Lenihan and Micheli: Biological effects of shellfish harvesting on oyster reefs 



89 



Statistical analyses 



The density of live and dead oysters, and the propor- 

 tion of the total number of oysters that were found 

 dead before harvesting, were compared among treat- 

 ments by using two-way, fixed factor analysis of vari- 

 ance (ANOVA) tests. The two main factors in the 

 ANOVAs were creeks (Pages and Whiskey Creeks) 

 and harvest treatments (clam harvesting, oyster har- 

 vesting, both, or neither). The same ANOVA model 

 was used to test for differences in 1 ) the density of live 

 and dead oysters, and the proportion of dead oysters 

 (i.e. number of dead oysters/live -i- dead oysters) after 

 harvesting, 2 ) the density of live and dead, and propor- 

 tion of dead, naturally occurring clams after harvest- 

 ing, and 3) the density of live and dead, and number of 

 missing transplanted, hatchery-raised juvenile clams 

 after harvesting. Before all ANOVAs, homogeneity 

 of variances was tested by using Cochran's test (at 

 0=0.05). When variances were heterogeneous, data 

 were log transformed and homogeneity was retested. 

 After ANOVAs, post hoc tests for differences among 

 treatment means were conducted with Student-New- 

 man-Keuls method (SNK) tests (at a=0.05). 



Results 



In July 1996, prior to the application of experimental 

 harvests, the number of live and dead oysters (those 

 observed with naked eye; >1 mm in length) and the 

 proportion of dead oysters in experimental plots did 

 not vary with the interaction of creeks and harvest 

 treatment (ANOVA, creek x harvest treatment inter- 

 action, P=0. 33-0.56; Table 1), nor among harvest 

 treatments (P=0.57-0.99), but differed significantly 

 between creeks (P=0.02-0.0001). Whiskey Creek had 

 greater numbers of live and dead oysters and propor- 

 tion of dead oysters than Pages Creek (Figs. 1 and 2). 



E 

 in 



80 



70- 



60 



50 



40 



30- 



20 



10 



80i 



70 



60 



50 



40 



30 



20 



10 



Pages Creek 



JL 



_Mi 



I, ^ 



i 



Whiskey Creek 



X 



n Live 

  Dead 



u 



:MMl(iti 



Harvest type: 



O 

 Before 



Oysters 



Oysters 



After 



Figure 1 



Mean density of live and dead oysters (>1 mm in length) 

 before (5-10 July 1996 ) and after ( 10-23 July 1997 ) applica- 

 tion of experimental harvest treatments in Pages and Whis- 

 key Creeks, NC. Data are means and one standard error 

 (n=4) of counts taken within 0.25-m^ quadrats. Results 

 of SNK post hoc comparisons are illustrated with letters 

 above bars (a>b at P<0.05). Separate ANOVAs and SNK 

 tests were used to compare numbers of live and dead oys- 

 ters both before and after harvesting. 



Experimental clam harvesting conducted from Aug- 

 ust 1996 to May 1997 removed only hard clams ft-om 

 experimental plots (Table 2). In contrast, a few clams 



