Development of a Novel, Parsimonious, Model-based Approach for Representing High-resolution Gravel Facies
Thursday, 18 December 2014
A precise, time-efficient, cost-effective method for quantifying riverbed roughness and sediment size distribution has hitherto eluded river scientists. Traditional techniques (e.g., Wolman counts) have high potential for error brought about by operator bias and subjectivity when presented with complex facies assemblages, poor spatial coverage, insufficient sample sizes, and misrepresentation of bedforms. The application of LiDAR facilitated accurate observation of micro-scale habitats, and has been successfully employed in quantifying sediment grain size at the local level. However, despite considerable success of LiDAR instruments in remotely sensing riverine landscapes, and the obvious benefits they offer – very high spatial and temporal resolution, rapid data acquisition, and minimal disturbance in the field – procurement of these apparatus and their respective computer software comes at high financial cost, and extensive user training is generally necessary in order to operate such devices. Recent developments in computer software have led to advancements in digital photogrammetry over a broad range of scales, with Structure from Motion (SfM) techniques enabling production of precise DEMs based on point-clouds analogous to, and even denser than, those produced by LiDAR, at significantly reduced cost and complexity during post-processing. This study has employed both an SfM-photogrammetry and Terrestrial Laser Scanning (TLS) approach in a comparative analysis of sediment grain size, where LiDAR-derived data has previously provided a reliable reference of grain size. Total Station EDM theodolite provided the parent coordinate system for both SfM and meshing of TLS point-clouds. For each data set, a 0.19 m moving window (consistent with the largest sediment clast b axis) was applied to the resulting point-clouds. Two times standard deviation of elevation was calculated in order to provide a surrogate measure of grain protrusion, from which sediment frequency distribution curves were drawn. Results through semi-variance analyses elucidated continuity of each data set. Where univariate statistics failed to reveal disparity between the two data sets, semi-variance analysis exposed considerable variability in roughness, thus revealing a greater degree of detail in SfM-derived data.