Sequential Video Frame Analysis of Fish Community Structure
on Lophelia Reefs - Northern GOM Continental Slope

Ken Sulak, Kirsten Luke, April Norem & Mike Randall
Coastal Ecology & Conservation Research Group, Gainesville, Florida

Presented at the FlSC Strategic Review in St. Petersburg, Florida, May 9-12, 2006

INTRODUCTION

       The deep-water matrix-building scleractinian coral, Lophelia pertusa (hereafter Lophelia in this paper) occurs circumglobally, including the Gulf of Mexico. This coral species builds large thickets and elevated banks that function as deep-water coral reefs, providing 3-dimensional habitat heterogeneity, shelter for invertebrates and fishes, feeding habitat for ambush predators and microvores, and probable spawning grounds for a few demersal fish species. Lophelia habitats function as oases of macrofaunal and megafaunal biodiversity amidst the otherwise monotonous open sedimented landscape of the continental slope. Lophelia reefs also appear to serve as focal points that concentrate megafaunal organisms otherwise occurring in low abundance on non-coral habitats. Additionally, such reefs may also concentrate particulate food resources, as the elevated coral matrix intercepts bottom currents, generating eddies that entrain plankton and organic particles.

       While the demersal fish fauna associated with Lophelia reefs has been relatively well investigated in the eastern North Atlantic, the fish fauna of those in the western North Atlantic remains essentially undocumented. This study represents the first quantitative analysis of fishes associated with Lophelia reefs and reef-associated biotopes in the Gulf of Mexico, and generally in the western North Atlantic.

METHODS

       The Harbor Branch Oceanographic Institution's manned Johnson-Sea-Link (JSL) submersibles were used to conduct two missions in July-August 2004 and September 2005. Both missions investigated the demersal fish faunas of two prominent elevated topographic features on the continental slope of the northern Gulf of Mexico (Fig. 1). Identified by reference to the MMS oil lease blocks in which they lie, these two features have been designated Viosca Knoll 826 (VK-826), and Viosca Knoll 906/907 (VK-906/907) by previous researchers. Viosca Knoll 826 rises to a minimum depth between 435-480 m; VK-906/907 to a minimum depth between 305-340 m. The two sites represent two biologically distinct depth horizons centered on depths of 500 m and 325 m, respectively, on the continental slope in terms of resident megafaunal fishes and invertebrates. Together, they provide a distinct 3-dimensional hard-substrate, live-bottom continental slope sub-biome, in contrast to the dominant, essentially 2-dimensional, soft-substrate, open slope biome.

       'BELT TRANSECTS' were conducted with the submersible cruising as slowly as possible (typically 0.3 kt) into the direction of bottom current to maintain constant speed and consistent course direction. Altitude was held as close above the substrate as possible (i.e., skimming over the bottom). Color video was obtained using a Sony DX2 3000A 3-chip CCD camera, with 6-48 mm zoom lens. The 'belt', or area of substrate being transected, was illuminated by two high intensity 5600 °K HMI work lights, and by four additional individually-selectable HMI lights surrounding the video camera. Video was recorded to a mini-DV tape recorder. During transecting, the extensible support arm supporting the video camera was kept fully retracted and focal length was maintained at 6 mm (i.e., full wide angle). The camera was panned inward (toward the transect centerline) 15º, and tilted downward 45º. Pan, tilt and zoom were held in this pre-determined configuration throughout designated transects. Transect duration was 5.0 min (sometimes truncated due to operational exigencies). During transecting, data including time (hr:min:sec) and depth (ft) data were continuously overlaid onto the video record.

       VIDEO ANALYSIS – In the laboratory, all original mini-DV tapes were copied onto DVD for analysis and archival. The entire video record of each dive (ca 2 hr) was converted to sequential still frames separated by an interval of approximately one second (0.996 sec) using VideoCharge 3.0 frame-grab software. Analysis of the resulting full-quality images (in uncompressed .tiff format) proceeded as follows:

       ANALYSIS I – Quantitative species abundance and rank order from standardized transects: Step 1) The original mini-DV was initially viewed using a Sony mini-DV VCR to establish identities of all demersal fish species recorded (identification by senior author), denote time segments for capture as still images to document species identifications, empirically define biotopes encountered, and record the starting and ending times of each quantitative transect. Separate high-quality frame grabs documenting individual fish species were assembled into a taxonomic identification reference library used by project personnel. Step 2) Using the DVD frame grab record, each designated transect was viewed on a high-resolution flat monitor, advancing frame-by-frame using the Windows software 'picture and fax viewer' utility. Each transect was analyzed by each sequential 1-second still frame, building an Excel spreadsheet file recording dive number, transect number, frame grab file number, date, time, depth, fish occurrence by species, major biotope, and sub-biotope designations. The number of frame grabs was totaled for each transect to yield the total time analyzed (i.e., total number of grabs, multiplied by 0.996 s grab^-1). Only individual fish specimens crossing either the lower, left, or right margins of the video frame (i.e., fish leaving the field of view as the submersible moved forward) were scored. This minimized recounts of individuals swimming along with the submersible, appearing in more than one frame grab, and/or re-entering the field of view. Each fish was counted only once, as it initially intersected the video margin. Species abundance scores were totaled per transect to determine rank order. Scores for all species were totaled to estimate population density per unit area.

       ANALYSIS II – Species occurrence and relative rank order from non-transect video segments: Step 1) Individual occurrences of each species (regardless of number of individuals in the field of view) were recorded per each 1-second frame over the entire frame grab record for that dive, excluding transects, but including time intervals when the submersible was stationary, and when the camera was panning or zoomed in upon the substrate and/or on sessile invertebrate assemblages. The occurrence of a species was positively scored if that species was present within the analyzed field of view (lower two-thirds of the total field of view). Step 2) Scores were summed by species to determine rank order by frequency of occurrence for the total pool of analyzed frames. For both abundance and occurrence analyses, each data entry included scoring of major biotope category, and the presence or absence of Lophelia coral. Taxon abundance and occurrence data were analyzed for all dives per each of the two Viosca Knoll depth horizons (325 m and 500 m).

RESULTS

       Twenty submersible dives were accomplished on target Lophelia sites; ten in 2004, and ten in 2005, with 12 on the VK-826 site (Fig. 2), and eight on the VK-906/907 site. Only relatively limited portions of the flanks and crests of selected ridges were found to be colonized by extensive Lophelia reef (Fig. – Bottom, Center Left). Remote sampling deployments included four fish traps, ten bottom trawl stations and one benthic sled station. Total submersible bottom time was 44 hrs, 45.4 min, all of which was used to document demersal fish species identifications. Only a limited portion of that time was available for dedicated moving video transects. Thirty-two transects from seven dives were accomplished to support Analysis I; total transect time was 141 minutes. Additional non-transect segments used for Analysis II totaled an additional 115 minutes of video.

     ANALYSIS I: Video records for 32 moving transects from seven USGS submersible dives were converted into 8,486 1-second frame grabs. Analysis of these frames documented at least 37 total demersal species identified from video during USGS submersible missions. Of the total frame grab record, 4,498 frames represented transects accomplished on the 325 m depth horizon (VK-906/907), yielding 648 individual fishes scored among seven species. Frames representing the five biotope categories were scored as follows: 'Open' – 383 frames; 'Plate' – 2,800; 'Rock' – 1,315; Rubble – 0, 'Thicket' – 0. Only 51 'Rock' biotope frames contained substantial (but <50%) Lophelia coral cover.  Thus 'Plate' and 'Rock' hard-substrate biotopes without Lophelia (but almost always populated by anemones, sponges, bamboo corals, and black corals) appeared to dominate the VK-906/907 landscape, accounting for all but three of 648 fish scores. The 3,988 frames representing the 500 m depth horizon (VK-826) yielded 38 individual fishes scored among ten species. Frames representing biotope categories were scored as follows: 'Open' – 1,671 frames; 'Plate' -1052; 'Rock' - 335; 'Rubble' – 0; 'Thicket' – 925; plus six frames over open space as the submersible crested a ridge top.

       ANALYSIS II: A total of 6,879 frame grabs were analyzed for demersal fish occurrence (presence of a given taxon in the analyzed field of view) from the 20 dives, documenting at least 37 distinct species.  The 325 m depth horizon analysis included 2,368 frame grabs, yielding 598 fish occurrences (622 individual fish) among 16 species. The 500 m depth horizon analysis included 4,512 frame grabs yielding 230 fish occurrences (233 individual fish) among 23 species.  All individuals of all identified species recorded during video analyses appeared to be adults or subadults; no obvious juveniles were observed. Nor were obvious juveniles observed in close-up imaging using the digital still camera.  Analysis of occurrence in non-transect segments yielded dominant taxa listed in Table 1. On the 325 m depth horizon, dominant taxa (orders and families) included the Perciformes (Stromateidae, Serranidae, Polyprionidae, Scorpaenidae), Beryciformes (Trachichthyidae), Zeiformes (Grammicolepidae, Zeidae), Gadiformes (Gadidae, Moridae), and Anguilliformes (Congridae). On the 500 m depth horizon, faunal composition by major taxa was similar, except that the large mobile foraging Stromateidae, Serranidae and Polyprionidae dropped out, and the open-slope Macrouridae (rattails) and Squalidae (dog sharks) entered as important contributors to the fauna. Despite the similarity in composition among major taxa at the two depth horizons, there was substantial faunal transition between these horizons at the species level (Table 1). Species dominance rank by occurrence during non-transect video frames roughly paralleled dominance rank by abundance during moving transect frames.

Table 1. Dominant Fish Taxa Determined from Video Analysis II

Species Newly Discovered in NEGOM During USGS Submersible Investigations

SUMMARY & CONCLUSIONS

• The demersal fish fauna of Lophelia pertusa coral reefs and associated hard-bottom biotopes was investigated at two depth horizons on Viosca Knoll in the northern Gulf of Mexico using a manned submersible and remote sampling – with quantitative transects analyzed via sequential frame grabs.
   
• The Viosca Knoll fauna consisted of at least 53 demersal fish species, 37 of which were documented by submersible video.
   
• On the 325 m horizon, dominant taxa determined from frame-by-frame video analysis included Stromateidae, Serranidae, Trachichthyidae, Congridae, Scorpaenidae and Gadiformes.
   
• On the 500 m horizon, large mobile visual macrocarnivores of families Stromateidae and Serranidae dropped out, while a zeiform microcarnivore assumed importance on reef 'Thicket' biotope, and the open-slope taxa Macrouridae and Squalidae gained in importance.
   
• The most consistent faunal groups at both depths included sit-and-wait and hover-and-wait strategists (Scorpaenidae, Congridae, Trachichthyidae), along with generalized mesocarnivores (Gadiformes). The specialized microcarnivore, Grammicolepis brachiusculus, appears to be highly associated with Lophelia reefs.
   
• Coral 'Thicket' biotope was extensively developed on the 500 m site, but fish abundance was low, only 95 fish/hectare.
   
• In contrast to Lophelia reefs from the eastern North Atlantic, coral 'Rubble' biotope was essentially absent.
   
• Lophelia reef and associated hard-substrate biotopes harbor several economic resource species, such as snowy grouper, barrelfish and conger eel. These live-bottom biotopes also harbor some species otherwise rare or previously unreported from the Gulf of Mexico, including the wreckfish, ragged-tooth sandtiger shark, and the thorny tinselfish.