96 
Fishery Bulletin 106(1 ) 
prevailing ocean conditions, but duration of tows never 
exceeded two minutes in an attempt to ensure that fish 
were collected from discrete habitat types. For each 
tow, a random selection of up to 20 individuals of each 
species were measured to the nearest millimeter fork 
length (FL) or total length (TL), and the remainder were 
counted. Beginning in 2001, bottom and surface salinity, 
temperature, dissolved oxygen, and pH were measured 
with a YSI model 85 handheld dissolved oxygen and 
conductivity instrument (Yellow Springs Instruments, 
Yellow Springs, OH) and recorded after the first tow. 
Depth and bottom topography were determined with a 
Furuno model 256 video depth recorder (Furuno, Hyogo, 
Japan). Water transparency was measured with a Secchi 
disk at each station. 
Habitat characteristics 
To assess the importance of habitat complexity to the 
structure of fish assemblages, the substrate was char- 
acterized qualitatively. Because Beach Haven Ridge and 
its vicinity have been intensively studied in the past, it 
was possible to identify habitat characteristics for each 
station from previous research where SCUBA, submers- 
ible, and remotely operated vehicle, or sidescan sonar 
were used, or where structural samples of the habitat 
were collected during trawl sampling. Additionally, any 
benthic material (clay and silt clods, sea stars, sand 
dollars, algae, shell hash, Diopatra tubes, Asabellides 
mounds) retained by the 2006 otter trawl sampling was 
categorized and quantified to evaluate if any changes in 
substrate had occurred over time. Stations were assigned 
an index of complexity of 1, 2, or 3 based upon the 
type and amount of substrate and macroalgae or other 
structural components present, with 1 being the least 
complex (bare sand) and 3 being the most complex (one 
or more substrate types present with multiple biogenic 
components). Habitat complexity was lowest inshore, 
peaked on the sides of the ridge, and was of an interme- 
diate value offshore (Table 1). Stations BHR-5, BHR-6, 
and BHR-7 had multiple substrate types and structural 
components and were given an index value of 3. Station 
BHRTOP and the inlet stations were all dominated by 
bare sand substrates and had no structures and were 
assigned a value of 1. The offshore stations varied in 
their substrates but had some complexity owing to bio- 
genic features and were assigned a value of 2. In addi- 
tion, the distance from a sampling location to the station 
located at the top of the ridge (BHRTOP), as well as the 
distance from a sampling location to the shore, were 
determined by using ArcGIS (Environmental Systems 
Research Institute, Inc., Redlands, CA). 
Data analyses 
A number of univariate and multivariate techniques 
were used to calculate population measures and assem- 
blage structure. Catch per unit of effort (CPUE), or the 
number of fish captured per tow, was determined for 
each species at each station. Significant differences in 
patterns of CPUE among stations were tested by using 
ANOVA procedures. CPUE data were log transformed 
before analysis to correct for heterogeneity of variance 
(Underwood, 1997). Mean species richness per unit of 
effort (RPUE), or the number of different species caught 
in a tow, was also calculated for each sampling sta- 
tion. ANOVA and Tukey multiple comparison tests were 
used to assess the differences in mean species richness 
(RPUE) between stations. Frequency of occurrence was 
calculated for each species across tows at each sampling 
station. All univariate statistics were performed with 
SAS (vers. 9.1, SAS Inst., Inc., Cary, NC). 
To determine the structure of the assemblages in the 
different habitats, two related ordination techniques 
were employed. Canonical correspondence analysis 
(CCA) is a constrained ordination technique in which 
the sample scores are constrained to be linear combina- 
tions of the explanatory variables (Van den Brink and 
Ter Braak, 1999). This technique is one of the most 
widely used gradient analysis tools in ecology because 
of its capacity to handle highly skewed species distribu- 
tions, high noise levels, complex sampling designs, and 
the fact that it does not create an artificial arch effect 
(Palmer, 1993). Canonical correspondence analysis was 
performed on a subset of the overall data matrix for 
which environmental information was available for all 
stations. For the beam trawl data this “subset” was the 
entire data set, whereas for the otter trawl samples, the 
subset was limited to Beach Haven Ridge from 2001 to 
2006. The data were arranged in a species-by-sample 
matrix, where the samples were the combination of all 
tows for a given location and date, and the CPUE rep- 
resented the value fields. The data were log (CPUE+ 1) 
transformed to reduce the influence of abundant spe- 
cies. Any species whose abundance did not exceed five 
percent from at least one station was removed from the 
matrix. 
Because CCA orders species only along gradients 
of the measured environmental variables, it may not, 
depending on the environmental variables available, 
be representative of the assemblages encountered. 
Therefore, the CCA was checked for bias by using cor- 
respondence analysis (CA), an unconstrained ordination 
method (McGarigal et ah, 2000). The species-by-sample 
matrices were treated in the same manner as in the 
CCA. 
Results 
Environmental gradients 
Summer temperature tended to decrease with increas- 
ing distance from the shoreline, and depth, salinity, and 
water transparency generally increased with increasing 
distance (Fig. 2). Average station depth increased with 
increasing distance from the shoreline (range: 2.8-19.9 
m), with the exception of the station located on the top of 
the ridge (BHRTOP) (Table 1). Within the Beach Haven 
Ridge transect there was a change in depth from BHR-4 
