332 
Fishery Bulletin 1 13(3) 
Crewe, UK) interfaced with a PDZ Europa 20-20 iso- 
tope ratio mass spectrometer (Sercon, Ltd.). 
Data analyses 
Field data Data on water quality, reef structure, and 
organism abundance (dependent variables) were ana- 
lyzed separately by season. For all analyses, a sig- 
nificance level of alpha=0.05 was used, and results 
are presented as means with standard errors. Unless 
otherwise indicated, SAS software, vers. 9.2 (SAS In- 
stitute, Inc., Cary, NC) with the GLIMMIX procedure 
was used for all analyses. Data for water quality, reef 
structure, and resident oyster reef community were 
analyzed by using separate generalized linear mixed 
models (GLMMs) to test for the effects of harvesting 
(independent variable), with station used as a nested 
random effect to remove effects of station variation. 
Bottom DO, salinity, temperature, chlorophyll-a, TPM, 
water depth, and index of integrity were examined by 
harvesting treatment (actively harvested and nonhar- 
vested) with GLMMs that used station as a nested 
random effect to remove effects of station variation. 
To examine differences in depth and index of integrity, 
GLMMs were run with a normal distribution. Densi- 
ties of market- and seed-size oysters, mussel density, 
volume of loose shells, and volume of shell clusters 
were also analyzed by harvest treatment with station 
as a nested random effect; in addition, a negative bi- 
nomial distribution with a log-link function was used 
to account for overdispersion. Significant results for 
water quality and reef structure parameters were de- 
termined with a type-III test of fixed effects. Vertical 
relief for the 2 harvesting treatments was compared 
with a 2-sample /-test. Specifically, vertical relief, the 
difference between the 2 extreme depth measures, was 
calculated at each station, for 3 stations per site (N=6 
[3 stationsx2 sites]). For resident oyster reef communi- 
ties, GLMMs with a negative binomial distribution and 
a log-link function to account for overdispersion were 
run on common species density (species representing 
>1% of total abundance), invertebrate density, fish den- 
sity, total nekton density (fishes and invertebrates com- 
bined), and total number of species. Significant results 
for resident community parameters were determined 
with a type-III test of fixed effects. 
For examination of species— environment relation- 
ships, canonical correspondence analysis (CCA) was 
performed with CANOCO software, vers. 4.5 (Wa- 
geningen UR, Netherlands; ter Braak and Smilauer, 
2002) to analyze the relationship between abun- 
dances of common resident species and environmen- 
tal variables (water quality and reef structure), by 
combining all summer and fall catches. Summer and 
fall catches were combined to increase the number 
of samples per species and to focus on species-envi- 
ronment relationships that held true, regardless of 
season. The number of environmental variables was 
reduced by using backward selection, sequentially re- 
moving the least influential variable until 4 variables 
remained. Species abundances were logU +1) trans- 
formed for the CCA to improve normality. A Monte 
Carlo simulation test was used to determine statisti- 
cal significance of canonical axes with 1000 simula- 
tions on the full model. 
Stable isotope data Isotope data were analyzed by sea- 
son. Isotope values of 8 15 N to 8 13 C were used to deter- 
mine contributions of basal food sources (BFSs; marsh 
plant and FPOM) and consumer trophic positions (de- 
pendent variables). Contributions of BFSs to naked 
goby, freckled blenny, skilletfish, grass shrimp, flatback 
mud crabs, eastern oysters, and CPOM were deter- 
mined for each site by using a 2-source mixing model 
(Fry, 2006), with the mean S 13 C values of dominant 
marsh plants and FPOM from each site. Trophic posi- 
tion (TP) was determined with the following equation: 
TP = 1 + (8 iri Norganism — 8 15 Ng ase )/T'F)F 1 (Post, 2002), 
where a trophic enrichment factor (TEF) of 2.54% was 
used (Vanderklift and Ponsard, 2003; Caut et ah, 2009). 
Separate 2 sample /-tests were used to test for dif- 
ferences between harvest treatments (independent 
variable) for the trophic position and BFS contributions 
of dominant species (Post, 2002; Layman et ah, 2007). 
The convex hull area (the smallest area that incorpo- 
rates all isotope biplot points for individual species or 
communities) were calculated and used as a means to 
represent the trophic diversity within a food web (Lay- 
man et al., 2007), but they were not statistically tested 
because there was only one set of convex hull areas 
per site (no replication). Convex hull areas were con- 
structed with the convex hull option in the XTools Pro 
toolbar in ArcMap, the central application of ArcGIS, 
vers. 9.3.1 (Esri, Redlands, CA). Data for harvest treat- 
ments that were not normally distributed were com- 
pared with a nonparametric Wilcoxon rank-sum test. 
Results 
Field data 
Water quality In the summer, there were no differ- 
ences in temperature, salinity, and DO between har- 
vested and nonharvested treatments (Table 1). Levels 
of TPM and chlorophyll-a were higher at the harvested 
site, Sister Lake (44.2 mg/L [SE 3.2] and 18.0 pg/L [SE 
2.0]), than at the nonharvested site, Sabine Lake (17.9 
mg/L [SE 3.5] and 8.2 pg/L [SE 0.4]). 
In the fall, there were no differences in temperature 
or DO between harvest treatments (Table 1). Chloro- 
phyll-a levels were higher at the harvested site, south- 
ern Calcasieu Lake (16.9 pg/L [SE 0.4]), than at the 
nonharvested site, northern Calcasieu Lake (8.6 pg/L 
[SE 0.3] ). Salinity was significantly higher at the non- 
harvested site (20.0 [SE 0.2]) than at the harvested 
site (19.2 [SE 0.2]), although the difference was prob- 
ably not ecologically important. 
