suite of complexity metrics was derived from the bathymetry surface in order to
highlight the differences between distinct habitat structures on the seafloor. These
metrics specifically included: (1) mean depth, (2) standard deviation of depth, (3)
curvature, (4) plan curvature, (5) profile curvature, (6) rugosity, (7) slope, and (8)
slope-of-slope. These metrics were collectively transformed into their principal
components to remove redundant information and retain unique information that
described the complexity and structure of the seafloor. Using this principal component image, discrete seafloor features were then segmented using edge detection
algorithms. A variety of spatial, spectral and textual attributes were next calculated
for each seafloor feature, describing its size, shape and color.
Using a supervised classification approach, the seafloor attributes, along with
known points denoting habitat types at particular locations, were used to train the
QUEST algorithm to classify all of the seafloor features extracted by the segmentation process. QUEST is a type of Classification and Regression Tree
(CART) algorithm (Breiman et al. 1984) that efficiently splits an image into two
parts recursively, until all seafloor features are classified or the algorithm is
instructed to stop. QUEST partitioned the acoustic imagery into 35 unique combinations of geomorphological structure, detailed biological cover and live coral
cover types. This classified habitat map was manually reviewed and edited before
its thematic accuracy was assessed. The thematic accuracies (corrected for proportional bias) for the major and detailed structure, major and detailed biological
cover, and live coral cover classes were: 95.7, 88.7, 95.0, 74.0 and 88.3 %,
respectively.
Fig. 9.8 Depicts the benthic habitat map created from MBES imagery for the moderate-depth
(30–60 m) areas in and around the Virgin Islands Coral Reef National Monument in St. John,
U.S. Virgin Islands. The map is symbolized to denote physical composition (i.e., geomorphological structure) of the seafloor
9 Acoustic Applications
239
highlight the differences between distinct habitat structures on the seafloor. These
metrics specifically included: (1) mean depth, (2) standard deviation of depth, (3)
curvature, (4) plan curvature, (5) profile curvature, (6) rugosity, (7) slope, and (8)
slope-of-slope. These metrics were collectively transformed into their principal
components to remove redundant information and retain unique information that
described the complexity and structure of the seafloor. Using this principal component image, discrete seafloor features were then segmented using edge detection
algorithms. A variety of spatial, spectral and textual attributes were next calculated
for each seafloor feature, describing its size, shape and color.
Using a supervised classification approach, the seafloor attributes, along with
known points denoting habitat types at particular locations, were used to train the
QUEST algorithm to classify all of the seafloor features extracted by the segmentation process. QUEST is a type of Classification and Regression Tree
(CART) algorithm (Breiman et al. 1984) that efficiently splits an image into two
parts recursively, until all seafloor features are classified or the algorithm is
instructed to stop. QUEST partitioned the acoustic imagery into 35 unique combinations of geomorphological structure, detailed biological cover and live coral
cover types. This classified habitat map was manually reviewed and edited before
its thematic accuracy was assessed. The thematic accuracies (corrected for proportional bias) for the major and detailed structure, major and detailed biological
cover, and live coral cover classes were: 95.7, 88.7, 95.0, 74.0 and 88.3 %,
respectively.
Fig. 9.8 Depicts the benthic habitat map created from MBES imagery for the moderate-depth
(30–60 m) areas in and around the Virgin Islands Coral Reef National Monument in St. John,
U.S. Virgin Islands. The map is symbolized to denote physical composition (i.e., geomorphological structure) of the seafloor
9 Acoustic Applications
239