Technical information, news, research, and opinion on avalanches, snow safety, and winter backcountry travel.

Monday, December 21, 2009

Snowpack Observations: The Search For Instability

An avalanche is defined most simply, as a block of snow cut out by fractures—McClung/Schaerer

NOTE: Backcountry avalanche forecasting is composed of four interconnected elements: goal, people, awareness, and uncertainty. Four types of observations, human, terrain, snowpack, and weather are used in backcountry avalanche forecasting.

In the prior article, I discussed terrain observations, and six universal qualities that indicate exposure. Now it's time to discuss snowpack observations. The relationship between terrain and snowpack is complex, but this article should provide useful context on how to think about the nature of their relationship. As with terrain observations, this article presents non-technical methods that are compatible with human behaviour.

NOTE: This discussion of snowpack observations is focused on slab avalanches. Most of this discussion departs from the pure shear fracture model of avalanche formation first proposed by McClung in the late 1970's. Recent avalanche research has revealed limitations in applying the pure shear fracture model to many types of avalanches.

The main difficulty with snowpack observations is not spatial variability, nor the interpretation of results—although experience in these matters is valuable. The main difficulty with snowpack observations is learning the right way to think about these observations. Snowpack observations can provide direct information about instability, but only if you understand how slab avalanches form, and only if you know how to think about snowpack observations. Thankfully, recent research, along with several helpful analogies, provides some answers.

Thinking About The Snowpack
The seasonal snowpack is a multi-layer composite material formed by the weather. The state of the snowpack, including the condition of its surface and the structure of its interior, are also determined by the weather. Therefore, the seasonal snowpack is a multi-layer composite material, the quality of which is controlled by the weather. Different layers have different strengths, and the strength of an individual layer varies from place to place, and with the passage of time. The region where these layers touch each other is referred to as the interface between the layers, and as with the layers themselves, the strength of these interfaces varies from place to place, and with the passage of time. So, in total, the strength and structure of the snowpack is extremely complicated, right?

Believing that the snowpack is complex does not contribute to the goal of backcountry avalanche forecasting. Therefore, rather than believing that the strength of these layers and interfaces is complicated, it's much more helpful to focus on the very simple truth: extensive variations in the snowpack create a lot of uncertainty. If you never learn anything about the snowpack, and never dig a single hole in the snow, remember that the snowpack is a source of significant uncertainty.

Snowpack observations increase your awareness of these variations and their strengths, but residual uncertainty always remains. This residual uncertainty is part of the reason why snowpack observations are often perceived as vague, when in fact, the often vague nature of snowpack test results simply reflects the fact that the structure and strength of the layers and interfaces are a source of significant uncertainty. This uncertainty is perfectly normal, and snowpack observations can help greatly, especially when you decide to travel in big terrain.

Why then, with the apparent widespread variation of layer and interface strength, are skier-triggered avalanches so rare? Well, the seasonal snowpack is stable most of the time because it is a heavy, multi-layer laminate material. While snow is the weakest natural material, and while snow is typically found at temperatures close to its melting point, multi-layer composite materials are known to exhibit extraordinary strength even when the individual layers are constructed from relatively weak materials. Some of the other physical properties of snow, such as its ability to deform elastically and settle onto a mountain side, along with its ability to bond with neighbouring layers under a variety of conditions, also help explain why skier-triggered avalanches are rare.

But forget about the relative strength of the snowpack for a moment because the strength of the snowpack isn't related to avalanche formation. It's the weak spots that matter.

Carbon fibre composites are used to construct aircraft components such as rudders and stabilisers. Cavities are a significant concern during the process of manufacturing carbon fibre composites because hollow cavities provide space for collapsing and cracking. When manufacturing carbon fibre laminates, great care is taken to ensure that cavities do not form in any part of the laminate because tiny flaws can turn into very large problems. For these reasons, carbon fibre laminates are usually manufactured in an absolutely dust-free environment, and highly-controlled procedures are used to prevent formation of cavities in any parts of the laminate. This might not seem relevant to avalanches, but the presence of cavities in the snowpack is just as important.

Avalanche formation is caused by a type of mechanical failure referred to as delamination. This means that the layers of the snowpack become detached from one another, and on a steep slope, gravity will pull the detached layer ( the slab ) downhill. It's helpful to think of this process as similar to paint flaking. Perhaps there is a small patch of dust that will cause a poor bond between the final coat and the primer. Later, as the paint dries, the outer layer pulls away from the inner layer ( delamination ) and cracks form at the margins ( detachment ). Eventually a paint chip falls onto the floor ( avalanche ).

Recent research by European and North American avalanche experts proves that buried surface hoar and facets indicate areas where the snowpack is effectively hollow. You can think of these regions as very large, thin cavities in the laminate structure of the snowpack. Weaknesses in new snow, such as low density sections of the new snow, operate on the exact same principle: low density snow, regardless of the form it takes, means that a portion of the snowpack is, essentially, hollow. As discussed above, the structural integrity of composite materials depends on the strength of the individual layers and the strength of the contact interfaces between the layers. Hollow regions indicate that, essentially, there is no interface between the layers, or any interface that does exist is very, very weak. Large hollow regions in any interface or in any part of any layer indicate the presence of a potentially serious flaw in the laminate structure of the snowpack.

Consider a layer of buried surface hoar, or a layer of facets, above a layer of hard, well-settled snow, and below a layer of soft, but well-settled new snow. It seems convenient to focus on the intrinsic weakness of the surface hoar crystals, and their well-known persistence in the snowpack, but I'm afraid that focusing on the crystals themselves is a bit of a red herring. Surface hoar and faceted crystals are dangerous not only because of their intrinsic weakness, but also because their presence indicates that some degree of mostly empty space is present in the snowpack. Such crystals are easy to crush, and the presence of empty space overlayed by a cohesive slab provides a means by which the crushing can spread on its own.

The inherent uncertainty in layer and interface structure and strength is one primary source of confusion when interpreting snowpack observations. But it's much easier to manage this uncertainty once you realise that significant uncertainty is normal and can be addressed by seeking additional information, making conservative choices, or simply by teaching yourself the difference between calculated risks and gambling. Gambling is a perfectly valid choice if you are prepared to accept significant losses, or if your friends and family are willing to pick up the tab.

Thinking About Snowpack Observations
After uncertainty, the second primary source of confusion is not knowing how to think about snowpack observations.

Since the seasonal snowpack is a material that is manufactured by the weather, quality control is a useful way to think about snowpack observations. Quality control is concerned with determining whether or not flaws are present in any material, including the extent of any such flaws and the potential outcomes and consequences.

Snowpack observations are simply a means by which you can perform quality control for snowpack by gathering information about the structure and strength of layers and interfaces. Since the structure and strength of layers and interfaces is highly related to avalanche formation, these observations can be very useful. In terms of snowpack observations for backcountry avalanche forecasting, slab avalanches are much more likely with some layer and interface configurations than with others. ( Learning about these configurations is one reason that you should take an avalanche class. )

Man-made composite materials, such as carbon fibre laminates used for aircraft construction are extremely strong because of the consistency of the layered construction, the strength of the individual layers, and the high-quality contact interfaces between each layer. Of course, manufacturing defects are much less likely in a highly-controlled manufacturing environment, but are still a source of concern. Manufacturing defects in the seasonal snowpack are a source of concern because such defects make avalanche formation possible. Weather is the source of the snowpack, and therefore, weather is the source of most manufacturing defects. The chaotic interaction between terrain and weather makes it very difficult to assess the state of the snowpack without close observations.

The difficulty in determining the location and extent of defects in the seasonal snowpack, including small changes from the continual influence of weather, is the biggest challenge in backcountry avalanche forecasting. So, the problems with snowpack quality control essentially boil down to a question of scale and consistency. It is definitely possible to thoroughly search a 4×4 metre square of carbon fibre laminate for the presence of flaws, and the process of quality control is made easier because of the consistency ensured by highly-controlled manufacturing conditions. Given the inconsistent and uncontrolled conditions during snowpack manufacturing, along with limitations in time and manpower, it is impossible to search an entire valley for flaws. That's why it's important to perform snowpack tests in areas where you are likely to find flaws.

To develop a useful mental model for evaluating snowpack observations, it is helpful to stop thinking about what an avalanche looks like on its way downhill, and think instead about why slab avalanches form.

Delamination Theory, Part I: Crushing & Collapsing
NOTE: Some of this material is derived from old and new research on the shear fracture model of avalanche formation proposed by McClung, and recent work involving anti-cracks. There is an element of theory to most models of avalanche formation.

Avalanche formation requires delamination, and crushing is one mechanism by which delamination occurs in layered composite materials. Let's examine some very simple snow profiles and determine whether or not delamination is possible.

Figure 1.1. Snow profiles for two different regions of the same slope. White pixels are air. Note the individual layers and interfaces in each image. The implication is fairly clear: the surface hoar crystals form a hollow cavity in the snowpack on the right, and the interface between the two layers is almost non-existent. Pretend for a moment that you are a "quality control engineer" out for a day of ski touring. Which product passes your quality control procedures and why? Which product does not pass your quality control procedures and why?

Figure 1.2. The snowpack laminate with simple crushing in progress. In this example, the layer on top is hard enough to provide a lot of energy for crushing as it falls, which means that the crushing is likely to spread on its own. Q: What happens if this crushing process spreads throughout the snowpack? A: delamination. Q: What happens if widespread delamination occurs on a steep slope? A: The delaminated portion of the snowpack detaches from its surroundings and moves downhill.

Figure 1.3. Old snow is dark gray, with layered storm snow above. The snowpack laminate with simple crushing in progress. In this example, the new snow layer on top is not hard enough to provide a lot of energy for crushing, but the layer immediately below the new snow is so hollow that the energy requirement is low. In this example, the crushing ends when it reaches the dense, old snow layer below. This is an example of a classic soft-slab avalanche in new snow. The most dangerous weakness is suspended in the middle of the new snow, but the avalanche appears to form on the interface of the old/new snow only because all the new, upper layers are swept out during the initial collapse. In reality, this avalanche started in the new snow, and the new snow could collapse no further than the dense old snow, which then served as the bed for the avalanche. New snow is often full of widespread weaknesses that require very little energy for avalanche formation. You would almost certainly hear whumpfing and cracking if you traveled on this snow.

The Avalanche Handbook helpfully points out that the very common nearly perpendicular orientation between the slab and the bed surface indicates a friction free relationship between the slab and the bed surface immediately prior to the avalanche. This is expected, because if significant friction existed between the slab and the bed surface after the collapse of the weak layer, then you would have a failed avalanche. Failed avalanches can occur and are usually indicated by a loud whumph. Most people correctly recognise this sound as an absolute indicator of high snowpack instability.

Delamination Theory, Part II: Shearing & Collapsing
Delamination can occur whenever there are hollow regions in the snowpack, and failures tend to spread when the snowpack layer above the hollow region is dense enough to generate energy as it falls onto the layer below. The potential for failure in the previous example was very obvious. However, recent avalanche research shows that you only need a hollow space approximately 1mm thick to provide 10× more energy than is actually required for the crushing process to sustain itself. In these cases, crushing may initiate from strong shear forces that crack the inter-crystalline bonds in the interface ( or at the base of the slab itself ) and shift the slab just enough to crush the ice crystals and start a self-sustaining collapse.

Figure 1.4. Snow profiles for two different regions of the same slope. White pixels are air. Note the individual snowpack layers and interfaces in each image. The flaw in this example could be much more difficult to find. In fact, it may be almost impossible to find this flaw with the naked eye. You may have to use a magnifying glass and examine the snowpack very carefully.

Figure 1.5. You can also initiate crushing in very fine, extremely thin layers ( such as facets or a crust ) if you apply enough energy to cause a shearing fracture in part of the weak interface or inside the weak layer itself. This initial shear fracture causes local collapse and crushing of the thin, weak snowpack layer. This collapse spreads through the combined action of a propagating shear fracture, with some collapsing as shown above. It's easy to visualise the results of collapsing, whether or not the collapse is initiated primarily by simple crushing or by the more complicated process of shearing. Either way, the result is the same: delamination, detachment, avalanche.

Detachment occurs during and after delamination. Once delaminated, the layer of snow is no longer attached to the layer of snow below. Gravity is always in effect, so immediately following delamination, the layer of snow begins to pull on its margins and visible cracks form at the edge of the delaminated region. Now the layer of snow is completely detached from the snowpack and it begins to move downslope. In terms of human perception, this process is instantaneous, which is understandable since research has proved that disturbances can travel at speeds approaching 20 metres per second.

Quality Control
Interpreting snowpack observations is a primary source of confusion for many people because of high uncertainty related to the strengths and weaknesses of layers and interfaces, and because many people are not taught to approach the problem with the correct frame of mind.

In the backcountry, you serve as quality control. Snowpack observations, such as snow profiles, compression tests, and ski pole tests, are important and useful quality control procedures that help determine whether or not dangerous flaws exist in the snowpack.

In conclusion:

  • The seasonal snowpack is a multi-layered composite material manufactured in poorly controlled conditions.
  • The length of the manufacturing process is the entire winter.
  • Flaws introduced during manufacturing make avalanches possible.
  • Variations in the layers and interfaces originate from the chaotic interaction of terrain and weather.
  • These variations are a primary and ongoing source of complexity uncertainty.
  • This uncertainty is normal, and must be acknowleged and addressed prior to decisions.
  • A highly uncertain mind is a breeding ground for speculation, half-truths, denial, and other undesired human factors.
  • A highly uncertain mind makes excessively conservative decisions when avalanche danger is actually low.
  • You can address uncertainty by seeking additional information, making conservative choices, or by gambling.
  • Quality control procedures such as snowpack tests are designed to find and evaluate flaws in materials.

Thursday, December 3, 2009

Terrain Observations: Choosing Terrain Appropriate For Conditions

I'm learning to fly, around the clouds ... what goes up, must come down—Tom Petty

NOTE: Backcountry avalanche forecasting is composed of four interconnected elements: goal, people, awareness, and uncertainty. Four types of observations, people, terrain, snowpack, and weather are used in backcountry avalanche forecasting.

CAPSULE: Recreational backcountry skiers are taught to analyse terrain for sources of exposure by using static variables that don't account for the dynamic nature of avalanche phenomena. This article discusses a more general model for analysing terrain that accounts for the dynamic nature of avalanche phenomena with respect to size, phase, and velocity. This model is non-technical and can be effectively applied by most people without any specialised training.

THANKS: To my father for his invaluable input.

Why would you ski in one place instead of another place? How do you choose terrain appropriate for conditions?

Recreational backcountry skiers are taught to choose terrain using a simple list of shape parameters such as slope shape, angle, openness and size. This information is then combined with a simple list of terrain traps such as trees, rocks, cliffs, gullies, benches, and depressions. Collectively, we can refer to all these elements as slope shape parameters because ultimately all these elements do refer to the static shape of the terrain.

Unfortunately, the use of static parameters has some significant drawbacks. For example, using a simple list of slope shape parameters can limit perception, and that can lead to dangerous gaps in analysis. Doesn't something very specific come to mind when you hear the word gully or convexity or cliff? Perceptual traps are often associated with rules of thumb, and the relationship between rules of thumb and serious errors is well documented. In the real world, using static slope shape parameters teaches recreational backcountry skiers to find very specific hazards, but it also produces analysis that contains many blank spots. I refer to this phenomenon as pin the hazard on the donkey, or with the more technical term hazard diffusion.

Hazard diffusion occurs because a list of static slope shape parameters does not account for the existence of multiple spatial scales in the relationship between terrain and avalanches. To this point, consider the Strathcona-Tweedsmuir tragedy which occured in the Connaught Valley drainage of Glacier National Park, British Columbia in February 2003. For an avalanche of sufficient size and velocity, the Connaught Valley is a giant terrain trap. If you are willing to indulge in some semantics, then we can say that for a large, fast-moving avalanche, the term Connaught Gully would be a much more appropriate description.

The heart of the problem is that determining the degree of exposure for a given section of terrain depends largely in part on the size, phase, and velocity of an avalanche. Since avalanches occur at a variety of spatial scales, and since terrain exists at a variety of spatial scales, analysis of exposure must account for a variety of spatial scales. This type of analysis requires a more general set of parameters that integrate the spatial and temporal elements of avalanching with multiple scales of terrain. Such integration allows recreationists to very effectively analyse exposure across multiple spatial scales of terrain for a variety of important aspects of avalanche phenomena.

In The Beginning
You might be wondering how this model arose and whether or not it's backed up by any hard evidence? It's a good question with a very complicated answer, but I'll try to be as concise as possible. For the past couple of years I've been working on developing computer-based tools to analyse terrain for exposure to avalanches. It's rough business, especially since computers are mind-numbingly unaware. When a human looks at a digital elevation model, all the terrain features are staggeringly obvious. Unfortunately, a computer sees nothing but a list of elevation values. To get an idea of the problem, imagine if you had to perform route selection from a grid of elevation values. It's quite literally like trying to ski with your eyes closed.

To get around these problems, you can develop algorithms that the computer can use to perform analysis on the elevation values. For example, you can tell the computer about the relationships between the values, and then the computer can use those relationships to reconstruct data such as slope angles. This is all very simple stuff. But let's take things a step further and imagine that you want the computer to identify an avalanche starting zone, or a runout zone linked to multiple avalanche paths. Modern computers are incredibly powerful, but they cannot easily be programmed to solve certain types of problems. This is doubly true when you are faced with real world constraints such as time and money. After a few months of dead ends, I began to realise the depth of the problem and the futility of approaching it head on.

I found the first part of the solution in an old document that discussed Peter Schaerer's work on avalanche terrain along the Trans-Canada highway. Peter Schaerer discusses surface area as an important attribute of an avalanche starting zone, and at that very moment I became intensely curious. Could surface area be used as a statistical signal of avalanche terrain? Needless to say, I returned immediately to my computer and began to calculate surface area for different sections of terrain. The values were startling. It turns out that surface area is a very useful statistical measure of exposure inherent to any section of terrain. Eventually, I developed other geostatistics that made the original problem very tractable. For this reason, I feel comfortable saying that this model is backed up by hard data, and in fact this model is currently being used successfully in the real world.

In the avalanche community, there is a spectrum of complexity, from complex mathematical models to textual hazard communication tools. Regardless, everyone agrees that simplicity is an extremely effective strategic-tactical approach to education and communication. The value of simplicity far exceeds the scope of this article, but it's safe to say that simplicity is an effective technique because people cannot absorb material they do not understand. In the words of an esteemed technical writer who is a friend of mine, it's a lot of work, making things simple.

When Models Collide
I can't find anyone who really knows where the list of shape parameters arose, but it seems likely that it coalesced over time from the collective wisdom of mountain folks. During my time in the mountains, I have observed a very interesting phenomenon with respect to terrain selection. The list of shape parameters provided to most recreationists is extremely useful, but most don't seem to realise that it's just a partial list. To this point, the Avalanche Handbook specifically states that it's important to consider combinations of features in addition to individual features. Yet there is a lack of simple tools that people can use to accomplish this type of multi-scale, multi-feature analysis.

Finally, it's incredibly important to note that this model is not designed to help people outsmart terrain, which is a form of abuse frequently undertaken with travel technique. This model is designed help recreational backcountry skiers by providing simple, scale-invariant tools to determine the real degree of exposure inherent to a route or objective.

The Model
LOCALE. A model for analysing exposure.

Line-Of-SightIs line-of-sight limited? Do local terrain features obstruct your line-of-sight to terrain above or below? Is your uphill view blocked by rocks or trees? Can you see all the terrain below or only some of it? Pay attention and double-check your decisions when line-of-sight is limited for any reason. Limitations to line-of-sight can reduce your reaction time to zero.
OpenIs the terrain open enough to produce small, medium, or large avalanches? Are there open breaks in the forest that allow snow to travel unobstructed to your location? How much open terrain is present and where is the open terrain relative to your location? Large quantities of snow can accumulate in open areas near mountain tops before an avalanche and in valley bottoms after an avalanche.
ConfinedIs the terrain confined relative to the size or speed of an avalanche of any size? Could a large, fast-moving avalanche overrun the valley floor? Are you in a gully where escape from a small but fast-moving avalanche could prove impossible? Small avalanches form deep deposits in confined terrain, such as hollows or depressions, where snow can accumulate. Estimate your reaction time before something goes wrong and double-check your decisions if reaction time is short. Reaction time is a very important part of your margin of safety.
AccumulationHas snow accumulated above or below? Avalanches often start where snow accumulates, and then run downhill where they deposit snow on the valley floor.
ObstacLesAre there obstacles above or below? This includes cliffs, crevasses, rocks, and trees. Obstacles above may block your line-of-sight, inducing hazard blindness, and can cause traumatic injuries during any phase of an avalanche. On the other hand, obstacles can block or redirect flowing snow and may offer protection at your current location.
StEepIs the terrain steeper than 30 degrees? Avalanches start and rapidly accelerate on steep terrain. Steep terrain produces fast-moving avalanches that can release above you and travel toward your location. If you trigger an avalanche on steep terrain, you may be unable to escape from the flowing snow. Steep terrain can limit your line-of-sight, and indicate locations suitable for sluffs, cornice drops, serac fall, or rock fall—all of which can start avalanches above or around you.

During Trip Planning
You can use this model during trip planning as well.

Two examples are provided, along with relevant markup. According to The Avalanche Handbook, it is important to consider combinations of features rather than simply identifying separate gullies, benches, and cliffs. These parameters find specific traps and diffuse sources of exposure for terrain of varying size. At the bottom, several unmarked examples are included for large, medium, and small scale terrain. Feel free to evaluate the unmarked images by using the LOCALE parameters.

Image 1.1. Goat Island Mountain. At a large scale, these parameters are mostly concerned with large avalanches. This type of analysis would be suitable for general route-finding. It's worth noting that areas with line-of-sight limitations and obstacles often present significant route-finding challenges.
CONCLUSION: This analysis helps you develop a big picture understanding of exposure and traps for this area. Travel through terrain on the left side of this image takes place above and below accumulators on steep, open slopes with numerous obstacles and many limitations to line of sight. Travel through terrain on the right side of this image takes place above and below accumulators, on steep, confined slopes with numerous obstacles, and many line of sight limitations.

Image 1.2. Chinook Pass. At a small scale, these parameters are mostly concerned with small avalanches. Several possible backcountry ski runs are illustrated. You can envision other backcountry ski runs in the context of the hazards. Despite its relative simplicity, some of the backcountry ski runs in this area have numerous obstacles, line-of-sight limitations, and and significant accumulation potential. Intricate terrain with complex groundcover makes beacon searches difficult.

CONCLUSION:The ski runs in this area are steep, with numerous obstacles, but line-of-sight is generally good, except in a few areas. Still, small avalanches could have very bad consequences in many of the ski runs. Intricate terrain increases time for search and rescue.

Try It Out
Use the LOCALE parameters to evaluate the terrain in the following photographs. Feel free to use your own photographs as well.

Image 1.3. Baker Backcountry. Line-of-sight, Open, Confined, Accumulators, Obstacles, Steep.
Image 1.4. Baker Backcountry. Line-of-sight, Open, Confined, Accumulators, Obstacles, Steep.

Image 1.5. Baker Backcountry. Line-of-sight, Open, Confined, Accumulators, Obstacles, Steep.

Image 1.6. Snoqualmie Pass Backcountry. Line-of-sight, Open, Confined, Accumulators, Obstacles, Steep.

Image 1.7. Snoqualmie Pass Backcountry. Line-of-sight, Open, Confined, Accumulators, Obstacles, Steep.

Image 1.8. Snoqualmie Pass Backcountry. Line-of-sight, Open, Confined, Accumulators, Obstacles, Steep.