Baumgartner et at: Cetacean habitats in the northern Gulf of Mexico 
229 
and sperm whales had distributions that extended from 
the upper continental slope to the deep Gulf. Mann-Whit- 
ney tests between Risso’s dolphins and Kogia spp. for 
each of the environmental variables indicated that only 
their distributions with respect to depth gradient ( {7=2.12, 
P<0.05) and zooplankton biomass ({7=1.69, P<0.05) were 
significantly different. Similar tests between pantropical 
spotted dolphins and sperm whales indicated that their 
distributions with respect to the depth of the 15°C iso- 
therm ({7=2.26, P<0.05) alone were significantly different. 
Differences between species were also detected with 
MANOVA and canonical linear discriminant function anal- 
ysis. Unfortunately, low sample size for both sea surface 
temperature and sea surface temperature variability pre- 
cluded their use in the multivariate analyses. Of the re- 
maining variables, the sample sizes for each species were 
as follows: bottlenose dolphins (n=18), Risso’s dolphins 
(n=35), Kogia spp. (n- 25), pantropical spotted dolphins 
(t?= 51 ), and sperm whales </? = 19). The null hypothesis of 
equal mean vectors was rejected in the MANOVA (Wilks’ 
A=0.446, PcO.OOOl). The first two canonical variables in 
the canonical LDF analysis accounted for 94.5% of the 
total variability, and likelihood ratio tests indicated that 
only the first two canonical variables were significant 
(P<0.01 for each). The structure correlations indicated that 
low depths and high zooplankton biomass were associated 
with positive values of the first canonical variable, where- 
as shallow occurrences of the depth of the 15°C isotherm 
and low surface chlorophyll concentration were associated 
with positive values of the second canonical variable (Fig. 
6A). The separation between groups along canonical axis 
1 supports the importance of depth in habitat partition- 
ing in the northern Gulf of Mexico. The significance of zoo- 
plankton biomass in this first canonical variable was due 
to the inclusion of the bottlenose dolphin in the analysis 
and the presence of high zooplankton biomass on the con- 
tinental shelf. Note that the bottlenose dolphin was clearly 
separated from the other species along canonical axis 1 
and the sperm whale was separated from the other species 
on both canonical axes. 
Because inclusion of the bottlenose dolphin strongly 
influenced the results of the multivariate analysis, a sec- 
ond analysis was conducted for just the oceanic species. 
The null hypothesis of equal mean vectors was rejected 
M t 1.0 
< O 0.5 
If 0.0 
§ I -°' 5 
55 - 1.0 
DP D15C PL 
DPG CHL 
04 
CO t 
x o 
< o 
1.0 
0.5 
0.0 
-0.5 
- 1.0 
TT E " ir g-r 
DP D15C PL 
DPG CHL 
m t 10 
3 o 
< o 0.5 
S I 0.0 
§ 1 -0-5 
.™ 55 - 1 .0 
DP D15C PL 
DPG CHL 
1.0 
0.5 
P - 0 0 
< o 
-S <u 
o 1 -0.5 \ 
^ -1 0 j 
DP D15C PL 
DPG CHL 
2 
B 
1 1 
• 
^ K 
-2 -1 0 1 2 3 _2 -1 o 1 2 
Canonical axis 1 
Figure 6 
Means and interquartile ranges (error bars) of the canonical linear discriminant function variables for (A) all species and 
(B> all species except the bottlenose dolphin. The structure correlations associated with each canonical axis represent the 
approximate correlations between the canonical variables and depth (DP), depth gradient (DPG), depth of the 15°C isotherm 
(D15C), surface chlorophyll concentration (CHL), and epipelagic zooplankton biomass (PL). Species abbreviations are the 
same as those shown in Figure 5. 
