Summary and Keywords
Avalanches have long been a natural threat to humans in mountainous areas. At the end of the Middle Ages, the population in Europe experienced significant growth, leading to an intensive exploitation of upper valleys. At almost the same time, Europe’s climate cooled down considerably and severe winters became more common. In the Alps, several villages were partly destroyed by avalanches, forcing inhabitants to develop the first mitigation strategies against the threat. By the late 19th century, the development of central administrations led to the creation of national forestry departments in each alpine country, principally to tackle the dangers posed by avalanches. As a result, forest engineers conceived not only the science of avalanches but also the first large-scale techniques to alleviate avalanche risks (such as reforestation). However, with the steady growth of transport, industry, tourism, and urbanization in high-altitude areas, these earlier measures soon reached their limits. A new impetus was then given to better forecasting avalanche activity and predicting the destructive potential of extreme avalanches. Avalanche zoning, snowfall forecasts, avalanche-dynamics models, and new protection systems for the protection of structures and inhabitants have become increasingly more common since World War II.
With the advent of personal computers and the increasing sophistication of computational resources, it has become easier to predict the behavior of avalanches and protect threatened areas accordingly. The success of this research and the protection policies implemented since World War II are reflected in the drastic reduction in the number of disasters affecting dwellings in the Alps (most deaths by avalanche now occur during recreational activities). Significant progress has been made since the 1980s, leading to a better understanding of avalanche behavior and the mediation of associated risks. Yet we should not assume that this progress is steady or that our capacity to control such hazards is more advanced than it was two decades ago. Efforts to predict avalanches contrast with work in other sciences such as meteorology, for which forecasts have become increasingly more reliable with advancements in computational power. Explaining the difference is simple: in meteorology, the material is air, a substance whose behavior is well known. The main difficulty lies in the computation of enormous volumes of air encountering various flow and temperature conditions. For avalanches, the material is snow, a subtle mixture of water (in different forms) and air, whose behavior is remarkably complex. Modern models of avalanche dynamics are able to predict this behavior with varying degrees of success.
Avalanches are masses of snow moving down mountain slopes. Most avalanches travel for a few hundred meters at lower velocities (a few meters per second), but some can travel as far as 15 km or even reach speeds as high as 360 km/h. They can also pack an incredible punch: up to several atmospheres of pressure. Generally, the depth of the flow does not exceed a few meters (typically 2 to 5 m), but some high-speed avalanches suck in, or entrain, massive amounts of ambient air, which leads to the formation of large snow clouds whose depths can exceed 100 m.
Avalanches Throughout the World
Because they are composed of snow and driven by gravity, avalanches are a threat typical of mountain areas, whether the snow cover is seasonal or permanent. These areas account for about 12% of the land surface of the earth, as shown by Figure 1. Avalanches take mostly place in the northern hemisphere, but southern countries such as Peru, Argentina, Chile, and New Zealand are also affected.
Worldwide, the death toll and damage to infrastructure caused by avalanches are certainly far less significant than other natural hazards such as floods or earthquakes. On the whole, avalanches account for 0.1% of the total death toll and economic losses caused by natural disasters from 1900 to 2015 (see Table 1). Most major natural disasters are large-scale events that cause death and wreak havoc on scales that can extend beyond national borders. By contrast, avalanches are small-scale events that are localized to specific areas, whose size infrequently exceeds a few square kilometers and are often thinly populated. To get some sense of the destructive power of avalanches, the reader can refer to Table 2, which enumerates some of the deadliest avalanches throughout the world. By comparison, the natural disaster that stands out above all in recent history is the 2004 Indian Ocean tsunami, which claimed the lives of more than 230,000 people around the Indian Ocean. The deadliest avalanches are reported to have occurred in December 1916, when a snowstorm and an artillery duel between Italian and Austrian troops caused the release of many avalanches that caused devastation and death in military camps near Mount Marmolada (9,000–10,000 soldiers killed).
Table 1. Number of Major Events, People Killed, and Economic Losses (in million USD) for Each Type of Hazard from 1900 to 2015 Source: Centre for Research on the Epidemiology of Disasters (http://www.emdat.be).
Table 2. List of Avalanche Disasters Ranked by Number of Deaths Since the 16th Century Worldwide The list is not comprehensive, and the death toll may be approximate. The events reported correspond to single events that afflicted a village or a series of avalanches that occurred within the same region on close dates. Ice avalanches and other mass movements, including rocks and snow (e.g., the Huascarán avalanche in Peru in May 1970), are sometimes considered among snow avalanche disasters, but this is not the case here. Most sources have no access to reliable information on the number and nature of casualties. For developing countries such as Turkey, it is beyond the current capacity of the state (or any other organization) to generate such data.
Marmolada avalanches (Dolomites)
Lahaul Valley avalanche
Mitsumata (Niigata prefecture)
Vorarlberg and Tyrol avalanches
western Norway avalanches
Vorarlberg and Tyrol avalanches
Grisons and Ticino avalanches
Anzonico and Fusio-Mogno (Ticino)
Saint-Martin et Chèze (Pyrénées)
El Teniente mines (Cachapoal Province)
Kohistan Valley avalanches
Görmeç (Gümüşhane Province)
avalanches in the Cuneo valley (71 deaths in Frassino)
Pragelato mines (Piemonte)
Livigno (Province of Sondrio)
Palm Sunday avalanche (Alaska)
Rogers Pass avalanche (British Columbia)
Üzengili avalanche (Bayburt Province)
Møre og Romsdal avalanches
Queyras and Ubaye avalanches (villages of Costeroux and Fouillouse)
Lenin Peak (Trans-Alay Range)
Puente del Inca (Mendoza Province)
The economic importance of activities in mountain ranges (e.g., mining, forestry, electricity production, or tourism), as well as the potential significance of the transportation arteries through them, have long been sufficient grounds for the study of avalanche dynamics and the development of mitigation techniques.
Today, as a result of mitigation measures undertaken since the early 20th century (see Avalanches in the Past), the death toll due to avalanches in European populated areas has substantially decreased, and most reported fatalities occur during recreational activities, mainly off-piste skiing and ski touring, and on rare occasions they involve transportation corridors.
With regard to avalanches involving skiers and alpinists, the mean number of deaths due to avalanches has been fairly stable over the last 20 years in the Alps, with 31 victims in France, 22 in Switzerland, 26 in Austria, 20 in Italy, 2 in Spain, and 10 in Germany. In Japan, the death toll is reported to be an average of 30 each year, with 24 in Turkey, 30 in the United States (although this trend is increasing), and 7 in Canada.
Avalanches in the Past
This section will address how people have struggled with avalanches and how specific techniques were developed to survive the winter and its dangers. The article’s main area of study covers the Alps not only because they were the cradle of avalanche science (see The Creation and Development of Avalanche Science) but also because the history of the integration of people into their environment is well documented. While the Alps have been permanently occupied since the Neolithic period, engineered avalanche protections appeared late (after the 16th century), and not everywhere in the Alps. This raises the questions as to how people coped with avalanches before the 17th century and why some communities started to tackle the threat posed by avalanches differently in the 17th century.
In ancient times, the Alps were far from being a barrier impervious to human exchanges. People lived in high-altitude areas in order to exploit mines and then transported alpine flint and ores to the north and south of the range. Hannibal’s army and Rome’s legions crossed the Alps, showing that enormous numbers of humans were able to walk through snow-capped mountain ranges at the same time. Even after the conquests of Caesar and Augustus, road construction and the foundation of new cities in Alpine valleys, snow, and avalanches made the Alps an inaccessible and dreadful place according to Greek and Roman literature. It is often said that Romans were great builders, and it was not until the 19th century that European cities reached the same level of comfort as Roman cities. While Romans took mitigation measures against floods, there is no evidence that they took a keen interest in avalanche defense.
The collapse of the Roman Empire was followed by a severe decline in the population, and as commercial relationships between different parts of the empire withered, mountain regions were particularly affected. At the end of the Middle Ages (after the 13th century), Alpine regions saw renewed economic development and population growth, and higher-altitude areas like Davos (1,550 m, Switzerland) or Bonneval-sur-Arc and Saint-Véran (1,750 m and 2,050 m, respectively, France) were occupied permanently. Local populations developed specialized strategies to survive the winter months and their dangers. These included protecting dwellings from avalanche impacts by reinforcing walls using mounds of earth, as well as the demarcation of dangerous areas, often placed under the protection of saints. When dwellings were protected from avalanches by forests, communities and feudal landlords laid down strict rules about where timber resources could be exploited: cutting down trees in forbidden forests (Bannwald, in German) was punishable by death. In Switzerland, travelers taking the road to the Gotthard Pass were struck by the forest protecting the tightly packed houses of the village of Andermatt; it was the only green spot remaining in the midst of desolate valleys where most of the steep, bare slopes experienced large avalanches in the winter.
A change in the climate, known as the Little Ice Age, occurred from the late 15th to the 19th century, with its climax in the late 17th century. Long, very cold winters became more common in the northern hemisphere. In high-altitude areas, these conditions favored large accumulations of snow and the propagation of devastating avalanches. Many villages were repeatedly hit by them, leading their inhabitants to abandon their dwellings and find safer places to live. Built in the late 13th century, Vallorcine (France) was a typical village of wooden chalets closely packed around the church—the only solid stone structure. In 1674, part of the village was swept away by an avalanche and the inhabitants subsequently decided to move away to more remote hamlets. However, they also decided that the church and the priest should stay in the same place. That choice may be surprising for us today, but for Christians of the time, nothing in the dynamics of natural forces was due to chance. Natural disasters were interpreted as manifestations of Providence, messages or punishment from God. Act of God is still the term used in insurance contracts to refer to an unforeseeable natural event. For the very religious Alpine communities of the time, like Vallorcine, God would surely spare their church. Unfortunately, their church was hit by another avalanche in 1720. The inhabitants did not change their opinion, they merely decided to build a massive stone wedge to protect the wall exposed to avalanches (see Fig. 2).
While avalanches are ubiquitous in the Alps, engineered protection structures against avalanches appeared late in the 17th century and only in a few valleys. They were used in heavily populated areas, while in sparsely populated mountain ranges, villages were deserted after they were hit by disastrous avalanches. For instance, the highest village of the Alps, Costeroux in southern France, hosted many Protestant families during the wars of religion. Severely damaged by avalanches in the 17th and 18th centuries, the village was finally abandoned by its last inhabitants in the 1820s. The examples of Vallorcine and Costeroux show that the mitigation strategy adopted by inhabitants was closely related to anthropic pressure: in heavily populated valleys, people had no choice but to stay at the same place and develop more efficient mitigation strategies against avalanches.
In the 20th century, many mountainous regions experienced considerable economic growth with the development of transport, industry, and tourism. New techniques evolved for mitigating the dangers of avalanches. In the early part of the century, the emphasis was on avalanche defense strategies on the upper slopes. These included reforestation and engineered structures to hold snow in place and prevent avalanche formation (see Fig. 3). Explosives were also used, but with the opposite goal of forcing avalanche release Tragic examples of the use of explosives have occurred during World War I in the Alps, between Italy and the Austro-Hungarian Empire, and in the more recent conflict in Kashmir, between India and Pakistan: military camps and positions were buried under huge avalanches triggered by intensive shelling. The 1960s and 1970s were marked by several significant avalanche disasters in the Alps, which contributed to an increased public awareness of the risks posed by them in increasingly inhabited valleys. Two of these tragedies in particular were highly significant because they exposed severe failings in how avalanche threats were being managed in newly populated areas. The villages of Davos (Switzerland) and Val d’Isère (France) are both centuries-old and have been transformed into attractive ski resorts. Despite their long traditions of avalanche mitigation, avalanches caused severe damage—24 fatalities in Davos in 1968 and 39 fatalities in Val d’Isère in 1970. In the aftermath of these disasters, the emphasis turned toward developing more holistic strategies for mitigating avalanche hazards. Apart from structural measures such as wall reinforcement, much more emphasis was given to nonstructural techniques such as avalanche risk mapping, land use planning, avalanche forecasting and warning, field monitoring, and the completion of historical databases, as well as computational tools for predicting avalanche run-out distances and the impact pressures of extreme avalanches (see Avalanche Mitigation).
In February 1999, the Alps were struck by a series of snowstorms, which caused disastrous avalanches in France (12 fatalities in Chamonix), Switzerland (17 fatalities), Austria (37 fatalities), and Italy (1 fatality). Figure 3 shows the emergency relief operations underway in Chamonix just after an avalanche swept through 20 chalets. The economic losses due to damaged dwellings and structures, as well as the indirect costs related to the tourist business, were quite significant. Although the protection systems installed failed to provide absolute safety in February 1999, they did help to avoid the occurrence of larger disasters during the peak tourist period that year. In Switzerland alone, protection systems prevented the release or propagation of more than 300 significant avalanches in inhabited areas in the winter of 1999. The direct costs incurred by insurance companies due to the damage caused by avalanche and snow loads exceeded 1 billion Swiss francs (the indirect cost due to economic losses was estimated at about 200 million Swiss francs). These figures are to be compared with the total investments in protection structures (1.5 billion Swiss francs from 1951 to 1999) and research costs (40 to 50 million Swiss Francs every year).
The Creation and Development of Avalanche Science
The science surrounding avalanches has developed in parallel with the gradually increasing economic weight of the human activities occurring in mountain ranges. The history of their study dates back several centuries. It initially only concerned a few countries—Switzerland, France, Austria, and Russia—which are still at the forefront of avalanche research. In recent decades, countries such as Italy, the United States, Canada, Japan, Iceland, Norway, and India, have played an increasing role in developing new techniques for the analysis of avalanches and extending our understanding of their dynamics.
The specific dangers threatening travelers have been mentioned by geographers and writers, such as Strabon, Titus Livius, and Silius Italicus, since antiquity. The word avalanche did not exist at that time, and writers used generic words such as ruina (which could refer to landslides or rock falls); however, their detailed descriptions leave no doubt about their awareness of the dangers related to snow. This absence of a specific term for avalanches lasted a long time—until national languages replaced Latin in the scientific literature of the 18th century. In the late Middle Ages and during the Renaissance, narratives still put an emphasis on the dangers to which Alpine travelers exposed themselves. In his book De Alpibus Commentarius (1574), Swiss theologian Josias Simler compiled all the available information on these dangers and wrote the first complete scientific description of avalanches (labina) by examining their causes and effects. However, Simler’s writing style did not differ much from that used by Strabon 1,500 years earlier, as he gave readers stern warnings about the dangers threatening travelers when they crossed the Alps in winter.
The word avalanche appeared when scientific literature started employing national languages. Its exact etymology is unknown, all the more so because each country used its own term: German writers thought that Lawine (avalanche) came from Löwin (lioness); French writers thought that labina (from the Latin verb labor, meaning to slide or to fall) was the root for avalanche; and others, including French and Italian writers, remarked that avalanche was close to their words for valley or closely related terms (avalanche to aval in French; valanga to valle in Italian).
As the Little Ice Age was drawing to a close in the 19th century, Europe was marked by several disastrous floods. During the same period, in the wake of the French Revolution and Napoleonic wars, European states were reinforcing central administration and trying to maintain strict control over all parts of their territories. Natural disasters were no longer seen as acts of God but as the consequences of local populations’ actions. In 1792, the French revolutionary government sent Ramond Lomet, a civil engineer, to Barèges, a thermal spa resort on the Spanish border. His mission was to find long-lasting solutions to protect its military hospital from avalanches. His report severely criticized the inhabitants of Barèges: “In the past, all the mountains surrounding Barèges were covered in oak woods . . . Men still alive today once looked upon the remains of those woods and then cut them down . . . The inhabitants of the plateaux ravaged everything (. . .) they forget that during the winter they trembled in their dwellings, scared of being swept away with them by the tumbling snow they had caused through their obstinacy.” Reports like this were common at the time. For politicians, these complaints arrived at the right time and gave them the opportunity to impose themselves and intervene in regions known for their fierce determination to retain independence.
The primary objective of the national forestry departments and forestry schools created in the first half of the 19th century was to improve the management of timber resources. Forestry engineers were also firmly convinced that floods and avalanches were the results of the massive deforestation of mountain areas. Reforestation was the solution, and where no trees could grow, such as above the timberline in high-altitude areas, forests could be substituted with rows of metallic piles. In 1860, military engineers tried out this absurd idea above Barèges, but it failed to achieve the desired results, and the whole project was abandoned. Foresters were more successful and enterprising than military engineers. Across Europe at that time, German forestry management was seen as exemplary, and it cultivated the idea that forests could be managed in accordance with scientific principles. In Alpine countries, it encouraged foresters to monitor avalanche activity, and this was the beginning of the snow and avalanche science as we know it today.
The pioneer was Swiss forestry engineer and topographer, Johann Coaz, a man of many talents who started to gather avalanche data in 1876. In 1881 and 1910, he published the first scientific monographs entirely devoted to avalanches and mitigation systems. National forestry services maintained frequent correspondence, and Austrian and French forestry engineers took inspiration from Swiss know-how, replicating it in their own countries in the late 19th century. In 1900, Paul Mougin developed the idea of permanently monitoring avalanches in France, especially near inhabited areas, and he created a huge avalanche database that is still in use today. Although he was a forestry engineer and a fervent defender of forests as a mitigation system, he also promoted the construction of defense structures such as stone walls and snow rakes. In 1923, Mougin was credited with building the first model of avalanche dynamics. In order to calculate avalanche velocities and forces, his simple model made an analogy between an avalanche and a sliding block. In Austria, a civil engineer, Vincenz Pollack, worked in close collaboration with Johann Coaz on avalanche defense structures for protecting the country’s rapidly expanding network of railway lines. In 1906, he published his authoritative work on avalanche defense structures. In the 1860s and later, railway construction in Austria’s Vorarlberg province and over the Brenner Pass (the line connecting Innsbruck, Austria, to Verona, Italy) provided perfect opportunities to test new defense structures. The Innsbruck Torrent Control Service gained considerable experience in this field, forging Austria’s current longstanding tradition and reputation in avalanche and torrent control. In the 1860s and later, railway construction in the Caucasus led to the first scientific studies and attempts to map avalanches in the Russian empire.
For decades, foresters controlled the scientific and technological developments pertaining to snow avalanches, but this situation changed in the early 20th century. A first crack in their monopoly appeared with the publication of books for alpinists and skiers. The biggest danger for practitioners of mountain sports is being caught in an avalanche of their making. Whereas books in the 19th century had simply reported on the fatalities due to avalanches, 20th-century books were manuals teaching mountaineers how to recognize avalanche terrain, how avalanches are released, and how meteorological conditions influence avalanche risk. A German geologist, Wilhelm Paulcke, was one of the pioneers in the development of Alpine skiing. Between 1899 and 1934, he spent considerable time studying avalanche formation, documenting avalanche events using film and photographs, and popularizing his findings in books form. In his book entitled Alpine Skiing at All Heights and Seasons (1921), the English skier and mountaineer Arnold Lunn summarized all of the practical information on snow and avalanches that skiers should be familiar with. This monograph was a great success and the first of a long series of books devoted to avalanche safety for a broad audience. Today, most of the books published on avalanches are handbooks written by mountain sports professionals for skiers and climbers. In Europe, Swiss mountain guide Werner Munter became famous by developing decision-making techniques for assessing avalanche risk in situ. In the United States, Bruce Tremper, an avalanche forecaster from the Forest Service Utah Avalanche Center, has been the author of bestselling books on good practices in avalanche terrain.
A more serious breach in forestry engineering doctrine appeared in the 1920s when part of the scientific community called into question the efficiency of forests for reducing the occurrence of floods and avalanches. There was growing evidence that the frequency and intensity of floods in reforested watersheds did not change to a significant degree. Mountain forests were not only very demanding in terms of maintenance, they were insufficient to prevent extreme events alone. Moreover, defense structures (such as walls and snow rakes), used as a complement to reforestation, suffered from poor design. The forces exerted on a structure by a creeping snowpack were clearly underestimated, which led to massive and costly damage to wooden or stone structures. In Europe, especially in Russia and in France, academics revisited all of the problems related to snow and avalanches. Geographers like Raoul Blanchard and André Allix produced a great deal of literature on the topic. A decisive step toward the scientific quantification of avalanche processes took place in 1936 with the creation of the Federal Institute for Snow and Avalanche Research (SLF) in Davos (Switzerland). For the first time, a laboratory was built to investigate the physical properties of snow. Robert Haefeli, a trained geotechnical engineer, developed the first theories on the mechanical and physical behavior of snow. His works resulted in new design methods for defense structures, all of which were integrated into a single document, referred to as the Swiss Guidelines. These guidelines have been steadily updated and are now used worldwide in the design of avalanche mitigation systems. With his colleagues, Henri Bader and Edwin Bucher, Haefeli also worked on the thermodynamic transformations that snow undergoes (these are called metamorphisms as they induce substantial changes in the shape of the snow grains). In 1939, Haefeli and his coworkers published their book entitled Snow and Its Metamorphisms, which remained the authoritative reference until Samuel Colbeck revised snow metamorphism theory in the 1970s. In 1955, another Swiss engineer, Adolf Voellmy, elaborated on Mougin’s avalanche-dynamics model and proposed a complete framework for calculating the forces that avalanches exert on obstacles, as well as the important features of avalanche motion (velocity, distance traveled). Voellmy’s model was hugely successful and was used by engineers until the late 1990s. In the decades following the creation of the SLF, Haefeli’s group was the leader in avalanche research. In the 1960s and 1970s, avalanche research in the West was quite dormant. Behind the Iron Curtain, however, Soviet researchers were working on a new generation of avalanche-dynamics models inspired by hydraulics. By developing an analogy with flash floods, Sergei Grigorian and Margarita Eglit adapted shallow water equations to describe the motion of snow floods, whereas Andrei Kulikovskii proposed the first model for powder avalanches, which still inspires most current models.
The disastrous winters of 1968, 1970, and 1972 revealed the numerous shortcomings in European avalanche mitigation strategies. They gave new impetus to avalanche research, with an emphasis given to computational methods, risk mapping, and avalanche monitoring. In the absence of sophisticated computer modeling, Bruno Salm from the SLF developed a simple analytical framework based on Voellmy’s model, which made it possible to estimate the run-out distances and velocities of extreme avalanches. In the late 1970s, French engineers Rémy Pochat and Gérard Brugnot, collaborating with mathematician Jean-Paul Vila, took an innovative approach by working on numerical simulations of the shallow water equations. With increased computing power has come the increasing use of numerical models, especially from 2000 onward, for solving engineering problems. Almost all the numerical models in use today are based on Vila’s concepts, which can be linked back to gas dynamics (whose mathematical structure is very close to that of the shallow water equations).
Today, most countries affected by avalanches have their own dedicated avalanche research centers. Some institutions (e.g., the SLF in Switzerland and Météo-France in France) perform operational tasks such as providing regional avalanche bulletins and conduct their own research. Since the 1980s, universities (such as the University of British Columbia, in Canada, or the University of Moscow, in Russia) and private organizations (such as the Norwegian Geotechnical Institute) have also been increasingly involved. In several countries, well-documented avalanche tracks have been equipped with high-technology sensors (e.g., radars, pressure gauges, or force transducers) for monitoring avalanche activity and gaining new insights into the dynamics of large avalanches. Test sites include Vallée de la Sionne in Switzerland, Col du Lautaret in France, Ryggfonn in Norway, Monte Pizzac in Italy, Rogers Pass in Canada, and Kurobe Canyon in Japan.
Avalanche science has also benefited from contributions from related sciences. For instance, the physics of granular flows and the dynamics of density currents have shed new light on the fundamental physical processes involved in snow avalanches. For this reason, while the community of snow avalanche researchers is only made up of a few hundred people around the world, there is a wider community of scientists in the fields of fluid mechanics, physics, and physical geography whose work has substantially contributed to avalanche science in recent decades.
Avalanche risk is managed through both its temporal and spatial dimensions. The temporal dimension refers to the capacity to forecast avalanche activity in a given place or area in the near future. In North America and Europe, national weather services publish regional avalanche bulletins every day during the winter season. These provide useful assessments of the avalanche danger for the following day for a wide audience, including people who make a living in the mountains, local authorities, and practitioners of mountain sports.
The spatial dimension involves different aspects of avalanche mitigation, the most important of which being avalanche zoning. In many Western countries, municipalities use avalanche zoning as a legal means of controlling land development or construction and an informative, detailed safety tool for areas concerned about avalanches. Information is synthesized using the intensity-frequency principle: the less frequent avalanches are, the more destructive their potential. Intensity is the expected impact pressure exerted by an avalanche against a rigid wall; this is measured using the kilopascal (kPa). To give a sense of what this unit of measurement is equal to, we can also use atmospheric pressure (1 kPa = 0.01 atm) or a corresponding measurement of mass per unit surface (10 kPa = 1 t/m²). Frequency is expressed in terms of the period of return (see Snowfall and Snowpack). Three or four levels of risk are then derived depending on the combination of frequency and intensity, and zones at risk are color-coded (red, blue, yellow, and white). For instance, a red zone corresponds to a high risk of frequent avalanches with impact pressures ranging from 3 to 30 kPa or rare avalanches (whose period of return exceeds 100 years) with high impact pressures (in excess of 30 kPa). These would cause substantial damage to dwellings, should they occur. Thus, the construction of new houses in red zones is prohibited, although existing buildings can be used but cannot be extended. Other zones include the blue (medium risk—reinforced constructions possible), yellow (low risk—possible evacuation in emergency situations), and white (no risk or residual risk—no regulation) zones. Figure 5 shows part of the avalanche risk map for Chamonix (France).
Artificial release has become commonplace in ski resorts to protect skiers and infrastructures from avalanches. It is also routinely used for protecting roads. The idea is that frequently triggering the release of small volumes of snow avoids the natural release of larger avalanches. The method involves explosives, projectiles launched by pneumatic cannon or military weapon (e.g., mortar), and gas exploders (i.e., avalanche-release systems that create shock waves by igniting a propane and oxygen mixture). It is not recommended for urbanized areas. Indeed, triggered avalanches may be much larger than anticipated and cause damage. Under certain snow conditions (e.g., wet snow), the blast caused by explosives or exploders may be not sufficiently strong to destabilize snow.
Engineered structures and forests can be used to control avalanche release and motion. Different strategies have been developed over the years to offer a high level of safety, especially for urbanized areas and large infrastructures:
• Forests, supporting structures (e.g., snow nets, bridges and rakes; see Figure 3), and terraces dug in the starting zone aim to prevent avalanche release by retaining the snow or, at the very least, to reduce avalanches’ initial volume.
• Reinforced walls, snow sheds, and splitters (see Figure 2) are designed to reduce avalanche impact or deviate avalanches in the close vicinity of the structure to be protected.
• Catching structures such as dams and mounds (see Figure 6) are meant to slow down avalanche velocity and enforce deposition.
• Deviating walls (deflector) deflect avalanches from their natural path and redirect them in a safer direction. They can also be used to control avalanche width and limit lateral spread by confining avalanches between two parallel dykes.
Prevention is not always sufficient to avoid fatalities, especially during recreational activities. In recent years, emphasis has been given to educating and raising awareness among skiers and mountaineers so that they are more familiar with search-and-rescue techniques for finding victims caught in an avalanche and providing first aid. The survival rate drops significantly as time passes: while 80% of people buried for less than 5 minutes survive, the chances of survival fall to 40% after 15 minutes, and 20% after 1 hour. The victim’s life depends crucially on the immediate action of survivors or witnesses. In Western countries, most ski resorts and municipalities in mountain areas have prepared search-and-rescue plans as well as back-up and contingency plans involving well-trained rescue teams (including specialized rescue dogs), helicopters, and special equipment for finding victims and providing medical assistance. Regional rescue plans also exist if a larger-scale avalanche hits dwellings or roads.
Snow Mechanics and Avalanche Formation
Avalanches follow the same basic principle: snow accumulates on a mountain slope until the gravitational force at the top of the slope exceeds the binding force holding the snow together. A layer of snow, or sometimes the whole snowpack, can then slide its way across the underlying layer, resulting in an avalanche.
Snowfall and Snowpack
During snowstorms, snow accumulates on slopes. The amount of precipitation depends on many parameters, including global climatic conditions, local topography, and specific features of the snow storm. In the Alps, where the climate is continental or temperate, daily snowfall rarely exceeds 130 cm. In the Chamonix valley (France), for instance, the maximum daily snowfall ranges from 72 cm to 106 cm depending on the altitude (from 1,050 m to 2,100 m, respectively). That maximum reaches 130 cm over the Bernina Pass in Switzerland (altitude 2,328 m). In areas with a maritime climate (such as the Pacific coast of the United States, Norway, or northern Japan), the shock of great masses of warm and cold air hitting each other leads to striking amounts of snow along the coastline. At Crestview in California (altitude 2,300 m), in January 1952, records were broken when 2.1 m of snow fell within 24 hours.
Yearly totals show significant spatial and temporal variations. For instance, in the Chamonix valley, the average total annual snowfall depends on the altitude: at 1,050 m, it ranges from 14 cm to 5.2 m, with an average annual snowfall of 2.5 m. At 1,500 m, it ranges from 3 m to 12 m, with an average of 5.8 m. The effect of altitude on snowfall totals is less marked at higher elevations: an increase in elevation from 1,500 m to 2,000 m leads to a 15% rise in total annual snowfall (average 6.6 m, 3.8–13 m range). In maritime zones, coastal mountain ranges are associated with higher levels of precipitation. In 1999, the annual total snowfall exceeded 28 m on Mount Baker (California, altitude 3,286 m).
In contrast, over shorter periods, the amount of snow precipitation is very dependent on altitude, mostly because of the fluctuations in the altitude of the freezing point during storms. For the Chamonix valley, the record precipitation over 10 days is 1.3 m at 1,050 m, 2.6 m at 1,500 m, but 4.15 m at 2,000 m. This increasing trend is likely to break down at higher elevations as a result of the decrease in air humidity and wind effects.
The determination of extreme snow precipitation is of great importance because the probability of avalanche release increases with the amount of new snow. Like many other hydrological variables, the annual maximum of daily snow precipitation varies randomly, and these variations can be well captured by an extreme-value distribution such as the Gumbel law. This probability distribution links precipitation intensities with their frequency of occurrence. The occurrence probability P of an extreme event is very small, and a common practice for distinguishing between rare events is to introduce the period of return, T = 1/P, expressed in years. The precipitation associated with a period of return of 100 years has P = 0.01 chances of occurring (or of being exceeded) in any one year. Figure 7 shows the variation in the annual maximum of daily snowfall in Chamonix town center (altitude 1,050 m). The extreme-value distribution matches the recorded maxima fairly closely, but the adjustment is not perfect: the likely cause of this lack of fit is a slight nonstationarity in the time series resulting from the different climate patterns over the last few decades.
It is very tempting to want to apply extreme-value theory to avalanches, particularly to define the period of return of an avalanche event. The key problem is determining the random variable that describes the intensity of the phenomenon. In avalanche zoning (see Avalanches Mitigation), impact pressure is used in the definition of the intensity–frequency relationship, but its estimation from in situ observations is difficult. Scientists have thus used the run-out distance (point of farthest reach), which can be easily determined in situ when the avalanche deposit is visible. For regions with smooth tracks (like in Canada and Norway), Dave McClung and Karsten Lied (1987) showed how the period of return can be defined from the record of run-out distances. For mountain ranges with a complex topography (like in the Alps), extreme changes in the local landscape lead to significant changes in the distribution of avalanche run-out distances in areas that are relatively close together, which precludes the use of extreme value distributions. In the absence of more relevant variables, the period of return has been associated with the snow volume mobilized by the avalanche. Snowfall over three days (or the snowpack growth in three days) is used as a proxy for avalanche volume, and, therefore, the periods of return for the snowfall and its resulting avalanche are assumed to be the same. This is a very crude approximation of reality, but it has the advantage of being easily usable for a broad range of applications in engineering and risk mapping.
Avalanche release depends not only on the amount of new snow but also on how the snow cover has built up from successive snowfalls as well as on other meteorological parameters (such as wind speed and air temperature). An avalanche usually results from a combination of factors rather than just a single process (e.g., heavy snowfall). This is why accurate forecasting of avalanche activity in a given area remains so difficult. Many avalanches, especially large avalanches, occur after snow precipitation. If the snowfall exceeds 50 cm in three days, then the risk of avalanche release is marked, whereas an accumulation of more than 1 m of new snow within three days is usually associated with widespread avalanche activity. Not all avalanches are triggered by an increasing snow load. Mild spells or rainfall can also lead to a significant decrease in snow’s strength, followed by large deformations of the snow cover, which ultimately breaks and forms an avalanche.
Terrain is also a major contributing factor, which explains why some slopes are prone to frequent avalanche releases whereas others experience no or weak avalanche activity. Slope angle is the most important topographic factor: most avalanche accidents triggered by humans occur in starting zones with slope angles ranging from 30° to 45°. On rare occasions, natural avalanches, and even those triggered by humans, can be observed on gentler slopes (lower than 25°). Other topographic features (e.g., terrain convexity, path roughness, vegetation, orientation to the sun or altitude) play essential parts in changes in the snowpack. In avalanche mapping, potential starting zones and avalanche paths can be identified using photogrammetry (now superseded by Geographical Information Systems), and thus terrain inspection was used for many years to get a qualitative picture of a given region’s potential avalanche activity.
Snow is a highly variable and complex material. Its physical properties show great diversity, and they are prone to significant variations over the seasons. For instance, the density of fresh snow typically ranges from 100 to 200 kg/m3. Snow densifies as a result of metamorphisms and weight-induced compaction. The density of a typical season’s snowpack ranges from 200 to 400 kg/m3. Late in the season, the density of firn (old compacted snow) can exceed 600 kg/m3. Shear and tensile strengths also exhibit substantial variations over time. For instance, under windy conditions, loose powder snow rapidly becomes cohesive and can easily sustain a man’s weight because ice bonds connect the closely packed arrangement of snow crystals. If the air temperature then increases sufficiently (under the effect of the sun or because of a mild spell), liquid water forms and percolates downwards. The contacts between the snow grains become lubricated by a thin film of liquid water and the bulk loses its cohesion.
The life cycle of seasonal snow cover usually passes through two major phases. In the early part of the season, snow accumulates layer by layer, making the snowpack look like a sandwich. Its physical properties (e.g., density, cohesion, and shear strength) may be quite different from one layer to another. The interface between layers also plays an important part. For instance, it may be composed of loose snow with faceted grains, which offers low shear strength, or of hard snow (sun-generated crust, frozen-snow crust), which forms an ideal sliding plane. These weak layers are a necessary prerequisite for avalanche formation. When they break (due to snow loading or human action), the upper layers start to slide downward over the weaker layers, in the form of a cohesive slab, but this quickly smashes into thousands of chunks. For this reason, the resulting avalanche is called a slab avalanche. The slab fractures along a broad front. The resulting broken line of fracture at the top of the avalanche, known as the crown, can be 10 to 1,000 longer than the slab’s thickness. One striking feature of slab avalanches is that when the snow cover is sufficiently unstable, the weight of a single man is sufficient to create the initial disturbance that breaks the static equilibrium of tons of snow. This explains numerous fatalities caused every year by skiers who trigger the avalanches that bury them. Figure 8 shows the typical zigzag cut of the upper crack or crown of a slab avalanche involving dry snow. There is still a lack of understanding of the exact mechanism whereby a snow slab breaks away. Most recent theories emphasize the part played by weak layers in dry-snow slab avalanche release: the collapse of the weak layer combined with the propagation of shear cracks (along this weak layer) is seen as the most plausible scenario of avalanche triggering.
When approaching the late part of the season, and as a result of the air temperature or rainfall, snow grains start melting and the structure of the snowpack changes radically under the different effects of liquid water. Firstly, water alters the bonds between ice grains; it particularly favors the formation of clusters of ice grains held together by the capillary forces of the water menisci at the contact points between the snow grains. Second, when the liquid water content is sufficiently high (above 3%), the liquid layer covering the grain surfaces can slowly drain away under the effect of gravity and reach the lowermost layers of the snowpack, altering its entire structure. Low liquid water contents usually consolidate the snow cover: weak layers are gradually transformed, and when freezing overnight, the liquid bonds between snow grains ensure strong cohesion. When the air temperature is too high and the snow no longer freezes overnight, it loses its cohesion. On most occasions, avalanches initiate at one specific point and grow larger as they go downhill. These avalanches are called loose snow avalanches. Figure 9 shows the starting and deposition zones of a loose snow avalanche.
Less frequently, liquid water lubricates the boundary between the snow cover and ground, leading to large internal deformations of the snowpack. The differences in glide rates (typically a few cm/day) between areas of snow cover close to one another, even if they are both slow, may generate significant tensile stresses within the snowpack, and once a critical level of stress is reached, the snow cover develops a glide crack. Viewed from above, glide cracks often form arc-shaped crevasses, which may extend over a few tens of meters. The lag time between crack initiation and slab release ranges from a few hours to a few weeks. Slab release seems to be highly correlated with the air temperature and the degree of snowpack settlement. The resulting avalanche is called a glide avalanche. Figure 8 shows a glide crack that initiated in January 2012 in a ski resort in France. In early March, during a mild spell, the deformations within the snow cover increased until a glide avalanche formed, descending the hill very slowly (the typical velocity of this glide avalanche was 1 m/s), but damaging a chairlift.
One particular difficulty in describing avalanche dynamics lies in the multitude of interlinked processes related to terrain, snowpack, and meteorological conditions that affect avalanche release and propagation. As with many other natural phenomena, almost any combination is possible, and so a great variety of words are used for referring to different types of avalanches. Several avalanche classifications have been proposed since the earliest developments of avalanche science. In 1925, French geography André Allix noted ironically that “several classifications have been suggested, which provides evidence that none of them is satisfactory.” Terms such as slab avalanches, wet-snow avalanches, and powder-snow avalanches are common, but they are ambiguous. For instance, the term powder-snow avalanche is often used to refer to a cloud of snow that descends at high speed, but on rare occasions wet snow can also form high-speed clouds. Powder snow can also form avalanches that do not develop clouds.
The former head of the SLF, Marcel de Quervain, was well aware of the risks of confusion induced by these terms. In the 1970s, he led an international committee on the classification of avalanche morphology, and this resulted in a multilingual book entitled Avalanche Atlas (1981). This scientific classification is especially useful for comprehensively characterizing actual events. Mountaineers prefer to use a classification describing the type of snow and release. Engineers are often more interested in the dynamic features of avalanches and the trajectories they follow. Apart from an avalanche’s volume, its dynamic features take little account of the initiation processes, so the classification used by engineers is based on flow dynamics. Their classification is usually broken down into two basic types:
• powder avalanches—high-speed, low-density flows that take the appearance of a snow cloud
• flow avalanches—dense flows of snow, which stick closely to the ground’s contours
This dichotomous reduction of avalanche dynamics makes it possible to conceptualize the behavior of snow avalanches and find answers to complex engineering and avalanche mapping problems. In practice, however, when referring to an actual event, the morphological variety of avalanches requires further refinement. For instance, many large avalanches have two components, which may interact or move independently of each other: a dense core (which behaves like a flow avalanche) and an aerial component (composed of snow in suspension in the air) that often hides the dense core. In cases like this, it makes more sense to refer to each component of the avalanche separately rather than to see it as a single entity. The morphology of an avalanche can also vary as it descends a slope. A powder airborne avalanche may undergo an abrupt transformation in which the highly concentrated layers that form its base separate from the less compact upper layers. Due to the higher energy dissipation, the denser underflow decelerates rapidly and comes to a halt, whereas the less compact, diluted cloud detaches and continues its descent.
Among the numerous factors that affect flow dynamics, the following are especially significant:
• Avalanche trajectory depends to varying degrees on topography: the higher its velocity and the higher the center of its mass relative to the ground, the less influence the terrain has on the trajectory. Inertia-dominated avalanches refer to avalanches whose momentum is sufficiently strong for the flows to be weakly dependent on terrain. Gravity-driven avalanches have low-inertia flow regimes, in which motion is dependent upon the difference between frictional forces and the driving forces due to gravitational acceleration. The respective parts played by inertia and energy dissipation in avalanche dynamics are still studied today.
• The dynamics of snow avalanches depend not only on the volume initially released, but also on the volume of snow entrained or deposited during the flow phase. The final size, velocity, and destructive potential of avalanches can be strongly influenced by the amount of snow that is carried away from the snowpack. Documenting the mechanisms and effects of entrainment, as this is known, is one of the greatest challenges faced when monitoring avalanches in test sites such as the Vallée de la Sionne. In avalanche-dynamics models, the role of mass and momentum exchanges at boundaries between the flow and the snowpack are still open questions.
A powder avalanche (also called a powder-snow avalanche) is an inertia-dominated flow, which involves a cloud of snow particles maintained in suspension by air turbulence. It has a high velocity (50–100 m/s). Owing to its high momentum, it may be free, to a significant degree, from the influence or constraints of topography and follow paths that are straight and parallel to the line of steepest slope. Figure 11 shows a powder avalanche whose momentum carries it up the facing slope after reaching the valley bottom. The physics of powder avalanches is partially understood. The most recent investigations conducted by Betty Sovilla and her colleagues at the Vallée de la Sionne show that the internal structure of powder avalanches is quite complicated, and it is difficult to disentangle the various processes at play, which vary considerably spatially and temporally. So the description below tries to give the big picture by synthesizing field observations and measurements, laboratory experiments on density and turbidity currents, and theory.
The cloud of snow expands during the descent as it entrains ambient air. If the cloud also entrains snow from the snow cover, then the effect of dilution is counterbalanced and the cloud remains vigorous and potentially very destructive. In the absence of snow entrainment, the mean density inside the cloud decreases quickly. If this is the case, even though the cloud has a high velocity, its destructive potential is substantially lower. The typical cloud flow height ranges from 10 to 100 m. The mean density of the overall cloud is a few tens of kg/m3, but its local density varies significantly with depth (a phenomenon called density stratification).
Naturally occurring powder avalanches are difficult to observe because they are much less common than flow avalanches. Furthermore, their deposits extend over large areas, and, unless they are covered with plant debris, their surface is smooth. In the most common scenario, powder avalanches release during snow storms involving heavy snow precipitation and low temperatures. Occasionally, they can occur after a snow storm (in fine weather, as shown by Fig. 10). On rare occasions, old snow can form powder avalanches, especially on steep slopes. After heavy rainfall in May 1983, the northern face of the Aiguille du Midi, in Chamonix, discharged a large volume of wet snow into the steep gully above the entrance of the Mont-Blanc tunnel. The snow accelerated so fast that it formed a high-speed cloud that swept through the protecting forest.
Powder avalanches are often confused with dry-snow, high-speed avalanches, with which they have many features in common. For instance, after major disasters such as those at Maseguchi (Japan; 13 fatalities in January 1986) or Chamonix (France; 12 fatalities in February 1999), the media reported the occurrence of powder avalanches, but subsequent appraisals by experts concluded that the damage was caused by high-speed flow avalanches mobilizing dry snow. Typically, in the run-out phase, the flow avalanche’s speed drops below 25 m/s, whereas that of a powder avalanche still exceeds 50 m/s. At the same time, the mean bulk density increases in flow avalanches (due to snow compaction), whereas it decreases in powder avalanches (due to air entrainment). Consequently, the types of damage caused and the features of the deposition zones are quite different depending on the type of avalanche.
A powder avalanche passes through different processes from release to run-out:
• A powder avalanche starts as a shallow flow, but quite early, for reasons that still cannot be adequately explained, part of the snow volume is raised into suspension in the air. This process is sometimes called ignition.
• A denser head is formed at the leading edge of the flow, followed by a more diluted tail (also called turbulent wake). The tail is often separated from the head by a billow.
• Motion is produced by the contrasting densities between the suspension and ambient air. As air is entrained into the current, the snow concentration decreases inside the main body of the avalanche, leading in turn to a decrease in the buoyancy force unless the body is supplied by a sufficient input of particles. Snowpack erosion as the avalanche passes over it and air entrainment into the head are, therefore, the two main processes controlling the avalanche’s bulk dynamics.
• If the head entrains snow from the snow cover, it forms an acute, wedge-shaped edge. The typical flow depth is 10–20 m (theory predicts a front angle of 60° to the ground). The snow–air interface is sharp. In the absence of entrainment, the head inclination is much steeper (close to 90°) and sometimes overhanging. In such cases the flow depth can rapidly increase above 50 m; the air–snow interface is blurred.
• In order to counteract particle settling, sufficient turbulence must maintain snow particles (ranging from snowflakes to snowballs) in suspension. The key parameter for the formation and development of a powder cloud is that the fluctuating rising particle velocities exceed the settling particle velocities.
• In the run-out phase, the cloud collapses and settles to form a vast, thin deposit (thickness less than 1 m). In many events, the cloud has been observed separating from the denser basal layer, as this comes to a halt as soon as the slope gradient is too low (typically lower than 20–25%).
A powder avalanche’s potential for destruction lies primarily in its kinetic energy. Damage to obstacles is caused by the impact pressure; the magnitude of that pressure depends on the square of the avalanche’s velocity and on its density. Typically, a cloud with a mean density of 50 kg/m3 and a velocity of 50 m/s generates a mean impact pressure exceeding 60 kPa on a narrow obstacle such as a chairlift mast. Instantaneous impact pressures can be much higher as a result of density stratification (the basal layer is denser than the upper layers), as well as velocity and density fluctuations. An obstacle’s features (e.g., size or shape) also influence the impact pressure. In the Vallée de la Sionne, mean impact pressures close to 500 kPa have been measured, with peak values (over timespans as short as 5 ms) three times higher than that average. A pressure of 500 kPa is more than the static load exerted by a 40-ton truck on a square meter surface! The exact nature of the dense basal layer is still debated: Is it a dense core feeding the cloud as suggested by Dieter Issler? Is a dense flow produced by particle settling behind the front as shown by Betty Sovilla? Is it similar to the abrupt transition observed in turbulent particle-laden flows in the laboratory, for which part of the cloud collapses and forms an independent basal layer?
A flow avalanche is an avalanche with a high-density core at its base. The flow depth does not generally exceed a few meters. The typical mean velocity ranges from 5 to 25 m/s, but velocities ranging from 25 to 50 m/s can also be observed for inertia-dominated avalanches. On average, their densities are fairly high, ranging from 150 to 500 kg/m3. Gravity-driven avalanches closely follow ground contours. In the absence of topographical constraints (e.g., gullies or ridges), their lateral limits are poorly, but when they are confined to a main channel (talweg) or to a gully, they are channeled along its well-defined track. On occasions, the main channel can be cluttered with debris and/or former avalanche deposits and a new avalanche can be diverted toward a secondary channel or another slope. Except for in those circumstances, the trajectory of gravity-driven flow avalanches is easy to predict as it is imposed by the terrain’s features.
Inertia-dominated flow avalanches reach higher velocities and may overrun low-terrain obstacles. For this reason, it may be more difficult to anticipate their precise trajectories as they will depend on avalanche speed, snow consistency (e.g., density and moisture), and local terrain roughness. Figure 12 shows three deposits from flow avalanches near Geschinen (Switzerland) after a snow storm. The first avalanche was an inertia-driven avalanche mobilizing dry snow (perhaps with an airborne component). The avalanche was confined in the Geschinerbach gully for part of its flowing phase and then suddenly spread out on arriving on the alluvial fan. In spite of this spread, part of the flux went beyond the valley bottom, crossed the Rhone River, and deposited snow on the opposite bank. The very long run-out distance was not surprising due to the high velocity of the avalanche and reduced friction observed in extreme dry-snow avalanches.
In the starting zone, this avalanche crossed Lake Trützi and emptied it, with the important consequence that the liquid water content in the snow cover downhill of the lake increased dramatically. The second avalanche thus mobilized wet snow and came to a halt in the upper part of the alluvial fan. Interestingly, the third avalanche also involved wet snow but went much farther than the second avalanche. Its run-out distance was slightly shorter than that of the high-speed avalanche. This shows that even with higher friction and lower velocity, gravity-driven flows can travel long distances over shallow slopes (the mean slope from the alluvial-fan’s apex to the Rhone River is 8%). At least two mechanisms were at play: first, the first two avalanches smoothed out the ground’s irregularities, reducing flow resistance for the third avalanche; second, lateral deposits (levees) formed and confined the ensuing flow, limiting the lateral spread of the third avalanche. These levees were formed by grain sorting (a process called particle size segregation): wet snow forms rounded snowballs of different size; the largest snowballs concentrate in the fast-moving upper layers (next to the free surface of the avalanche) and are transported to the leading edge; from here they are pushed to the side by the core (made up of the finest snowballs), which creates static coarse-grained levees. This self-organization has a great influence on the flow’s behavior as it reduces the dissipative effects of frictional forces. The Geschinen example illustrates the variety in morphology and behavior of flow avalanches.
Most flow avalanches flow like very viscous fluids, and this analogy has inspired avalanche-dynamics models based on the Saint-Venant equations (also called the shallow-water equations). The scientific underpinning of these equations is the fundamental assumption that the fluid is homogeneous—an assumption that performs well with many environmental fluids—but this is not the case with snow. Figure 13 shows a sample of avalanche deposits, and the diversity of shapes and sizes of blocks, snowballs, slurry and slabs is clear to see. The bulk density can also vary considerably, between 300 and 700 kg/m3. Small volumes of dry snow usually lead to thin, low-density deposits, whereas large volumes of wet snow are more likely to produce thick interlocked masses of compact blocks. Reports from victims caught in wet-snow avalanches reveal that the interlocking chunks of snow made them feel as though they were trapped in a concrete cage. The diversity in snow’s features (e.g., cohesion, grain size, and density) and its thermodynamic sensitivity (to temperature or liquid water content) have two important consequences when modelling avalanche dynamics:
• Determining the mechanical properties of flowing snow is extremely difficult, if not impossible. Laboratory devices cannot be used because of size-scale effects and the severe difficulties inherent to snow (sample fracture during shearing tests, thermodynamic transformations, etc.). Large-scale outdoor experiments may hold more promise. For instance, Martin Kern and his colleagues (2004) used a chute to study small-scale avalanches. By measuring the avalanche’s velocity profile at the chute sidewall, they deduced that flowing snow behaves like a viscoplastic material. In most cases, snow friction simply cannot be measured in the laboratory or the field. Researchers subsequently tried to use empirical relationships to link snow friction to flow variables (flow depth and velocity) and to adjust the frictional parameters to field data such as observed run-out distances and, when available, front velocity. Adjusting these frictional parameters provides decent results for high-speed dry-snow avalanches, so avalanche-dynamics models perform fairly well for them. The models run into trouble with wet-snow avalanches: because these can travel long distances at low velocities, most (if not all) empirical models of snow friction fail to capture the avalanche dynamics of their run-out phase. As the frictional parameters cannot be measured but can only be estimated by matching the model’s predictions to field data (with varying degrees of success), avalanche-dynamics models are based on speculative foundations. Their accuracy may thus be difficult to assess depending on the avalanche’s path and features.
• Avalanche mobility depends not only on snow friction but also on the flow structures created by the avalanche as it descends the hill. As shown by the Geschinen example (see Fig. 10), there is clear evidence that avalanches, especially wet-snow avalanches, can develop complex internal structures (e.g., levees that channel the flow, coarse-grained fronts that may retain the flow behind them, elongated lobes, fingering instabilities or density stratification) that are not included in current avalanche-dynamics models and may be out of reach for a while. One major research hurdle to overcome in the coming years will thus be coupling flow structures and flow dynamics.
The destructive potential of flow avalanches depends on their characteristics. Inertia-dominated flow avalanches behave like powder avalanches: they cause damage to structures by hitting them at high velocity. The force of impact is thus directly proportional to the avalanche’s density and the square of its velocity. Gravity-driven avalanches exert compressive hydrostatic-like forces on obstacles. The resulting force is thus proportional to the avalanche’s density and the square of its flow depth. For a wet-snow avalanche with a density of 450 kg/m3 and a velocity of 10 m/s, Betty Sovilla and her colleagues (2010) measured pressures ranging from 100 to 300 kPa, which shows that contrary to common belief, wet-snow avalanches can pack an incredible punch.
Conclusions and Perspectives
The study of avalanche phenomena, the knowledge developed, and the modern modeling of snow avalanches have all essentially been driven by land management issues and engineering applications. Indeed, predicting run-out distances, impact forces, and the frequency of occurrence of rare events is of paramount importance for mapping risks and protecting against natural hazards. Snow mechanics and avalanche release are also problems particularly relevant to the safety of skiers and mountaineers. They have been at the core of research since the earliest developments of the avalanche science.
Even though other tools, such as statistical techniques (e.g., data correlation, extreme value theory, or Bayesian simulation), have been used for predicting extreme events, the fluid-mechanics approach has emerged as the most successful way to calculate the salient characteristics of avalanches. New generations of models have developed since the early 1920s, with ever-increasing levels of sophistication. The first avalanche-dynamics models conceptualized avalanches as rigid blocks whose motion was controlled by the difference between the gravitational and frictional forces. The second generation of models viewed avalanches as analogous to water floods, which led to models governed by the Saint-Venant equations. Although the idea dated back to the 1960s, it was not until the 1980s that computers and numerical techniques were sufficiently powerful to solve the Saint-Venant equations. In the early 2000s, the first commercial products based on these equations were made available to engineers. With the steady increase in computational resources since then, it has become easier to predict the behavior of avalanches.
Strikingly, whereas the last 30 years have provided substantial progress in terms of the physical understanding of avalanches, improvements in land management and engineering applications appear much more limited. Indeed, a number of problems (e.g., model calibration or the values of input parameters) that already existed in the sliding-block models have yet to be fixed; they persist, often hidden by the level of complexity of current numerical models. The frictional parameters cannot be measured and are thus fitted to field data. These parameters are clearly more conceptual than physical: they do not represent a single physical process; rather they combine many different physical processes into a single, simple mathematical expression.
The lack of progress in avalanche modeling is in contrast to the advances made in other scientific fields, such as meteorology. Since the 1970s, weather forecasts have become ever-more reliable in parallel with improvements in computational power. The explanation for this difference is simple: in meteorology, the material is air, a fluid whose behavior is well known. The main difficulty lies in computing the enormous volumes of air encountering various flow and temperature conditions. For avalanches, the material is snow—subtle mixtures of water (in different forms) and air exhibiting remarkably complex behavior. Avalanches involve processes that are currently impossible to fully reproduce in the laboratory; field surveys provide only fragmentary information, and historical records are often patchy. Consequently, there are still many problems to be solved in the field of avalanche dynamics and, as yet, few solutions. The common sense and field experience of experts who have to deal with avalanches day-to-day are still needed to properly interpret numerical data from models, notably by pinpointing and removing the most evident errors.
All these unknowns in avalanche science raise questions about how much trust can be placed in state-of-the-art avalanche prediction and forecasting. Predictability is a key issue in science, but it can also become central to legal debates in courts of justice when decisions must be taken on whether individuals or institutions are responsible for a disaster. After the fatal avalanches of February 1999 in Chamonix (France) and Évolène (Switzerland), the courts ruled against local authorities. They decided that the authorities had failed to foresee their respective avalanches despite regional avalanche bulletins and avalanche risk maps. On the whole, experiences over the last 40 years have shown that standard engineering methods (e.g., the Swiss guidelines, statistical tools, and avalanche-dynamics models) provide decent predictions on most occasions. There are, however, events for which the predictive capacity of current models is clearly insufficient. Some failures can be explained by the shortcomings of current knowledge and best practices, but others provide evidence for a more subtle shortfall in today’s approaches to predicting extreme events. Indeed, the statistical and fluid-mechanics models fitted to historical events make implicit use of the assumption that the past sheds light on what is yet to come: extreme avalanches are just bigger avalanches than their older siblings. Yet, just as with stock-market crashes, there are also outliers, i.e., avalanches with dynamic features that share little in common with those observed for previous avalanches on the same path. So, in order to avoid unpleasant surprises in the future, computational frameworks should be constantly refined by integrating all possible elements of complexity. In this respect, it should not be forgotten that since its very beginning, avalanche science has been a cross-disciplinary subject combining forestry science, physical geography, physics, statistics, and other fields. Although there is a clear demand for better quantitative tools for risk assessment, especially numerical models, the importance of qualitative techniques (based on interpretations of historical records, field observations, and the analysis of aerial photographs) should not be forgotten. The key to better avalanche safety may be to put the emphasis on a balanced combination of qualitative and quantitative tools.
Ancey, C. (2012). Gravity flow on steep slopes. In E. Chassignet, C. Cenedese, & J Verron (Eds.), Buoyancy driven flows (pp. 372–432). New York: Cambridge University Press.Find this resource:
Ancey, C., & Bain, V. (2015). Dynamics of glide avalanches and snow gliding. Reviews of Geophysics, 53.Find this resource:
Aydın, A., Bühler, Y., Christen, M., & Gürer, I. (2014). Avalanche situation in Turkey and back-calculation of selected events. Natural Hazards and Earth System Sciences Discussions, 2, 581–611.Find this resource:
Bozhinskiy, N., & Losev, K. S. (1998). The fundamentals of avalanche science. (C. E. Bartelt, Trans.). Davos: EISFL.Find this resource:
Brugnot, G., & Pochat, R. (1981). Numerical simulation study of avalanches. Journal of Glaciology, 27, 77–88.Find this resource:
Bründl, M., Romang, H.E., Bischof, N., & Rheinberger, C.M. (2009). The risk concept and its application in natural hazard risk management in Switzerland. Natural Hazards and Earth System Sciences, 9, 801–813.Find this resource:
de Quervain, M. (1957). Avalanche classification. In Program Commissions of Snow and Glaciers International Association of Scientific Hydrology, pp. 387–392. Toronto: IAHS.Find this resource:
de Quervain, M. (Ed.). (1981). Avalanche atlas. Paris: Unesco.Find this resource:
Haegeli, P., Falk, M., Brugger, H., Etter, H.-J., & Boyd, J. (2011). Comparison of avalanche survival patterns in Canada and Switzerland. Canadian Medical Association Journal, 183, 789–795.Find this resource:
Heierli, J., Birkeland, K.W., Simenhois, R., & Gumbsch, P. (2011). Anticrack model for skier triggering of slab avalanches. Cold Regions Science and Technology, 65, 372–381.Find this resource:
Höller, P. (2007). Avalanche hazards and mitigation in Austria: A review. Natural Hazards, 43, 81–101.Find this resource:
Issler, D. (2003). Experimental information on the dynamics of dry-snow avalanches. In K. Hutter and N. Kirchner (Eds.), Dynamic response of granular and porous materials under large and catastrophic deformation (pp. 109–160). Berlin: Springer.Find this resource:
Kern, M. A., Tiefenbacher, F., & McElwaine, J. N. (2004). The rheology of snow in large chute flows. Cold Regions Science and Technology, 39, 181–192.Find this resource:
Laternser, M., & Pfister, C. (1997). Avalanches in Switzerland 1500–1990. In J. Matthews (Ed.), Rapid mass movements since the Holocene (pp. 241–266). Stuttgart: G. Fischer.Find this resource:
McClung, D. M., & Lied, K. (1987). Statistical and geometrical definition of snow avalanche run-out. Cold Regions Science and Technology, 13, 107–119.Find this resource:
McClung, D. M., & Schaerer, P. A. (1993). The Avalanche Handbook. Seattle: The Mountaineers.Find this resource:
Mears, A. I. (1992). Snow-avalanche hazard analysis for land-use planning and engineering. Denver: Colorado Geological Survey.Find this resource:
Pudasaini, S. P., & Hutter, K. (2006). Avalanche dynamics. Berlin: Springer.Find this resource:
Rudolf-Miklau, F., Sauermoser, S., & Mears, A. I. (2014). The technical avalanche protection handbook. Berlin: Ernst & Sohn.Find this resource:
Salm, B. (2004). A short and personal history of snow avalanche dynamics. Cold Regions Science and Technology, 39, 8–92.Find this resource:
Schweizer, J., Jamieson, J. B., & Schneebeli, M. (2003). Snow avalanche formation. Reviews of Geophysics, 41, 1016.Find this resource:
Simpson, J. E. (1997). Gravity currents in the environment and the laboratory. Cambridge, U.K.: Cambridge University Press.Find this resource:
Sovilla, B., Kern, M. A., & Schaer, M. (2010). Slow drag in wet-snow avalanche flow. Journal of Glaciology, 56, 587–592.Find this resource:
Sovilla, B., McElwaine, J., & Louge, M. Y. (2015). The structure of powder snow avalanches. Comptes Rendus Physique, 16, 97–104.Find this resource:
Stethem, C., Jamieson, J. B., Schaerer, P., Liverman, D., Germain, D., & Walker, S. (2003). Snow avalanche hazard in Canada—a review. Natural Hazards, 28, 487–515.Find this resource:
Tremper, B. (2008). Staying alive in avalanche terrain. Seattle: The Mountaineers Books.Find this resource: