4D Analysis of Slope Monitoring Data from Terrestrial Laser Scanning

Thursday, 18 December 2014
Jack Williams1, Nick J Rosser2, Richard J Hardy2 and Ashraf Afana2, (1)University of Durham, Durham, DH1, United Kingdom, (2)University of Durham, Durham, United Kingdom
Analysis of deformation from actively failing slopes is essential for gaining insight into the rates, mechanisms and controls on failure. Recent models have focussed upon the temporal evolution of failures, the validation of which requires increasingly high-resolution, high-frequency monitoring data. Since its introduction to geomorphological study, Terrestrial Laser Scanning (TLS) has become a frequently used means of characterising change to failing slopes. The most computationally efficient approach represents change on a pixel-by-pixel basis using rasterised 2.5D DEMs of Difference; however, the level of detail reduces on steep surfaces and the use of a fixed grid spacing limits the ability to resolve fine-scaled features, both of which may underpin failure mechanisms. A number of algorithms and software packages have been developed to better characterise surface and joint structures using ‘true 3D’ point clouds; however, 3D change detection with a large number of scans remains limited. In addition to developments in geometric change detection, TLS systems now provide radiometric information by digitising the energy-time structure of the reflected laser pulse, sensitive to surface moisture amongst other variables.

This study draws upon a unique dataset of > 800 sequential scans captured across a failing rock slope. Our algorithm extracts change between a large number of scans, using a Moving Least Squares adjustment to filter data through time and space. The analysis explores optimal kernel structures for retaining spatial resolution and temporal responsiveness to articulate the nature of change in rock slopes, distinguishing discrete failures (e.g. rockfalls) from ongoing deformation (e.g. creep). The code segments successive clouds into an octree structure of planar surfaces and provides 3D change metrics through time. We use the code to test the ability to separate movement at various scales, with the aim of capturing movements suited for failure-time prediction. In addition, rasterised amplitude data from a coastal cliff is presented as a means of identifying regions of ground water seepage and tide inundation, both of which may predispose a slope to failure. The approach holds implications for analysing 3D data in order to extract geomorphologically intuitive deformation measurements.