104 



Fishery Bulletin 104(1) 



num.), manatee grass {Syringodium filliforme), and shoal 

 gprass (Halodule ivrightii). Percent coverage was estimat- 

 ed visually, or if water clarity was insufficient to visually 

 inspect the bottom, bottom samples were collected at 

 3-m intervals during deployment of the seine. For ana- 

 lytical purposes, areas sampled had to contain at least 

 ten percent seagrass to he considered seagrass habitat. 

 Samples were classified as "vegetated" or "unvegetated" 

 after sampling; and monthly collections varied from one 

 to seven hauls per month depending on seagrass cov- 

 erage. Seagrass habitats were sampled by pulling the 

 seine into the current or wind, whichever was strongest. 

 We kept the opening of the net at a constant width by 

 maintaining tension on a 15.5-m line that was attached 

 between each end pole of the seine while the seine was 

 hauled for a distance of 9.1 m. The distance covered by 

 the seine was measured by a weighted line from the 

 starting point. The net was retrieved by bringing the end 

 poles together and pulling the net at an angle around a 

 vertical pole that closed the wings of the net and forced 

 the catch into the bag. A typical seine haul over seagrass 

 habitat covered approximately 140 m-. 



In the field, all fish were identified to the lowest pos- 

 sible taxon. counted, and released. Up to 40 individuals 

 of species of special interest (important to the commer- 

 cial or recreational fishery) and 10 individuals of all 

 other species were measured to the nearest millimeter 

 standard length (SL). For quality-control purposes, 

 three specimens of each species collected were returned 

 to the laboratory so that species identification could be 

 confirmed. At each site, Secchi depth and water depth 

 were measured, and water temperature (°C), salinity, 

 dissolved oxygen level (mg/L), and pH were measured 

 by using a Hydrolab Surveyors'- water-quality instru- 

 ment (Hach Environmental, Loveland, CO). 



Data analysis 



Multivariate analyses were used to compare fish com- 

 munity structures along tidal-creek shorelines to those 

 found in seagrass habitats (Field et al., 1982). Average 

 monthly abundance estimates (number of fish divided 

 by the number of hauls) were calculated separately for 

 each fish species in each habitat type. Average monthly 

 abundance estimates were then converted to percent 

 composition to correct for bias introduced by the two 

 different net-deployment methods and for the different 

 levels of effort in each habitat. Fishes that were not 

 identified to species were eliminated (<0.1% of total 

 fish collected) except where species complexes, such as 

 silversides (Menidia spp.), mojarras (Eucinostomus spp. 

 <50 mm SL), menhaden {Brevoortia spp.), and minnows 

 {Notropis spp.) were substituted. Species complexes 

 were used when meristic characters for juveniles were 

 insufficient to distinguish between two or more pos- 

 sible species (Eucinostomus spp. and Notropis spp.) or 

 where there was possible hybridization (Menidia spp. 

 and Brevoortia spp.). 



Fish-community comparisons based on percent spe- 

 cies composition by habitat and month were conducted 



by using algorithms in PRIMER (Plymouth Routines 

 in Multivariate Ecological Research, vers. 5, Plymouth 

 Marine Laboratory, UK) for the study of community 

 structure (Clarke and Warwick, 1994). To identify 

 fish assemblages, hierarchical agglomerative cluster 

 analysis was performed with the Bray-Curtis similar- 

 ity matrix calculated on fourth-root transformed per- 

 centage data. The fourth-root transformation reduced 

 the dominance of abundant species and increased the 

 influence of less abundant species in the community 

 analysis. The cluster analysis was run on one matrix 

 consisting of all transformed fish abundance esti- 

 mates collected in tidal-creek and seagrass habitats 

 combined. 



To identify species that were responsible for the pat- 

 terns observed in the cluster analysis, similarity and 

 dissimilarity percentage breakdowns were conducted 

 by using the SIMPER procedure in PRIMER (Clarke, 

 1993). Average similarities between assemblages were 

 analyzed to determine the contribution of each species 

 to the overall similarity. This procedure reduced the 

 number of species required to explain the patterns 

 observed in the cluster diagram and allowed for a sim- 

 plified interpretation of the species assemblages. The 

 higher the similarity value, the more alike samples 

 were within assemblages. Alternatively, dissimilarity 

 values were examined to identify species that were 

 characteristic of a particular assemblage. Species that 

 have high average dissimilarity values and low stan- 

 dard deviations are those that contribute consistently 

 to samples within their group, with the result that they 

 can be used to distinguish between groups. 



The relationship between environmental variables 

 and fish community structure was examined by us- 

 ing the BIO-ENV procedure. A Spearman rank cor- 

 relation test was used to compare ranked values 

 from the aforementioned biota similarity matrix to 

 ranked values from an environmental similarity ma- 

 trix, which was created from environmental variables 

 measured in this study. Comparisons were based on 

 normalized Euclidean distance. The environmental 

 variables used to create the environmental similar- 

 ity matrix included pH, water temperature, salinity, 

 water depth, and Secchi depth. Dissolved oxygen was 

 strongly correlated with water temperature and was 

 therefore not included in the analysis because it would 

 produce results similar to those produced by water 

 temperature. Abiotic variables were standardized by 

 subtracting each mean and dividing by the standard 

 error to remove any bias associated with the different 

 measurement scales. 



The influence of recruitment of YOY fishes in defining 

 fish assemblages was examined. By relating increases 

 in abundance of species that were identified to be im- 

 portant contributors through the SIMPER procedure to 

 decreases in their average length in both habitats, we 

 were able to identify the timing of juvenile recruitment. 

 Length-frequency distributions showed that the seine 

 continued to catch larger individuals and therefore the 

 decrease in average length was not due to a decrease 



