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

Transcription

1 DRONE-BASED UWB RADAR TO MEASURE SNOW LAYERING IN AVALANCHE STARTING ZONES Rolf Ole Rydeng Jenssen *1,2, Markus Eckerstorfer 1, Hannah Vickers 1, Kjell-Arild Høgda 1, Eirik Malnes 1, Svein Ketil Jacobsen 2,1 1 Norut, Northern Research Institute, Tromsø, Norway 2 UiT The Arctic University of Norway, Department of Physics and Technology, Tromsø, Norway ABSTRACT: Slab avalanches release due to failure in a weak snow layer. Determining the spatial distribution and depth of weak layers in avalanche starting zones are high-risk tasks. Moreover, by manually digging snow pits, the occurrence of a weak layer can only be identified on a pit scale (meters). We therefore propose a technical solution to this problem by mounting an Ultra Wide Band (UWB) radar system onto a drone to obtain information about the occurrence, depth, and spatial distribution of weak layers over a larger area in order to improve safety for avalanche professionals. Here, we present the testing of an UWB radar system and show its capabilities of detecting snow stratigraphy. To simulate airborne operations, we have during the spring 2016 operated the radar system via a stationary rig 1 m above the snow, along 4.2 m long transects. For verification, we dug a full snow profile pit, identifying snow stratigraphy, liquid water content and snow density using traditional methodology as well as the Avatech SP2 and the Toikka SnowFork. Preliminary results were promising and showed the potential of an airborne UWB radar in detecting distinct snow layers. In the coming winter, the radar will be mounted on a drone to perform further airborne measurements. Future work will also include more comprehensive data analysis methods to improve snow layer identification and classification. KEYWORDS: UWB radar, UAV, Snow stratigraphy, Avalanche risk assessment, Avatech. 1. INTRODUCTION The release of a dry snow slab avalanche requires the formation of failure within a so-called weak layer buried below a cohesive, denser slab layer (Gaume et al., 2016). To determine the occurrence and spatial distribution of weak and slab layers, pit scale observations are upscaled using process thinking. This traditional field methodology is both time consuming, risky, and prone to observer bias. We therefore propose a noninvasive method of obtaining spatially upscaled information about snow stratigraphy in avalanche starting zones, by using an UWB radar mounted underneath an unmanned aerial vehicle (UAV) or drone. This method has to some extent been explored with semi-stationary, ground based rigs before (Gogineni et al., 2013; Kanagaratnam et al., 2007; Marshall et al., 2007). A similar approach has also been tested from airplanes over arctic sea ice (Kwok et al., 2011). Here we present the first results obtained using a stationary rig where the radar was mounted in order to * Corresponding author address: Rolf Ole Rydeng, Norut, Pb 6434, Tromsø Science Park, 9294 Tromsø; tel: ; rolfolejenssen@gmail.com simulate airborne data acquisition. 2. METHODS 2.1 Ultra Wideband Snow Sensor (UWiBaSS) The Ultra Wide Band Snow Sounder (UWiBaSS) is a UIT developed radar system (Fig. 1) specifically designed to detect snow stratigraphy. It consist of an Ilmsens m:explore 1 sensor connected to two Archimedean spiral antennas (one transmit (TX), one receive (RX)) and a single board computer running the Ilmsens developed software to acquire and log the impulse responses. The constructed antennas have opposite polarization for RX and TX; where RX is Right Hand Circularly Polarized (RHCP) and TX is Left Hand Circularly Polarized (LHCP) (Fig 2). These antennas have a measured frequency range of approximately 950 MHz - 11 GHz (ca. 10 GHz bandwidth) (Tbl 1). However, the full range of the antennas is not used since the radar sensor operates at GHz (5.9 GHz bandwidth) (Tbl 1). The antennas were placed in a housing with absorbing material in the backing cavity to remove the rear lobe of the antenna (Fig. 1). This causes a

2 reduction of 0.5 (50 %) in antenna efficiency (i.e.3 db gain loss), but the alternative is to use a reflective backing, which is a very difficult task to accomplish given the large bandwidth. With such a large bandwidth, a reflector needs to facilitate for all frequencies to avoid cancelling when returning from the reflector. There are several suitable antenna designs for snow measurement applications (Mosy, 2009), however the Archimedean spiral antenna have a impedance stability over a very large bandwidth that is hard to match with other antenna designs. HPBW Data acquisition rig Before the radar is mounted on a UAV, we simulated its intended elevated position above the snow surface with a dolly rig (Fig. 3). To allow the radar to move continuously and controlled in a single direction a self-made camera dolly (mostly used in film production to allow cameras to move smoothly) was used, imposed on a rig using aluminum ladders (Fig. 3). This proved to save both time and funds, and yielded a smooth, controlled, continuous data acquisition. Step ladders where used as feet for the rig, with some extra clamps to hold the suspended ladder between them. This assembly operated satisfactory and was easily moved to different locations if necessary. The total measurement length became effectively the length of the ladder chosen, which was approximately 4.3 m and provided a measurement length of 4.2 m due to the fastening at the end, and the length of the dolly. Fig 1: UWiBaSS system with antennas at the bottom. UWB sensor and control PC are mounted inside the box Fig 2: Archimedean spiral antennas with opposite polarization. Tbl 1: UWiBaSS characteristics, Sensor Value Bandwidth 5.9 GHz Range Resolution cm Sampling rate GHz Measurement rate 1 khz MLBS order 9 (511 points) Antenna Value Bandwidth 10 GHz Efficiency 0.5 Fig 3: Experimental setup: Dolly rig with mounted radar systems and PC to log data. In Fig. 3 there are two mounted radar systems visible. A Novelda radar 2 and the UWiBaSS radar, which were tested simultaneously for comparison. The data from the Novelda radar will not be presented in this paper. Some experimentation with different speeds while scanning was conducted to give some estimation of the maximal speed one could use while still acquiring a satisfactory image of the snowpack. The maximum speed will depend on the desired longitudinal resolution of the scan, but in our case where centimeter resolution was desired the

3 maximum speed was found to be about 0.02 m/s. There are many possibilities to how the scanning might be performed from a UAV. For instance, the UAV might be stationary during scanning and move to selected locations in the avalanche starting zone to perform data acquisition. 2.3 Data processing Dry snowpacks have relatively low electromagnetic attenuation and dielectric constant. This allows the signal to penetrate with ease and reach the bottom of the snowpack, which is desirable. However, this also implies that the reflected signal from each layer will be very weak. Presently the processing of data is very time consuming as even the smaller datasets contained around 10 data points. We firstly loaded the radar data and subtracted the reference signal, before we modeled phase response and compensated for non-linear phase. Then we reformatted the data into an image and performed image processing methods. These included histogram equalization, wiener filtering and sobel mask filtering. We then rectified the pulse and adjusted for propagation speed through the snow. 2.4 Snow pit observations The snow profile measurements consisted of identifying snow stratigraphy and taking a hardness profile with the Avatech SP2 snow probe, as well as measuring the density and liquid water content with the Toikka Snow fork. The density measurements with the Snow fork were also correlated with weight measurements. During both campaigns in February and March 2016, snow conditions were dry. 3. RESULTS & DISCUSSION 3.1 Comparison between snow pit observations and radar measurements In the following, we present preliminary results that will give a first impression on how capable the system is in resolving snow stratigraphy. In Fig 4 we show the results obtained during the first field campaign on 25 February 2016, compared to a stratigraphy profile obtained with the AvaTech SP2. Due to recent accumulation of low-density snow on the surface, we placed a Perfect Electrical Conductor (PEC) plate at the snow surface to mark the snow surface. This is seen as a large red area on the right side. Low-density snow is difficult to detect with this radar due to the low relative permittivity of snow, which decreases significantly with lower densities. The liquid water content of snow also greatly affects the relative permittivity (Tiuri et al., 1984). In our test snow profiles, the snow layers had no significant amount of liquid water. The density of the snow ranged from about 0.1 g/cm 3 to 0.35 g/cm 3, which was the largest contributor to horizontal reflections in the radar image. In general, however, these first results are very satisfying as several thin layers of both low and high strength and density were resolved in the radar image. In Fig 5, we present a comparison between an obtained radar image on 11 March 2016 and a density profile obtained by the Toikka Snow Fork. Note the response from the aluminum ladder in the upper part of the radar images. This is due to the wide radiation pattern of the Archimedean spiral antennas used. This needs to be taken in to consideration when mounting on a UAV. Again, correlation between measured snow stratigraphy and radar response is visible. Clearly, with increasing density, the radar reflectance increased as well. 3.2 Future work The Ilmsens system will as a next step be mounted underneath an octocopter drone and flown over snow. Based on the experiences from the experimental setup we anticipate some challenges with imaging snow depth and stratigraphy from a UAV. Especially the distance to the target will have to increase from 1 m to 3-10 m.10m for safe flying. According to the radar equation the need for more radiated power increases with the fourth power for increasing distance to the target (Taylor, 2012). A possible improvement is to mount a more directional antenna in the UWiBaSS system (e.g. UWB Vivaldi antennas). This will limit clutter from the frame of the UAV and increase the power directed down to the snowpack. Future work will also include finding better methods for extracting the interesting part of the data. Some focus is directed at the Hough transform as it should provide a good measure for horizontal lines. This will hopefully create improved result in terms of resolving the snowpack stratigraphy. Additionally, more work should be done regarding the in situ measurements. The work done so far shows acceptable correlation with the radar measurements, but several more factors need to be examined e.g. grain size, snow temperature and detailed density for each layer. 575

4 Fig 4: Comparison between AvaTech SP2 hardness profile, and radar image from Fig 5: Comparison between AvaTech SP2 hardness profile, density profile derived from the Toikka Snow Fork and radar image from CONCLUSION In this study we have investigated the possibility of detecting snow stratigraphy with a radar operating at some distance above the snow surface. In preliminary tests using a dolly rig, our developed UWiBaSS system seems to perform well. Prominent snow layers are visible in the radar images and with more sophisticated processing methods, it should become possible to distinguish snow layers for inexperienced users as well. In the coming months, we will work on getting the radar airborne with a drone, in order to be able to fly it to avalanche starting zones, where it should seek for weak layer slab layer combinations. 576

5 5. REFERENCES Gaume, J., van Herwijnen, A., Chambon, G. and Schweizer, J., Snow fracture in relation to slab avalanche release: critical state for the onset of crack propagation. The Cryosphere.doi: /tc , 2016 Gogineni, S., Yan, J.B., Gomez, D., Rodriguez-Morales, F., Paden, J. and Leuschen, C., Ultra-wideband radars for remote sensing of snow and ice, IEEE MTT-S International Microwave and RF Conference, pp Kanagaratnam, P., Markus, T., Lytle, V., Heavey, B., Jansen, P., Prescott, G. and Gogineni, S.P., Ultrawideband Radar Measurements of Thickness of Snow Over Sea Ice. IEEE Transactions on Geoscience and Remote Sensing, 45(9): /TGRS Kwok, R., Panzer, B., Leuschen, C., Pang, S., Markus, T., Holt, B. and Gogineni, S., Airborne surveys of snow depth over Arctic sea ice. Journal of Geophysical Research: Oceans, 116(C11): n/a-n/a /2011jc Marshall, H.-P., Schneebeli, M. and Koh, G., Snow stratigraphy measurements with high-frequency FMCW radar: Comparison with snow micro-penetrometer. Cold Regions Science and Technology, 47(1-2): doi: doi: /j.coldregions Mosy, J.S., Ultra wideband radar antenna design for snow measurement applications, Montana State University. Taylor, J.D., Ultrawideband radar. Applications and design, Bosa Roca. Tiuri, M., Sihvola, A., Nyfors, E. and Hallikaiken, M., The complex dielectric constant of snow at microwave frequencies. IEEE Journal of Oceanic Engineering, 9(5): doi: /JOE