Computational and methodological aspects of terrestrial surface analysis based on point clouds

Processing of high-resolution terrestrial laser scanning (TLS) point clouds presents methodological and computational challenges before a geomorphological analysis can be carried out. We present a software library that effectively deals with billions of points and implements a simple methodology to study the surface profile and roughness. Adequate performance and scalability were achieved through the use of 64-bit memory mapped files, regular 2D grid sorting, and parallel processing. The plethora of the spatial scales found in a TLS dataset were grouped into the “ground” model at the grid scale and per cell, sub-grid surface roughness. We used centroid-thinning to build a piecewise linear ground model, and studied “detrended” standard deviation of relative elevations as a measure of surface roughness. Two applications to the point clouds from gravel river bed surveys are described. Linking empirically the standard deviation to the grain size allowed us to retrieve morphological and sedimentological models of channel topology evolution and movement of the gravel with richer quantitative results and deeper insights than the previous survey techniques.