224 
Fishery Bulletin 99(2) 
the canonical LDF analysis describes which environmen- 
tal factors contribute most to these group differences. The 
canonical LDF analysis is accomplished by finding a linear 
combination of the environmental variables that best dis- 
criminates between the species groups. These linear com- 
binations (canonical variables) are then examined by using 
the LDF structure correlations (Fluberty, 1994) to assess 
their ecological meaning and significance. The structure 
correlations are essentially the correlations between the 
canonical variables and the original environmental vari- 
ables and their interpretation is analogous to the interpre- 
tation of factor loadings in factor analysis. 
The second analysis uses univariate and bivariate chi- 
squared (^ 2 ) tests, Mann- Whitney tests, Monte Carlo tests, 
and equal-effort sighting rate distribution plots to deter- 
mine the specific relationships between the distribution of 
each species and each of the environmental variables. For 
the x 2 analysis, the effort data were used to compute ex- 
pected uniform distributions for each species with respect 
to the individual environmental variables. Classes were 
chosen such that each contained an equal amount of effort 
(Kendall and Stuart, 1967). This approach "normalized” the 
sighting rates by creating class sizes of equal sighting prob- 
ability based on the effort and guaranteed that the anal- 
ysis would not be distorted by classes with exceptionally 
low or high amounts of effort. For a complete description of 
the methods used to compute the uniform distribution, see 
Baumgartner (1997). The actual distributions were then 
compared with the predicted uniform distributions by us- 
ing the x 2 statistic. Equal-effort sighting rate distribution 
plots were constructed directly from the contingency tables 
used in the x 2 analyses. In some cases, the sample size was 
lower than the minimum required for a conservative x 2 test 
( 72 =25 ), therefore the species’ and effort distributions were 
compared by using a Mann-Whitney test. 
Of the five species examined here, each had a distribu- 
tion with respect to depth that was significantly different 
from a uniform distribution. Further analyses with Monte 
Carlo (randomization) tests were conducted to determine if 
the distribution of a particular species with respect to the 
other environmental variables was an artifact of that spe- 
cies’ distribution with depth. For example, consider a hypo- 
thetical species that is only found on the continental shelf. 
The continental shelf in the northern Gulf of Mexico is char- 
acterized by low depth gradients, whereas the continental 
slope has high depth gradients and the abyssal plains of the 
deep Gulf have low depth gradients. Because this species 
occurs on the continental shelf, it would have distributions 
with respect to both depth and depth gradient that were 
significantly different from a uniform distribution. Howev- 
er, this species’ distribution with respect to depth gradient 
is merely an artifact of its distribution with respect to depth 
because of a correspondence between shallow depths and 
low depth gradients over the continental shelf. 
The Monte Carlo tests consisted of randomly choosing 
n transect sections from the effort data set that had the 
same depth distribution as the n sightings of the species 
of interest. These transect sections represent n “virtual” 
cetacean sightings that have the same depth distribution 
as the species of interest but have a random distribution 
with respect to all of the other environmental variables. 
A x 2 analysis was then conducted to determine if the dis- 
tribution of the “virtual” sightings with respect to the par- 
ticular environmental variable of interest (e.g. depth gra- 
dient in the example above) was different from a uniform 
distribution predicted by the effort. The process of choos- 
ing ?i “virtual” sightings and of conducting the x 2 analysis 
was performed 10,000 times. The proportion of the result- 
ing 10,000 x 2 statistics that exceeded the x 2 statistic as- 
sociated with the species’ actual distribution with respect 
to the environmental variable of interest was considered 
a P-value. This P-value represented the probability that 
the actual x 2 statistic could have been observed by chance 
and was used to test the null hypothesis that the species’ 
distribution with respect to the environmental variable of 
interest was the same as a uniform distribution given its 
distribution with respect to depth. 
Results 
NOAA Ship Oi egon II completed 113 days of effort during 
the spring surveys from 1992 to 1994 and sampled the 
entire oceanic northern Gulf of Mexico once each year. 
A total of 9101 1-km transect sections (units of effort) 
were completed during adequate sighting conditions. The 
amount of environmental data available for each transect 
section was dependent on survey design, on instrument 
availability and performance, and, in the case of the re- 
motely sensed sea surface temperature variability, on sat- 
ellite orbital parameters and cloud conditions (Table 2). 
The Loop Current penetrated into the eastern Gulf to 
at least 27.5°N during each of the surveys and warm-core 
eddies could usually be found in the central and western 
Gulf (Fargion et al. 10 ). Both the Loop Current and the 
warm-core eddies were often accompanied by cold-core fea- 
tures at their peripheries. Examples of the major oceano- 
graphic features of the northern Gulf are shown in the 
composite AVHRR sea surface temperature image and the 
contoured depth of the 15°C isotherm (Fig. 2). The Loop 
Current is easily identifiable as the broad region in the 
eastern Gulf where the 15°C isotherm was at depths be- 
low 250 to 300 m and sea surface temperatures reached 
a local maximum. The remnants of a warm-core eddy 
(Eddy V) are evident in the northwestern Gulf centered 
at about 27.0°N, 95.5°W (Jockens et al., 1994; Fargion 
et al. 10 ). Warm-core features like the Loop Current were 
characterized by depressed isotherms and were often ac- 
companied by warm surface temperatures and low zoo- 
plankton biomass (Fig. 3). Surface temperature gradients 
were high at the edge of these mesoscale features when 
10 Fargion, G. S.. L. N. May, T. D. Leming, and C. Schroeder. 1996. 
Oceanographic surveys. In Distribution and abundance of 
cetaceans in the north-central and western Gulf of Mexico: 
final report, vol.I I: technical report (R.W. Davis and G.S. Far- 
gion, eds.), p. 207-269. U.S. Department of the Interior, Miner- 
als Management Service, OCS Study MMS 96-007. [Available 
from Public Information Office, MS 5034, Gulf of Mexico Region, 
Minerals Management Service, 1201 Elmwood Park Blvd., New 
Orleans, LA 70123-2394.1 
