Processing time-series point clouds to reveal strain conditions of the Helheim Glacier terminus and its adjacent mélange

Wednesday, 17 December 2014
Logan C Byers1, Leigh A Stearns1, David C Finnegan2, Adam L LeWinter2, Peter J Gadomski2 and Gordon S Hamilton3, (1)University of Kansas, Department of Geology, Lawrence, KS, United States, (2)U.S. Army Cold Regions Research and Engineering Laboratory, Hanover, NH, United States, (3)University of Maine, Climate Change Institute, Orono, ME, United States
Flow near the termini of tidewater glaciers varies over short time-scales due to mechanisms that are poorly understood. Repeat observations with high temporal and spatial resolution, recorded around the terminus, are required to better understand the processes that control flow variability. Progress in light detection and ranging (LiDAR) technology permit such observations of the near-terminus and the pro-glacial ice mélange, though standard workflows for quantifying deformation from point clouds currently do not exist. Here, we test and develop methods for processing displacements from LiDAR data of complexly deforming bodies. We use data collected at 30-minute intervals over three-days in August 2013 at Helheim Glacier, Greenland by a long-range (6-10 km), 1064 nm wavelength Terrestrial LiDAR Scanner (TLS). The total area of coverage was ~25 km2.

Distributed shear in glaciers prevents a simple transformation for aligning repeat point clouds, but within small regions (~100 m2) strain is assumed to be minor between scans. Registering a large number of these individual regions, subset from the full point-cloud, results in reduced alignment errors. By subsetting in a regular grid, rasters of velocities between scans are created. However, using data-dependent properties such as point density causes the generation of unevenly spaced velocity estimations, which can locally improve resolution or decrease registration errors. The choice of subsets therefore controls the output product's resolution and accuracy. We test how the spatial segmentation scheme affects the displacement results and alignment errors, finding that displacements can be quantified with limited assumption of the true value of displacement for the subset, barring great morphological changes. By identifying areas that do not deform over the temporal domain of the dataset, and using these as the subsets to align, it should be possible to deduce which structures are accommodating strain. This allows for a characterization of the deformation style at the surface that is not attainable with other existing technologies.