An Automatic Procedure for the Quantitative Characterization of Submarine Bedforms

A model for the extraction and quantitative characterization of submarine landforms from high-resolution digital bathymetry is presented. The procedure is fully automated and comprises two parts. The first part consists of an analytical model which extracts quantitative information from a Digital Elevation Model in the form of objects with similar parametric characteristics (terrain objects). The second part is a rule-based model where the terrain objects are reclassified into distinct landforms with well-defined three dimensional characteristics. For the focus of this work, the quantitative characterization of isolated dunes (height greater than 2 m) is used to exemplify the process. The primary metrics used to extract terrain objects are the flatness threshold and the search radius, which are then used by the analytical model to identify the feature type. Once identified as dunes, a sequence of spatial analysis routines is applied to identify and compute metrics for each dune including length, height, width, ray of curvature, slope analysis for each stoss and lee side, and dune symmetry. Dividing the model into two parts, one scale-dependent and another centered around the shape of the landform, makes the model applicable to other submarine landforms like ripples, mega-ripples, and coral reefs, which also have well-defined three-dimensional characteristics.