PETERSON ET AL.: IMPACT OF MECHANICAL CLAM HARVESTING 



SEAGRASS 



3 

 O 



o 

 "o 



o 



to 

 o 



E 



o 



V— 



> 

 <I 



BEFORE 



AFTER 



CONTROLS RAKING LIGHT INTENSE 



KICKING KICKING 



TREATMENTS 



Figure 3.— Average density oi Mercenaria recruits (<2.5 cm in 

 length) before han'est treatments in spring 1980 and after in Octo- 

 ber 1980 and 1981 (averaged together). ANOVA's showed no 

 significant effect in the sand flat but several significant changes 

 after treatment in the seagrass bed (see Table 4). Seagrass 

 matrices are grouped together for illustration of effects on the basis 

 of results of Duncan's tests performed on 1980 and 1981 data ad- 

 justed for spring 1980 differences in recruit densities. Conse- 

 quently, these groupings separate those seagrass matrices that 

 changed in recruitment pattern after treatment. 



among seagrass matrices in M. mercenaria recruit- 

 ment, we analyzed the posttreatment data by both 

 simple ANOVA to test for differences in each post- 

 harvest sampling and by ANOVA on adjusted data 

 to test for significant changes away from the ini- 

 tial differences. The results of these two different 

 sorts of analysis were inconsistent. ANOVA's on 

 simple recruit densities [log {x + l)-transformed, 

 which homogenized variances in Cochran's tests] 

 demonstrated significant differences among ma- 

 trices in October 1981 and 1982, but not in 1980 or 

 1983. Duncan's tests on the 1981 and 1982 results 

 showed few significant differences and no consis- 

 tent difference in these 2 years (Table 4). The un- 

 weighted means suggest that M. mercenaria recruit- 

 ment may have been less in the 2 intensely kicked 

 matrices, but the two-way ANOVA had a significant 

 date by treatment interaction preventing applica- 

 tion of Duncan's test. 



Despite an indication of lower M. mercenaria 

 recruitment in the 2 intensely kicked matrices (Table 

 4), ANOVA's performed on recruit data adjusted for 

 initial differences among matrices to test whether 

 those differences changed after treatment revealed 



a different pattern. Only the 1980 and 1981 results 

 were significant (both at P < 0.001). The patterns 

 of change in recruitment among matrices were the 

 same in Duncan's tests on both 1980 and 1981 data 

 (Fig. 3). The shelly control I exhibited over 4 times 

 as much recruitment in October 1980 and 1981 as 

 in spring 1980, while control II exhibited about a 

 40% decrease after harvest (Fig. 3). Raking and 

 light-kicking matrices behaved similarly, showing 

 about the same value after harvest as before. The 

 2 intense-kicking matrices showed about a 30% in- 

 crease in M. mercenaria recruitment after harvest 

 (Fig. 3). Thus, the ANOVA's on adjusted data pro- 

 duce results dependent upon whether control I is 

 discarded or averaged together with control II. 



This demonstrates that conclusions about how 

 clam harvest affects M. mercenaria recruitment are 

 not robust to the decision of how to treat the shelly 

 control or to the relaxation of the assumption that 

 matrices are expected to repeat any initial differ- 

 ences in recruitment in the absence of treatment as 

 an intrinsic characteristic. The choice of analysis 

 might be made by examining whether matrices that 

 are treated identically show similar or dissimilar pat- 

 terns of recruitment in different years. A compari- 

 son of all posttreatment recruit data in the 2 intense- 

 ly kicked matrices (Table 4) reveals that they never 

 differed from one another significantly, although the 

 mean difference and even ranking between them 

 varied. The 2 control matrices diverged radically 

 from one another (Table 4), but unpredictable illegal 

 clamming in matrix I may be at least partly respon- 

 sible. Because of the ambiguities in these data, it 

 is impossible to draw any firm conclusion on how 

 treatments affected M. mercenaria recruitment in 

 the seagrass. 



Seagrass Biomass 



Because of substantial and significant differences 

 among seagrass matrices in seagrass biomass in 

 spring 1980 before application of any treatment, we 

 analyzed the posttreatment data by both simple 

 ANOVA to identify significant differences among 

 matrices after treatment and also by ANOVA on ad- 

 justed observations to test for significant change in 

 the initial pattern of biomass differences among 

 matrices. The results of these 2 types of analysis are 

 qualitatively identical, so we present only the results 

 on adjusted data. We prefer this analysis because 

 the Zostera marina and Halodule wrightii in North 

 Carolina are perenials that do not readily and quick- 

 ly spread into new areas (Thayer et al. 1985), so that 

 initial patterns of difference in seagrass biomass 



289 



