Automated Detection and Characterization of Submarine and Subaerial Channels Using Wavelet Analysis

The form, extent, and plan-view geometry of natural channels are used to infer bank-full discharge, sinuosity, and relations between these parameters and catchment area, flow width, and flow depth. The extraction of channel morphometry usually requires field measurement or interpretation of plan-view and cross-sectional data from maps, air photos, and other remote-sensing generated imagery. We developed an automated method for channel detection from plan view data of various types, including air photos, seismic imagery, and digital elevation models (DEMs). The proposed method identifies channels by comparing the directional derivative of the data with that of an idealized channel using wavelet analysis. The best-fitting form provides channel orientation, width, relief, and sharpness of the channel boundary. In cases where diffusive processes operate to smooth channel boundaries through time, relative dating of channel abandonment can be carried out based on the smoothness of these boundaries. Application of the proposed method over a high resolution bathymetry (comparable to ALSM data) of the Lucia Chica channel system, a submarine distributary system off-shore California, shows good agreement between the automatically detected channels and those detected by interpretation of the associated chirp data (2D seismic) and traditional seafloor mapping. Automatically extracted relative channel ages are partially in agreement with those interpreted from seismic data by traditional techniques, suggesting that the submarine topography is smoothed through time by diffusion-like processes. These relative ages are combined with 14C-derived estimates of the most recent channel activity to determine the absolute ages of different channel segments, the temporal frequency of avulsions between these segments, and the diffusion coefficient for surface smoothing processes at this submarine environment. While additional experiments are required to test the sensitivity of the method and the extent of its applicability to various datasets, this methodology offers an efficient screening of large datasets to detect channels and extract spatial and temporal variations of their morphologies.