Automatic Processing of Bathymetric Data from Multibeam Echosounders

Multibeam Echosounder data is a powerful method for rapidly mapping the seabed, either from the water surface or from a remote (or autonomous) vehicle. However, the data density of such systems can make data handling difficult, and most conventional processing schemes are subjective due to their reliance on human experience. We describe an automatic scheme for converting raw data to a preliminary gridded product without human interaction, using the estimated measurement error of the data to adjudge the reliability of each sounding and hence the weight that it should be given in the grid. Along with a propagation of information argument which implicitly defines the local bathymetry, the algorithm is able to track multiple estimates of depth to provide extra robustness (e.g., in the face of burst-mode instrument failure, or outlier data) and can continually update its estimate of the bathymetry as new data is presented, potentially from multiple sources or measurement systems. All depth hypotheses are maintained on-line so that human revision of the estimates is possible after the algorithm's best resolution is determined. We illustrate the algorithm on two datasets: a shallow water survey in Portsmouth, NH, and an intermediate depth survey on the west Florida shelf, Gulf of Mexico, where we also compare the surfaces generated by conventional processing methodologies. We conclude that our automatic algorithm provides robust processing, with the `value-added' products of an estimate of uncertainty in the bathymetry, and an indication of any problem regions in the data. The algorithm can integrate data much more quickly that it can currently be gathered, and provides an objective `best' assessment of depths. It also generates surfaces which are statistically similar to conventionally constructed products.