Before Helene’s remnants passed through western North Carolina, the boulder-strewn area in the photo above was covered with trees and buildings. A small stream flowed behind the wrecked buildings on the left of the photo. The damage seen here occurred suddenly on the morning of Friday, September 27, 2024, as a huge wave of boulders, trees, and mud surged down the small stream’s channel. This wasn’t a flash flood–it was a debris flow, a type of fast moving, fluidized landslide associated with heavy rainfall. The extent of the damage from the debris flow is visible in the before/after GIF below. The photo above was taken near the top of the GIF images, looking towards their bottom. The large building labeled above is visible near the bottom of the GIF images.

The small stream is visible trickling through the damage swath; water flooding alone from a stream this small could never approach the level of damage caused by the debris flow. Fortunately, no one was seriously injured in this particular debris flow, but many lives were lost in similar events elsewhere during Helene. Understanding these particularly dangerous landslides is a big part of storm safety in southern Appalachia. So, what are debris flows, where do they start, and what makes them so dangerous?

What are debris flows?

Debris flows are fast-moving, highly mobile, fluidized landslides transporting saturated soil, boulders, and trees downslope. They are specifically associated with saturated soil, which results from heavy rainfall in our region. Debris flows move like a liquid but contain large amounts of solid material–65% solids (rock, soil, and wood) is an average composition, with the rest of the flow volume being water mixed into the soil and rock. Often called “mudslides,” debris flows also carry large trees and boulders. Because of their solid content, debris flows are approximately twice as dense as water, so they hit with destructive force. They follow ravines and small stream channels downslope due to their fluid consistency, spreading over wider areas at the base of slopes until they lose their momentum. The sketch below gives a general idea of a debris flow’s start-to-finish journey downslope, from its beginning in a steep hollow to its damaging end on flatter ground at the base of the mountain.

Where and how do debris flows start?

Debris flows frequently start in steep hollows, or slightly concave slope areas, above the headwaters of small streams. They can also begin on road embankments, or any other landscape feature that can initiate a small landslide on steep ground. Debris flows are specifically associated with heavy precipitation and saturated ground. When a landslide starts in saturated soil, it liquefies and accelerates. When the liquefied landslide hits the saturated soil in its path, this soil liquefies as well and is added to the debris flow. Much like a snowball, debris flows accumulate more and more debris moving downslope, adding to their volume and destructive power. The GIF below shows the basic idea of debris flow initiation in an area like the one indicated by the red box in the sketch above. Note that the debris flow starts in rocky soil beneath a cliff, where rock fragments have accumulated to the point of instability. Due to saturation, once the slide starts, it liquefies, and then liquefies the soil in its path.

How do debris flows move?

Debris flows follow ravines and stream channels due to their liquefied condition, typically scouring large amounts of soil and stream sediment on their way down. Moving within the stream channel keeps the flow confined and intact. Collisions between soil and rock particles help keep the flow liquefied. A confined, thicker flow also loses less of the water trapped within it. Debris flows can move very quickly, often at speeds of 20 mph or more. They frequently run up onto the slope on the outsides of bends due to their speed. Their width greatly exceeds the width of the stream or creek whose channel they follow. The conceptual sketch below illustrates the scouring process as well as debris flow size relative to the “usual” stream in the channel, even during high water flow.

The scouring created by debris flows can be quite impressive; it greatly exceeds the potential for water erosion by small headwater streams. The picture below (taken upslope of the first photo in this post) gives an idea of what the effects of the scouring look like.

The leading edge of the debris flow contains trees and larger boulders picked up by the debris flow through its scouring action. Smaller cobbles and mud trail behind. Even small debris flows can transport surprisingly large boulders and trees due to the density of the fluidized soil (it “floats” boulders), making a debris flow strike on a structure incredibly destructive.

When debris flows exit tighter channels or ravines onto flatter ground, they often spread out, but remain mobile and destructive for some distance. In western North Carolina, many tight stream channels open onto flatter areas at the base of the steep slope. These flatter areas are older, accumulated debris flow deposits. The GIF below shows debris flows exiting a stream valley and spreading onto a flat deposit area, where buildings are destroyed. Though an unpleasant thought, this sequence of events played out many times during Helene (as well as during many other storms in our region’s history). The satellite photo below the GIF shows a bird’s eye view of the debris flow where the first photo in the post was taken.

Debris flow deposits are an indicator that an area can experience debris flows and should be developed cautiously, if at all. This often seems counterintuitive, as the flatter slopes suggest safety from landslides. In reality, these flat deposit areas are a main indicator of possible debris flow hazard. People already living in such areas should be aware of the hazard during heavy rainfall. In southern Appalachia, about 5 inches of rain over 24 hours produces conditions necessary to make debris flows possible.

This is the first post in a series of Helene-related posts discussing landslide events during the storm. Posts will be a combination of remote sensing interpretation and first-hand, on-the-ground experience. Our goal is to increase understanding of what happened during this event and help folks plan for future hazard.

Additional discussion of how debris flows fluidize can be found in this video. It’s an interesting process that isn’t fully understood, but the basics are outlined here in greater detail.

Philip S. Prince

In 1902, E.W. Myers published an account of a 1901 flood (the “May Fresh”) and its effects on the mountain landscapes of western North Carolina. Included in the report is the following sketch of a debris flow, sourced in digital form from Anne Witt’s ResearchGate page. Notably, no indication of cardinal directions or distance to nearby landmarks are included in the sketch, making the exact location of the debris flow (whose indicated size is impressive!) uncertain.

Also excerpted in Anne’s work is Myers’ text description of the feature he drew.

I thought it would be interesting to try to find this debris flow based on the information Myers provides. With the sketch and text, Myers indicates that the debris flow entered Gouges Creek from a downstream-left direction, left a “heart-shaped” scar on the hillside, and was near a ridgeline and summit and generally very high within the surrounding landscape. In the era of high-resolution LiDAR imagery, these details turned out to be sufficient to locate the debris flow on the Mitchell County-Avery County line north of Spruce Pine, North Carolina. A LiDAR hillshade image of the slide, as well as a Google Maps image of its location, are shown below. The slide is not large in the context of the entire mountainside, but the heart-shaped scar from the initial failure (end of yellow arrow), downslope scouring, and proximity to the ridgecrest/summit are clearly visible.

Locating the debris flow was easier than expected. It is the only significant debris flow on Gouges Creek (not to be confused with Gouge Creek and many other Gouge place names in Mitchell County), and its distinct heart-shaped initiation zone distinguishes it from other debris flow features on nearby slopes. A close look at the “heart shaped” scar as revealed by LiDAR hillshade makes clear what Myers hoped to capture with his sketch.

A 2023 geologist of the Google Earth and drone era has to remember that Myers drew his 1901 or 1902 map sketch by interpreting from the ground, and he was quite successful in capturing the general layout of the landscape around the debris flow. No indication of his vantage point for making his sketch is given in the drawing of the text excerpt; he may have been along Gouges Creek or viewing from the ridge to the south. The images below compare the actual topography to Myers’ sketch, with labels on notable features. The light blue lines in the LiDAR image indicate faint channels with may generally locate the many small failure Myers illustrated. No significant trace of these failures is visible in today’s landscape.

Myers did overestimate the size of the failed area, but he didn’t do badly assuming he was visually estimating from a distance. The actual scar is a bit smaller than he suggested, but is still quite significant. The LiDAR detail below shows the real dimensions of the scar, assuming Myers length estimate was made from the top of the “heart” to where the failed area narrows to the scoured channel.

Myers also reported that the material which moved during the debris flow seemed to have “disappeared utterly,” indicating that the flow was quite mobile and did not deposit significant material in Gouges Creek near the failure itself. This observation is reflected in LiDAR imagery, where no significant deposit is visible in the area shown; instead, scouring continues down the Gouges Creek channel. The yellow arrow indicates the heart-shaped initiation zone.

Myers did not use the term “debris flow” (I don’t think it existed at the time!), but his description captured some principal components of debris flows, included the “excavated” initiation zone (“evacuated” would be used today), erosion or scouring of a track into the soil profile below the initiation zone, and the high mobility of the moving material. The simple physical model below shows this scouring process, during which the once-solid material that detached to produce the debris flow fluidizes and erodes the slope as it travels downhill. Compare the pre-flow “normal” channel texture to the post-flow scour. This erosion adds material to the flow, making it grow in volume. A key distinction between the model and a real debris flow is that the model uses dry, low friction materials; a real debris is a very “wet” feature that occurs due to soil saturation as a result of extreme rainfall.

LiDAR reveals other past debris flows in the area, including an impressive feature (left arrow, below) northwest of the Gouges Creek flow. The GIF shows the outline of the flow and its path from the initiation zone downslope.

This debris flow has a less impressive scar/initiation zone, but the flow did leave traces of a process known as “superelevation,” which describes a flow sloshing up the sides of bends in the channel carrying it due to its momentum.

The GIF below shows another physical model, which demonstrates superelevation. Note how the flow runs up the channel banks where bends are present.

I haven’t read the full Myers report, so I don’t know if the superelevating debris flow is discussed or if it is even from the 1901 “May Fresh”–there are plenty of other extreme precipitation events to choose from. In any case, the physical characteristics of the superelevating flow are sufficiently distinct from the Gouges Creek debris flow to distinguish them with LiDAR imagery with the help of Myers’ sketch and description.

The scar and track of the Gouges Creek debris flow are now located in an entirely forested landscape, and would be visible only from the ground due to tree cover. With LiDAR imagery, however, the feature is impressively visible and fresh-looking despite being 116 years old at the time of LiDAR data collection.

The Gouges Creek debris flow did destroy a log cabin; apparently it was unoccupied or uninhabited, and no one was hurt or killed. While modern land use patterns are different in the North Carolina Mountains, debris flows still represent a significant threat to life and property due to their extreme mobility and fast speed (you can’t outrun one). Were a debris on the scale of the Gouges Creek event to occur in a more populated part of the mountains today, the potential for very negative outcomes is quite high. Mountain residents need to be aware of the weather conditions that produce debris flows, as well as their potential paths, to stay out of harm’s way. Records of past events like Myers’ report are useful reminders of the potential for extreme rainfall and debris flows throughout the North Carolina Mountains, though events may be separated by years or decades in a given area.

by Philip S. Prince

*Note: There is indeed a sand model in this post, but it’s near the end…

Well, it’s more of a hilltop, but it has dropped several feet since iron mining was active on its lower slopes in the early 20th century! The summit of this unnamed hill southwest of Low Moor, Virginia, hosts conspicuous scarps that define a graben, or downthrown block, that has sunk about 6 ft (just under 2 m) relative to the surrounding hilltop. The graben is a bit over 300 ft (100 m) wide. The scarps are subtle but very clearly visible in 1-meter resolution, lidar-derived hillshade imagery. Results of iron mining are visible at lower right in the upper image below (more on this later, for sure).

The scarps (which face outwards and downhill) and counter-scarps (which face inward, or towards the upslope direction) do not appear to cut road grades on the slope. They do, however, offset a prospect trench within the subsided graben. The slope movement thus post-dates the trench, but was likely gradual and episodic over the course of months or years. I don’t know if movement would have been discernible during the period that mining was active, or if it completely post-dates work at the site.

Disturbance of the prospect trench puts movement on the scarps within, or after, the iron mining period, which is consistent with other iron-mine related features in Alleghany and Botetourt Counties in Virginia. I have written about these other iron mine-related landslides before (click here), but the summit graben shown here is unique because of its position on the slope and its 300 ft / 100 m width relative to the overall scale of the affected hill. Because of the graben’s position on the hill and its width, deformation is likely seated very deep within the hill. The slow-moving GIF below shows an interpretive cutaway of the slope, with shear planes sketched in from the shape of the slope and reasonable angular relationships.

Interpretive is the key word here…I have not even been to this site in person, so I am just inferring a possible subsurface structure from surface deformation and the size and shape of the hill and graben. That said, the graben is quite large with respect to the size of the hill, and its position so far from the toe of the slope definitely implies very deep-seated movement. This is certainly not a thin, translational bedrock-involved slide of the type that is common within the Appalachian Valley and Ridge. The outwardly-moving hillslope is not a dip slope in the first place, and the scale and position of the graben make a shallow translational explanation impossible. A slightly zoomed version of the subsurface interpretation is shown below. Note that the cut of the diagram is a bit out of plane of the oblique image and slightly exaggerates relief and the thickness of the moving area.

Where does the iron mining factor into the story? Highwall surface mining for limonite ores occurred along the foot of the hill in the early 20th century. The workings were not exceptionally large, but they did tend to make contour-parallel cuts along the foot of the slope. The largest of these (the one visible at lower right in the post’s first image) appears to have experienced a major collapse. The other cuts also show indication of slope movement, though it is less pronounced. Mined areas and their location with respect to the graben are shown below.

Even though the mined areas aren’t exceptionally extensive, they obviously destabilized the slope in proximity to the cuts; small parallel scarps are visible upslope of the mined areas in the image above. Presumably, this destabilization and associated slope movement turned out to be very far-reaching, allowing the whole hill to expand a bit laterally and force graben subsidence. The following GIF offers a simplified, cartoon-style interpretation of the process.

This style of slope movement suggests that hillslopes in the area may be near a deep-seated stability threshold that makes them susceptible to modest but volumetrically significant movement in response to very minor surface modification. This susceptibility may be a reflection of bedrock geology and structure. This area is underlain by a complexly folded and faulted sandstone and limestone interval with thick shales above and below. The result is shale-cored mountain topography that is “propped up” by the strength of armoring sandstones. Additionally, the area has experienced considerable geologically-recent relief production due to its location in the James River basin, so slopes have grown taller and steeper as valleys have lowered. Given these background conditions, potential slope behavior under disturbance could be expected to be different from the behavior of a more geologically consistent, lower relief landscape.

I think the most interesting aspect of this human-induced topographic “adjustment” is how a very modest excavation produced an equally modest slope movement, at least in terms of overall displacement of the moving mass of hillslope. None of the mining-related slope movements in the area appear to have produced rapid, long-runout slides or flows; all resulted in slight shifts that are really only discernible with lidar-derived imagery. All are reforested, and lidar imagery does not suggest any are subject to imminent significant movement. I tried to set up a simple physical model to capture this behavior. Two trials are shown below. The model hillslope is constructed of a cohesive sand-and-flour “shell” over a weak glass microbead core. With disturbance of the toe of the slope, a small, lurching failure occurs, reestablishing equilibrium and stability with a tiny but obvious movement.

Setting these models up to produce this style of failure was a worthy challenge. Their behavior is definitely material-specific; cohesive or loose sand alone (or microbeads) won’t do it. The slope has to be just unstable enough–and mechanically appropriate–to settle slightly with minor disturbance to the toe of the slope. The failures are not as deep-seated with respect to the slope as the real graben scenario, but they are deep enough to produce back rotation of the head of the slide with very little displacement, as shown below. Scarps and counterscarps are marked with hachured lines.

While these models experience rapid, one-time movement, the amount of movement necessary to reestablish stability is very small. This was the goal of the models, specifically as a way of connecting the slope’s mechanical architecture to failure behavior. A very brief YouTube video linked below shows the actual failures at full speed. It seems ridiculous to go to great lengths for such tiny movement, but it was much more challenging than producing a high-runout model, for sure.

While the summit graben feature is an extreme example, mine-related failures abound in this area. The image below shows other failures just up the valley from the graben feature. Again, cuts on contour appear to be the issue, but slopes responded with modest movement. In the image below, a large area of slight displacement is clearly visible above a highwall cut. The hill with the graben is just outside the field of view to the right. More examples exist just southwest of this one.

The location of these slope has largely spared them from modern-day engineering, but it is interesting to consider how they might respond to more carefully planned modification! It also invites comparison to the well-documented Rattlesnake Ridge landslide in Washington.

by Philip S. Prince

The two images below show the same model landslide at different points during its evolution. The lower image, of course, looks much uglier, with a broken-up slide block and huge, looming headscarp. Notably, the thick toe labeled in the upper image is conspicuously missing…(spoiler: you can watch this model evolve at 1:04 in the video link at the end).

So, how did the slide progress from a not-too-impressive feature to a mangled mess? The answer relates to that toe, or rather its absence in the lower image, with a few details of the slide’s shape thrown in (more on that in a bit). Toe removal, however, is the main idea, and removing a landslide’s toe invites trouble in the form of probable continued movement. A slide like the one shown above (that is, not a rapid, flow-type slide) can move until the toe becomes large enough to resist the driving load that led to slide movement, assuming that friction on the sliding surface remains low. Removing the toe allows driving load to overwhelm resistance again if friction on the slide plane is sufficiently low, leading to renewed movement of the slide. The GIF below illustrates this idea in basic terms. I-26 drivers in western North Carolina might appreciate the pink excavator…

A well-developed landslide toe can therefore be regarded as the “prop” that supports the rest of the upslope slide mass atop the weak failure surface. Altering a toe is often a fast track to instability and slide reactivation. The simple model shown in the GIF below demonstrates toe development, removal, and reactivation in a slide that behaves much like the one drawn above. When toe material is removed, the slide advances until enough supporting toe is restored. Note that this model is much “neater” than the model in the first images!

Continued toe removal leads to continued motion, particularly in the model above, as the slide mass does not break apart and continues to exert a strong driving load on its base. Once the slide mass is broken along its edges (lateral scarps), it no longer has the small amount of support provided by the cohesion (“stickiness”) of the model material. Real rock and many soils have some cohesive strength as well. It is never very much, but it does provide a component of resistance to sliding. Once the material is broken, however, this tiny bit of resistance is lost, and ongoing movement of the slide due to toe removal is that much easier. In a model like the one above, the entire slide block can be consumed by removing the initial toe and continuing to remove material pushed out into the new toe.

Weakness and potential instability associated with a landslide can persist for a long time after the slide becomes dormant and is not obvious in the landscape. Nearly “invisible” old slides that aren’t currently moving can be reactivated long after they stopped moving by disturbing the slide toe. Using lidar to identify such old slides is a key part of pre-construction site assessment.

Destabilization of landslides by toe removal is a very real thing, with actual scenarios–both large and small–arising from construction and natural stream or wave erosion. Many are slow and manageable, but they can be ongoing, hard to stop, and quite expensive over time. A famous example I am fond of is shown below. This is the “Galloping Highway” slide in Giles County, Virginia, along US 460. The toe of an older, dormant slide was removed for the highway cut, initiating literally decades of reactivated–and slowly ongoing–movement. This slide is, in effect, a real-world example of the previous GIF where an “invisible” slide came back to life. The slide block is nearly 1,000 feet across along US 460.

The slides shown below are in McDowell County, North Carolina. Both are situated for possible reactivation due to the road cut that has impacted their toes as well as steady stream erosion of the toes, which would pre-date the road cut. I imagine both have a reactivation history due to stream erosion; the smaller slide on the right appears to have reactivated more recently due to the road cut.

The next impressive example is along a railroad grade in western North Carolina. Lumpy areas just above the tracks look very much like recent reactivation, and a faint toe seems to be pushing out of the base of the cut. Two logging roads at the right of the slide are offset by scarps, indicating recent slide movement. The relative impact of the railroad cut and possible removal of material from the slide toe as it slowly pushes outward can’t be discerned from lidar, but the offset roads and ugly toe indicate this one definitely deserves engineering attention.

The slide above looks quite different from the first few examples and provides a nice way to circle back to the very first model in this post. The shape of a landslide’s failure (sliding) surface strongly impact what happens to the sliding mass as it moves downhill. A slide on a failure plane with constant slope can move intact, but a failure plane with changing slope requires the slide mass to constantly change its shape as it moves. If the failure plane becomes less steep, the top of the slide has to stretch to match its shape, forcing cracks or new internal scarps to form. The uppermost part of the slide can actually tilt backward until a depression forms (see above and below). The more the slide moves, the more exaggerated the cracking, internal breakup, and back-rotation of the upper part of the slide become. The GIF below illustrates the initial phases of the concept.

Obviously the gap below the slide block can’t exist…the collapse and stretching of the block are constantly developing as it moves. Physical models offer a nice way to show this process in action. Note how the cracking accompanies the backwards tilting of the upper part of the slide. This model is the last one shown in the video link.

So what happens to these early cracks if the toe is removed and movement continues? With the right failure surface shape, something like the very first model develops. This is where the video comes in. Linked below are more reactivated landslide models, including all those shown above. Some really fall apart with toe removal and continued movement; in others, a smaller and smaller block continues downslope with little or no internal break-up. I assure you that this is the ONLY place you will see this many model landslides in one place, whatever that may mean! The two styles of slide described in this post are represented. They can be distinguished by how much the slides break apart as they continue to move after toe removal. I think it’s cool to watch their motion all at once. It beats trying to piece together months or years of episodic movement!

by Philip S. Prince

Debris flows move fast, making them an exceedingly dangerous type of landslide. “Fast” is always a relative term, but in the case of debris flows, speeds of 30 or 40 miles per hour (~50-60 kilometers per hour) are quite reasonable. Debris flow speed is often sufficient to cause the flow to “run up” the slope on the outside of bends in the channel the flow is following. The result is a scoured track notably higher on the outside of the bend than on the inside. This is “superelevation”–the “sloshing” of a flow up a slope due to its speed and momentum when it encounters a channel bend. The sketch below illustrates the idea.

The sketch actually contains a number of superelevated areas due to the bends in the channel. Each is characterized by asymmetry of the damaged, scoured area–higher on the outside of the bend than the inside. The GIF below puts the process in motion. The key point to look for is the flow running up the slopes as it rounds the bends in the channel.

The GIF exaggerates the process for illustrative purposes, but real-world examples of the superelevation concept can be quite dramatic. The image below shows the results of a Jun 27, 1995, debris flow in Madison County, Virginia, near Graves Mill. The scoured area clearly “swerves” away from the blue stream channel due to significant superelevations around bends. Notably, scour from the flow reached 290 feet from the stream channel in the foreground, where moderate slope offered less resistance to superelevation. At the end of the yellow arrow, the flow superelevated about 52 ft (16 m) up a steeper slope.

The ~52 ft superelevation site was photographed soon after the event. A particularly impressive photo is shown below, and the source report (Eaton et al., 2004) can be found at this link. The small red dot in the image above shows the approximate location from which the photo was taken. Note that the flow scoured trees and soil away, leaving bare bedrock. Obviously, structures in the path of a flow like this would not stand a chance.

As indicated by the caption of the photo, superelevation height is used to estimate debris flow velocity, with the 52 ft run-up suggesting a ~45 mph speed. At this highest superelevation point, the flow had traveled about 3,100 ft (940 m) and descended 885 ft (270 m) in elevation. The image below shows the 52 ft superelevation point (yellow arrow) in relation to the initiation zone of the debris flow. Note the number of flows in the area, a result of the extreme but localized precipitation associated with an atypical thunderstorm (check out this link).

As shown in the earlier GIF, this flow produced a number of superelevations due to slight bends in the stream valley the flow followed. Bracketed yellow lines show them in the image below, with the 52 ft feature near the center of the image.

To the east of this flow, a smaller flow produced another impressive superelevation. The following images show the flow’s track with and without a bracketed yellow line to indicate the superelevated zone. I am not sure exactly how high the flow reached here, but Google Earth suggests the height approached 48 ft (15 m).

A notable aspect of the imagery shown thus far is that it was taken soon after the actual debris flow events, likely in 1995 or 1996, as no vegetation has returned to the scoured flow tracks. The area looks much different today, but 1-meter resolution lidar imagery preserves evidence of the superelevations. Seeing the scoured areas may take a bit of focus, but they are definitely there, and can be nicely matched to the 1995 (?) imagery. The GIF below transitions between the 1995 imagery, current imagery, and a lidar hillshade/slopeshade image.

Traces of the superelevation to the east are visible in lidar imagery as well. The GIF below cycles through the same imagery types. This example is equally subtle but readily identifiable once you know it’s there.

In areas affected by debris flows sufficiently long ago that forest recovery largely obscures scoured flow tracks, lidar imagery can provide useful information about flow paths and superelevation potential. All of mountainous Appalachia can experience destructive and potentially lethal debris flow events, but the details of flow behavior can vary between sub-regions with different rock and soil types and different slope geometries. Studying past flow behavior can help geologists increase understanding about potential slope behavior during extreme rainfall events. This information can, in turn, inform planning and land use decisions, particularly when human safety is a consideration.

The 1995 Madison County debris flows are also an excellent illustration of how localized extreme rainfall and debris flow occurrence can be. The image below shows a larger view of the setting of the two superelevated flows described above, which are indicated by yellow arrows. Notably, debris flows are restricted to the southeast slopes of the mountain, where the superelevated flows occurred. Just across the ridge, no flows or landslides of any kind occurred.

The extreme rainfall (~30 in or 76 cm in 8 hours or less) that produced the debris flows resulted from upslope flow of moist air into a small but very intense thunderstorm. The details of the process are described in the second paper linked above (and again here), but a takeaway is that wind direction, altitude of winds, and the shape of topography “funneled” moist air upslope and into the storm to focus the catastrophic rainfall on a small area. The interplay of topography, weather, and slope geometry in this Madison County, Virginia, example nicely illustrates how meteorologists and geologists can support each other’s work to help people avoid debris flow hazards.