 2). East of the Cascade divide, quakes have mostly been small, shallow, and weakly clustered (Piety and others, 1990; Gomberg and others, 2012). Most epicenters do not fall along prominent scarps or mapped faults of the Yakima Fold Belt (Miller and others, 2001). No spatial correspondence exists between historic-period epicenters and the clusters of large clastic dikes in the study area; dikes are neither larger nor more numerous in areas with more seismic activity (i.e., eastern Saddle Mountains, Entiat area, White Bluffs, MSH-MH corridor, etc.). YFTB = Yakima Fold Thrust Belt, D = The Dalles, El = Ellensburg, GRZ = Goat Rocks fault zone zone, HA = Hanford Site, MA = Mount Adams fault zone, MH = Mount Hood fault zone, Pa = Pasco, SHZ = St. Helens fault zone, UL = Umtanum lineation, P = Portland, W = Wenatchee, WRZ = Western Rainier fault zone.

Review of the 1872 North Cascades Earthquake

Written reports on the December 15, 1872 North Cascades (Chelan) earthquake deserve scrutiny and a bit of historical context. While the quake is the largest on record for Washington State, all contemporary written accounts of the event are found in newspapers. Witnesses to the quake observed water spouts, ground cracks, small landslides, and the collapse of the roof of an old cabin (Washington Standard Newspaper 11 Jan 1873; Coombs and others, 1976; Brocher and others, 2018, Appendix B). Certainly, to the locals it was a memorable event, but no one died and no building suffered major damage. The press interviewed no geologists about the calamity and the story quickly fades from prominent, both signs that it probably wasn't a big deal.

The town nearest the epicenter was Wenatchee, which in 1872 was a frontier outpost consisting of a few storefronts constructed of wood and unreinforced masonry. The city would not be officially incorporated for another 20 years when the Great Northern Railroad was completed in 1893. At the time of the quake, neither the light bulb nor the telephone had been invented. Ulysses S. Grant was President and Washington, Idaho, Colorado, Wyoming, Utah, New Mexico, and Arizona were not yet States of the Union.

In 1872, just 6 rudimentary seismographs monitored ground motions for the entire Pacific Northwest, including parts of Canada. Archival datasets for the North Cascades and the 1936 Milton-Freewater/Stateline events were reprocessed using modern tools and their magnitudes down-rated as a result (Brocher and Sherrod, 2018; Gutenberg and Richter, 1954; Coffman and others, 1982; Noson and others, 1988). The North Cascades quake went from >7.2 to 6.8. The Milton-Freewater quake went from 6.1 to 5.8. Given the antiquity of the sensors, archaic processing methods, and a documented history of disagreement among modern seismologists regarding the magnitudes of old quakes (Milne, 1956; Malone and Bor, 1979; Hopper and others, 1982; Bakun and others, 2002), why wouldn't all pre-WWII earthquake records be simply flagged as "raw" or "provisional" by default? Wouldn't the prudent manager at USGS or PANGA simply down-rate all such quakes to the lowest possible magnitude and intensity given the uncertainties inherent in old data? The modern summary of the 1872 quake should read something like, "the quake had a magnitude of at least 6.5", not, "the 7.4 temblor sent shockwaves across a vast region".

Washington State's paleoseismology group, housed mainly within the USGS and UW does some great work. But from time to time some of the folks there seem to make shit up. For example, Brocher and others (2017) somehow divine a connection between the aftershock hazard associated with the 1872 North Cascades event and that associated with an M 7.5 subduction zone quake at Nobi, Japan. The two ruptures share absolutely nothing in common - not geologic setting, not rock type, not stress regime, not rupture length, not even decade (1872 vs. 1891). Similar nonsense appears in a few other reports. For example, "widespread liquefaction" the USGS claims to have found at Starthistle, Finley Quarry, and Kittitas Valley (Mahon and others, 2022) simply does not exist. Evidence for strong shaking at these sites is as flimsy as it gets, yet they charge ahead and publish with confidence (Sherrod and others, 2016; Angster and others, 2023). The fact is, no significant seismically-generated liquefaction has been observed anywhere in Eastern Washington despite more than a dozen trench investigations.

This group of overenthusiastic, mostly desk-bound hazards researchers consistently punches up the spectacle regardless of the data they obtain, faults they trench, or features they observe. Read their stuff. Its heavy on narrative, light on field evidence. While newspapermen are free to splash "Earthquake!" across the front page anytime they choose, no one has tasked the USGS with anything more than good honest science and clear communication. Spectacle sells newspapers, not science. If a dollop of spectacle is necessary to keep the lights on in the Hazards Program office, then its time for a change. Redirect funding elsewhere and trim the staff to one post-doc running an M5 and a few AI agents. We'd all be better served.

Saddle Mountains Fault. The Saddle Mountains Fault is actually six separate segments each with its own rupture history. Many geologists believe growth of the anticline has slowed since the Miocene, while others find evidence of a two-phase uplift history with considerable relief generated late, during the Late Pleistocene. Wherever you fall, liquefaction has not a big part of the story. Very few liquefaction features, contorted beds, or clastic dikes have been found in Miocene or Pliocene sediments cut by the fault at elevations above the level of Missoula flooding. The sandy-silty interbeds are sandwiched between tilted basalt flows in many places along the 600m-high north flank. Some contain thin, light-colored fractures (photo at right). These are not liquefaction features, but bleached shear bands common in deformed sandstones worldwide. Because of its prominence and accessibility, the ridge has been studied and mapped in detail, mostly by Steve Reidel, and serves as a model for other anticlines in the fold belt. Should it? S.W. Cooley photos.

Saddle Mountains crest. I've worked methodically along the entire crest of the Saddle Mountains, Smryna Bench, and Taunton Bench, examining Pliocene and Pleistocene sediments preserved in cliffs and gullies there (Cooley, 2023). I found no sheeted clastic dikes in sedimentary interbeds above 366 m elevation, the upper limit of Missoula flooding. In fact, surprisingly little evidence of strong shaking is found in numerous alluvial sections exposed along the 85 km-long ridge. Is it possible that Saddle Mountains rose one M 5.9 quake at a time? S.W. Cooley photo.

Minimally deformed Miocene interbeds. Ebinghaus and others (2012) examined Miocene-age sedimentary interbeds in the Columbia River Basalt at 14 sites in Pasco and Quincy Basins near the Saddle Mountains and Frenchman Hills Faults. Small flame structures and load casts were noted at certain contacts between mudstones and sandstones. Such features are common where channel sands spill through levees onto overbank muds. No liquefaction or clastic dikes were reported. The Ebinghaus findings are consistent with those of previous workers (Schmincke, 1964; Hays and Schuster, 1983; Smith, 1988a,b; Humphrey, 1996) and my own.

Mostly undeformed White Bluffs. The White Bluffs of the Columbia River north of Richland, WA stretch for 50 km and expose Pliocene Ringold Fm, Pleistocene flood deposits, and Holocene dunes. The nearly continuous section is gently tilted and partially dissected by recent landslides, but otherwise undeformed. Slides along the river's east bank are exacerbated by irrigation practices on the benches above. Most of the White Bluffs is public land and readily accessible by foot via the Old Ringold Road Trail. Google Earth image.

Lost near Lyons Ferry. About 15 Touchet Beds overlie a thick bar gravel in the Snake River canyon near Lyons Ferry, WA. John Whitmer photo (WGS Archive No. 03144).

Style of Deformation Changes Along the Floodway

The same floods produced different types of soft sediment deformation features in different parts of the floodway. Grainsize and rheology of the flooded substrate explain most of the variation. Same stimulus, different response. Different substrates, different deformation features. In silt-sand rhythmites near Walla Walla, Lewiston, Cecil, and Zillah, we find slender, sheeted, wedge-shaped dikes that number in the thousands. In varved lake beds of the upper Columbia Valley, we find abundant t-shaped mud squirts, rubbly injectites, a few sheeted dikes, and other features associated with mass wasting. In coarse-grained eddy bars near Umatilla and Washtucna, we find a few stubby, gravel-filled dikes with crude vertical sheeting. In silty, gravel-free silt rhythmites and silt-pebble diamicts near the upper limit of flooding where floodwaters lapped against loess-covered hills (i.e., eastern and southern margins of the Palouse Hills), we find a few thin dikes here and there. In basalt-floored valleys scoured by energetic flows and filled by deep lakes (i.e., Snake and Columbia gorges), we find a few sheeted dikes intruding jointed bedrock. Coarse, laminated sands like those at at Qualchan, the mouth of Rock Creek, and quarries near Corfu are nearly devoid of dikes because they lack silt.

Deformation style varies with sediment type. The style of soft sediment deformation differs throughout the floodway region (light blue area). Wedge-shaped sheeted dikes are common in slackwater deposits in the southern half where sand-silt rhythmites dominate, while t-shaped mud squirts and flame structures are common in the northern half, where abundant varved lacustrine sediments occur. Each map symbol represents multiple outcrops. The divide between the northern floodway and southern floodway is approximately Moses Lake, WA (Lake Lewis shoreline). The dashed blue line follows a longitudinal profile of the Columbia River drawn by Atwater (1987, Figure 2) and used in O'Connor and others (2020, Figure 8).

Sheeted Dikes Without Earthquakes

Examples of sheeted, wedge-shaped dikes formed by aseismic processes in a variety of geologic settings are reviewed below. In all cases, overloading, rapid sedimentation, silty sediment, and hydrofracture were involved. Downward-injected dikes are found in lahar deposits in the Aleutians (Herriott and others, 2014), ash flows in calderas (Brocard and Moran-Ical, 2014), deepwater turbidite-fan systems (Jenkins, 1930; Duranti and Hurst, 2004; Huuse and others, 2007; Monnier and others, 2015; Cobain and others, 2015, 2016), and debris flow deposits at Black Dragon Canyon in the San Rafael Swell, UT (author's field notes). The literature on subglacial dike injection is particularly robust and does a really good job explaining how sheeted injection works in wet, possibly frozen, sediments that are rapidly overridden by ice (Kruger, 1938; Boulton et al., 1974; Dionne and Shilts, 1974; Amark, 1986; Von Brunn and Talbot, 1986; Burbidge et al., 1988; Larson and Mangerud, 1992; Kumpulainen, 1994; Dreimanis and Rappol, 1997; Hindmarsh, 1997; Rijsdijk et al., 1999; Van der Meer et al., 1999; Jolly and Lonergan, 2002; Le Heron and Etienne, 2005; Boulton, 2006; Boulton and Zatsepin, 2006; Gozdzik and van Loon, 2007; Phillips et al., 2007; Denis et al., 2009; Van der Meer et al., 2009; Schomaker et al., 2010; Knight, 2012; Phillips et al., 2013; Phillips and Hughes, 2014; Ravier et al., 2015; Sutherland and others, 2022). Glacial overloading is the closest analog to megaflood loading I've found to explain sheeted, wedge-shaped dikes in the Channeled Scabland. Phillips and others (2013) in Boreas provides one of the most complete and approachable treatments to date.

Lahar at Mount Spurr, AK. Sheeted dikes with characteristics identical to those in the Channeled Scablands were discovered by Herriott (2014) in sandy lahar deposits on the side of an Aleutian volcano. Rapid deposition, surface loading, wet over dry sediments, and hydraulic fracturing were all involved. The red arrows are Herriott's and point to silt skins. Image courtesy of Herriott.

Till dikes on Whidbey Island. Gravel-filled dikes descend from the base of a stratified, cobbly till into a underlying diamict. The wedge-shaped dikes all dip in the same direction and form a loose polygonal network. Deformation occurred during the Pleistocene, though younger landslides and scarps dissect the bluffs nearby. Possession Point State Park, Whidbey Island, WA.  S.W. Cooley photo taken June 2026.

Subglacial dikes in Iceland. Sheeted dikes identical to those in the Touchet Beds were injected subglacially and interpreted as water escape structures triggered by ice loading. Water escaped out of the till into more permeable outwash below, partially following bedding. Myrdalsjokull forefield (Slettjokull), Iceland. Photo by Jaap van der Meer in Schomaker et al. (2010, Fig. 8.10).

Subglacial dikes at Voss, Norway. Sheeted dikes identical to those in the Touchet Beds cut sandy outwash sand at Voss, Norway. Similarly, the wall-parallel laminations formed by "a repetitive process", namely cyclic reinjection (Mangerud and Skreden, 1972; Mangerud and others, 1981; Larsen and Mangerud, 1992).

Subglacial dikes in British Columbia. A gravel dike descends from bouldery stratified till into outwash sand below. Hat Creek, British Columbia (Broster and Clague, 1987).

Sheeted dikes at Hat Creek. Sheeted gravel-sand dikes resembling to those at Starbuck, WA and other classic localities in the Channeled Scablands cut the underlying sandy outwash (Broster, 1991, Fig. 9b).

Subglacial dikes in Wrangell-St. Elias National Park, AK. A sheeted dike intrudes sandy recessional outwash of the Bering Glacier. Photo by Crossen (2014).

Subglacial dike at Vatnajokull. A sheeted dike formed by the surging of Breidamerkurjokull during the Little Ice Age cuts soft subglacial sediments (sand-silt-clay-peat) in SE Iceland. Photo by Denis and others (2009, Fig. 6).

Subglacial dikes in Ontario. A sheeted, wedge-shaped clastic dike formed by downward injection measure nearly a meter wide and extends >16 m through late Wisconsin drift exposed along the shoreline of Lake Erie near Bradtville, Ontario. Its laminated fill is composed of remobilized diamict, pebbly sand, silt, and clay. Figure by Dreimanis and Rappol (1997, Fig. 5).

Wedge-shaped dikes beneath a debris flow. Wedge-shaped injections originating from the base of an overriding debris flow cut the cross-bedded sandstone below. Black Dragon Canyon, San Rafael Swell, UT. S.W. Cooley photo.

Downward dikes at Big Pumice Cut, CA. Gravel- and sand-filled dikes sourced in a boulder gravel at the top of the exposure descend several meters through ash flow and ash fall facies of the Bishop Tuff and terminate in the underlying Sherwin till. (Wahrhaftig, 1965; Sharp, 1968; Lipshie, 1976). Hwy 395 at Sherwin Summit, CA.

In Patagonia. A sheeted dike cuts varved glaciolacustrine sediments in the northern Patagonian Andes (Perucca and Bastias, 2008, Fig.11; Van der Meer and others, 2009, Fig. 1d). Pocket knife in shadow.

Dikes by pressurized meltwater under ice. Sheeted dikes formed by hydraulic fracture in a subglacial setting. The dikes dip away from the ice load and down the pressure gradient. Numbers are from the original figure by Phillips and others (2013, Fig. 2A) and indicate the relative age of intrusion based on crosscutting relationships. The glacial-age dikes intrude Devonian fluvial sandstones and conglomerates (Old Red Sandstone) exposed near Inverness, Scotland.

Tidewater setting can promote subglacial injection of wedge-shaped dikes. A grounding tidewater glacier in Sweden forces wedge-shaped, till-filled dikes into the muddy substrate below. Figure by Von Brunn and Talbot (1986, Fig. 16).

Sheeted dikes in a Polish coal mine. Two sets of wedge-shaped, sheeted dikes fill extensional fractures at the crest of a small anticline near Kleszczow, Poland (Haluszczak and others, 2007, Fig. 6e). This example is nearly a meter wide and its fill sourced in undeformed Quaternary deposits that overlie folded Miocene-Pliocene bedrock. A slightly older set of dikes also cuts the underlying bedrock, and is sourced in slightly older, slightly folded Quaternary deposits.

Idaho vs. Utah. On the left is a sketch of a clastic dike at Lewiston, ID. On the right is a cartoon of a much larger deep sea sand injectite from a slideshow by Dr. Lansing Taylor, formerly of the University of Utah's Energy & Geosciences Institute. I flipped the injectite image upside down. In both examples, injection followed efficient, low-resistance pathways that alternate between horizontal and vertical.

Deep sea sand injectites. Many similarities exist between clastic dikes in the Channeled Scablands and sand injectites formed in deep sea turbidite channel-fan complexes (Jenssen and others, 1993; Braccini and others, 2008; Hurst and others, 2011). Figure by Parize and Fries (2003).

Rubbly Injectites at Indian Creek, WA

In November 2017, I discovered several breccia-filled dikes that cut varved Glacial Lake Columbia beds along lower Indian Creek Rd (Hawk Creek) east of Lincoln, WA. They intrude sediments in response to subaqueous slumping of house-sized blocks of sediment. Glacial Lake Columbia was likely present at the time. The breccia fills are unsheeted and contain broken, stratified clasts of the host material. The dikes intrude the lower portion of the >20 m-thick section comprised of at least 24 rhythmites (alternating lake varves and flood sand intervals). In 2017, rotational slump blocks were exposed in fresh, recently-bladed slopes, but are now covered by erosion control matting and grass.

Rubbly injectites. Pleistocene injectites cut slump blocks lakebeds in Glacial Lake Columbia, which locally occupies the northern margin of the floodway region.

Rubbly fills. Rubbly injectite crosscuts clay-rich varves at a low angle.

Rip-ups. Rubbly, unsheeted fills contain stratified rip-up clasts - chunks liberated from the surrounding material.

Parallel and crosscutting. A light-colored sand-filled injectite intrudes clayey varves of Glacial Lake Columbia. Both crosscutting (stair-stepping base) and bedding-parallel (separates bedding) relationships are clear. Indian Creek is located north of Davenport, WA and also goes by Olson Canyon and Hawk Creek.

Field Work Matters

The origin of clastic dikes in sedimentary sequences can be ambiguous, especially at first glance. Earthquakes are not always required. In fact, clastic dikes occur in many settings where seismic shaking played no role whatsoever in their formation (Shanmugam, 2016). Lessons learned in coastal California or the Wabash Valley do not apply universally.

Clastic dikes are threshold features that, if interpreted one way, may prompt policy makers to brand a landscape hazardous and unfit for occupation or development. Interpreted another way, the same dikes become Ice Age relicts of little importance to anyone other than academics and megaflood enthusiasts.

Careful field work that involves a significant number of observations, measurements, samples. Perhaps most important is a study area properly scaled to the geological phenomenon at issue. Overuse of "seismite", shoddy field notetaking, and methods poorly suited to the region are unacceptable practices.

Project planning is the responsibility of the Field Geologist. The assisting Paleoseismologist should also be field-trained. Data gathered in the course of a paleoseismic investigation (fault slip rates, event dates, and shaking effects) are critical inputs to building codes, hazard planning documents, and land use policies. Data from the field certainly informs and often drives land use policymaking, which affects the lives of real people. Academics can sometimes lose track of the real world. Unlike esoteric articles and pretty tables of recurrence probabilities, maps constructed from field measurements are easily appreciated by all audiences. They are uniquely influential and tend to find their way into land use policy documents, which persist for decades.

Basic geometries of dikes in outcrop. Roadcuts rarely provide a complete view of dikes. Connections to source beds and pinchouts are commonly concealed. Reduce ambiguity by zeroing in on key geometries shown here. (A) Twin-tapering forms that do not look like typical dikes. They are axe blade-shaped fractures propagating laterally and emerging from the face of the outcrop. A trick of geometry in the third dimension (see Arris and Aperture figure earlier in article). (B) Three dikes that lack a taper direction are truncated at their tops by erosional surfaces (bedding contacts). (C) Buried sediment remobilized in response to shaking vents sand upward to a higher stratigraphic position (sill) or to the surface via a feeder dike (sand blow). May be sourced from above or below. (D) Upward and downward tapering dikes. Local shearing may have offset a single dike, causing it to appear as two with opposite tapers. A trick of limited exposure. Excavate features continue your search to find more conclusive relationships nearby. (E) Downward tapering dike-sill geometry with upward-curving intersections are sourced from above. (F) Upward tapering dike-sill geometry with upward (tree branch-like curving intersections are sourced from below.

The details matter. How a dike transitions to a sill is important. Here, a sheeted dike filled with sandy flood sediment intrudes a tuffaceous mudstone of the Ringold Fm. A portion of the dike splits away, follows bedding, and becomes a sill that quickly pinches out. The shape of the transition from vertical to horizontal is important. This geometry can only form by downward intrusion (i.e., example E in the figure above). White Bluffs above Columbia River Road not far from the Hanford Offices at Richland, WA.

Know your SSDs. Many soft sediment deformation features are distinctive, but many others can look alike. Where ductile material is involved, which mean more than one process can produce similar-looking features. Also features at an early stage of development may morph into very different shapes over time. Careful observation is usually the key to sorting things out correctly.

Deformation caused by rapid sedimentation. The swirls, flames, and dike-like features shown here formed during Missoula flooding, not during an earthquake. Two interpretations are possible: Pulses within a single flood or by two separate floods where deposits from the first remained soupy until the next flood arrived. Combined loading and viscous drag by fast-moving currents produced the features seen here at White Bluffs, WA and many other locations in the Channeled Scablands. S.W. Cooley photo.

Flood sand deforms lake beds. Flame structures in the light-colored mud formed during a flood. The dense, sand-choked current moved left to right over the unconsolidated, rippled lakebed. Some of the sediment was swept upward and into the flow, forming these spectacular flames. White Bluffs, WA. S.W. Cooley photo.

T-shaped mudsquirts are syndepositional structures. Rapid deposition of a sand bed on top of soupy lake bottom muds triggered the rise of t-shaped mudsquirts. Note how sand swirls with the mud dikes. The deformation is a result of rapid deposition and loading during a flood, not strong shaking. The gray sand on top, dumped by a Missoula flood, temporarily disrupted quiet-water deposition in Glacial Lake Columbia. Sanpoil Valley, WA. S.W. Cooley photo.

Flood and repeat. Three flood rhythmites exhibiting an identical bedform progression provide clear indication of repeated, syn-depositional fluidization in the soupy, upper portions of Touchet Beds. Ringold Rd, WA. S.W. Cooley photo.

Wrinkled lakebeds of Glacial Priest Lake. The wrinkled zone is deformation caused by bedding parallel slip, the result of mass wasting, not worrisome seismicity. Priest River, ID. S.W. Cooley photo.

Silt-sealed Cracks and Hydrofracture

Sand-propped hydraulic fractures are used to stimulate oil and gas reservoirs, a procedure commonly known as "fracking". Fracturing is induced by shutting in a portion of the well, adding a proppant slurry (sand + water + chemicals), and using pumps to jack up the fluid pressure until the formation yields. When the rock surrounding the well bore fails fluid-driven fractures propagate outward. The pressurized proppant slurry fills the expanding fractures and holds them open, permitting hydrocarbons to flow back to the well. The network of propped fractures exponentially increases the surface area of a well.

Unlike fracked formations at depths of hundreds to thousands of meters, the sediments that host the dikes are surficial deposits, unconsolidated and sandy that lack low-permeability layers that might act as a seal. Two key factors explain the formation of the dikes in Missoula flood deposits: high strain rate (rapid loading) and the rapid formation of silt skin walls on fracture walls (sealed pressure vessels).

• Rapid loading - When loaded by a catastrophic flood, pressure in the shallow subsurface built so rapidly that hydraulic fracturing was induced. The normally loose (ductile) material failed in the brittle mode at the high strain rate. Pressure rose above that required for fracture and exceeded the sediment's capacity to dissipate pressurized fluid through its pore network. Rapid loading alone appears adequate to initiate fracture in a low tensile strength material with no true seal such as the Touchet Beds.

• Silt skin walls - Once fractures began to form and fill, silt skin walls began to build. The sealing effect of the low-permeability skins delayed leakoff and facilitated further fracture. New fractures as well as natural flaws (cracks, soil macropores, burrows, etc.) provided low-resistance routes for new fractures to follow. Fractures immediately filled with sediment, the injected slurry (a natural proppant) sourced from within the overriding flood. Dewatering (leakoff), integral to the formation of the skin wall, begins the moment sediment enters a fracture. The skin-sealed crack begins to behave as a pressure vessel almost immediately. Pressure inside of the sealed fracture (pore fluid pressure, Pf) rises until it exceeds the confining strength of the material (Pf > 03) and the crack tip advances or, if leakoff loss exceeds Pf, the fracture closes. The fracture propagates in the 01–02 plane (vertical), widening against 03. As the fluid pressure equilibrates to the confining pressure (Pf = 03), the fracture tip halts, filling ceases, the crack closes down on its proppant, and pressure begins to build again if the load is still present. Each time Pf exceeds 03, the tip jumps forward or a new fracture initiates nearby. With each increment of widening, fluid pressure drops (volume increase = pressure decrease), but soon rebuilds. This loading-driven crack-fill-seal cycling created the dikes’ vertically sheeted fabric. 

Sheeted infill illustrated. Fluid pressure-driven crack and fill (crack volume cycling) is shown at the scale of a dike (nearfield scale) during a flood loading event. Time steps 1 through 12 in the pressure-time curve correspond with crack tip locations. During overloading, dike growth corresponds with pressure-volume cycling where fluid pressure remains between the minimum and maximum principal stress values. Silt-sealed fractures become sheeted dikes in my study area and in other geologic settings where silt is present and similar overloading has occurred (lahar, grounding glacier, debris flow, etc.). Diking seems to have occurred twice during a flood-load event. The first is the initial onrush of the overland flood (or backflood). The second occurs once a slackwater lake has formed (sustained load). Gravelly or sandy dikes are likely produced by the overland flood, while silty-sandy dikes result from the lake.While hydraulic fracture is well understood, the development of wall-parallel laminae (sheeting) in clastic dikes by a combination of rapid loading, hydraulic fracture, and silt wall seals, as I've illustrated here, has not previously been described in detail. Figure 9 in LeHeron and Etienne (2005) is the closest I've seen. My illustrations are not copied from anyone; I created them to clarify my thoughts and convey my argument to others.

Sheeting forms pulse by pulse. Cyclic fluid-driven fracturing produces vertically-sheeted dikes. In this example, four pulses produce 3 dikes that pioneer new paths (1a, 2b, 4a) and 5 dikes that intrude alongside them, exploiting established paths (2a, 3a, 3b-c, 4b, 4c). Sheeting may develop during a single flood (compound dikes sourced from one flood bed) and by repeated flooding over time (composite dikes sourced from multiple beds). Dikes in this example are all filled by 'Bottom Sediment' of a single flood and, therefore, are single-fill and compound dikes.

Near field and far field fracture? I modified the standard stress-strain curve to illustrate my tentative concept of diking at two scales, near-field and far-field. Near-field = Scale of a single dike (centimeters). Far-field = Scale of an outcrop containing multiple dikes. The red curve describes the flood load imposed over a broad area (meters). The blue curve describes cyclic pressure pulses that occur at a finer scale (millimeters to centimeters). The gradual drop in the red curve reflects the declining force of a passing flood and, to a lesser extent, leakoff to the formation. Abrupt drops in the blue curve reflect the cyclic hydraulic fracturing (opening and filling) during the peak of loading. Volume increase and decrease cycles = repeated injection. Leakoff continues after hydrofracturing ceases as the entire system (dikes and host) equilibrates to the formation pressure.

Flood counts and the development of vertical sheeting. Stacks of rhythmites (Touchet Beds) accumulated to different thicknesses in different parts of the Channeled Scablands. Many scabland floods followed different paths from others, therefore, rhythmite counts vary depending on location in the floodway. The most complete rhythmite sections occur in protected slackwater valleys inundated by Lake Lewis, Lake Condon, and Lake Allison. All floods flowed through Wallula Gap and slackwater lakes that ponded there were the deepest anywhere in the region (>200m). Consequently, full rhythmite sections in the southern Pasco Basin, easternmost Umatilla Basin, and western Walla Walla Valley contain the largest composite dikes. Dikes widen by the addition of new sheets of sediment, so their widest portions occur lower in the section rather than near their tops. Large dikes can appear to taper upward because younger fills tapping successively younger flood beds intruded alongside older fills. Sheet counts on composite dikes do not match flood counts, but scale proportionally with them. Sheets counts up to ~10 are common for compound dikes (single flood, one rhythmite). Sheet counts for large composite dikes can exceed 100 (multiple floods, many rhythmites). Lake Lewis filled the Walla Walla Valley (Waitt, 1980; 1985), Lewiston Basin (Bretz, 1929 field notes; Webster and others 1982), and Tucannon Valley (Smith, 1993) to ~366 m elevation. Lake Condon filled the Umatilla Valley (Benito and O'Connor, 2003), Willow Creek Valley (Cooley, 2015), and Sixmile Valley. Lake Allison filled the Willamette Valley (Glenn, 1965). Lake Columbia filled the Upper Columbia Valley (Kiver and Stradling, 1982; Hanson and Clague, 2012), Sanpoil Valley (Atwater, 1986), Latah Creek Valley (Rigby, 1982; Kiver and Stradling, 1982; Waitt, 1983; Meyer, 1999), and Foster Coulee (Russell, 1893). Floods that ponded in glacial Priest Lake deposited rhythmites today exposed at Peninsula Road near Priest River, ID (Walker, 1967; Breckenridge, 1989). Varved beds in glacial Lake Missoula are found at Lightning Creek near Clark Fork, ID (Breckenridge and Othberg, 1998), Clark Fork Valley (Smith, 2004, 2017), Missoula area (Chambers, 1971; Hanson and others, 2012), and Mission Valley (Levish, 1997). An excellent summary of varved deposits is provided by Waitt and Atwater (2023, Fig. 17).

Evaluating Proposed Origins

In this section, I briefly evaluate seven proposed origins based on my observations and the literature.

(A) Desiccation hypothesis - Grolier and Bingham (1978) speculated that the dikes might be large mud cracks, filled by sediment that was blown in by wind or washed in by water. The idea was not revisited in subsequent reports by them or anyone else. Rightly so. Little evidence supports the dikes formed by desiccation. Their shape and nearly all internal characteristics are fundamentally at odds with the gravity-driven infilling of meters-deep cracks.

(B) Ground ice hypothesis - Many believe Eastern Washington's climate during the Last Glacial was 'periglacial' and the landscape was 'tundra'. But what do these terms mean? Periglacial is defined by the presence of ground ice (permafrost). Tundra by the absence of trees. Pleistocene paleosols are abundant in the region and indicate soils were never perennially frozen. There is no evidence permafrost or an active layer ever existed. Palouse soils contain abundant evidence of continuous life from ~2 Ma in the form of plant roots, rodent burrows, cicada burrows. Backfilled burrows in the Touchet Beds attests to rapid recolonization between outburst floods. No mention of relict soil wedges, frost stirring, or gelifluction is found in NRCS Soil Surveys for the Colville Indian Reservation (NRCS, 2002), Okanogan County (NRCS, 2010), Chelan County (USDA, 1975), Douglas County (NRCS, 2008), Grant County (USDA, 1984), or Lincoln County (USDA, 1981). Frost wedges in varved lake beds of the Glacial Lake Missoula basin (Chambers, 1984; Chambers and Currey, 1989; Levish, 1997; Hanson and others, 2012; Hanson, 2013; Smith, 2014, 2021) have not be found in the Channeled Scabland. Pollen samples from lake bottom cores indicate cold-tolerant plant species and conifers persisted during the Late Wisconsin (Blinnikov and others, 2002; Whitlock and Brunelle, 2006). Mammoth that roamed Columbia Basin were nourished by steppe-grassland forage, not tundra plants (Fry, 1969; Last and Barton, 2014).

No Pleistocene permafrost in Washington. Compiled climate-proxy information indicates permafrost never formed in the Channelled Scablands or Okanogan Highlands during the Late Wisconsin despite their proximity to glacial ice. Pewe (1983) imprecisely grouped the Palouse Hills into an "alpine permafrost zone" on the basis of loess, not ground ice features. Whatever zone of periglacial activity existed along the former margin of the Okanogan Lobe (Withrow moraine area), it was remarkably narrow and nearly absent of ground ice features (Murton, 2020). The orange region is one I defined from historic data. It covers the area where temperature has dropped to -21 degrees C (-6 deg F). Pleistocene cirque elevations in the Rocky Mountains (Pierce, 2003, Fig. 1) project well above the crests of Yakima Fold Belt ridges. Mima mounds, abundant in the Scablands, are not unique to periglacial landscapes or diagnostic of deep cold (Busacca and others, 2004). Mima mounds are found from central Mexico to the Arctic and some mound fields in Washington date to the Holocene. Mima mounds indicate wind, dust, sparse tree cover, and aridity, but not cold. Thin, frost-formed cracks are occasionally found on exposed ridges of the Palouse and Umatilla Plateau and a few rock glaciers still linger in cold hollows east of the Cascade divide (Lillquist and Weidenaar, 2021), but neither indicate Pleistocene temperatures approached those of the modern Arctic.

Ice wedges confirm permafrost and tundra. Hundreds of studies have been published on active ice wedges and fossil ice wedge casts in Arctic permafrost (Lachenbruch, 1962; Pewe, 1973; Romanovskij, 1973; Mears, 1987; Yershov, 1998; Bockheim, 2002; Murton, 2020, etc.). Fossil wedges found in the Lemhi Range of Idaho (Butler, 1984; D.R. Butler photos and written communication), in lava caves of the Snake River Plain (Dort, 1968; Butler, 1969), in terrace gravels near Lewistown, MT (Schafer, 1949), in prairie soils near Browning, MT (unpublished field notes by the author) and High Plains near Laramie, WY (Grasso, 1979; Mears, 1981, 1987; Nissen and Mears, 1990; Munn and Spackman, 1991; Dillon and Sorenson, 2007) are not present in the Columbia Basin, Palouse, Cascade Range, or Okanogan Highlands. In fact, all climate archives describe the region at LGM as a 'cold steppe', with widespread bunchgrass, sagebrush, and pockets of pine forest (Spencer and Knapp, 2010). Eastern Washington never lost its trees. The terms 'periglacial' (ground ice) and 'tundra' (no trees) do not really apply, though some continue their use (O'Geen and Busacca, 2001; Gaylord and others, 2003).

Ice wedges active and relict. Ice wedge (left) in thick silt at the USACE Permafrost Tunnel Research Facility near Fairbanks, AK (www.erdc.usace.army.mil/CRREL/Permafrost-Tunnel-Research-Facility). A fossil ice wedge cast (right) in sandy alluvium in northern Europe. Photo by Richter/Freiberg Instruments.

Wedges in western Montana. Fossil frost wedges in Glacial Lake Missoula lakebeds in the Clark Fork River Valley, MT (Chambers, 1971, 1984; Chambers and Curry, 1989; Levish, 1997; Hanson and others, 2012; Hanson, 2013 Photo B; Smith, 2014, 2021). These small wedges formed on the lake bottom during lowstand periods (i.e., shoaling, subaerial exposure).

Soil wedges in eastern Montana. Soil wedges are occasionally found in treeless prairie soils near Browning, MT. Relict features like these bear no resemblance to Touchet-type clastic dikes, but are useful in defining the former limits of perennially frozen ground along the terminus of the Laurentide Ice Sheet (French, 2017; Murton, 2020). Similar soil wedges have not been found along the former margin of the Cordilleran Ice Sheet in Washington, some 450 km to the west. Roadcut along Hwy 89 between the Two Medicine River and Badger Creek. S.W. Cooley photo.

(C) Lateral spreading hypothesis - Lateral spreading involves liquefaction at depth, horizontal block slides, and extensional cracking of the ground surface. The two figures below are classic examples of surface cracking resulting from spreading to a 'free face' (Cruden and Varnes, 1996). Cracks formed by spreading can fill with sediment and form dikes. Such dikes tend to parallel to the headscarp or nearby drainage. Steep-sided channels are not found in Touchet Beds sections and little evidence of low-angle sliding as depicted by Cruden and Varnes exists.

Lateral spreading model. The standard landslide classification system used by geologists today was developed by Varnes (1978) and updated by Cruden and Varnes (1996). The system formalizes terminology associated with the various modes and features of mass wasting, including lateral spreads. Two classic examples of lateral spreading are shown above. A.) Blocks of sandstone sliding atop a weak shale layer tilt and separate along vertical fractures containing rubble (Zaruba and Mencl, 1969). Ostensibly, dikes would fill those fractures. B.) Spreading of clay-rich glaciomarine sediments at Anchorage, AK triggered by a massive earthquake in 1964 (Seed and Wilson, 1967). By this model, clastic dikes would rise as liquefied sands from below to fill vertical extension fractures between the more or less coherent blocks.

Lateral spreading triggered by a major earthquake. Large portions of the Turnagain Heights neighborhood located near downtown Anchorage slid into Cook Inlet during the magnitude 9.2 Great Alaska Earthquake in 1964. Rooftops of houses at upper left provide scale. The event lasted nearly 5 minutes and produced a rupture >950 km long. Note the NE-SW orientation of the large fractures and scarps. The 1964 quake struck on Good Friday and remains the largest ever recorded in North America. Major changes in zoning and construction standards were implemented following this quake (see USGS Professional Papers 541 through 546). Nothing comparable has struck Eastern Washington. Aerial photo from the W.G. Sprowls archive at University of Alaska (Number UAA-HMC-1464-F1-41).

Spreading vs. Liquefaction vs. Injection. (A) Lateral spreading scenario - Channel incision removes support and creates a free face that facilitates lateral spreading. Sliding and block separation atop a weak layer at depth creates wedge-shaped, vertical fractures that fill with sediment. (B) Liquefaction scenario - Seismic shaking triggers liquefaction in a wet, sandy layer and clastic dikes that taper upward. Some dikes reach the surface and vent liquefied sand in cone-like sheets (sand blows). (C) Loading-triggered hydrofracture scenario - A large vertical load imposed by a megaflood (overland flood, slackwater lake) increases pore fluid pressures in the underlying sediments and triggers hydraulic fracture. Fractures propagate downward through the vadose zone and immediately fill with sediment sourced from the base of the circulating flow. Fractures are propped by sandy fills become wedge-shaped clastic dikes. Internal sheeting develops during each flood event (pressure-volume cycling) and new dikes form over time during subsequent floods. New dikes follow and merge with older ones.

Dikes formed by spreading along Lake Roosevelt. A set of wedge-shaped, gravel-filled dikes that unambiguously formed by lateral spreading to a free face is exposed near Hunters, WA. Unstable glacial lake sediments capped by outwash gravel comprise steep shoreline bluffs and show evidence of block topple into the Columbia River channel. Spreading opened wedge-shaped fractures between blocks that were passively filled from above with outwash gravel. These dikes check all the boxes for lateral spreading - free face accommodation, block translation, and a slide plane at depth. The dikes at Hunters are a local phenomenon. They are unusual and not representative of most dikes in the Channeled Scablands. Similar dikes are described by Montenat and others (2007, Fig. 20). More about the geology of the shoreline bluffs near Hunters Campground HERE.

Lake Roosevelt shoreline. Wedge-shaped dikes filled with sand and gravel sourced in the overlying outwash intrude varved beds of Glacial Lake Columbia near Hunters, WA. S.W. Cooley photo.

(D) Seismic shaking and liquefaction hypothesis - According to some, earthquakes along the Olympic-Wallowa Lineament created the clastic dikes in the Channeled Scablands. This requires the dikes to have formed by liquefaction, in whole or in part. The dikes are therefor feeder conduits to sand blows. However, the dikes bear no resemblance to liquefaction features described in textbooks (Allen, 1982; McAlpin, 2009), review articles on earthquake-caused soft sediment deformation (Dzulzynsky and Walton, 1965; Van Loon and Brodzikowski, 1987; Nichols and others, 1994; Obermeier, 1996; Montenat and others, 2007; Owen and others, 2011; Shanmugam, 2017), or >200 papers on liquefaction published in journals over the past two centuries. No liquefaction features have been found in more than a dozen seismic trenches excavated across active faults in Eastern Washington. Liquefaction-related clastic dikes in Washington have only been found in Holocene floodplain sediments west of the Cascade divide and fluvial-tidal deposits along Pacific beaches (Craig and others, 1993; Dickenson, 1997; Obermeier and Dickenson, 1997; Peterson and Madin, 1997; Atwater, 2000; Atwater and others, 2005, 2015).

(E) Flood-generated vibration hypothesis - The Missoula floods would have produced a tremendous rumble as they coursed through the landscape. The repeated cataclysms must have terrified humans and animals who witnessed their passage. The audible roar was likely accompanied by a vibratory resonance established in the bedrock and sedimentary cover, affecting broad areas beyond the advancing flood. Fracturing of sediment and bedrock by vibration is purely speculative as Icelandic jokulhlaups of the past century do not appear to produce damaging vibrations and no evidence of flood-related vibrations is known in geologic record of glacial regions more broadly. Research into seismicity generated by large floods is purportedly underway at Université de Grenoble-Alpes, France by Kristen Cook, Florent Gimbert, and Alain Recking.

(G) Hydraulic fracture triggered by floodwater loading (overloading) hypothesis - The field evidence is most consistent with rapid loading by cataclysmic floods. Fracturing and perhaps a bit of horizontal drag were triggered by the weight and velocity of the floods and slackwater lakes. Diking and flooding appear closely linked in time and space.

Rapid overloading and hydraulic fracturing go together. Soft sediment deformation features, including clastic dikes, are known to form during rapid sedimentation and overloading. Dikes are widely documented in both seismically-active and low-seismicity areas and in a variety of geologic settings and deposits. While earthquakes commonly trigger liquefaction and can a produce clastic dikes, they are most commonly unsheeted, upward-tapering, and small (i.e., feeder dikes to sand blows). Dike morphology differentiates liquefaction-type dikes from other types. Geologists in the field can use taper direction, internal characteristics, and connection to source bed to distinguish each type, trigger, and origin. Not all clastic dikes are seismites. Figure modified from Shanmugam and others (2016, Figure 16).

Dike-sill-dike. Evidence of both fluid-driven fracture (hydraulic fracture) and Darcian flow (matrix flow) is present in clastic dikes in the Touchet Beds. An understanding of how hydraulic fractures initiate and propagate is necessary to understanding the clastic dikes in the megaflood region. Loading-triggered hydraulic fracture can be taught at the undergraduate level, even by professors who must learn it for the first time themselves. Read Jolly and Lonergan (2002), Ravier and others (2015), Bons and others (2022), and articles on sand injectites from the folks at Aberdeen. Students will pick it up. If you avoid the physics, then you will forever flounder about with stories involving earthquakes, liquefaction, and the OWL. S.W. Cooley photo.

Conclusions

This article summarizes my work on sheeted clastic dikes in the Channeled Scablands over the past 30 years. I've documented and measured thousands of dikes at hundreds of outcrops throughout the floodway region. All of the dikes occur within the margins of Ice Age floodways and are identical with respect to size, shape, sedimentology, and age. All formed by the same mechanism: rapid loading and hydraulic fracture triggered by tsunami-like megafloods moving overland and ponding against bedrock constrictions for month-long periods. Deep, fast-moving floods and slow-draining slackwater lakes imposed enormous loads on sedimentary and bedrock substrates, opening wedge-shaped fractures that rapidly filled with sediment sourced from circulating bottom currents and the soupy lake bottom. Vertical sheeting, the result of crack-and-fill cycling, occurred during each flood event (producing compound dikes) and during successive flood events over time (producing composite dikes). Fluted skin walls, downward taper, and lack of a connection to underlying beds indicate dikes were filled from the top and leakoff began the instant sediment entered the fracture. The largest dikes occur where floodwaters were deepest and rhythmite stacks thickest. Dike widths scale with rhythmite counts and sheet counts. Unlike most clastic dikes in the literature, the features described here did not form by liquefaction or seismic shaking; they are not feeder conduits to sand blows. Clastic dikes in the Channeled Scablands are flood injectites, not seismites. This study confirms a hydrofracture origin (Pogue, 1998), validates most field observations by Jenkins (1925), Lupher (1944), and Black (1979), but rejects slumping (Baker, 1973), lateral spreading (Cooley et al., 1995), and lake drainage (Newcomb, 1962) as viable origins.

References

See THIS POST

---

This article expands on one I published in Northwest Geology v. 49 in August 2020. Northwest Geology is the yearly journal published by the Montana-based Tobacco Root Geological Society in conjunction with the TRGS field conference (www.trgs.org). Online version was first posted here 15 Sept 2020.

Keywords: clastic dike, sand dike, sand injectite, hydraulic fracture, Channeled Scablands, Missoula floods, megafloods, Columbia Basin, Cordilleran Ice Sheet, Touchet Beds

LAST UPDATED: Aug 2026