on a site-by-site basis as a function of the overall characteristics of each site. Both
of these approaches are described in detail in the sections below, as well as an
assessment of the resolution necessary to properly map cold-water coral habitats in
the Straits of Florida.
10.3.3 Cold-Water Coral Mound Characterization
Cold-water coral fields in the Straits of Florida have been mainly reported as
individual mound build-ups that can reach up to 50 m in relief and 1,000 m in
footprint area (Neumann et al. 1977; Mullins et al. 1981; Messing et al. 1990; Paul
et al. 2000; Reed et al. 2006). The distribution and spatial characteristics (e.g., size,
morphology, complexity) of the mounds in these fields are poorly described due to
the coarse resolution of available maps. To determine the minimum level of resolution required for accurately assessing cold-water coral mounds in the Straits of
Florida, this study analyzed mound size-frequency distribution of three different
DEMs (50, 20, and 3 m grid-size resolution; Table 10.1) collected over the
47 km
2 -surveyed area of the GBB site.
Mounds are visible on all three DEM maps of the GBB site (Fig. 10.4). Mound
perimeters are, however, complex and difficult to consistently define. To systematically assess mound size-frequency distributions for each DEM, an automated mound extraction approach was developed. This approach relies on the
change of slope angle between mound and surrounding area. First, slope angle
maps are generated from each DEM (Fig. 10.5a), and closed polygons are created
along the contour line where the slope angle exceeds 8° (Fig. 10.5b). This 8° cutoff
was determined following an attempt to delineate mound perimeters manually.
This manual delineation indicated that the majority of the mounds rise out of the
surrounding seabed with a cutoff plane of *8°. Because the slope angle can vary
within a mound feature by more than 8°, the algorithm can also create new
polygons within a given mound (Fig. 10.5b). Therefore, all polygons within
another polygon were filtered out, so that only the outermost polygons would
represent mound perimeters (i.e., mound footprint; Fig. 10.5c). Data from each
original DEM is then removed within the areas enclosed by mound perimeter
(Fig. 10.5d). The DEM is re-gridded to generate new bathymetric maps without
the mounds themselves, where the vertical relief within each removed mound was
interpolated from the mound perimeters (Fig. 10.5e). The newly gridded surfaces
are then subtracted from the original DEMs to produce maps in which only the
vertical relief within mound perimeters is displayed (Fig. 10.5f). Finally, a Matlab
routine calculates the maximum thickness (i.e., height) within each mound polygon. In this study, a mound feature in any given DEM is defined as every closed
polygon that is [1 m in height and has a footprint area [81 m
2 . This minimum
mound area for calculating morphometrics is based on the fact that a 3 9 3 pixel
matrix is 81 m
2 (given that each pixel = 3 m). Smaller matrices (e.g., 2 9 2) do
not contain sufficient pixels to represent free-form mound footprints.
264
T. B. S. Correa et al.
of these approaches are described in detail in the sections below, as well as an
assessment of the resolution necessary to properly map cold-water coral habitats in
the Straits of Florida.
10.3.3 Cold-Water Coral Mound Characterization
Cold-water coral fields in the Straits of Florida have been mainly reported as
individual mound build-ups that can reach up to 50 m in relief and 1,000 m in
footprint area (Neumann et al. 1977; Mullins et al. 1981; Messing et al. 1990; Paul
et al. 2000; Reed et al. 2006). The distribution and spatial characteristics (e.g., size,
morphology, complexity) of the mounds in these fields are poorly described due to
the coarse resolution of available maps. To determine the minimum level of resolution required for accurately assessing cold-water coral mounds in the Straits of
Florida, this study analyzed mound size-frequency distribution of three different
DEMs (50, 20, and 3 m grid-size resolution; Table 10.1) collected over the
47 km
2 -surveyed area of the GBB site.
Mounds are visible on all three DEM maps of the GBB site (Fig. 10.4). Mound
perimeters are, however, complex and difficult to consistently define. To systematically assess mound size-frequency distributions for each DEM, an automated mound extraction approach was developed. This approach relies on the
change of slope angle between mound and surrounding area. First, slope angle
maps are generated from each DEM (Fig. 10.5a), and closed polygons are created
along the contour line where the slope angle exceeds 8° (Fig. 10.5b). This 8° cutoff
was determined following an attempt to delineate mound perimeters manually.
This manual delineation indicated that the majority of the mounds rise out of the
surrounding seabed with a cutoff plane of *8°. Because the slope angle can vary
within a mound feature by more than 8°, the algorithm can also create new
polygons within a given mound (Fig. 10.5b). Therefore, all polygons within
another polygon were filtered out, so that only the outermost polygons would
represent mound perimeters (i.e., mound footprint; Fig. 10.5c). Data from each
original DEM is then removed within the areas enclosed by mound perimeter
(Fig. 10.5d). The DEM is re-gridded to generate new bathymetric maps without
the mounds themselves, where the vertical relief within each removed mound was
interpolated from the mound perimeters (Fig. 10.5e). The newly gridded surfaces
are then subtracted from the original DEMs to produce maps in which only the
vertical relief within mound perimeters is displayed (Fig. 10.5f). Finally, a Matlab
routine calculates the maximum thickness (i.e., height) within each mound polygon. In this study, a mound feature in any given DEM is defined as every closed
polygon that is [1 m in height and has a footprint area [81 m
2 . This minimum
mound area for calculating morphometrics is based on the fact that a 3 9 3 pixel
matrix is 81 m
2 (given that each pixel = 3 m). Smaller matrices (e.g., 2 9 2) do
not contain sufficient pixels to represent free-form mound footprints.
264
T. B. S. Correa et al.