On stability sampling strategy at the slope scale

https://doi.org/10.1016/j.coldregions.2010.02.013Get rights and content

Abstract

Snow slope stability evaluation is often based on a single test location within a slope. However, we know that snow cover properties and stability may vary at the slope scale. Reliably estimating the slope-scale variability requires many samples, ideally more than 100. As this is unpractical, it has been proposed to perform at least two tests — about 10 m apart — on a given slope. In addition, if small column stability tests are used (such as the compression test), it seems reasonable to perform two tests at each of the two locations. Differences between the two tests at one location allow one to assess the small (or pit-) scale variability (and/or the test uncertainty), whereas differences between the pairs at different locations may hint at the slope-scale variability. We analyzed 22 small slopes each with four pairs of stability tests. In 61-75% of the cases the two stability tests at a specific location provided consistent results, depending whether we focused on the CT score or the fracture character (which was less variable). Comparing the different sampling locations on a given slope (∼ 10–15 m apart) showed that at the slope-scale the differences between sampling locations (59–75%) were similar to the differences found at the pit-scale. Rather stable slopes tended to have more pit-scale variation than rather unstable slopes. Based on our analysis, we suggest an interpretation scheme and an adjusted sampling procedure. In particular, a second pit on a slope seems only necessary if the first pit does not indicate instability. In all other cases, a second pit can reduce the number of false-stable predictions.

Introduction

For assessing avalanche risk in backcountry operations and for public forecasting of the regional avalanche danger information on snowpack instability is of crucial importance. Snowpack instability manifests itself by recent avalanche activity, whumpfs or shooting cracks. In the absence of these obvious signs of instability manual observations of snow stratigraphy combined with stability tests are presently the method of choice to seek instability data. The problem with these measurements is twofold: (1) Their availability is limited since the measurements are time consuming and in addition may expose the field crew to an undesirably elevated level of risk. Consequently, these snow stability data are available only with low resolution in space and time. In the future, snow micro-penetrometer measurements may speed up sampling — provided stability can be derived from the SMP signal (Bellaire et al., 2009, Pielmeier and Marshall, 2009) and simulated snow stratigraphy data may complement manual observations (e.g. Schirmer et al., 2010-this issue). (2) The validity of the measurements is limited, since test results need to be extrapolated to the surrounding terrain. This limitation follows from the fact that the mountain snowpack is inherently variable at various scales (e.g. Schweizer et al., 2008). The uncertainty resulting from the extrapolation is difficult to quantify, but depends on the specific snow conditions. In general, the snowpack variability is often not such that observations are useless; they may at least reveal relevant weak layer and slab properties, and whether they may interact in a critical manner; stratigraphy tends to be spatially more uniform than stability test results. On average, point stability observations performed by very experienced forecasters are fairly reliable. Schweizer and Jamieson (2010) estimated the error rate to about 5–10%.

Obviously, we are not simply interested in the snowpack stability at a specific location, but rather would like to know whether an avalanche may release (or may be triggered) on slopes similar to the sampling site. The slope where the measurements are taken is typically safe as otherwise the safety of the field crew would be compromised. Due to spatial stability variations, point stability (which we observe) may not be the same as slope stability (which we wish to know). Variations in point stability at the slope scale can affect release probability as is sometimes exemplified when an avalanche is triggered by a skier who was not the first to enter the slope. This prompted various attempts to relate avalanche release probability to variability at the slope scale — at least conceptually (e.g. Kronholm and Schweizer, 2003). However, reliably measuring slope scale variability is too time consuming as a large number of measurements is required — even if using, for example, a snow micro-penetrometer. If the measurements should be amenable to a geostatistical analysis and the analysis has to yield accurate estimates, the number of measurements required seems to be on the order of 50–100 (Webster and Oliver, 2007, p. 125). Obviously, such slope scale measurements of weak layer and slab properties can only be collected on safe slopes, either not steep enough to slide or at times of generally fair (or even higher) stability. Measurements at the slope scale (e.g. Kronholm et al., 2004) have revealed that the correlation length was often on the order of ≤ 10 m.

As it is not practical for operational purposes to completely sample a slope, the question arises how a single point stability measurement should be supplemented so that slope stability (or avalanche release probability) can more reliably be extrapolated. Accordingly, it has been proposed (e.g. Birkeland and Chabot, 2006) to do a second observation at a representative site beyond the correlation length from the first test and choosing the least stable of the two test results. As the correlation length is unknown, at least about 10 m have been proposed as the distance between two independent tests (Jamieson and Johnston, 1993, Schweizer et al., 2008). Furthermore, performing two tests side by side at one location can decrease the uncertainty of test results and may indicate the small scale variability (0.1–1 m).

Independent estimates suggest that the critical size for a self-propagating fracture is on the order of 1–10 m (Schweizer et al., 2003). The lower range is probably more relevant for skier-triggered avalanches, i.e. the critical size may be on the order of the slab thickness (McClung and Schweizer, 2006). We assume that a fracture can propagate (and avalanche release is possible) if at one of the two test locations (∼ 10 m apart) test results suggest that initiation and propagation is possible. On the other hand, significant variations at the scale of 0.1–1 m rather indicate conditions unfavorable for fracture propagation.

The aim of the present study is to assess whether performing two pairs of tests about 10 m apart improves our ability to predict snow slope stability, and if so, to suggest a procedure for sampling.

Section snippets

Data

Our study area is located in the Eastern Swiss Alps near Davos (9°47.5 E, 46°48 N). We collected compression test (CT) results (Jamieson, 1999) on 22 slopes located above tree line, i.e. most slopes were not sheltered. The mean elevation of the test slopes was about 2400 masl and the mean slope angle was 25°. About half of the slopes had northerly, the other half mainly south-westerly aspect. On each slope, four pairs of CTs were performed. Pit locations where two adjacent tests were done,

Methods

Compression tests were performed according to Jamieson (1999). The loading step at failure, failure depth and fracture character (van Herwijnen and Jamieson, 2007) were recorded. On 18 out of 22 slopes multiple failures occurred, i.e. the CT indicated several potential weak layers.

In the centre of the four pits with each a pair of CTs, a manual snow profile was observed complemented with a rutschblock (RB) test and another pair of CTs.

For assessing the variability within pairs (or pits) as well

Results

On each of the 22 slopes, we evaluated the within-pair variability (pit-scale) for the four locations, and the between-pairs variability (slope-scale) for the six combinations existing between the four pairs (Table 1).

Discussion

We investigated a relatively small area (∼ 150 m2) on a given slope. The reason for restricting ourselves to a small area is — apart from the advantage that sampling on the same slope is possible several times during the winter — based on the fact that most spatial variability studies that used geostatistics for analysis found correlation lengths ≤ 10 m (e.g. Kronholm, 2004). We can therefore assume that beyond a distance of about 10 m measurements will be un-correlated. In other words, sampling 10 

Conclusions

We explored the value of a simple sampling scheme that attempts to capture the small (∼ 1 m) as well as the large (slope-) scale variability (∼ 10–15 m). We analyzed the variations between two compression tests in a single pit and between two pits each with a pair of tests. Hence this is the first study — to our knowledge — that explicitly analyzes the value of a second pit on the same slope.

The present study has been performed on non-sheltered slopes above tree line. The area investigated on a

Acknowledgements

We would like to thank Alec van Herwijnen, Christoph Mitterer, Sina Schneider, Michael Schirmer and Charles Fierz for their help with the field work. The manuscript was much improved as a result of reviews by two anonymous referees who provided very helpful and extremely constructive comments. We gratefully acknowledge financial support by the European Commission (FP6-STRep-NEST, project no. 043386: TRIGS).

References (26)

  • A. van Herwijnen et al.

    Comparison of micro-structural snowpack parameters derived from penetration resistance measurements with fracture character observations from compression tests

    Cold Reg. Sci. Technol.

    (2009)
  • K. Winkler et al.

    Comparison of snow stability tests: extended column test, rutschblock test and compression test

    Cold Reg. Sci. Technol.

    (2009)
  • S. Bellaire et al.

    Deriving spatial stability variations from penetration resistance measurements

  • Cited by (4)

    • Measuring spatial variations of weak layer and slab properties with regard to snow slope stability

      2011, Cold Regions Science and Technology
      Citation Excerpt :

      We hypothesize that variations and their length scales are only relevant when the variations are around the threshold between rather unstable and rather stable conditions. Otherwise, it seems that the slope is either stable or unstable, independent of spatial variations, and point stability measurements might well be indicative — as exemplified by the average classification accuracy of stability tests (70–90%) (e.g. Schweizer and Jamieson, 2010; Schweizer and Bellaire, 2010). In other words, spatial variations might be most critical for slope stability evaluation when stability is “fair”.

    • Applied snow and avalanche research

      2010, Cold Regions Science and Technology
    • Comparison of avalanche survival patterns in Canada and Switzerland

      2011, CMAJ. Canadian Medical Association Journal
      Citation Excerpt :

      We therefore conclude that differences in snow climate accounted for the remaining differences in the first 35 minutes of burial. This conclusion is supported by the similarity of the early stages of the survival curves in the Swiss sample and the continental snow climate in the Canadian sample, since the snow climate of Switzerland has been described as transitional to partly continental.16–18 These results highlight the importance of prompt extrication by companions,19,20 especially in areas with a more maritime snow climate.

    • On snow stability interpretation of extended column test results

      2020, Natural Hazards and Earth System Sciences
    View full text