Detecting snow depth changes in avalanche path starting zones using uninhabited aerial systems and structure from motion photogrammetry

Peitzsch, E., Fagre, D., Hendrikx, J., & Birkeland, K. (2018). Detecting snow depth changes in avalanche path starting zones using uninhabited aerial systems and structure from motion photogrammetry. In Proceedings of the International Snow Science Workshop. https://arc.lib.montana.edu/snow-science/objects/ISSW2018_P04.16.pdf

This paper is cited by the USGS's Snow and Avalanche Project as providing an excellent description of how structure-from-motion photogrammetry and uninhabited aerial systems are being used for high resolution mapping of snow depth in complex terrain. Snow depth can be highly variable in the alpine and is an important factor in avalanche forecasting and mitigation. However, this input data is missing from many current GIS Avalanche models because it is difficult, dangerous, and expensive to collect enough accurate snow depth data. Most current methodology relies on in-situ snow depth measurements or point measurements from automated weather stations which estimate snow depths for larger areas using interpolation algorithms. The authors note that even dense weather station networks are unable to accurately gather snow depth because of the complex variability inherent in alpine terrain.

Structure from Motion is a remote sensing photogrammetry technique that uses a series of overlapped images collected from a wide array of positions to generate high-resolution (4.97cm/pix in this instance) topographic reconstruction of a target area that can assess changes in snow depth. It calculates snowpack depth by subtracting bare-Earth elevations from subsequent snow covered elevation scans.

This study evaluated the effectiveness of SfM methodology on a 0.1km² slope in northwest Montana by comparing data produced by SfM with in-situ measurements. They found the error between the SfM digital surface models and in-situ measurements was 14-27cm, which they attributed to sampling difficulties created from GCPs. With the ability to capture accumulation differences greater than 10cm between days, as well as the spatial variability of those increases, this research shows that SfM is an effective, low-cost tool for avalanche forecasting and mitigation efforts. This is especially important because, as many studies I read have indicated, snowfall accumulation data from weather telemetry is not a good predictor of snowpack depth in alpine terrain where wind transport is the dominant factor in accumulation distributions.

Prior to reading this article, I was only familiar with the use of structure from motion photogrammetry related to the mapping of landscapes and structures. I did not realize the technology is accurate enough to be used for measuring snow depth changes. This seems revolutionary for the field of snow and avalanche science as it removes the need for people to be on potentially dangerous slopes to take measurements. Also, because of its ability to capture fine scale differences in elevation, it is an excellent technique for mapping snow depth which is highly variable in alpine terrain.