About the Data
Rationale for Acquisition
Benthic habitat maps around the US Virgin Islands and southwest of Puerto Rico are being produced by the National Ocean Service's (NOS) Center for Coastal Monitoring and Assessment (CCMA). These maps use acoustic data from multibeam sonar to differentiate and ultimately delineate distinct benthic habitats (e.g., mud, sand, pavement, patch reef) based on spatial patterns in bathymetry and sonar reflectance properties (backscatter). To ensure the maps are accurate, in situ visual classifications of the seafloor are used to ground-truth benthic habitats and to assess map accuracy.
Imagery Collection Methods
Imagery of the seafloor was collected using remotely operated vehicles (ROVs) and drop camera systems due to the range of depths involved (20-900m). In 2005 and 2006 data was collected using ROVs which ran along predefined transects. In 2007 data was collected using a drop camera system which was dropped at predefined locations and dragged several meters underwater. Hereafter, transects and drop locations are called clusters. All data was collected off the NOAA ship Nancy Foster.
Two independent sets of clusters were used to collect imagery. One set of clusters was collected to ground-truth acoustic data. These transects were purposefully placed over as many distinct habitat features and habitat transitions as possible. Distinct habitat features and transitional areas were determined from spatial patterns in fine-scale NOS bathymetry data. In rare cases when fine-scale data was not available, coarser-scale bathymetry data from NOAA's GEOphysical DAta System (GEODAS) was used. A second set of transects was collected to assess the accuracy of benthic maps. Transects in this set were randomly positioned in mapped areas. Due to logistical constraints all transects were typically positioned in line with predominant current and wind vectors. In 2006 all transects ran along bathymetric contours to avoid large variations in depth which are dangerous below 500m.
Each year different platforms were used to acquire data. Collection methodologies are described below by year of acquisition.
A video camera and high-resolution digital still camera mounted on a Spectrum Phantom S2 Remotely Operated Vehicle (ROV) was used to collect video and still image data, respectively. Data was collected using the NOAA ship Nancy Foster from February 1 to 12, 2005 predominantly during daylight hours to ensure adequate ambient light levels. The spatial coverage of these data correspond with areas surveyed with multibeam in 2004 and 2005. High powered strobe lights mounted on the ROV were used to supplement ambient light levels during the day and served as the only source of light during night operations. An Ultra Short Baseline (USBL) system was used to determine the relative position of the ROV to the Nancy Foster, and in conjunction with the ship's dynamic positioning system, the geographic position. The positional accuracy was estimated to be within 5m for the maximum depth surveyed (200m).
Video data was collected throughout the duration of a transect and photo stills were collected first at 1 minute (first 2 transects) and later at 30 second intervals. The forward-facing video camera was pointed at a 45 degree downward angle to give ROV pilots a view of upcoming obstacles and researchers a view of the benthic habitat. The ROV height above the substrate and speed were approximately 2m and 1m/s, respectively. The ROV pilot attempted to keep the ROV height and speed as constant as possible to standardize the field of view and spatial resolution of characterizations. Two downward pointing parallel lasers separated by 5 cm and the scale of habitat features and organisms were used to estimate height off the bottom. Still photo images were acquired using a downward pointed camera. The uniform distance between lasers was used in photo characterizations as a scale reference.
Video images of benthic habitat were taken during a March 21 – April 2, 2006 cruise on the NOAA ship Nancy Foster. A SeaEye Falcon ROV was used to acquire video at depths between 20m and 800m. A forward mounted, ±90° tilting camera was used to both drive the ROV and collect images for subsequent characterization. The ROV collected video data along predefined transects that typically ran parallel to isobaths, because of the time needed to change the ROV's position in the water column. ROV speed was variable, but always under 1m/s. ROV distance from the bottom was on average 1m. The video camera was predominantly positioned at −60 from horizontal to provide the best view for benthic habitat characterization and driving, but was positioned horizontally when the ROV moved sideways along steep (>70°) cliffs. High powered lights provided illumination. A USBL system was used to determine the relative position of the ROV to the Nancy Foster, and in conjunction with the ship's dynamic positioning system, the geographic position. The positional accuracy of the USBL is dependent on depth and therefore accuracy was estimated to range between 5m at a depth of 200m to approximately 50m at a depth of 800m. Still captures of video were taken every 30 seconds.
The depth, slope, rugosity and geographical zone for each image were determined from bathymetry data collected during each mission (click here for bathymetry data). Geographical zone was assessed based on the relative position of each image to broad-scale geological features according to Green et al. 1999 (e.g., continental shelf, continental slope, abyssal plain).
Each image was characterized by an observer trained in distinguishing benthic habitats at depths between 20m and 1000m. A hierarchy of geomorphological structure types was used to classify benthic habitats. At its coarsest spatial scale, benthic habitats are divided into “hard bottom”, “soft bottom” and “other” types. Each of these is further divided into numerous subcategories which describe greater detail at finer spatial scales. The list of geomorphological structure types incorporate contributions from Green et al. (1999), Kendall et al. (2004) and a preliminary assessment of video data to identify which benthic habitat types were present.
A biological assessment was also carried out for each image to provide complementary data on the biotic cover. Numerous categories of biotic cover were used to classify organisms, including hard coral, soft coral, sponges, coralline algae, turf/filamentous algae, and uncolonized area. The spatial coverage of biotic cover in each category was quantified according to the following ordinal scale:
- Absent (0%)
- Rare (1%-10%)
- Sparse (11%-50%)
- Patchy (51%-90%)
- Continuous (91%-100%)
All observer characterizations corresponded to dimensions which define the envisioned benthic habitat map's minimum mapping unit (MMU) (i.e. map spatial resolution). A range of MMUs, which increased in size with depth were used to accurately map benthic habitats. A single MMU was not employed, because either spatial resolution would be lost at shallower depths or characterizations would be inappropriate for deeper sites. The MMUs were determined by the spatial resolution of the multibeam sonar signals used to make the map. In depths shallower than 50m the MMU was 2m by 2m, between 50m and 250m the MMU was 5m by 5m and in depths greater than 250m the MMU was 10m by 10m. At depths below 200m and during the 2007 mission, when still photos were taken from video data, much of the characterization was completed using video data.