Remote seabed characterization: Statistical tools for correlating maps and samples

Remote characterization of seafloor is a useful geophysical tool, as it allows data collection while ship is moving, being more expedite than traditional methods, which require stopped ships for grabbing sediment samples. This work evaluates techniques applied for seabed characterization using high frequency sonars. Programs analyze the backscatter signal through a series of methodologies (e.g. image texture and angular response analysis) to build classification maps. But, varied configurations used during processing are able to generate different maps, causing uncertainties about the quality of results. Some recent studies discussed about classification maps, but usually checking processing programs separately, using qualitative analysis to correlate maps and grabbed samples. This study combines results of several classification programs simultaneously, being required for that to establish an appropriate statistical variable to quantitatively represent correlation results. To solve this problem, a statistical tool termed as mutual association coefficient is proposed herein. Future studies would benefit from using this method, because a standard value is more appropriate to designate classification results and facilitate inter comparison between several approaches, being tested in the same geographic area. In this study, two areas in Guanabara Bay, in Rio de Janeiro, has been researched, where data from singlebeam echosounder, multibeam echosounder and sidescan sonar was collected, in addition to grabbed sediment samples obtained with Van Veen sampler.