Multivariate analysis of habitat selection by rorqual whales PDF

Multivariate analysis of habitat selection by rorqual whales
RATIONALE
• The Gulf of St Lawrence (GSL) is a rich summer feeding ground
for many North Atlantic whale species
• Rorquals in the GSL face several anthropogenic threats (e.g.,
entanglement in fishing gear, ship strikes, chemical pollution)
• Fine-scale information on habitat use is needed for better
management decisions
• This is also a rare opportunity to study how four sympatric,
closely related species of whales with overlapping diet preferences
partition their habitat
OBJECTIVES
• Quantify the influence of environmental variables on whale
habitat selection in the northern GSL
• Propose and test prediction models of habitat use
• Study niche overlap and resource partitioning of the 4 rorqual
species
DATA COLLECTION
• Boat surveys (1996-2002); GPS locations of whale sightings
• Computer ocean model for environmental variables (horizontal
resolution 5km, vertical 5m, time 30 min)
Environmental variables:
Distance to shore
Salinity (10 m)
Temperature (10 m)
Temperature (bottom)
Depth
Slope
Horizontal & Vertical currents
Effort = 10,338 km / year
Blue = 84 sightings
Mingan
Finback = 2065
Humpback = 1196
Minke = 1886
b
Berteaux
& Richard
ANALYSIS
• Random points plotted to represent available habitat
• Discriminant Function Analysis (DFA) to find which variables
best separated used from available habitat  Resource Selection
Functions (RSFs)
• RSFs were tested on a previously selected part of the dataset to
assess successful reclassification rate and prediction power
• DFA to identify variables that best discriminate among species
HABITAT SELECTION (Fig.1)
• Strong selection towards some combinations of variables that
differ from available habitat
• Humpback and finback habitats were best described by static
factors but models had poor prediction power
• Blue and minke habitats were best predicted by a combination of
static and dynamic factors  better prediction power
c
Sears
Humpback
Minke
minke
blue
finback
humpback
centroids
Function 2 (can.corr.=0.28)
warmer, less saline
Thomas
a,b,c
Doniol-Valcroze , Dominique
Finback
Function 1 (can.corr.=0.44)
deeper, flatter, less current, farther from shore
NICHE COMPARISON (Fig.2)
• Minke whales preferred shallow waters close to shore with
strong currents; these conditions might help them hunt for fish or
represent habitat with less competition from bigger species
• Humpback & finback used deeper, steep habitats located
relatively close to shore that could correspond to the edges of
underwater banks and of the Anticosti shelf
• Blue whale were found in deeper, farther and colder areas that
could correspond to local upwelling plumes
• These patterns could be explained in part by differences in diet,
body size and social structures
• Humpback & finback use extremely similar habitat  they might
partition their resource use at a finer scale in space/time to avoid
competition
Figure 2. Scatterplot of DF scores for sightings of four rorqual species
CONCLUSIONS
• Importance of dynamic environmental variables for blue and
minke whales
• Clear habitat specialization for blue and minke whales but strong
competition potential for humpback and finback whales
• The endangered blue whale appears to have the most specialized
habitat preferences  better prediction model but more
vulnerable to environment change?
• Other models have fairly low prediction power and remain very
area-specific, indicating that other factors should be taken in
consideration (social dynamics?)
ACKNOWLEDGMENTS
F. Saucier for the ocean model; MICS team members for data
collection and all-weather fun times; M. Humphries & WEEL for
support and advice; J.W. McConnell Foundation for financial support
Blue whale drawing: Daniel Grenier
DF score (+/- 1 SE)
0
-1
-2
SPECIES • DIET • GROUP SIZE
blue whale (Balaenoptera musculus) • only krill • solitary
finback whale (B. physalus) • krill & fish • form groups
humpback whale (Megaptera novaeangliae) • krill & fish • form groups
minke whale (B. acutorostrata) • krill & (mostly?) fish • solitary
1
0
-1
-2
random
blue
p<0.001 • can. corr.=0.69
correct classification: 87%
2
shallower, steeper,
colder, stronger currents
1
2
deeper, steeper, closer to shore
blue whale
deeper, colder, more saline
random point
2
deeper, steeper, closer to shore
2
Anticosti Island
Blue
1
0
-1
finback
p<0.001 • can. corr.=0.26
correct classification: 63%
AFFILIATIONS
a. Department of Natural Resource Sciences, Macdonald Campus,
McGill University
b. Canada Research Chair in Conservation of Northern Ecosystems,
Université du Québec à Rimouski
c. Mingan Island Cetacean Study
0
-1
e-mail: [email protected]
-2
random
1
-2
random humpback
p<0.001 • can. corr.=0.21
correct classification: 58%
random minke
p<0.001 • can. corr.=0.33
correct classification: 71%
Figure 1. Discriminant function scores of random points and whale sightings;
arrows indicate direction of selection for principal habitat characteristics
Mingan Island Cetacean Study