the highest recruitment was then found in the presence of the fishes, which cropped 

 the algal competitor without harming the corals. Also, damself ishes , maintaining 

 territories that form algal mats, can have either positive or negative effects on 

 corals, depending upon the location of the coral relative to the territory (see Vine 

 1974, Kaufman 1977, Wellington 1982). 



Observing the browsing activities of Caribbean acanthurid and scarid fishes, 

 Birkeland (1977) found evidence that fishes avoided corals as small as 3mm; yet Bak 

 and Engel (1979) saw no such avoidance. Moreover, Neudecker (1977, 1979) observed 

 fishes selectively preying on corals transplanted to different areas on Pacific 

 reefs. The extent to which urchins and fishes prey upon corals may depend to a 

 great extent on the availability of the preferred food resource. Diadema , for 

 instance, shift to feeding on corals as the preferred algal food is diminished 

 through grazing (Carpenter 1981). The differing observations reported above may be 

 a reflection of different grazer densities and/or algal biomass in the various study 

 sites. Manipulative studies such as Sammarco's (1980) and Brock's (1979) indicate 

 the existence of optimal levels of grazing for coral survival, with these grazer 

 densities apparently being closer to those commonly found on the reefs. 



We sought to determine the influence that grazers naturally present at interme- 

 diate depths have on populations of juvenile corals. As a large number of cement 

 artificial reefs were constructed and censused, extensive underwater time was re- 

 quired and the work would not have been feasible without the saturation diving 

 facilities of NOAA's underwater habitat, NULS 1 (Hydrolab). 



The study area is off the north coast of St. Croix, US Virgin Islands, in Salt 

 River Submarine Canyon. The canyon extends seaward from a small estuary, with the 

 canyon walls being variously covered by hard corals and coral rubble, sponges, gor- 

 gonians, etc. (see Rogers et aj. 1983). The artificial reefs were established on 

 the sandy floor of the canyon, well removed from the reef complex of the walls. 

 This is an area of shifting sand uncolonized by benthic macroflora; during periods 

 of heavy swells on this exposed coastline, sand scouring around the reefs is evi- 

 dent. 



METHODS 



Artificial reefs were established during NOAA Hydrolab Mission #81-1 in January 

 1981. Each reef structure was identical, consisting of 11 cement cinder blocks 

 stacked and lashed together in three layers of 6, 4, and 1 block-per layer. Total 

 censused surface area of each reef was 0.93m vertical and 0.46m horizontal sub- 

 strate. Corals inside the holes of the cinder blocks were not censused. Cage 

 enclosures were composed of 2.5 X 5cm wire mesh over a 1.5 X 3 X 1.5m frame. 



A total of fifteen reefs were constructed at a depth of 20m on the canyon floor 

 along a line a constant 15m distance from the junction of the sandy floor and the 

 east coral wall. The reefs were divided equally into three treatment types, with 

 each reef along the line being an alternate treatment. Uncaged reefs were exposed 

 completely to fishes and invertebrates; caged reefs were totally enclosed by cages; 

 and cage controls had cages with two sides and the top partly-open, allowing access 

 by macrofauna. 



In January 1982 during Hydrolab Mission #82-1, all of the reefs were censused j_n 

 situ for juvenile scleractinian corals. Maps of each reef were employed to record 

 the location of the corals, its identification to genus and sometimes species, and 

 its length and width (± 0.5-lmm). Deterioration of all the cages during storms in 

 February 1982 did not allow repeat censuses of the corals on the caged reefs. 



After initial establishment of the reefs, periodic fish censuses were performed 

 to quantify the number of resident fishes on the reefs. Transient fishes (including 

 herbivorous fishes around the reefs) were not quantified. Numbers of Diadema antil - 

 larum , being low, were recorded during only one census in February 1982. 



90 



