Mapping benthic habitat using acoustic remote sensing

Backscatter imagery from multibeam echosounders (MBES) is increasingly used for benthic habitat mapping. This research explores MBES backscatter classification using QTC-Multiview on data from Stanton Banks (UK) and Cashes Ledge (USA). Image-processing algorithms are used to extract values from samples of backscatter data, which are reduced by principal components analysis and are objectively clustered. This process is initially evaluated using 2005 data from Stanton Banks and compared with ground-truth data to determine their biological validity. Low-levels of agreement are observed between acoustic class and ground- truth data «35%); video is determined to be the most spatially appropriate method for comparison. Subsequently, the area was resurveyed in 2006 using the same MBES with different operational parameters, acquiring low- and high-density data coverage. Percentage agreement between classifications was 78%, determined to be due to operational parameters as opposed to environmental change. Agreement with ground truth data improved from 71 % to 77% with increased data density. In 2008, a 2 km2 area was resurveyed at two different orientations and vessel speeds within the same 24 hr period. Classification revealed 53% similarity at 4 rns-1 and 49% at 2 rns-1 from opposing orientations. The same orientations surveyed at different speeds were between 68% (k=0.583) and 53% (k=0.384) similar. These results suggest that both orientation and speed are significant considerations in image-based classification. Finally, the significance of water-column biomass in backscatter classification was examined at Cashes Ledge using MBES data from kelp beds. Two approaches were examined for detecting the presence of macrophytes; image-based and manual picking. Comparison with video data revealed comparable success, with both methods most successful at predicting Laminaria sp. (77.3%-82.6% correct) in shallow water «30m). This research demonstrates the significance of MBES backscatter and image-based classification as potential tools for the emergent discipline of benthic habitat mapping.EThOS - Electronic Theses Online ServiceGBUnited Kingdo