Drymon et al.: Factors that affect the distribution of sharks throughout the northern Gulf of Mexico 
373 
on species and size of a shark at capture (Kohler and 
Turner, 2001). 
Additional data sets 
To determine whether the patterns that characterize 
the shark community assemblage in our study region 
(coasts of Alabama and Mississippi, hereafter referred 
to as small scale ) were applicable across the northern 
Gulf of Mexico (hereafter called large scale), we ob- 
tained bottom-longline data from the SEFSC Mississip- 
pi Laboratories. This information included catch, fork 
length, and environmental data collected across Gulf of 
Mexico statistical zones 4-21 (Fig. 2A) in 2006-09. Dur- 
ing that period bottom-longline sets were conducted by 
the Mississippi Laboratories in August and September. 
The methods used for the bottom-longline survey were 
identical at the small and large scales, and a complete 
description of these methods is provided in Driggers et 
al. (2008). 
To examine factors that potentially influence the 
distribution of sharks on both small and large scales, 
we analyzed the relationships between longline shark 
data and a set of environmental factors, including trawl 
data and abiotic parameters. Biotic trawl data were 
obtained from the Southeast Area Monitoring and As- 
sessment Program (SEAMAP) database (http://seamap. 
gsmfc.org, November 2010) of the Gulf States Marine 
Fisheries Commission. We restricted our analysis of 
SEAMAP data to those years for which trawling was 
conducted across the entire northern Gulf of Mexico 
(2007-09). The data from those years that were used 
in our analysis originated from both state (Louisiana, 
Alabama, Mississippi, and Florida) and federal (NOAA 
Fisheries) regulatory agencies. All data archived in the 
SEAMAP trawl database were collected according to 
standard SEAMAP trawl protocols (Rester, 2012). Two 
biotic variables, representative of the availability of 
potential prey for sharks, were selected for inclusion 
in our analysis of SEAMAP trawl data: fish biomass 
and crustacean biomass per station in kilograms. All 
biomass data from the trawl data set were standard- 
ized to kilogram per minute. The abiotic variables tem- 
perature (degrees Celsius), salinity (practical salinity 
unit), dissolved oxygen (milligrams per liter) and depth 
(meters) were collected with conductivity, temperature, 
and depth (CTD) instruments (SBE 911p/zzs and SBE 
25 plus Sealogger, Sea-Bird Electronics, Inc., Bellevue, 
WA) during bottom-longline sampling at both the large 
and small scales. 
To include a proxy for primary production in our 
analysis, we used data on chlorophyll-a (chl-a) concen- 
tration as a measure of phytoplankton biomass (Can- 
ion, 2008; Martinez-Lopez and Zavala-Hidalgo, 2009). 
The satellite-based ocean color data used in this study 
were derived from the moderate resolution imaging 
spectroradiometer (MODIS) on the Aqua satellite (for 
a detailed sensor description go to the MODIS mis- 
sion website at http://modis.gsfc.nasa.gov). The data 
on chl-a concentration used for analyses in our study 
were downloaded from the Ocean Color website (http:// 
oceancolor.gsfc.nasa.gov, accessed April 2012). For this 
study, annual binned level-3 chl-a data (Campbell et 
al., 1995) at a spatial resolution of 4 km were used 
from 2006 to 2009. The annual composites are produced 
by averaging all valid, cloud-free acquisitions for each 
ocean pixel. The valid pixels are determined by using 
an extensive quality control process that tests for nu- 
merous factors known to degrade data accuracy. Addi- 
tional details for the level-3 chl-a data can be found at 
http://modis.gsfc.nasa.gov/data/atbd/index.php. Despite 
that extensive quality control process, the optically 
complex nature of the coastal zone can still present dif- 
ficulties for ocean color algorithms. In the case of data 
on chl-a concentration, algorithms are known to over- 
estimate concentrations in coastal zones, particularly 
in regions that are influenced by a river, because of 
estuarine materials, such as suspended sediment and 
concentrations of dissolved organic material. However, 
this phenomenon occurs primarily at depths <10 m 
(Martinez-Lopez and Zavala-Hidalgo, 2009); therefore, 
we obtained data on chl-a concentration from the 25-m 
isobath to limit the effect of these degrading influences. 
Data analyses 
Bottom-longline data sets were limited to those spe- 
cies observed in both the small- and large-scale bot- 
tom-longline surveys. Data of catch per unit of effort 
(CPUE), measured as sharks 100 hooks -1 h -1 , were 
iog(x+l)-transformed to reduce the influence of the 
most common species and to standardize the data (Leg- 
endre and Legendre, 1998). All sets, including those 
with zero catches, were included in our analyses. Mean 
CPUE data were then analyzed as a function of block 
(blocks 1-8) across the small scale (Fig. 1A) and as a 
function of statistical zone (zones 4-21, minus zone 12) 
across the large scale (Fig. 2A). 
A centered principal component analysis (PCA) was 
performed on mean transformed shark CPUE data for 
both small- and large-scale data. The data collected 
with the CTD instruments and the MODIS satellite 
data (collectively hereafter referred to as environmen- 
tal data) at both small and large scales were analyzed 
with a normed PCA. At each spatial scale, centered 
(for shark CPUE data) and normed (for environmental 
data) PCAs allowed for the identification of the major 
spatial patterns that characterize shark assemblages 
and environmental conditions and for the visualiza- 
tion of covariances between shark species and of cor- 
relations between environmental factors (Legendre and 
Legendre, 1998). 
The relationships between values of shark CPUE 
and environmental factors were then analyzed at each 
spatial scale with co-inertia analyses (COIA). Co-in- 
ertia analysis is a flexible, multivariate method that 
couples environmental and faunal data and measures 
the degree of agreement between them (Doledec and 
