HIGH RESOLUTION (SPATIAL AND TEMPORAL) TERRESTRIAL LASER SCANNING IN AVALANCHE RELEASE AREAS

Benjamin Meier1*, Jeffrey S. Deems2 and Walter Steinkogler 1

1 Wyssen Avalanche Control, Reichenbach, Switzerland 2 National Snow and Ice Data Center, Boulder, Colorado, USA

ABSTRACT: Previous efforts measuring spatial snow height distributions with terrestrial laser scanning have been conducted from a distance and at irregular intervals. We present results from a laser scanner directly mounted on the deployment box of a remote avalanche control system (RACS) in an avalanche release area. This in-situ installation allows retrieval of relative snow height at a large number of points in the avalanche release area. The scanner maps the spatial snow distribution in a ≤40 m radius around the tower, and continuously measures the snow cover evolution at different points with centimeter-scale accuracy. The laser was operated for three months in Winter 2017/18 with one scan every three or six hours. We used the laser scanner setup in two modes: a) forecasting mode, monitoring changes in snow height, and b) avalanche detection mode, during avalanche control operations. Unfortunately, no avalanche was artificially released during the measurement period. Nevertheless, the system allowed to verify these “negative” avalanche control results. Our presentation will also describe other challenges such as signal dilution due to ambient light, range limitations during snowfall events, and angular imprecision in the mechanical pan/tilt mechanism. Finally, we present our work to-date in an operational context and illustrate how a combined 3D laser scanner and remote avalanche control system can lead to increased efficiencies and reduce costs. A setup as such can provide 1) additional information for decision making directly from the release area, e.g. absolute and relative snow height, 2) a measure to control the hazard, i.e. using the remote avalanche control system, and 3) verification of control mission results, i.e. an avalanche was released or not plus information about the fracture height.

KEYWORDS: Laser scanning, spatial and temporal snow depth mapping

1. INTRODUCTION

Avalanche control personnel often face the challenge that only limited information is available directly from the release areas. In most cases information, such as automatic weather stations, needs to be extrapolated to the local scale. This makes the decision-making process for measures, such as closures or avalanche control, a difficult task. Snow height information is an especially critical variable in operational avalanche forecasting on the local, i.e. path, scale.

Recent work has demonstrated the use of terrestrial laser scanning (TLS) for mapping snow depth distributions in avalanche starting zones to support control operations, infrastructure planning, and post-control assessment (e.g. Deems et al., 2015; Prokop et al., 2016; Hancock et al., 2018). The very high spatial resolutions achievable with TLS systems allows an assessment of snow depth, snow depth change, and avalanche release properties on a scale that captures variations in snow accumulation critical to avalanche dynamics.

While mapping starting zone snow depths with TLS from a remote location provides safety to the observers, a number of challenges are induced by weather and logistics. During times of obscured visibility (e.g. snowfall or low clouds) the laser view of the starting zone is commonly obstructed. Even in good conditions, it is common for release areas to lack well-defined, snow-free targets that can be used for scan registration, adding time, complexity, and uncertainty to the data post-processing. The ability to collect these data in closer proximity to the starting zone and from a fixed location would improve performance in poor-visibility conditions and minimize scan-to-scan registration challenges.

With this motivation we have installed a TLS system directly on the deployment box of a remote avalanche control system (RACS) in an avalanche release area. Providing avalanche control personnel with information directly from the release area (Figure 1) allows reduced uncertainty in their decision making.

Proceedings, International Snow Science Workshop, Innsbruck, Austria, 2018

388

Figure 1: Illustration of a snow depth distribution measured by a laser scanner mounted on a Wyssen avalanche tower in the release area.

2. METHODS

We used a laser scanning which is mounted at the bottom of a deployment box of the RACS. It is protected from the explosion by a steel housing which can be opened and closed mechanically. The laser is attached to the RACS power supply and automatically takes measurements at a wavelength of 850 nm in 38 lateral layers with a tilting of 2.5 degrees (Figure 2). Each individual layer scan provided data with an angular resolution of 0.25 degrees resulting in a maximum of 720 surface point measurements. This results in a maximum of 27360 points of measurements per complete scan. During the measurements only a certain subset of these points yielded good results. The laser was operated for three months in Winter 2017/18 with one scan every three or six hours. The laser sends out a short light pulse and then measures the echo by counting photons at the given wavelength. Multiplying the time delay between sending and receiving with the speed of light results in the distance between sensor and object.

Figure 2: Example data set for laterally scanned layers. Units in metres from laser unit position..

3. RESULTS and DISCUSSION

The measured lateral scans were then interpolated and resulted in spatial maps of absolute snow depth (Figure 3) or snow height differences between two scans, such as new snow height for the last 24 hours. For operational use the resulting measurements are directly incorporated and visualized in the Wyssen Avalanche Control Center WAC.3

The laser scanner was operated in two modes: a) forecasting mode, monitoring changes in snow depth, and b) avalanche detection mode, during avalanche control operations. Unfortunately, no avalanche was artificially released during the measurement period. Nevertheless, the system allowed to verify these “negative” avalanche control results.

Figure 3: Interpolated values for one full scan.

The plot about daylight dependency is the reach of the sensor (furthest point measured) for each complete scan vs. hour of the day the scan was taken. The sensor unit returns a distance for each point of measurement. Or it returns "nothing detected" which basically means the echo was not clear enough. Since our measured "object" (the snow) doesn't really move that much it can be assumed that the maximum measurement should be about the same in each subsequent measurement. Yet, it has been observed that the ability of the sensor to measure at larger distances showed a variation during the day. Figure 4 shows the day daily variation at 3 hour measurement intervals of the maximum reach, i.e. the furthest points measured, for each complete scan. A lower distance can be observed around midday due to solar input reducing the signal-to-noise ratio and thereby reducing detection range. .

Proceedings, International Snow Science Workshop, Innsbruck, Austria, 2018

389

Figure 4: Box plot of distance measurements for three hour interval scans on a day with no clouds and direct sunlight during the day.

To prevent this effect two things can be done: Send out a stronger impulse. This would lead to a stronger echo and therefore a larger signal. This of course enhances the SNR by simply raising the signal power. Another option is to reduce the noise. The noise in our case is ambient light. The amount of light which can reach the sensor can be limited by mechanical blinders which limit the visibility field of the sensor to the field where laser pulses are sent to and the echoes are expected from. Use of a laser system that operates at a longer wavelength would also serve to reduce noise as solar input decreases with increasing wavelength.

Another measurement difficulty we encountered was due to the angle dependency due to the position of the mechanical opening mechanism. Therefore, we know the accuracy of the measured distances but we are uncertain about the angle angle accuracy of the measurement. However, the angular error is the same for one line of measurement. Therefore, it would technically be possible to eliminate this error mathematically but only to a certain degree. Since the source of the error is clearly the hardware and we are under full control of the hardware the aim is to eliminate the mechanical backlash with enhanced hardware.

4. CONCLUSIONS

For the first time it is possible to receive spatial and temporal information with high resolution directly from the release area. This provides essential information for the avalanche control teams and allows for decision making with less uncertainty, i.e. to define the hazard more accurate.

The laser, in its first operational version, has shown to work reliable and with sufficient accuracy for operational purposes. Yet, some system improvements will be necessary to improve the setup such as the presented diurnal variations in maximum measurement distance or angle accuracies due to the mechanical setup.

5. OUTLOOK

The measured data can serve as detailed input for engineering applications (Deems et al. 2016) and manual (Prokop et al., 2013) or automatic avalanche dynamics calculations. Vera et al. (2016) coupled a snow cover model with an avalanche dynamics model and ran this chain in an operational setting. They showed that the input data was one of the main sources of uncertainty in their calculations and the laser measurements could help to reduce the accuracy for snow cover modeling, e.g. depth or release and associated layers, as well as dynamics calculations, e.g. release area width and depth.

REFERENCES Deems, J.S., Painter, T.H., 2006. Lidar measurement of snow

depth: accuracy and error sources, in: Proceedings of the International Snow Science Workshop, Telluride, CO. pp. 1–6.

Deems, J.S., Gadomski, P.J., Vellone, D., Evanczyk, R., LeWinter, A.L., Birkeland, K.W., Finnegan, D.C., 2015. Mapping starting zone snow depth with a ground-based lidar to assist avalanche control and forecasting. Cold Reg. Sci. Technol. 120, 197–204. doi:10.1016/j.coldregions.2015.09.002

Deems, J.S., R. Evanczyk, D. Vellone, E. Greene, T. Weldon, D.C. Finnegan, P.J. Gadomski, A.LeWinter, 2016. SUPPORTING, EVALUATING, AND PLANNING AVALANCHE CONTROL EFFORTS WITH LIDAR-DERIVED SNOW DEPTH MAPS, in: Proceedings of the International Snow Science Workshop. Breckenridge, CO, pp. 205–212.

Hancock, H., Prokop, A., Eckerstorfer, M., Hendrikx, J., 2018. Combining high spatial resolution snow mapping and meteorological analyses to improve forecasting of destructive avalanches in Longyearbyen, Svalbard. Cold Reg. Sci. Technol. 154, 120–132. doi:10.1016/j.coldregions.2018.05.011

Proceedings, International Snow Science Workshop, Innsbruck, Austria, 2018

390

Prokop, A., Schön, P., Singer, F., Pulfer, G., Thibert, E., 2013. Determining Avalanche Modelling Input Parameters using Terrestrial Laser Scanning Technology, in: Proceedings International Snow Science Workshop, Grenoble – Chamonix Mont-Blanc, France. pp. 770–774.

Prokop, A., Procter, E.S., 2016. A new methodology for planning snow drift fences in alpine terrain. Cold Reg. Sci. Technol. 132, 33–43. doi:10.1016/j.coldregions.2016.09.010

Vera, C., N. Wever, Y. Bühler, L. Stoffel, S. Margreth, and P. Bartelt (2016), Modelling wet snow avalanche runout to assess road safety at a high-altitude mine in the central Andes, Nat. Hazards Earth Syst. Sci., 16 (11), 2303–2323, doi:10.5194/nhess-16-2303-2016.

Vera, C., Wever, N., Langeland, S. Oyvind, L. 2018. Automatic Dynamic modeling - an Example of its application in an operational setting in Norway, in: Proceedings International Snow Science Workshop 2018, Innsbruck, Austria

Proceedings, International Snow Science Workshop, Innsbruck, Austria, 2018

391