On March 22nd, a massive landslide buried a town in the state of Washington. It is the most deadly landslide within the United States in a decade, and we knew it could happen. Living in the path of impending catastrophe is a choice we all make daily, but that doesn't make it easy.

The Steelhead landslide partly blocked the river and buried a community. Image credit AP Photo/Ted S. Warren.


The Steelhead Landslide

The Steelhead Landslide (also called the Hazel Landslide or the Oso Mudslide) is located about 90 km northeast of Seattle, in the community of Oso, Washington.

The landslide is composed of glacial sediments, a mix of sands, silts, and rocks that turned into mud in the recent heavy rains. The landslide ripped up trees, entrained saturated soils and potentially even mixed river water into its mass, increasing volume and mobility as it ran downhill.


Anatomy of a rotational failure. Image credit: United States Geological Survey (USGS)

The curved scarp at the crown of the landslide and internal structures (minor scarps, cracks, intact blocks) suggest that the top portion is rotational failure, sliding along a curved failure surface. Below the river, the debris spreads into tongues where it transitioned into the fluid-motion of a flow.


By looking at satellite imagery, the failure area is roughly 450 meters wide and up to 500 m long. Judging from where the river cut through the debris, the landslide might be 10 m thick. As a very rough estimate, that puts the volume of this landslide at 2 million cubic meters, into the territory of a catastrophically large landslide.

The landslide briefly dammed the river before being cut through. Image credit: AP Photo/Washington State Dept of Transportation


From the head scarp of the failure area to the distant toe of the deposit, the runout distance is somewhere over 1.7 km long. The landslide split at the river, spreading to around 1.3 km wide.

The landslide briefly dammed the North Fork Stillaguamish River. A USGS stream gague about 20 km downstream measures the river level in near-realtime. The landslide occurred around 11 am on Saturday; the gage reported an abruptly drop in water level at 1:30 pm. The discharge decreased by about 34 cubic meters per second, with all that water building up behind the dam.


The dam held for about 30 hours before the stream eroded a new path through the deposit. At its peak, the drop in discharge suggests the dam was holding back over 3 million cubic meters of water; the USGS described the upstream pond as up to 10 meters deep. Fortunately, it looks like the dam is releasing the trapped water slowly, carrying debris and sediment downstream but not failing catastrophically with an outburst flood. A flash flood warning will remain in effect for downstream communities until the barrier lake finishes draining.

A Mobile History

The Steelhead Landslide is part of a long history of failures in the valley. At least five catastrophic-scale landslides have occurred in the valley in the past 12,000 years. More recently, in 1949 a large section 300 meters long and 800 meters wide slid out and impacted the river bank. Three years later in 1951, another large failure partly blocked the river. In 1988, the first movement of the current failure surface slipped out, followed by a reactivation in January of 2006, blocking the river again and diverting it into a community.


This history of instability was well-known. By 1967, local media were calling the "Slide Hill," while two of the local creeks sport the ominous titles "Slide Creek" and "Mud Flow Creek." By 1999, the US Army Corps of Engineers wrote a report directly citing the potential for a large, catastrophic failure.

This information wasn't ignored. The area has an extensive history of geomorphic reports and interventions to mitigate the hazards, even for this specific hill. To avoid reactivating the landslide, the toe of weak slope was shielded from undercutting by the river through building protective structures. There was even talk of diverting the river closer to the town, protecting the landslide but increasing risk of floods.


The landslide ran out over and deposited in the community of Oso. Traces of older landslide scarps and deposits are visible in the hills. Image credit: AP Photo/Washington State Dept of Transportation

So, if the history of mobility was known, and the potential for a large landslide was laid out in a geotechnical report, why was a community built in the runout zone? Because risk assessment is a tricky process with no right answers.

On Very Large Landslides

For a small landslide, the mechanics of where the material will go is simple, easily determined by friction and gravity. Although the actual topography and shapes involved are more complicated, the mechanics are the same "block sliding down an inclined slope" model used in high school physics classes. But, when landslides reach a certain size (somewhere around one million cubic meters), they start behaving strangely. How? By being more mobile than they should be by simple sliding-block physics.


The "Why?" is under academic debate, and has been for over a century. Scooping up of additional material and fluid along the travel path. Bulking up of volume as rock broke and fragmented. Lubrication of the base layer by water, fine rock dust, or even cushions of air. Positive feedback loop between the rumble of crashing rock and acoustic vibration enhancing flow velocity. Aside from eliminating the cushions-of-air theory after spotting highly-mobile landslides on thin-atmosphere Mars, all those theories and more are still in play as researchers argue about which mechanism is most plausible for explaining the high mobility of extremely large landslides.

The Steelhead Landslide is one of those catastrophically large landslides, with enough volume to run out farther than would be predicted by a simple sliding-block. The initial USGS map of the area outlines the failure scarp (top edge of the landslide) with a thick red line. The runout area where material was eroded but not deposited is left bare on the map. Finally, the deposit area where material came to a rest is shaded with green hatching. The background image is a 2013 LiDAR topography map of the area. From intuition, it doesn't seem like material should have spilled out nearly as far as it has, yet it did.


LiDAR map of the landslide area. Image credit: USGS

This excess mobility makes hazard forecasting tricky, as it is difficult to predict how far a potential landslide could run out. Although research is progressing on pinpointing what characteristics of a landslide will most influence its mobility, the scarcity and poor documentation of large events means that we just don't have enough data for reliable probabilistic forecasting. Even if someone had predicted exactly how much material would slough off in one event, pinpointing the landslide volume, their ability to predict how it would move and flow would have been imperfect.

Learning from Disaster

The Steelhead Landslide is a catastrophe that cost human lives, and is being extremely well-documented by local media. This data will help researchers better understand the mechanics and motion of large landslides, improving risk prediction at other locations. It will be included on local geotechnical field trips to help our students understand the importance of site investigations, and the very real tragedy that occurs if they don't pay attention and do their job to the best of their abilities.


Geohazard researchers investigating the mercifully casualty-free 2010 Mount Meager landslide in British Columbia. Image credit: Mika McKinnon

This is the ugly side of disaster research: to do our jobs and better understand the disasters that take lives and destroy homes, we need to take those personal tragedies and turn them into case studies and data. Those visiting researchers and students are almost inevitably going to come into conflict with the survivors. The survivors don't want their devastation over the loss of their friends and family incorporated into an article or a teachable moment, yet without looking, we won't learn from this event and pass those lessons on to the future engineers charged with protecting public safety.


Living in a Disaster Zone

As of midnight on Tuesday, 24 people are confirmed killed by the Steelhead Landslide, with over a hundred more reported missing. Over 40 homes and other structures were destroyed by the landslide, and search-and-rescue efforts are ongoing.

If the residents of Oso knew they lived near unstable ground, somewhere with the potential for a major landslide even if they couldn't predict the exact runout path, why didn't they just move before this disaster happened?


Search-and-rescue volunteers look for survivors in a destroyed structure. Image credit: AP Photo/Elaine Thompson

Because disasters are everywhere. You live in a hazard zone. Landslides, hurricanes, earthquakes, fires, severe storms, tornados, floods, volcanic eruptions, tsunami — everyone, everywhere, is at risk for some type of catastrophe. We've got the risk of global catastrophe from asteroid impact, or solar storms knocking out the power grid. Move away from the geohazards, and there's epidemics, nuclear meltdowns, war, terrorism, murder... the number of things that can go horribly wrong are beyond listing.


We all live somewhere at risk for something. But different areas have different types of disasters, different awareness of their hazard, and different levels of mitigation and preparation. Beyond that, governments have different levels of transparency in communicating disaster-risk to residents, and individuals have different levels of awareness of and acceptance for risk.


Globally, the risk of death by landslide is almost universally one in a million per year. This isn't because of some magical property of landslides, but because most governments have agreed that's an acceptable risk, and design their mitigation measures to that degree of safety. A few exceptions exist — Hong Kong combats landslides more vigorously, reducing their odds to 1: 10 million, while residents of the Philippines have a higher risk of death-by-landslide at 1: 100,000 annually.

So yes, the risk of landslide in northwest Washington was known, and yes, the risk of this particular landslide reactivating was also known. Yet based on what we knew, mitigation measures were taken until the risk to the community was deemed acceptably low. But one-in-million isn't impossible, and either we missed something in the risk assessments, or that rare chance came true. Either way, people paid with their lives. They weren't inherently foolish for living where they did; they were unlucky to be on the wrong side of the odds.

The Earth Observatory has a before/after satellite image interactive of the site. For regular updates on all the hydrogeological, geological, and geotechnical aspects of the slide, check out geoblog Reading the Washington Landscape by Dan McShane. If you want to learn more about landslide mechanics, read my thesis! ...or start a conversation in the Group Discussion to ask me questions.


Update March 26: here's an analysis of a seismic signal produced by the landslide.

Update March 28: more on the seismic signal, a preliminary USGS report on the landslide, & a first interpretation on slide mechanisms.

Update April 1: before and after photographs of the landslide.

Update July 23: post-mortem on the landslide following field investigation.