Grizzle et at: Effects of a large fishing closure on benthic communities in the western Gulf of Maine 
311 
images (n=13 to 94 per transect) along the length of 
each transect was produced. Each still image was ana- 
lyzed for bottom characteristics (predominant sediment 
type), visible burrow characteristics (size, density), and 
epifauna (taxa, density). A trio of laser beams fixed at 
known distances apart allowed the total area of each 
image to be determined. 
Data analysis 
The major focus of this study was to determine the effects 
of the WGOM closure by comparing samples taken inside 
the closure with samples taken outside the closure. Thus, 
potential confounding factors (e.g., sediment type, water 
depth, and sediment organic content) that might have 
affected comparisons between samples taken inside the 
closure with those taken outside (hereafter referred to 
as “in vs. out” comparisons) were assessed. Also, for the 
infauna analyses, data from the two sampling devices 
(grab and box corer) were analyzed separately to avoid 
complications with different sample sizes and selectivity 
of the gear. 
Infauna (sampled by grab and box corer) 
For both sampling devices, the role of sediment grain 
size (mud, sand, or gravel), water depth, sediment 
organic content (% LOI), and relative penetration of the 
sampling device were assessed, along with the factor 
of primary interest: whether the sample was taken 
from inside or outside the closure. Separate analyses 
were performed to examine each of the three dependent 
univariate infauna community variables: 1) density, 2) 
biomass, and 3) taxonomic richness. 
All analyses were performed by using generalized 
linear models (GLM; McCullagh and Nelder, 1989) 
in S-PLUS 7.0 (Insightful Corporation, Seattle, WA). 
For normally distributed data or data that could be 
transformed to approximate normality, significance for 
main effects and interactions were examined by using 
a combination of forward and backward model selection 
based on the C p statistic and sequential F-tests in an 
analysis of deviance with alpha levels of 0.05 and 0.10, 
respectively. For non-normally distributed data, signifi- 
cance for main effects and interactions were examined 
by using sequential chi-square tests in an analysis of 
deviance with alpha levels of 0.05 and 0.10, respectively. 
For each model, the inclusion of either Gaussian or 
exponential spatial correlation was examined with the 
extended generalized linear modeling capabilities in the 
S-PLUS correlated data library. 
Epifauna (documented along video transects) 
A generalized linear mixed-effects (GLME) model with 
Poisson error and log link was used to analyze the video 
count data of epifauna taxa and density (Breslow and 
Clayton, 1993). The mixed-effects portion of the model 
was necessary to account for multiple images within 
each transect and the images that may have been cor- 
related. The use of GLME allowed individual transects 
to differ from one another for reasons unaccounted for 
by the data and directly accounted for the repeated- 
measures nature of the data. S-PLUS 7.0 and the GLME 
extension from the S+ Correlated Data library (vers. 1.0, 
release 1) were used for all analyses. The significance 
of each fixed-effect, both main effects and interactions, 
was tested in an ANOVA framework by using mar- 
ginal F-tests (Pinheiro and Bates, 2000) based on the 
(restricted) penalized quasi-likelihood with an alpha- 
level of 0.05 for main effects and 0.10 for interaction 
effects (Sokal and Rohlf, 1981). In addition to the main 
effect of inside or outside of the closure, water depth and 
sediment type were assessed as potential confounding 
factors. Offsets were included in each model to account 
for the fact that the total surface area observed varied 
from image to image because of the variations in the 
height of the camera. Separate analyses were performed 
on the two dependent variables: total density and taxo- 
nomic richness. 
Results 
Preclosure and postclosure plots of the report data from 
fishing vessel trips showed several relevant patterns 
(Fig. 3). First, both preclosure and postclosure trip 
data verified the general expected pattern that trawls 
are mainly used on soft sediments in deeper water 
(greater than 60 m), and gillnets are used mainly on 
rocky bottoms in shallower areas (less than 60 m). These 
patterns indicate that a major effect of the closure was 
the removal of trawl impacts from the deeper, mainly 
finer sediments, and the removal of gillnet impacts 
from rocky areas along the top and southeast flank of 
Jeffreys Ledge. 
The report data indicated that total gillnet fishing 
intensity in the overall 515-km 2 study area was simi- 
lar before closure (2056 trips) and after closure (1812 
trips); however, gillnet intensity nearly doubled in the 
area outside of the closure after closure (761 trips be- 
fore closure compared to 1494 trips after closure). This 
pattern indicates that any data interpreted as showing 
recovery of benthic communities in rocky areas where 
gillnets were the major gear type needs to be tempered 
because of increased postclosure gillnet fishing intensity 
in the “control” area outside the closure. 
In contrast, trawling intensity in the overall 515-km 2 
study area decreased from 1103 trips before closure to 
581 after closure (Fig. 3). There was a 39% decrease 
for trawl intensity outside the closure, from 894 trips 
before closure to 544 trips after closure. Moreover, the 
only portion of the study area inside the closure that 
was likely strongly affected by trawls before the closure 
was the deeper area north of Jeffreys Ledge. This infor- 
mation indicates that any data interpreted as showing 
recovery of benthic communities in soft sediment areas 
needs to be tempered because of the decreased trawl- 
ing intensity after closure in the control area outside 
the closure. 
