Seafloor habitat characterization, classification, and maps for the lower Piscataqua River estuary

Data from multibeam echosounders were used to implement segmentations and classifications of the seafloor, and measurements from underwater images and physical samples were used to relate segmentations and predictions to observed seafloor characteristics. Texture analysis, using local Fourier histogram (LFH) texture features, was applied to multibeam bathymetry data in unsupervised- and supervised-classification modes. Unsupervised classification of bathymetry using texture features produced segmentations that corresponded to known spatial distributions of seafloor sediments, but required arbitrary choices for some parameter values and, therefore, included potential bias. Supervised classification of bathymetric texture overcame bias related to arbitrarily-chosen parameters and produced classifications that corresponded well with identified seafloor habitats, but accuracy of supervised-classification results depended on detailed and accurately-georeferenced ground-truth data. The LFH texture feature classification technique, using only gridded bathymetric data, was generally effective for predicting spatial distributions of seafloor morphologies and habitat structure classes on a per-grid-cell basis and was robust to data noise. In some cases, different substrates had similar morphologies, and in these cases, texture alone was inadequate for discriminating habitats. Using acoustic backscatter strength in addition to texture or roughness sometimes facilitated discrimination of habitats with distinct substrates and similar morphologies. Microtopographical roughness influences high frequency acoustic scattering, and roughness measurements can facilitate modeling and interpretation of backscatter data. Spectra model parameters (slope and intercept) were calculated to describe roughness of seafloor microtopography in sediment profile images (SPI). SPI spectral-model parameters were consistent with published estimates for data from other devices such as stereophotographs, and values varied by sedimentary facies and bioturbational regime. Traditional methods of ground truthing were not always sufficient for characterizing attributes of features seen in shallow-water multibeam data. Seafloor video-image mosaics were used to characterize biogenic features and verify transitions between habitats and allowed descriptions of features that were not determinable from other imagery