FISHERY BULLETIN: VOL. 85, NO. 2 



Table 1 .—Temporal design of data collections and of experimental treatments for both habitats, 1980-84. Entries are numbers of samples^ 



taken per matrix. 



36 



36 

 36 

 36 



36 



'In all cases where 36 or 9 samples were taken per matrix, these were 'A m^ samples distributed uniformly across the matrix such that no sample fell within 

 1 m of any previous sample location. Where 6 or 3 samples were taken, these were chosen at random from a group of 9 uniformly distributed samples positioned 

 in a similar way to avoid any overlaps. All sediment samples were cores of 5 cm diameter x 20 cm deep. Macroinvertebrate samples were cores of 10 cm 

 diameter x 25 cm deep. 



^Data taken from only the seagrass habitat on these dates. 



To collect a repeatable sample, we first inserted 

 a 0.25 m^ circular metal sampling frame pene- 

 trating to a depth of 15 cm and used an hydraulic 

 suction dredge to excavate the complete contents 

 to that same depth. The material was collected in 

 a 3 mm nylon mesh bag (for description and sam- 

 pling efficiency, see Peterson et al. 1983b). All living 

 M. mercenaria and A. irradians were removed from 

 the mesh bag and placed in separate, labeled plastic 

 bags for return to the laboratory. For all M. mercen- 

 aria we measured length in the longest antero- 

 posterior dimension, and for all A. irradians we 

 measured the distance from the flat top of the hinge 

 to the ventral margin using vernier calipers. Sea- 

 grass material from the mesh bag was packaged in 

 marked plastic bags in the field and returned to the 

 laboratory, where it was gently rinsed in freshwater 

 to remove attached salt and sediments, and dried 

 to constant weight (2-4 days) at 105°C. 



To estimate densities of small benthic macroinver- 

 tebrates, we took 9 uniformly distributed samples 

 from each matrix in each habitat on 4 sampling dates 

 (Table 1). We processed and analyzed a randomly 

 chosen subset of 6 of these 9 samples for each 

 matrix. The strategy of taking more samples than 

 one expects to analyze is optimal when marginal 

 costs of additional sampling are low, because extra 

 replicates are then available for later analysis if 

 among-sample variation proves so unexpectedly 

 high as to reduce statistical power to an unaccept- 

 able level. Benthic invertebrates were collected 



using 10 cm diameter cores taken to a depth of 25 

 cm. Complete contents of each core were placed in 

 separate plastic bags and gently sieved, in the lab- 

 oratory, through 1 mm mesh. Sieve contents were 

 held in bottles containing rose bengal in 10% buf- 

 fered formalin until animal tissues were adequate- 

 ly stained and hardened. We later picked and iden- 

 tified to class (and to species in a subset of the 

 samples) all animals in each sample. 



In spring 1980, we also took 8 randomly located 

 sediment cores (5 cm in diameter to a depth of 20 

 cm) from each matrix to characterize initial sedi- 

 ment conditions. Cores were transferred into in- 

 dividual plastic bags and frozen at - 10°C until 

 analysis of sediment size distribution by weight. We 

 split each sample by coning and quartering (Ingram 

 1971) and then used standard Rotap dry sieving and 

 pipetting procedures (Folk 1974) to estimate dry 

 weights of sediments in each of several size classes. 

 In addition, percent organic content was measured 

 by weight loss on ignition at 550 °C for 4 h (Gross 

 1981). Because our (customary) use of small-diam- 

 eter cores to sample sediments failed to include large 

 shell fragments and because such biogenic calcium 

 carbonate appeared to be extremely common in 1 

 seagrass matrix, we designed a sampling procedure 

 to estimate the relative degree of coarse shell. In 

 October 1985, we used the suction dredge to ex- 

 cavate 3 haphazardly located 0.25 m^ quadrats to 

 a depth of 12 cm in each of the 6 matrices in each 

 habitat. All shell fragments collected on a 3 mm 



284 



