Baumgartner et al.: Cetacean habitats in the northern Gulf of Mexico 
223 
30°N 
98 96 94 92 90 88 86 84 82°W 
Figure 2 
Sea surface temperature of the northern Gulf of Mexico derived from remotely sensed AVHRR data collected between 
21 and 23 May 1993. Image is a histogram-equalized, warmest-pixel composite of data derived from three satellite 
passes with some cloud contamination south of 24.5°N and also west of 93°W. CTD and XBT stations are indicated 
as filled circles and the contours represent the depth of the 15°C isotherm computed from the CTD and XBT casts 
collected between 19 May and 1 June 1993. The line along 27°N indicates the parallel from which data were extracted 
for Figure 3. The 200-m isobath is shown. 
26 
32 ' 
30. 
26 1 
24 8 
03 
22 3 
20 g 
co 
with mesoscale oceanographic features is much larger 
than these errors and therefore the interpolated fields rep- 
resent these features reasonably well (e.g. Fig. 2). 
The base unit of effort for this study was defined as 1 km 
of actively surveyed transect during adequate sighting 
conditions. To conform to this definition, each contiguous 
transect in the effort data set was broken into 1-km linear 
sections and all the environmental variables measured 
along each 1-km section were averaged. This provided a 
single set of observed environmental variables for each 
unit of effort. Only those 1-km sections that were actively 
surveyed (i.e. those where the observers were on-effort) 
during adequate sighting conditions (defined as Beaufort 
sea states of 3 or less) were used for analysis. Similarly, 
only those cetacean sightings that occurred while observ- 
ers were on-effort and in Beaufort sea states of 3 or less 
were used for analysis. All of the following analyses were 
conducted on cetacean group sightings and therefore do 
not account for group size. 
Some portions of the described data have been previ- 
ously published by Davis et al. (1998) and Baumgartner 
(1997). Davis et al. (1998) examined cetacean habitat in 
the northwestern Gulf of Mexico with respect to a variety 
of physical oceanographic and physiographic variables. We 
have included the sighting data and some of the environ- 
mental data from that study here (less than 40% of our 
total data set) to examine cetacean habitat throughout 
the entire northern Gulf of Mexico with an expanded set 
of environmental variables and new statistical analyses. 
With regard to Risso’s dolphin habitat, we have used the 
same sighting, depth, and depth gradient data presented 
in Baumgartner (1997). To these, we have added physical 
and biological oceanographic variables to test and extend 
the conclusions of Baumgartner ( 1997) and to strengthen 
the univariate and multivariate interspecies comparisons 
described below. 
Analytical methods 
The analysis of the sighting and effort data sets was 
conducted in two parts: 1) univariate and multivariate 
interspecies comparisons of the environmental variables 
measured at each cetacean sighting and 2) comparisons 
of each species’ distribution with respect to the environ- 
mental variables to that of the effort. The former analysis 
examined the null hypothesis that each species had simi- 
lar distributions with respect to each of the environmen- 
tal variables. This was tested with Mood’s median test 
(Conover, 1980) and the Kruskal-Wallis test (Sokal and 
Rohlf, 1981) as nonparametric substitutes for a one-way 
analysis of variance. Multivariate analysis of variance 
(MANOVA) and canonical linear discriminant function 
(LDF) analysis (Huberty, 1994; Johnson, 1998) with rank- 
transformed environmental variables were used to further 
examine interspecies differences. These analyses were con- 
ducted with the CANDISC procedure of the Statistical 
Analysis System (SAS, 1989), version 6.12. The MANOVA 
detects species group differences in multivariate space and 
