Comparing Two Methods of Surface Change Detection on an Evolving Thermokarst Using High-Temporal-Frequency Terrestrial Laser Scanning, Selawik River, Alaska

• Published in: Remote Sensing
• Main Authors: Theodore B. Barnhart, Benjamin T. Crosby
• Format: Article in Journal/Newspaper
• Language: English
• Published: MDPI AG 2013
• Subjects: terrestrial laser scanning; topographic change detection; error analysis; thermokarst; Arctic; permafrost; Science; Q; Thermokarst; Alaska
• Online Access: https://doi.org/10.3390/rs5062813; https://doaj.org/article/b39690f119a14b96bbfa6fda2a6831f6

Terrestrial laser scanners (TLS) allow large and complex landforms to be rapidly surveyed at previously unattainable point densities. Many change detection methods have been employed to make use of these rich data sets, including cloud to mesh (C2M) comparisons and Multiscale Model to Model Cloud Comparison (M3C2). Rather than use simulated point cloud data, we utilized a 58 scan TLS survey data set of the Selawik retrogressive thaw slump (RTS) to compare C2M and M3C2. The Selawik RTS is a rapidly evolving permafrost degradation feature in northwest Alaska that presents challenging survey conditions and a unique opportunity to compare change detection methods in a difficult surveying environment. Additionally, this study considers several error analysis techniques, investigates the spatial variability of topographic change across the feature and explores visualization techniques that enable the analysis of this spatiotemporal data set. C2M reports a higher magnitude of topographic change over short periods of time (~12 h) and reports a lower magnitude of topographic change over long periods of time (~four weeks) when compared to M3C2. We found that M3C2 provides a better accounting of the sources of uncertainty in TLS change detection than C2M, because it considers the uncertainty due to surface roughness and scan registration. We also found that localized areas of the RTS do not always approximate the overall retreat of the feature and show considerable spatial variability during inclement weather; however, when averaged together, the spatial subsets approximate the retreat of the entire feature. New data visualization techniques are explored to leverage temporally and spatially continuous data sets. Spatially binning the data into vertical strips along the headwall reduced the spatial complexity of the data and revealed spatiotemporal patterns of change.