Habitat mapping on a deep-water coral reef off Norway, with a comparison of visual and computer-assisted sonar imagery interpretation

During a research cruise in 1999, the Sula Ridge, an almost 14 km long and up to 35 m high deep-water coral reef structure on the Mid-Norwegian shelf mainly built by Lophelia pertusa, was entirely mapped with high-resolution sidescan sonar. In addition, a dense echosounding grid, underwater video observations and dives with the manned research submersible JAGO provided precise high-quality ground-truthing and allowed a detailed interpretation of the reef structure and its surrounding geological features from the sidescan sonar imagery. The result of this visual sidescan sonar interpretation is a facies/potential habitat map that delineates different regions of coral environment, e.g. live coral reef, dead coral structure, sediment covered coral etc. In an attempt to improve this interpretation, computer-assisted image analysis was applied to a representative section of the sonar data, trying to reveal patterns ?invisible? to the human eye (using the University of Bath TexAn software). The technique of texture analysis uses Grey Level Co-occurrence Matrices (GLCMs) to calculate statistical indices quantifying the distribution of grey levels and their spatial relationship within the image. For example, regions of rough texture (coral mounds) can be distinguished from areas of smooth background sediment or zones of heterogeneous texture resulting from sediment-covered coral debris and dropstones colonised by sponges. The results of this computer-assisted approach were carefully compared with the earlier visual interpretation to identify the differences and to see where the interpretation could be improved. Overall, it is shown that texture analysis certainly can be a useful tool to make facies/habitat mapping from sidescan sonar easier and faster, in the meantime revealing details overlooked during visual interpretation. However, validation of certain details by an experienced interpreter might still be necessary, and therefore visual and computer-assisted interpretation should be used as complementary tools.