A stratified random sampling design was employed to sample fish on the coral cap communities of the East and West Flower Garden Banks. The survey domain at each coral cap was designated as all areas shallower than 110 feet below sea level. Fine-scale (0.5 m) bathymetric data provided by the Flower Garden Banks National Marine Sanctuary was used to ascertain depths and slope. In 2006, the survey domain was partitioned into flat/gentle sloping (0-30 degrees) and steep sloping (>30 degrees) area and into East and West Banks. A continuous surface of slope estimates was produced for each Bank using a nearest neighbor filter on an aggregated resolution (5m X 5m) bathymetric grid derived from the original bathymetric dataset in ArcGIS (ESRI, 2006). The combination of two slope and two location groupings resulted in four strata. Survey sites were then randomly positioned in each stratum in ArcGIS (ESRI, 2006). Data were collected using 100m2 transects. Transects radiated from each point at a random bearing. In 2007, a sample frame consisting of mutually-exclusive 50m X 50m grid cells was produced and overlaid on each coral cap to exhaustively cover all diveable areas (i.e <110ft). Each grid cell was considered a sample unit and units were divided into six strata. Strata were defined using benthic habitat, location and depth. High (dominated by plate and head corals) and low relief (dominated by Madracis and rubble) coral categories were determined using a benthic habitat map generated by visual interpretation of multibeam data. Units were divided into deep (>105 ft) and shallow (<105ft) groups based on the location of sampling unit centroids on multibeam bathymetric models. Location and high relief coral categories were divided by depth, low relief coral was not. Data were collected within 100m2 transects. Transects radiated from the centroid of each sample unit at a random bearing.
Once in the field, the boat captain navigated to previously selected sites using a handheld GPS unit. On-site, divers were deployed and maintained contact with each other throughout the entire census. One diver was responsible for collecting data on the fish communities utilizing the belt-transect visual census technique over an area of 100m2 (25m length X 4m width). The belt-transect diver obtained a random compass heading for the transect prior to entering the water and recorded the compass bearing (0-360o) on the data sheet. Visibility at each site must be sufficient to allow for identification of fish at a minimum of 2m away. Once reasonable visibility was ascertained, the diver attached a tape measure to the substrate and allowed it to roll out for 25m while they collected data.
As a rule, the habitat was not altered in any manner by lifting or moving structure, however, the observer did record fish seen in holes, under ledges and in the water column. To identify, enumerate, or locate new individuals, divers moved off the centerline of the transect as long as they stayed within the 4m transect width and did not look back along area already covered. The diver was allowed to look forward toward the end of the transect for the distance remaining (i.e. if the diver was at meter 15, he can look 10 meters distant, but if he was at meter 23, he could only look 2 meters ahead).
On-site, no attempt to avoid structural features within a habitat such as a sand patch or an anchor was made as these features affect fish communities and are "real" features of the habitats. The only instance where the transect deviated from the designated path was to stay above 110 ft. Transect lasted 15 minutes regardless of habitat type or number of animals present. This allowed more mobile animals the opportunity to swim through the transect, and standardized the samples collected to allow for comparisons.
Data were collected on the following:
1) Logistic information - diver name, dive buddy, date, time of survey, site code, transect bearing.
2) Taxa presence - as the tape roles out at a relatively constant speed, the diver records all fish species to the lowest taxonomic level possible that come within 2m of either side of the transect. To decrease the total time spent writing, four letter codes are used that consist of the first two letters of the genus name followed by the first two letters of the species name. In the rare case that two species have the same four-letter code, letters are added to the species name until a difference occurs. If the fish can only be identified to the family or genus level then this is all that is recorded. If the fish cannot be identified to the family level then no entry is necessary.
3) Abundance & size - the number of individuals per species is tallied in 5cm size class increments up to 35cm using visual estimation of fork length. If an individual is greater than 35cm, then an estimate of the actual fork length is recorded.
4) Photos ? individuals too difficult to identify or unique in some manner may be photographed for later clarification.
Data caveats: The mission in 2007 was aborted early due to Hurricane Humberto (September 2007). No sites were surveyed on the West Bank and only a subset of the sample of random sites selected on the East Bank were surveyed. Randomly selected sites on the East Bank were not sampled randomly. These data should not be aggregated to make synoptic population or community estimates for the East Bank.