Avalanche dams

8 February 2010

Dam design to defend communities against avalanches has been advancing, but there is still great scope for research. Patrick Reynolds reports

In northern Europe and the Alps, where communities and infrastructure can face snow avalanche hazards, the research efforts have increased in recent years into the effective design of dams as planned defence structures.

While dams and associated braking mounds can often mitigate the impacts in the avalanche run-out zones, there remains some uncertainty about their effectiveness in deflecting and, in particular, stopping the flows. The bottom line is they can reduce risk but cannot totally guarantee safety.

The European Commission (EC) recently published a report on the theory and practice in the field – “The Design of Avalanche Protection Dams: recent practical and theoretical developments” – which is recognized as representative of the state of the art, establishing a new design framework. Some of the research that informs the most recent developments and production of the report was helped by EC projects CADZIE and SATSIE, and support also came the Icelandic Avalanche and Landslide Fund.

Researchers also note, though, the parallel but independent work in Switzerland that has produced a new set of design rules for catching dams. The Swiss work and EC report – which covers both catching and deflecting dams – are compatible, the researchers add. They are being used alongside one another and both are to be employed in a forthcoming commercial avalanche dam design software tool, which is being developed in Austria.

However, despite the progress, there are still many areas of broad research to be advanced as well as uncertainties to be reduced.

Snow avalanches

The main focus of research and dam design has been in relation to dry-snow avalanches, the conceptual model of which comprises a dense core (approximately 300kg/m3) with a fluidised layer on top, and in front too, and the mass is enveloped by a suspension layer called the powder cloud.

In the moving mass, transitions between the layers are variable and take place over a range of flow depths. The three layers are progressively less dense, and the model takes the avalanche core to have persistent contact between particles. It is estimated to be approximately 1m-3m in depth, and possibly greater where there are channelling effects in the running flow.

The fluidised layer has less contact between particles and the range of density is approximately 10kg/m3-100kg/m3. It typical depth is 2m-5m and its front can precede the dense core by tens of metres. The flow depth of the powder cloud is at least this magnitude though can extend to 100m or more.

Wet-snow avalanches differ in several ways from dry-snow events, not least in having more moisture and being slower, which also leads them to follow depressions in the terrain. The flows may have a long tail, can spread and split around obstructions, and they also carry loose materials and rocks, which may be deposited near dam and flow channels.

When it comes to powder-snow avalanches, however, they have such large run-up heights and speeds – observed to be hundreds of metres high and 50m/sec-70m/sec, respectively – that they cannot be stopped by man-made dams.

Dams and mounds

Two principle types of dams are employed as avalanche defence structures – deflecting dams, to divert flow, and catching dams to halt it as much as possible. In addition, braking (or retarding) mounds are used widely where there are dense, wet-snow avalanches.

Deflecting dams need to be well positioned on steep terrain in sufficiently long run-out zones, which should provide plenty of distance and space to spill the diverted snow flows. Optimally, they will divert an avalanche without slowing the flow by much to both keep deposits to a minimum and keep the dam mostly clear for the next event.

Where space is limited, such as at the foot of slopes, and there is the risk of run-out from dense avalanches or the dense parts of mixed flows, then it may be possible to build catching dams instead. The internal friction angle of the snow, so far as can be known, should be greater than the slope of the terrain. The storage volume upstream of the dam may need to be big enough to take more than one avalanche per season.

The effectiveness of catching dams, however, depends on the scale of events as well as the size and location of the structures. It is not unknown for them to be overtopped by large, fast avalanches.

Braking mounds are built uphill from dams. They are used to protect against wet-snow events and are thought to have little effect against faster, dry snow avalanches. The main aim of braking mounds is to help slow and break avalanche flows by dissipating some of the kinetic energy.

Properly built, traditional avalanche dams comprising loose materials (rock, gravel, sand) are usually 10m-25m high and can be 50m-500m in length, or even longer, resulting in volumes in the order of 104m3-105m3.

The deflecting dam at Seljalandsmuli in Isafjordur, in Iceland, is 650m long and its braking mounds are 7m high, resulting in total construction volume of 360,000m3. Elsewhere on the island, above Neskaupstadur, the defence structures comprise two rows of 10m high mounds and a steep 17m high by 400m long catching dam.

In addition to loose deposits, dams and mounds may also be built using reinforced earth or reinforced concrete (RC) where there are steep faces. In Norway, for example, there is use of a 8m high, buttressed RC wall at Odda and a 10m high dam of six RC shells anchored to bedrock.

Approaches to analysis

Traditional analytical principles are based on simple or dynamic point-mass calculations, or dynamics of the leading edge of the avalanche. The weakness in the point-mass approach is seen as its simplifying an avalanche mass so much that the transverse width is neglected, which leads to lateral and longitudinal interactions being ignored.

Modelling an avalanche is also a challenge when the mass might not strike a dam perpendicularly and this is the design case for deflecting dams. The simplified approach also means there is no objective method to calculate momentum loss within an avalanche upon impact with a dam.

Advances in theoretical, experimental and field research over the last decade have led to improvements in design criteria for deflecting and catching dams. The criteria are the more dynamic concepts of supercritical overflow and flow depth downstream of ‘shock’ – sharp gradients or discontinuities in flows. Though incomplete, and based on some subjective and partly justifiable considerations, note researchers, the criteria represent a new basis for avalanche dam design.

The updated approach takes the dynamic equations for the gravity flow of a shallow, free-surface layer of granular material on a slope, producing a range of Froude Numbers (Fr), and uses them in a depth-averaged calculation. Only the dynamics of the dense core of the avalanche are modelled, Froude numbers for natural dry-snow flows is approximately 5-10, which is well within the supercritical range as defined by Fr>1.

Changes in upstream flow depth and velocity are of greater importance to how the dense core interacts with obstacles than previously thought. Sharp gradients or discontinuities, in depth-averaged analysis, are believed to be an approximation to real physical responses to changes in shallow, supercritical flows. But, the shocks are generally local effects in avalanches.

Some parameters and processes, however, are not often represented in the depth-averaged analysis, such as: variable terrain and dam shapes; initial ‘splash’ impacts; shearing in overflow at the dam crest; and, overflow of other, non-core layers (fluidised, powder), and related dynamic and air pressure effects in the interstitial air.

Further processes not represented include: impacts that transfer snow from the core into suspension; transfer between the avalanche and snow cover, including deposition near the dam; and, snow drift.

Another parameter that is not well understood in terms of avalanche dynamics is terrain friction. It is expected to be relatively unimportant for dry-snow events as there are rapid, sequential impacts of avalanche layers, but where there is prolonged contact, such as at long deflecting dams with acute diversion angles, it may play a greater role.

But a cautionary note is sounded for this analysis, for the utility of the hydraulic equations can only be taken so far in the analogy; snow avalanches are not true fluids but inelastic granular materials. Therefore, a supplemental approach to help gain further, empirical information is to run small-scale tests, although modelling is, of course, limited by how well the dimensional similarity can be achieved between the natural, flowing snow mass and the reduced-scale experiment.

Tests that have been done include field chutes with snow and laboratory experiments with granular materials. They have, for instance, given some measurable understanding of the effectiveness of braking mounds with data indicating that, despite experimental uncertainties, they do substantially slow supercritical granular flows. The results are little affected by the scale of chutes, velocities and materials.

Design of dams

Once the location for a dam has been decided, the forgoing analyses are important to help determine the Design Avalanche with which to help estimate the appropriate size and shape of a dam or mound.

In applying design criteria, the sensitivity of results to various parameters and awareness of where uncertainties lie in theory, analysis methods and data should all be borne in mind. The design process can use the avalanche analysis calculations for an initial, simplified estimate of dam height and to help develop early ideas of geometry. To that end, the profusion of possible design variable should be omitted, initially.

Afterward, more complex design modelling might be undertaken by gradually, and increasingly, including the variables to build the numerical simulations that are based on depth-average equations and shock-capturing algorithms. The design work can help establish the dam height, length, width and geometry of a dam. For catching dams it is also important to estimate the storage capacity available on the upstream slope, on top of the existing snow cover.

Design work will also establish the ‘critical dam height’, defined as the maximum height of the obstacle (above snow cover) that changes the avalanche from supercritical to subcritical flow state as the snow passes over the crest. By adding the upstream flow depth then the minimum physical height of the dam to arrest the flow can be estimated. Designs based on such a shallow fluid dynamics approach of the avalanche core should be viewed as the minimum requirement, say researchers.

The design, though, assumes no loss of momentum in the avalanche during its impact with the dam. Yet the reduction in kinetic energy has been shown to be significant in tests, such as chute experiments with granular materials, and snow. The tests with glass beads show that where the upstream faces of dams have angles of at least 60 degrees to the terrain they have similar efficiency in dissipating energy to those built perpendicular to the slope. Angles of 30 degrees were found to be less efficient.

However, when seeking a basis for assessing momentum reduction in a design then there is considerable uncertainty in applying results from chute tests to full-scale dams when examination the reduced in velocity up the faces. Test data seem to overestimate the reduction. Whatever value of the momentum reduction factor is adopted in supercritical flow analysis, it is catching dam design that has most sensitivity to the parameter, according to chute tests.

Design of braking mounds

Designs of braking mounds draws upon limited field evidence, so far, to describe their effectiveness but laboratory experiments show that, at least using granular materials, they can reduce avalanche speed and also shorten run-out distances. However, there are no accepted design guidelines for the structures.

Experiments with braking mounds have only got underway seriously in recent years but they are viewed as having a similar effect to baffle blocks and piers in stilling basins at impounding dams.

Researchers suggest that, measured above snow cover, mounds should be up to three times as thick as the dense core of a wet-snow avalanche. They further suggest the mounds should be designed with steep upstream and sides faces, and the aspect ratio kept close to unity. A row of short mounds is taken to be more effective for energy dissipation than fewer, wider structures. Subsequent rows should have their mounds in staggered patterns, though data suggest they are less effective in reducing velocity than the first, uphill line.

Aside from their safety functions, of course, the avalanche defence structures need to fit well into the local landscape. They can have large environmental and visual impacts, notes the EC report, and so integration is important when designing such structures to be near settlements, in particular.


The EC report has a data set with 22 events, mostly from Norway, of the run-up of snow avalanches to man-made dams (six) and natural obstacles. Avalanche events ranged in volumes estimate at 15,000m3 to 800,000m3, the impact velocities were calculated to be 20m/sec-70m/sec and the vertical run-up the structures was 7m-90m.

Many of the obstacles are on rather steep terrain. While the run-up for many of the events fall within the ranges predicted by supercritical flow analyses there are several that were higher. However, the verification exercise also showed the uncertainty in the estimates from only moderate ranges in flow depth (1m-3m) and velocity (+/- 15%) producing large variations in run-up of the dense core. The greatest sensitivity is to velocity.

The report notes that ‘the rather wide spread of the data points compared with the assumed uncertainty of the theoretical predictions clearly indicates an incomplete understanding of the dynamics of the impact process.’ Damage may occur considerably higher than the run-up of the dense core, and some field data marks are believed to have been caused by the impacts of fluidised layers or powder clouds. There are virtually no observations of the interaction of powder snow avalanches with man-made dams, however, though inferences are drawn from impacts on natural obstacles.

Further research

Further research is needed on a number of fronts to take forward the understanding and design of avalanche dams and braking mounds.

The EC report identifies the following key areas for study: momentum reduction from impacts; maximum deflecting angle; the effect of terrain slope on the design; entrainment and deposition effects; and, the effects of the fluidised layer and powder cloud, which are ignored in analysis focused on the dense core.

In addition, study is needed into inconsistencies in field data, or at least greater variability than anticipated. More complete physical descriptions of lateral and longitudinal interactions within the avalanche snow mass during impact also need to be developed.

For deflecting dams, a larger data set is needed to judge if empirical adjustments of calculated run-up can help improve estimates. There is also a view among researchers, which can be examined further, that less flexibility should be allowed in selecting model parameters as this will help reduce subjectivity.

A copy of the EC report “The Design of Avalanche Protection Dams: recent practical and theoretical developments” can be downloaded from:


Privacy Policy
We have updated our privacy policy. In the latest update it explains what cookies are and how we use them on our site. To learn more about cookies and their benefits, please view our privacy policy. Please be aware that parts of this site will not function correctly if you disable cookies. By continuing to use this site, you consent to our use of cookies in accordance with our privacy policy unless you have disabled them.