classification. The advancements here are twofold: statistical and other machine
learning techniques allow for further automation and increased efficiency in the
production of management-ready habitat maps of coral reef ecosystems as well
as reduction in biases and process-based errors in interpretation. The applications presented in this chapter are predominantly departures from subjective
interpretative methodologies and are advances towards the goal of repeatable
classification techniques.
• Spatial scales and sampling resolution: Hierarchical benthic habitat classification schemes have recently been developed to meet the needs for resource
management over large spatial scales (Madley et al. 2002; Costa et al. 2009a).
Although acoustic bathymetry has long sufficed for the largest scales of seafloor
classification (e.g., shelves and basins), more information and advanced methods are needed for classifying coral reef habitats. Acoustic studies are in the
early stages of demonstrating whether and how such fine degrees of discrimination can be extracted from backscatter and textural properties.
• Ground verification scale: Consistently matching the scale of verification with
that of insonfication within and between studies can be a challenge given the
range of beam and swath widths. The acoustic footprint of an ASC operated in
10 m of water can range from 3 m
2 for a 6.4
o beam operated at 5 Hz, versus
200 m
2 for a 42
o beam operated at 1 Hz. In contrast, multi-beam and interferometric sonar offer beam footprints an order of magnitude higher resolution than
ASC, while achieving swath coverage greater than 39 water depth. In both
cases, ground verification needs to strike the balance between effort in the field
and minimizing uncertainty.
• Temporal variability: The potential for temporal variability in topography,
reflectivity, and biological attributes have rarely been accounted for in acoustic
remote sensing studies. This could be especially problematic in ephemeral
nearshore hardbottom habitats or seagrass beds, due to sediment transport by
winter storms and annual periods of expansion and retreat, respectively. Remote
sensing platforms such as acoustics can be applied to detect these changes when
coupled with objective classification techniques. Indeed, the ability to conduct
repeated and repeatable surveys is a notable strength of acoustic systems and
remote sensing systems in general.
• Reference areas: While resampling a known area of seabed is often used for
internal calibration within or between surveys, it is only truly useful if it is
accompanied by proper ground verification to determine the extent to which the
reference patch might have changed between samplings. Reference patches do
not address the current lack of an universal reference standard, which impedes
corroborative research efforts.
• Calibration of acoustic systems: The degree to which acoustic systems can be
calibrated differs both within and between acoustic platforms. Calibration varies
widely between ASC systems, varying from turn-key configurations to numerous manual and automatic gain adjustments. Moreover, commercial ASC vary
widely in their approaches to removing depth dependency via time-varied gain
and normalization of echo length to a reference depth. Calibration of fishery
9 Acoustic Applications
247
learning techniques allow for further automation and increased efficiency in the
production of management-ready habitat maps of coral reef ecosystems as well
as reduction in biases and process-based errors in interpretation. The applications presented in this chapter are predominantly departures from subjective
interpretative methodologies and are advances towards the goal of repeatable
classification techniques.
• Spatial scales and sampling resolution: Hierarchical benthic habitat classification schemes have recently been developed to meet the needs for resource
management over large spatial scales (Madley et al. 2002; Costa et al. 2009a).
Although acoustic bathymetry has long sufficed for the largest scales of seafloor
classification (e.g., shelves and basins), more information and advanced methods are needed for classifying coral reef habitats. Acoustic studies are in the
early stages of demonstrating whether and how such fine degrees of discrimination can be extracted from backscatter and textural properties.
• Ground verification scale: Consistently matching the scale of verification with
that of insonfication within and between studies can be a challenge given the
range of beam and swath widths. The acoustic footprint of an ASC operated in
10 m of water can range from 3 m
2 for a 6.4
o beam operated at 5 Hz, versus
200 m
2 for a 42
o beam operated at 1 Hz. In contrast, multi-beam and interferometric sonar offer beam footprints an order of magnitude higher resolution than
ASC, while achieving swath coverage greater than 39 water depth. In both
cases, ground verification needs to strike the balance between effort in the field
and minimizing uncertainty.
• Temporal variability: The potential for temporal variability in topography,
reflectivity, and biological attributes have rarely been accounted for in acoustic
remote sensing studies. This could be especially problematic in ephemeral
nearshore hardbottom habitats or seagrass beds, due to sediment transport by
winter storms and annual periods of expansion and retreat, respectively. Remote
sensing platforms such as acoustics can be applied to detect these changes when
coupled with objective classification techniques. Indeed, the ability to conduct
repeated and repeatable surveys is a notable strength of acoustic systems and
remote sensing systems in general.
• Reference areas: While resampling a known area of seabed is often used for
internal calibration within or between surveys, it is only truly useful if it is
accompanied by proper ground verification to determine the extent to which the
reference patch might have changed between samplings. Reference patches do
not address the current lack of an universal reference standard, which impedes
corroborative research efforts.
• Calibration of acoustic systems: The degree to which acoustic systems can be
calibrated differs both within and between acoustic platforms. Calibration varies
widely between ASC systems, varying from turn-key configurations to numerous manual and automatic gain adjustments. Moreover, commercial ASC vary
widely in their approaches to removing depth dependency via time-varied gain
and normalization of echo length to a reference depth. Calibration of fishery
9 Acoustic Applications
247