Proceedings, International Snow Science Workshop, Breckenridge, Colorado, 2016

Transcription

1 Proceedings, International Snow Science Workshop, Breckenridge, Colorado, 1 MONITORING THE STAVBREKKA GLIDE AVALANCHE INTERNATIONAL SNOW SCIENCE WORKSHOP 1 IN BRECKENRIDGE, CO Tore Humstad 1 *, Martin Venås 1, Halgeir Dahle 1, Knut I. Orset 1, Ingrid Skrede 1 Norwegian Public Roads Administration, Oslo/Tromsø/Molde, Norway Norwegian Water Resources and Energy Directorate, Stranda, Norway ABSTRACT: The yearly Stavbrekka glide avalanche is problematic for the summer opening of the tourist road between Breidalen and Geiranger in Norway. Several efforts have been made to protect road users from the avalanche, some of which to expedite the avalanche in the spring prior to the seasonal opening. Examples of these are adding slag to the snow surface, water bombing in the glide crack and blasting of the stauchwall. In addition, test of snow anchors in the glide zone, and evaluation of moving the road into a tunnel are performed. None of these efforts proved to find sustainable and cost beneficial solutions. Therefore, different monitoring systems of the glide have been tested the three last winter seasons with the aim to get a better understanding of snow glide, and to make a better prediction of this specific avalanche. The tested methods are; glide shoes on the snow base, photo tracking of the glide crack, time lapse and InSAR studies of the snow slab together with snowpack temperature monitoring. The results shows that glide fracturing initiates after a few days of isothermal snowpack, and the snow forms into two gliding lobes. After 3 to weeks of gliding, the avalanches usually take place. The glide velocity prior to avalanche depends on the temperatures in snow and air. Three hours before the avalanche, it is measured to 3 cm/hour. The tests revealed challenges to the equipment from severe weather conditions and from moving slab forces. The most useful method for real-time monitoring, seems to be the InSAR, however glide shoe measurements and photo analysis is useful for validation and supplementary analysis. KEYWORDS: wet snow, glide avalanche, monitoring, instrumentation 1. INTRODUCTION The national tourist road Fv3 between Breidalen and Geiranger in Norway, which reaches up to 13 m a.s.l., is closed during winter due heavy snow fall, drifting snow and avalanches. Even in the spring, the Norwegian Public Roads Administration (NPRA) is challenged by some avalanche paths that are yet not triggered. One of these is the Stavbrekka avalanche in Breidalen. This is a wet full depth avalanche and can be classified as a glide avalanche according to recent publications (e.g. Mitterer and Schweizer, 1, Höller, 13). The release area consists of a 15 m wide, m long, -35 steep and fairly smooth and planar rock surface. In the spring, usually in mid March, a glide crack appears at the upper boundary of the slab, at 13 m a.s.l. This is the first visual prove that the glide has begun. If the conditions are favorable, the slab continues to glide on the rock surface and forms a compression zone further downslope. Because of an obstacle in the middle of the rock plane, the slab splits into two lobes as seen in Fig. 1. * Corresponding author address: Tore Humstad, Norwegian Public Roads Administration, Fylkeshuset, Molde; tel: ; tore.humstad@vegvesen.no Fig. 1: Gliding slab in the release area a few days before the avalanche. The western lobe is to the left, the eastern lobe to the right (photo: H. Dahle). 51

2 Proceedings, International Snow Science Workshop, Breckenridge, Colorado, 1 In order to open the road as early as possibly in the spring, a few efforts have been tested over the years to speed up the gliding process. Slag was once added to the surface to enhance the snow melt, and in 5 explosives and water bombing was used to release the avalanche artificially, with no results (Hustad, 5). Two times, in 5 and, snow anchors were put up in the release area in order to prevent snow gliding and to obtain controlled melting and thus avoiding the slab to develop to an avalanche. The forces from the snow proved, however, to be too large, especially the lateral ones, and the anchors were removed permanently 1 (Norem, ; Nilssen, 13). Lately, in lack of any satisfactory method to protect the road from the avalanche, the focus has changed to find the best way to predict the event. Glide avalanches seem less influenced by recent snow fall, wind and air temperatures compared to other avalanches. Thus, well established indices to forecast these are not applicable (e.g. Jones, ). On the other hand, many glide avalanches usually happen one time per year, as the entire snow pack disappears in the event. Therefore, the cost-benefit potential of introducing precise alert criteria is large compared to expensive protection structures. A work package in the Norwegian interdepartmental research program on natural hazards, called NIFS, was focusing on introducing methods to monitor and forecast avalanche danger. Through this program, the authors of this paper worked on the design for monitoring the Stavbrekka avalanches, and other similar glide avalanches. The approach was to use Stavbrekka as a test site for several methods with the following objectives: Better understanding of glide avalanches Use of instruments in mountainous climate Suggest a future monitor design for this specific avalanche The following methods were tested and described further in section through : Review of historical events Glide shoe at the snow base Photo analysis of the glide Ground based InSAR radar Snow temperature monitoring A situation map of the avalanche release area and positions for some applied instruments, is shown in Fig.. Fig. : Situation map of the road, avalanche release area and the location of the applied equipment 515

3 Proceedings, International Snow Science Workshop, Breckenridge, Colorado, 1. REWIEW OF HISTORICAL EVENTS In order to make an appropriate test design, it was crucial to gather information from the typical avalanches in Stavbrekka. Through interviews, old newspapers and own experiences, Hustad (5) collected a number of dates and statistics for the summer opening from the 195s and up to. Although weak, these statistics give at least an indication of the date for the avalanches, as they usually happened some days or weeks before the road opening, rather than after. The authors of this paper has added new statistics through own work from to present time. From the latest ten years, we know more about the exact dates of the avalanche. Also, since 13, the date for the first glide crack observation is logged. Some of these statistics are visualized in Fig. 3. Mar Apr May Jun Glide crack Avalanche, east slab Avalance, west slab Avalanche, both slabs Road opening Road opening (no avalanche) Jul Fig. 3: Statistics from Hustad (5) and the authors of this paper on glide avalanche formation and date for road opening The data show that the road openings in 1997, 5 and where much later than usual as no avalanche struck, and NPRA had to wait for the snow to melt sufficiently. In 11, the road was in fact opened ten days prior to the avalanche and a non-predicted avalanche led to a new closure. The contractor usually needs -3 working days to prepare the road for opening after the avalanche events. In terms of the avalanche formation from the first sign of glide, through build-up of a compression zone and the following acceleration phase, the time series are too short to make an adequate statistical analysis. However, observations clearly indicates that east and west lobe turns on dropping first, and it could be up to two weeks between them. The data also show that the time it takes for the slab to develop into an avalanche, from the first observed glide crack, ranges from 3 to weeks. Hence, the Stavbrekka avalanche seems to develop slower than the majority in previous investigations, where release normally took place within 3 days after the glide crack formation (Feick et al., 1). An explanation could be that other glide avalanches occurs in steeper terrain and after more sudden wetting. The relative slow glide in Stavbrekka is crucial when choosing the monitoring design, logging interval and the strategy for alerting relevant personnel. Some average dates for Stavbrekka are presented in Tbl. 1. Tbl. 1: Average dates for repeated events Event Average date Number of registrations Glide crack observation Mar 3 Avalanche, west slab May Avalanche, east slab May 7 Road opening May 7 3. GLIDE SHOE 3.1 Background and field set-up A "glide shoe", as a method for measuring snow glide at the base of the snow pack, was first introduced by in der Gand and Zupančič (19) and later modified by Clarke and McClung (1999) and Ceaglio et al. (1). For Stavbrekka, the chosen design consisted of a steel skid with two transversal plates to ensure a good snow slab anchorage. A draw wire connected the glide shoe to a coil with a potentiometer in a steel box. The potentiometer monitors the increased voltage as the electrical resistance increases when the slab pulls the wire out from the coil (see Fig. ). The original field set-up consisted of three glide shoes spread out on the glide base. From the potentiometer, data cables transferred the signals to a data logger and GSM transmitter situated some 75 m away from the sensors, outside the glide zone. These were powered from a solar panel. This set-up proved to be vulnerable to the local environment. In 13, the data cables where broken by the stresses from the gliding snow. Therefore, before 1/15, the cables were enhanced to withstand these forces, and the most exposed stretches were put in traces that were cut down in the rock surface. However, the solar panel blew down during a hurricane, and a voltage peak lead through the cables, is believed to have caused 51

4 Proceedings, International Snow Science Workshop, Breckenridge, Colorado, 1 damage to the sensors. These problems gave limited data form the first winter (13/1) and no data the second winter (1/15). glide deformation [m] Fig. : Glide shoe, draw wire and steel box containing a potentiometer glide velocity [cm/hour] 3 1 Spring 1 For the third winter (15/1), we gave up the attempts to have a real-time data transfer from the sensors. The GSM transmitter and the solar panel were removed, as well as one of the glide shoes. Data loggers were moved into each of the two remaining steel boxes together with the potentiometers powered on internal batteries. Now, the test worked more as an analytic and scientific verification of other methods (see section -), rather than an independent operative monitoring system. The locations of the glide shoes for the third season (15/1) are shown in Fig.. 3. Results The first test season, only one of the glide shoes gave data. The glide began on April 7, 1, but unfortunately the cables were broken after three days and 1. m deformation. At that time, the glide velocity was 9 cm/hour (Venås, 15). From the third season in 1, data from glide shoe 1 are presented in Fig. 5 and Fig.. The first glide shoe displacement is detected on March 5, that is 11 days after the first observation of the glide crack. This delay is probably due to the distance of nearly 1 m from the glide shoe up to the glide crack where the deformation started (Fig. ). The glide accelerates until April, then retards until April 7 (Fig. 5). The slab lies still until the final acceleration starts at May 5 and the avalanche occur on May 7 (Fig. ). Fig. 5: Total glide displacement (upper) and glide velocity (lower) measured for glide shoe 1 glide velocity [cm/hour] May 5-7, 1 Fig. : Glide velocity the last hours before the Stavbrekka avalanche until the last reading on May 7, 1, UTC 9: Glide shoe data were logged once an hour, and the last registration is from May 7 at 9: (UTC), probably 5- minutes before the avalanche (section.). At this time, the total displacement was 1.75 m with 9.5 m draw wire left at the coil. The glide velocity was then 7.5 cm/hour. Glide shoe is suspected to be placed lower than the compression zone, i.e. outside the slab, and no displacement was recorded prior to the avalanche. 517

5 Proceedings, International Snow Science Workshop, Breckenridge, Colorado, 1. PHOTO ANALYSIS.1 Bakground Previous authors have described optical methods such as video tape recording of avalanche behavior (Akitaya, 19), glide crack detection from satellite images (Feick et al., 1) and local time lapse cameras (van Herwijnen and Simenhois, 1) as well as photo and GPS tracking of stakes (Hendrikx et al., 1) of slopes prone to snow gliding. In this project, attempts were made to document the glide behavior visually and to track pixels automatically. One web camera situated down at the road was used together with two trail cameras close to the gliding slab (Fig. ).. Recurring photos and time lapse The images from the three cameras were used to take a day-to-day check of status of the slab. By just comparing the present day s image with previous ones, one could easily get an impression whether there was gliding going on or not. To enhance the contrast in the image, seven wooden sticks were put down in the snow in the early spring. Images from the web camera were presented on a web page and downloaded from an ftp-server. One image from the trail cameras was sent automatically each morning, or could be ordered through a MMS service at any time. Images from each tenth minutes were saved locally on SD cards in the cameras. These images were later used to make time lapse videos and further analysis described in section.3 and...3 Dark Pixel Analysis A method to discriminate dark pixels from the glide crack and light pixels form the surrounding snow was introduced by van Herwijnen og Simenhois (1) and also used by Hendrikx et al. (1) and van Herwijnen et al. (13). The idea was to count the proportion of dark pixels of the total pixels, since the number of dark pixels was expected to increase as the glide crack developed. Venås (15) did this analysis for the images from 1/15. However, repetitive rounds of snow fall that superseded clear days, also in the melt season, gave challenges to discriminate the not-sodark pixels in the glide crack. Because of the amount of manual interpretation work, this method was skipped for the 15/1 season.. Pixel Tracking Wooden sticks that ware placed on the slab on March 1, 1 (section. and.3), could be tracked semi-automatically in different software. Here, the software ImageJ (Schneider et al., 1) with the extension MTrackJ (Meijering et al., 1) were used to identify the sticks positions. The positions were notified in terms of coordinates on the recurring images, as the image frame was somewhat fixed whereas the sticks were moving on it. Data from 15 are previously presented by Venås (15), and data from 1 are presented in Fig. 7 and Fig. below. glide shoe displacement [m] Fig. 7: Displacement of sticks within picture frame measured in pixels compared to displacement of glide shoe 1 (data from 1) glide shoe velocity [cm/hour] glide shoe stick 7 stick stick 5 glide shoe stick 7 stick stick 5 Spring 1 (UTC) 1 1 Fig. : Velocity of sticks within picture frame measured in pixels per hour compared to velocity of glide shoe 1 (data from 1) The results show a fairly good correspondence to the results from glide shoe 1, especially for the stagnation phase April -7 and the last acceleration towards May 7. However, since all sticks are situated higher than the glide shoe, the first acceleration seems to happen earlier for the sticks than stick displacement in photo [pixels] stick velocity in photo [pixels/hour] 51

6 Proceedings, International Snow Science Workshop, Breckenridge, Colorado, 1 for the glide shoe (Fig. 7). A disadvantage is that small changes in the camera position due to wind etc. give unrealistic notches on the graphs. Also, when the sticks moves away from the camera, the angle gets smaller and the ratio of the pixel size to the actual displacement gets bigger. It is therefore difficult to use this method as an absolute description of the glide, unless the same pattern is reproduced each year. This requires the exact same set-up of camera position, angle and focal length, as well as the same positions of the sticks from year to year. 5. GROUND BASED INSAR RADAR Ground based InSAR technology was tested both to evaluate the application of InSAR on gliding snow in general, and to document the glide prior to the avalanche in 15 and 1 (Skrede et al., 1). In 1, movement in radar pixels from coordinates corresponding to the start position of stick, and and glide shoe 1 and (Fig. ) were tracked from April 1. The total displacement at these coordinates is estimated from this date (Fig. 9). Displacement that took place before this date is not visible in the InSAR interpretation. displacement [m] InSAR at stick InSAR at stick InSAR at stick InSAR at glideshoe 1 InSAR at glide shoe glideshoe 1 Spring 1 Fig. 9: Displacement estimated from pixels in radar image, starting on April 1 The results show that the InSAR measurements at the surface above glide shoe 1 correspond very well with the glide shoe measurements at the base in the same position. It is also documented that the displacement higher up on the slab was greater (1-1 m for stick and for from April 1 to May 7 ) than for the lower part, where glide shoe 1 is situated. A fracture at the very upper part of the slab, lead to stagnation of the glide above this fracture on April 1 (stick ). The In- SAR data also indicates that glide shoe probably was situated below the compression zone, as discussed in section 3... SNOW PACK TEMPERATURES A sensor called SM consists of a series of digital thermistors mounted with a fixed distance of cm on a vertical pole. This sensor was used to log the temperatures in air and snow. Measurements from the thermistors were logged with 1 minutes interval, saved to an internal memory and GSM transferred to a central computer. The SM measures snow depth indirectly by identifying thermistors buried in the snow, since the insulating effect of the snow reduces the temperature fluctuations compared to the air temperatures (Ingólfsson et al., 1). The results from spring temperatures in 1, presented in Fig. 1, show that the transition from a cold to an isothermal snow pack takes place around March 1-1. This correspond well to the period when the glide crack appeared in the camera images. In this period, the nightly air temperatures also exceeded above C. Later, in April - 7, a cooling of the air and in the highest snow layers were observed during the stagnation phase of the slab, as described in section 3.. A new air temperature increase to above C, corresponds the final acceleration phase in May 5-7. Temperatures in snow and air [ C] T snow cm T snow cm T snow cm T snow 1 cm T snow 1 cm T snow cm Glideshoe 1 Fig. 1: Snow temperatures measured by thermistors - cm into the snow (from the ground) and air temperatures from 3 cm above the ground (data form 1) 1 1 Glideshoe displacement [m] T snow cm T snow cm T snow 1 cm T snow 1 cm T snow 1 cm T air, daily average 519

7 Proceedings, International Snow Science Workshop, Breckenridge, Colorado, 1 7. DISCUSSION AND CONCLUSOIN Through a study of historic events and recent measurements, updated knowledge is gained on the behavior of the Stavbrekka glide avalanche. This knowledge can be used both to design a monitoring system and to update general operation procedures for the road opening in the spring. From experience and data described in this paper, one could probably suggest some indicators and thresholds for whether the avalanche is expected to occur in a few weeks, in a few days, in a few hours or at any moment. Based on the specific results from Stavbrekka, the avalanche is expected to take place in a few weeks when the snow gets isothermal, the glide crack appears and the velocity starts to be detectable (> cm/day). There will probably be a few days left, when the air temperature exceeds C both day and night and when velocity exceeds 1 cm/h. It is likely to be a few hours left when the velocity exceeds 5 cm/h, and finally, the avalanche will probably occur at any moment when the velocity exceeds 15 cm/h. When choosing the monitor system for the future, the glide shoe seems to give the most precise data and requires the least manual interpretation. However, this method has proved to be difficult to use in a real-time warning system as long as reliable data transfer is not established. Hence, developing wireless data transfer through the snow pack would be an interesting assignment for the future. Ground based InSAR radar seems to be the method that gives the best real-time measurement. It is independent of visibility and daylight, and time series can be generated from any pixel on the entire glide zone. High costs and demand of some support from specialists are among the disadvantages. Timelapse series of the slab, pixel tracking and temperature measurements through the snow pack are all cost-efficient methods. They are useful both alone, when no other methods are affordable, but also as a supplement to validate and better understand data from other sources. Also, in a robust monitor system, it is always a strength to have access to more than one data source. ACKNOWLEDGEMENTS Thanks to Örn Ingólfsson at POLS Engineering, Iceland, Lars Krangnes and Atle Gerhardsen at Cautus Geo AS, Norway, for cooperation and assistance. Thanks also to our own organizations and to the Norwegian National Railway Administration for assistance and financing through the NIFS program ( REFERENCES Akitaya, E., 19. Observations of ground avalanches with a video tape recorder (VTR). Journal of Glaciology, (9): Ceaglio E., Freppaz M., Filippa G., Ferraris S., Zanini E., Segor V.: A characterization of snow gliding and potential predisposing factors in a full-depth slab avalanche release area (Valle D aosta, NW Italian Alps). Proceedings of International snow science workshop, Anchorage, Alaska, pp 51 5, 1. Clarke, J. & MCClung, D Full-depth avalanche occurrences caused by snow gliding, Coquihalla, British Columbia, Canada. Journal of Glaciology, 5, Hendrikx, J., Peitzsch, E. & Fagre, D. Time-lapse photography as an approach to understanding glide avalanche activity. Proceedings of the 1 International Snow Science Workshop, Anchorage, Alaska, USA,, 1. Hustad, A., 5. Rv 3 Geirangervegen, Vinterstengd veg - Vurdering av vegopning til fast dato. Report from Statens vegvesen, 1-. Höller, P. 13. Snow gliding and glide avalanches: a review. Natural hazards, 71, in der Gand, H.R. and Zupancic, M., 19. Snow gliding and avalanches, Symposium at Davos Scientific Aspects of Snow and Ice Avalanches, IAHS Publication, 9. Int. Assoc.Hydrol. Sci., Wallingford, U.K., pp. 3-. Ingólfsson, Ö., Grímsdóttir, H.:. & Jónsson, M. H. Monitoring Snowpack Temperature Gradient Using Automatic Snow Depth Sensor. Proceedings Anchorage 1 International Snow Science Workshop September, Meijering, E., Dzyubachyk, O. & Smal, I. 1. Methods for cell and particle tracking. Methods Enzymol, 5, 13-. Mitterer, C., Schweizer, J., 1. Glide Snow Avalanches Revisited. The avalanche journal, vol.11, p-71. Nilssen, K. 13. Notat Stavbrekka 11/1-sesongen. SIN- TEF Byggforsk report 5313, 1-1 Norem, H.. Rv 3 Geirangervegen. Evaluering av snøankrene i Stavbrekka.: SINTEF Byggforsk. Schneider, C. A., Rasband, W. S. & Eliceiri, K. W. 1. NIH Image to ImageJ: 5 years of image analysis. Nature methods, 9, Skrede, I., Kristensen, L., Rivolta, C.: Use of ground based radar (InSAR) to monitor glide avalanches. Proceedings of International Snow Science Workshop, Breckenridge, Co, U.S.A., October 3-7, pp. in press (see this issue), 1. van Herwijnen, A. & Simenhois, R. Monitoring glide avalanches using time-lapse photography.international Snow Science Workshop ISSW, 1. van Herwijnen, A., Berthod. N., Simenhois, R., Mitterer C.: Using time-lapse photography in avalanche research. Proceedings of International Snow Science Workshop Grenoble-Chamonix Mont-Blanc, 13. p Venås, M., 15: Overvåkning og varsling av glideskred ved Stavbrekka i Skjåk. Master thesis at Norwegian University of Science and Technology,