Stronghold of the Ice Crawler

Author: Steve Stampfli, White Salmon, WA, USA,

The occurrence of currently active geologic and weather processes that result in formation of both summer cold and winter warm temperature geological zones (and biological habitats) in Shellrock Mountain talus slopes is certain.   But in addition to that, it now appears possible that a second western Columbia River Gorge area could harbor several acres of relict periglacial terrain that dates back many thousands of years.  That seemingly bold statement is supported by growing biogeographic evidence.

Evidence of the second periglacial area arose via a March 6, 2018 email from biologist Jim Kirk, who had recently read the first article in the GorgeScienceShare blog “Ice Mountain – A Theory of Why Pikas Exist in the Columbia River Gorge”.   Kirk is an experienced field biologist who worked during the early 1980s describing the distribution of plethodontid salamanders in the western Gorge.  His email concluded with the amazing statement that he had trapped grylloblattid insects (ice crawlers) just above the Columbia River and west of Shellrock Mountain in February 1983… some 36 years ago.

It appeared that Kirk had discovered one of nature’s rarest and elusive insects.  But even more remarkable was the fact that the discovery was made far from their typical mountain, snowfield and glacial haunts, at an almost sea-level elevation in the western Columbia River Gorge.

Figure 1.  Recently molted immature ice crawler (Grylloblatta) from a lava tube cave near Mt. St. Helens in the state of Washington (photo courtesy of Joe Warfel).

~ The Ice Crawler ~

Ice crawlers are members of one of the world’s oldest insect orders, Notoptera, which date back 200 -250 million years to the Permian period of the late Paleozoic era.   When Notoptera first appear in the fossil record, the earth’s climate was typified by high carbon dioxide levels, and much warmer temperatures than today.  It was also a time when a huge amount of forest vegetation was being produced, and its carbon photosynthetically fixed and geologically sequestered.

The fossil record shows that early members of the order had two pairs of large wings that enabled them to travel widely throughout the hot and tropical forests, feeding on pollens from early conifers and other early plant types.  They filled a pollinator role in the primordial forests, much in the same role as today’s bees and wasps.  But then, a series of changes in the earth’s climate and biology occurred that would change the planet’s evolutionary path forever… the advent of flowering plants.

As Harvard biogeographer, writer and photographer Piotr Naskrecki wrote in his 2010 “The Smaller Majority” blog, “gradually, they (ice crawlers) disappeared from the fossil record. Strangely, there is not a single ice crawler known from the period after mid-Cretaceous. Their disappearance coincides roughly with the appearance of flowering plants, or angiosperms, and the nearly concurrent diversification of beetles and other plant pollinators. It seems that rather than allowing themselves to be outcompeted by this new army of more advanced plant-associated insects, the ancestors of ice crawlers found survival in a completely new lifestyle”.

Instead of being forced into extinction by the myriad new insect forms tailored to angiosperms, the ice crawler’s ancestors took advantage of previously unoccupied habitats that were forming on the then cooling earth.  These habitats amounted to the periglacial environments forming around the margins of glaciers and snowfields, plus the interior of frozen rock fields and icy caves.  In adapting to these cold and largely subterranean environs, the early Notoptera shed their wings and adopted a lifestyle based upon hunting other insects, and scavenging what food was carried into their range by wind and gravity.  Rapid decomposition and lack of inundation by fine particles in their new rock and snow haunts was not conducive to fossilization of their remains, hence the seeming disappearance of grylloblattids from the fossil record.

Today’s Grylloblattidae are a rare family of insects restricted to cold mountainous areas in western North America, and parts of northeastern Asia including Japan, both Koreas, China, far eastern Russia and south central Siberia.  As of 2018, the family exhibited just 33 species and 4 subspecies, within 5 genera worldwide. It belongs to the second-smallest insect order, Notoptera, along with the family Mantophasmatidae (rock crawlers).  All grylloblattid species are highly endemic, thus have very small geographic ranges (i.e., a median geographic area of only 179 square kilometers, or 79 square miles).  In North America, there are now 15 documented species and 3 subspecies, plus a number awaiting verification via physical and genetic analysis. North American distribution is limited to the states of Washington, Idaho, Montana, Oregon and California, and the provinces of Alberta and British Columbia.  (Note:  much of the technical information on grylloblattids presented in this article is from Schoville and Graening, 2013. “Updated checklist of the ice-crawlers of North America, with notes on their natural history, biogeography and conservation”).

The chief reason ice crawlers display such small endemic ranges and have such poor dispersal abilities is their evolutionary adaption to very narrow habitat temperatures (i.e., stenothermalism).  Almost all ice crawlers are found in habitats with rocky retreats that maintain steady cool temperatures and humidities throughout the year.  Fifty years ago, a pioneering grylloblattid researcher named Bill Kamp recognized that the cryophilic (i.e., “cold loving”) nature of the genus Grylloblatta along with its unusual distribution correlated with Pleistocene glacial advances.  He proposed that surviving populations would be limited to areas that were previously glaciated or at the edge of glaciers (i.e., periglacial environments) since the last glacial maximum some 21,500 years ago.  Ice crawlers are normally active above-ground during the night, but only when temperatures hover around freezing.  They show a preferred temperature range of 0-1° C.  The range associated with an undescribed species on Mt. Rainier, for example, shows acute temperature thresholds between -8.5 and 15° C.   This supports the colorful statement that human touch can transmit enough heat energy to kill the insects.

~ The Stronghold ~

Jim Kirk’s announcement of his 36 year old discovery was notable from the biogeography standpoint, since it was signal of yet another cryophilic and stenothermal organism being harbored in the taluses of the western Gorge.  Proving the continued presence of the organism would also support the odd probability presented in the first article that low-elevation permafrost conditions still exist hidden in the Gorge.

With a little research, I soon determined that there are several prominent entomologists in North America and northeast Asia who specialize in the study of ice crawlers, and who may be interested in understanding the significance of the animal’s occurrence in the Gorge.  One of these is Dr. Sean Schoville of the University of Wisconsin Department Of Entomology, who further specializes in the tools of molecular ecology.  Molecular ecology is the exciting, relatively new field that utilizes DNA analysis to unlock how new species evolved on our planet, and how they fit into the modern-day ecological framework.  Just as any of us are now able to use emerging genetics labs to uncover our own human history, scientists like Schoville use similar sequencing equipment to analyze insect DNA for understanding species evolution and ecologies.   (See for a well-written description of ice crawlers and the folks who study them).  Other prominent ice crawler researchers are located right here in Pacific Northwest, in the persons of Dr. Chris Marshall and Dr. Dave Lytle, both entomologists in Oregon State University’s Department of Integrative Biology.  This team recently uncovered two new grylloblattid species in the mountains of Oregon, including the snowfields around Mt. Hood (see this OPB Oregon Field Guide video segment .

After contacting Sean, Chris and Dave with details of my own work, and the report of Jim Kirk’s original discovery, all three agreed to help sleuth whether ice crawlers were indeed harbored in the Gorge, and if so what species are present and how they fit into the broader North American distribution.  In addition to these three, others who agreed to provide advice during the project included Jim Kirk, Dr. Scott Hotaling of Washington State University School of Biological Sciences, and Dr. Jeff Holmquist of UCLA Institute of the Environment and Sustainability.  The US Forest Service also agreed to become a partner via issuing a special use permit for grylloblattid collection activities, effective September 2018.

The very first visit to the site’s on-ground coordinates supported expectations of what I would find.  Most of the north-facing slope was dominated by well-sorted talus that was relatively free of fine-grained rock and soil particles.  Such “open” talus slope conditions seem prerequisite to free air exchange, and active slope cooling / warming mechanisms that can result in temperature-moderated habitats (see Ice Mountain article for description of Balch and chimney effects).  Infrared imaging currently being used at Shellrock Mountain is indicating that summer surface temperatures at the bottom of certain talus slopes are approximately 10° C cooler than temperatures slightly upslope.  Similarly during winter, mid-slope warm vents areas are being shown to have temperatures about 5° C warmer than nearby cold zones. The existence of habitats that display cold and hot zones in close proximity could be vital to organisms like ice crawlers, which are physiologically limited by both hot and cold extremes in summer and winter.  Importantly, LWIR imagery is showing that talus hot and cold zones can be only tens-of-feet of one another, therefore seasonally traversable by organisms with limited migrational ability.

But beyond the talus substrate, I noticed something related to the topography that I’d never seen before.  Kirk’s original trapping site and the surrounding thirty acres had somehow been shaped into a series of five, symmetrical east-west trending ridges and trenches.  The appearance was reminiscent of the artificial earthen structures built as fortification around hillforts and castles during Europe’s prehistory and Middle Ages (see Figure 2 and Figure 3 below).  If still present, Kirk’s anomalous low-elevation ice crawler population seemed to be harbored within a natural fortress comprised of a series of ramparts.  There was irony in the situation, since grylloblattids appeared to be taking refuge in a landform resembling the fortresses used by our own ancestors long ago for repelling invaders.  Of course, in the ice crawler’s case the refuge was likely providing stable temperatures in a larger world of much broader temperature extremes.  But that colorful wondering aside, the chief task for the moment was determining whether ice crawlers still occupied the site.

Figure 2.  Artificial earth ditch-and-rampart defenses from the Bronze Age at the Ipf Hill Fort, Bopfingen, Germany (from Dark Avenger at de.wikipedia ).

Figure 3.  Interior of 4th natural rampart at the “Stronghold” grylloblattid survey site, viewed from the east end (Stampfli photo).

The Columbia River Gorge grylloblattid survey was initiated on-ground in late October of 2018, and within one month the first ice crawler was trapped at the “Stronghold” study site.  This confirmed the validity of Jim Kirk’s historic observation, and happily signaled that this uniquely cryophilic (cold loving) and stenothermic (narrow temperature range) insect still inhabited the apparently ancient landform.

Sampling is on-going at the 30 acre site, plus some nearby talus slopes.  As of December 31, 2018, I had collected a total of 76 individuals for morphologic examination by Marshall and Lytle at OSU, and accession into the Oregon State Arthropod Collection.  Tissue samples for genetic analysis and species determination are also being forwarded by the OSU team to Schoville at the University of Wisconsin.  The western Gorge grylloblattid project is still in its early stages.  Therefore, the identification of what species are present is incomplete, as is understanding of the broader Gorge distribution and how Gorge species may relate to the higher elevation populations found on Mt. Hood, Mt. Adams vicinity, and elsewhere in the western US.

~ Periglacial Landforms in the Columbia River Gorge ~

There is no evidence that either continental ice sheets or Cascades Range alpine glacial conditions extended to low elevations in the western Columbia River Gorge during the height of the Pleistocene or later Holocene epoch.  The Laurentide ice sheet reached a point about 150 miles north of the western Gorge some 16,000 – 20,000 years ago, thus did not directly impact the area.  However, the massive ice sheet’s influence on western US climate, geology and natural history was enormous, given two primary factors.  First, the North American jet stream had become split, and the southern branch (and therefore the location of winter storm tracks) shifted south.  This southerly shift robbed the Pacific Northwest of its marine-derived moisture and moderated temperatures.  Second, due to the anticyclonic (i.e., clockwise) direction of winds that prevailed over the ice sheet, easterly “continental” wind patterns accompanied by dryness and cold became dominant (see Kathy Whitlock’s 1992 paper entitled “Vegetational and Climatic History of the Pacific Northwest during the Last 20,000 Years: Implications for Understanding Present-day Biodiversity”).  As a result of these two factors, plus the already cooling global climate, processes that supported the formation of periglacial steppe environments and landforms became dominant in the Pacific Northwest interior at the end of the Pleistocene.  From the geomorphology standpoint, mechanical weathering processes such as the freeze-thaw responsible for the formation of fractured rock deposits and localized permafrost conditions would have become very active during this epoch and later glacial re- intensifications during the Holocene.

Perhaps the most obvious example of a once large-scale, likely inactive periglacial landform in the Gorge is found in the Catherine Creek area east of Bingen, Washington.   This location exhibits large areas of what has been termed “fractal patterned ground”, and also what I believe to be a dormant or fossil rock glacier.  This feature would have once transported rock in an icy matrix down to the Columbia River from the 1,000 foot elevation scarps to the north.  Similar large-scale, but now likely dormant or fossil periglacial features can be witnessed just across the river near Mosier, Oregon.  These mostly unstudied and poorly understood landforms will be the subject of a future blog article.

There are at least two examples of relict, likely active periglacial landforms in the western Columbia River Gorge.  Both of these occur within the 20 mile reach between Shellrock Mountain and Multnomah Falls.  Not surprisingly, this is also the reach that displays anomalous low-elevation populations of the cold-adapted and stenothermal American pikas and grylloblattid insects.  The first example is the Shellrock Mountain talus slope described in the Ice Mountain article.  The second likely active feature is the subject of this current writing.

Over the course of several visits to the Stronghold site, I tossed around various explanations for how the linear ridge-and-trench (rampart) topography could have come to exist.  The question was also posed to a few professional geologists for their interpretations. To enable better examination, it was apparent that having a broader “looking glass” via the use of lidar imagery would be helpful.  Thanks to assistance of Hood River County GIS Coordinator Mike Schrankel, the below lidar image was soon acquired.

Figure 4.  Lidar imagery of suspected Pleistocene and Holocene pro-nival ramparts, being overtaken by more recent alluvial, landslide and talus fans, west of Shellrock Mountain, Oregon.  These ramparts represent at least one of the landforms now known to support grylloblattid (ice crawler) populations in the western Columbia River Gorge.

The first of the eventually rejected explanations for formation of the landforms shown in Figure 4 is some type of bedrock-source movement, such as rock falls, rock avalanches and/or translational rock slides.  The source of material for all of these mechanisms would principally be mechanical weathering of the currently 800 foot tall head scarps (cliffs) uphill of the features.  Typically, as bedrock scarps mechanically erode, they gradually shed rock, and “march” backward and usually upslope (in this case, to the south and toward the bottom of the image).  This results in stable, tapering talus and debris fans, and not trenches and ridges.  This holds true in all climatic and geological environments I am aware of.  It is also true that such large-rock dominated fans display high internal friction, and are not prone to rotational debris slides (slump-type failures), which could theoretically result in irregular ridge-like features on receiving lower slopes.  Note that the rampart features are mostly composed of well-sorted talus, and not the very mixed materials observed at surrounding fans of alluvium and landslide debris, which conceivably could be prone to mass movement and irregular downslope deposition.

A second currently rejected explanation of the Stronghold’s ramparts is that they were formed by a common periglacial process known as thermokarst deflation.  Just as “sinks” or depressions are formed in limestone terrains due to the dissolution of soluble bedrock from below, depressions can also be formed in permafrost terrains via the melting of subsurface ice.  Elongated and concave (dish-shaped) thermokarst depressions can form at the base of frozen talus slopes, when their basal ice lenses disappear during persistent warm periods.  I have found likely examples of such thermokarst at the 4,000 foot elevation on the Washington side of the Columbia River.  While the Stronghold seems to exhibit characteristics favorable for the formation of permafrost and the accumulation of subsurface ice, it is not clear why a series of parallel thermokarst depressions would result upon thawing.

The third hypothesis that might explain the origin of the Stronghold’s talus ramparts relates to the series of glacial outburst floods that occurred at the end of the Pleistocene epoch some 13,000 to 15,000 years ago.  Although far-fetched, the hypothesis is interesting to consider, and could deserve more careful consideration by geologists.  Many travelers on I-84 westbound have noticed what looks like a westward trending side canyon on the south side of the interstate west of the Shellrock area.  This appears to be an ancient 3.8 mile long high-water passage “channel” formed during the floods.  The channel passage climbs 400 vertical feet to a crest of 570 feet before dropping back down to the Columbia River.  At least one of the flood events may have reached an elevation of some 750 feet in this reach of the Columbia, thus easily filling (and perhaps creating) this channel passage via massive river erosion.  Interestingly, this projected flood crest is at, or a little below, the elevation of the Stronghold ramparts.  It is therefore conceivable that large blocks of ice floated by the floods could have bulldozed deep gouges into the talus dominated shoreline, and resulted in the striated surface seen on the lidar image.  Although it is hard to flatly reject the possibility that ice age floods bulldozed the Gorge’s Pleistocene shorelines, it is unclear how such gouging action could have resulted in such a well-spaced series of relatively uniform gouges.

The fourth possible, and I think most likely explanation of the Stronghold terrain, is that it represents a series of relict periglacial landscape features known as pronival ramparts.  Such ramparts are simply ridges of moderately well-sorted talus debris that was originally shed from an overhanging headwall, landed atop a steep snowfield, rolled down the snow/ice slope, and was finally deposited at the base of the snowfield.  Sorted talus is a material that geologists typically see just below headwall sources, so the first clue to the rampart’s periglacial origin is its anomalous location well downslope of the main cliff pediments.  Pronival ramparts account for such anomalous location of graded talus given the fact that the snow field ramps act much like a combination “grizzly screen /conveyor belt”, which “screens”-out smaller particles as the mixed material works its way down the snow slope.  Eventually, only the larger rocks find their way to the base of the snow field for final deposit.  Rock sorting on snow slopes is therefore very similar to that occurring on an angle-of-repose receiving slope, where fine particles (the lesser volume, in this case) settle to the slope first, while larger diameters (the bulk of the volume) roll or slide to the bottom.  Here they can form rampart features in straight or arcuate ridges, well downslope of any cliff base talus accumulations.  Note, however, that rock-on-snow transport can occur at low gradients, and the eventual point of deposition can be anomalously far downslope.

Figure 5.  Genesis of a pronival rampart (from

To enable formation of multiple and large pronival ramparts in the western Gorge, the presence of intermittent perennial snowfields would have been necessary, perhaps beginning in the late Pleistocene and ending as recently as the Little Ice Age (about 1900 AD).  The possibility is not hard to accept when realizing that deep depths of hard firn or ice pellets covered the Oregon side of the western Gorge as recently as the winter of 1884-85 (near the end of the Little Ice Age), and again 1922 (see discussion and historic snowbank photo in Chapter 4 of the Ice Mountain blog article).  It is logical to assume that even those recent and relatively small snow or ice fields could have lasted throughout following summers, given up to 20 foot depths of ice and the high degree of topographical shading on the Oregon side of the Columbia.

But what could explain the multiple parallel ramparts at the Stronghold site, instead of there being only one?  One idea is that the ramparts were deposited sequentially during the series of cold climatic periods that occurred in the late Pleistocene and Holocene.  As episodic cold periods favorable to mechanical freeze-thaw action ensued, the hard snow ramps necessary for rampart formation may have formed and lasted for decades or even centuries. They may have also been insulated from summer melting by veneers of insulating rock and soil, much as can be witnessed protecting the rock-veneered glaciers on Mt. Hood today.

It might, therefore, be reasonable to hypothesize that rampart formation corresponded to the major periods of cooling that began as early as 20,000 years ago, and ended as late as 1,900AD.  If glacial flood elevations did not rise above the lowest rampart (and thereby wash it away), it is conceivable that it was deposited during the first Cordilleran glacial maxima event (Evans Creek Stade) some 17,000 -22,000 years ago in the late Pleistocene.   Following this, a subsequent rampart could have formed during the second Cordilleran glacial maxima known as the Vashon Stade, some 14,000 – 14,500 years ago.  Following the end of the Pleistocene (and glacial floods), the Pacific Northwest entered a 3,000  year span of warming during the early-mid Holocene epoch (Holocene Climatic Optimum) between 6,000 and 9,000 years ago.  At the end of this warm period, the region experienced a new series of recent cooling periods that could correspond to subsequent rampart depositions.  A first Holocene “new glacial maxima” period occurred some 5,000 -6,000 years ago, which was followed by a second such period 2,500 – 3,500 years ago.  A third and final Holocene glacial maxima, known as the Little Ice Age, started in 1,350 AD and lasted until the current era of glacial retreat began in 1,900 AD.  That said, there seems no current way to determine ages of the five ramparts, and thereby sequence their formation.   Sequencing might be possible if a person had the ability to date the cliff and talus features in consideration of the past climates.

While the north and west slopes of Shellrock Mountain likely display currently active periglacial processes and permafrost, the Stronghold situation is less certain.  The strongest evidence in favor of permafrost at the site is biological, and the fact that the landform supports a population of insects that are believed to have dispersed along the advancing and receding margins of the Pleistocene continental ice sheet.  Additionally, these animals display physiologies that require current day habitats that never stray far from the freezing point.

Acquiring physical evidence of permafrost conditions within the Stronghold ramparts is in its early stages.  So far, fall-season LWIR imagery has indeed shown anomalously cold zones at the bottom of the trench features.  Additionally, the persistence of tiny interior patches of snow was noted during late fall 2018.  Both of these could be evidence of active Balch and chimney-effect processes.  Better knowledge will be available in the future, given the fact that sub-surface temperature loggers were installed in fall 2018 to track bi-hourly temperatures for the next several years.


I thank Jim Kirk, Sean Schoville, Chris Marshall, Dave Lytle, Scott Hotaling and Jeff Holmquist for their considerable technical, curatorial, lab, and advisory work  related to this project.  I also thank Steve Castagnoli for providing use of the entomology lab at OSU’s Mid-Columbia Agricultural Research and Extension Center in Hood River, and Hood River County GIS Coordinator Mike Schrankel for producing lidar images .

Comments on this article can be posted below, or via email at

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s