Sub-meter sediment classification using 400 kHz multibeam acoustic backscatter

Bathymetric mapping in shallow water (up to a few hundred m water depth) involves high frequency (> 100 kHz) swath sonar which sound at high spatial resolution (beam footprints up to a few m). Most acoustical sediment classification methods proposed to date either 1) utilize only a subset of backscatter information or 2) require a substantial amount of spatial averaging. For example, the angular dependence in backscatter often means that only backscatter from one or a subset of beam angles is used for classification. Backscatter is inherently variable in shallow water, owing to small numbers of independent scatterers within beam footprints, leading to non-stationarity. These factors have imposed limits on spatial resolutions of sediment classifications (typically up to 10s of m grid sizes). However, there are situations where finer resolution is required. We detail a new empirical spatially continuous sediment classification method that very reliably distinguishes between clastic substrates (sand, gravel, rocks/boulders, and clastic mixtures) in shallow water (so far, tested up to 30m) with a 400 kHz (Reson 7125) multibeam sonar, down to sub-m grid scales. The advance relies on 1) computing backscatter strength from long-echo amplitudes recorded by a multibeam sonar, in a way that maximizes the amount of usable data, accounting for directivity and bed slope effects and 2) utilizing the spatial autocorrelation in backscatter to maximize the number of observations upon which the classification is based. We 1) present a form of the active sonar equation that expresses instantaneous backscattering strength as a function of interfacial target strength and beam footprint based on modeled sonar geometry; 2) report on backscatter strengths over known sediment types (determined from underwater video observations); and 3) present the classification method which uses a calibrated machine-learning approach on the substrate-specific spectral content of backscatter strengths.