Content uploaded by Rob Williams
Author content
All content in this area was uploaded by Rob Williams
Content may be subject to copyright.
Available via license: CC BY-NC 4.0
Content may be subject to copyright.
J. CETACEAN RES. MANAGE. 9(1):15–28, 2007 15
__________________________________________________________________________________
Distribution and abundance of marine mammals in the coastal
waters of British Columbia, Canada
ROB WILLIAMS* AND LEN THOMAS+
Contact ema il:r.williams@fisheries.ubc.ca
ABSTRACT
Information on animal distribution and abundance is integral to wildlife conservation and management. However abundance estimates have
not been available for many cetacean species inhabiting the coastal waters of Canada’s Pacific coast, including those species that were heavily
depleted by commercial whaling. Systematic sightings surveys were conducted in the inshore coastal waters of the Inside Passage, between the
British Columbia (BC) Washington and the BC-Alaska borders. A total of 4,400km (2,400 n.miles) of trackline were surveyed in the summers
of 2004 and 2005. Abundance estimates (with 95% confidence intervals) assuming certain trackline detection for seven cetacean species were
as follows: harbour porpoise, 9,120 (4,210-19,760); Dall’s porpoise, 4,910 (2,700-8,940); Pacific white-sided dolphin, 25,900 (12,900-
52,100); humpback whale, 1,310 (755-2,280); fin whale, 496 (201-1,220); common minke whale, 388 (222-680); and ‘northern resident’ killer
whale, 161 (45-574). The potential for responsive movement to have affected the accuracy and precision of these estimates is difficult to assess
in small-boat surveys. However, the analyses were designed to minimise this factor in the most obvious case (Pacific white-sided dolphins)
and pilot data collection has begun to assess the magnitude of the effect and to calculate correction factors for other species. The density of
harbour seals, both along the shoreline and at sea, was calculated and it was estimated that total abundance of harbour seals in the study area
was at least 19,400 (14,900-25,200). These are new abundance estimates for this region for all cetacean species except killer whales. The small
sample size makes the killer whale estimate tenuous, but one worth noting, as it is close to the known number of northern resident killer
whales (2004 census was 219 animals, Cetacean Research Program, Pacific Biological Station, Fisheries and Oceans Canada). The common
minke whale abundance estimate is similarly tentative, however the results do reveal that common minke whales were relatively rare in this
region. While the majority of harbour seals were found as expected in the southern straits and in the mainland inlets, a substantial number of
animals were on the north coast and in the Queen Charlotte Basin as well. These data provide a systematic snapshot of summertime
distribution and abundance of marine mammals in the Queen Charlotte Basin, where offshore oil and gas development and seismic surveys for
geophysical research have been proposed to take place. Similarly, the abundance estimates could be used to form the basis of a simulation
exercise to assess the sustainability of observed levels of incidental bycatch of small cetaceans in commercial fisheries. The results described
here provide a useful reference point to which future survey data can be compared.
KEYWORDS: SURVEY-VESSEL; NORTHEAST PACIFIC; ABUNDANCE ESTIMATE; DISTRIBUTION; HARBOUR PORPOISE;
HUMPBACK WHALE; PACIFIC WHITE-SIDED DOLPHIN; MINKE WHALE; DALL’S PORPOISE; KILLER WHALE
___________________________________________________________________________________
INTRODUCTION
Marine mammals in coastal British Columbia
Information on animal distribution and abundance is
integral to wildlife conservation and management, however
abundance estimates have not been available for many
species of cetaceans inhabiting coastal waters of Canada’s
Pacific coast, including those that were heavily depleted by
commercial whaling. More than 20 marine mammal species
can be found in the coastal waters of British Columbia
(BC), Canada. They vary widely in their fidelity to inshore
Canadian waters, history of exploitation, conservation
status, and the extent to which they have been studied.
Killer whales (Orcinus orca) in BC are the most carefully
studied cetacean populations in BC (Bigg et al., 1990; Ford
et al., 2000; Ford et al., 1998; Olesiuk et al., 1990). Baleen
whales were the subject of extensive pelagic and coastal
whaling in the northeast Pacific Ocean; the last coastal
whaling stations in BC closed in 1967 (Gregr et al., 2000).
Incidental bycatch of small cetaceans in commercial gillnet
fisheries does occur (Hall et al., 2002). Whalewatching,
once seen as an alternative to whaling, is now considered a
potential threat to some cetacean populations via masking
effects of boat noise, potential energetic cost of vessel
avoidance tactics (Williams et al., 2002a; Williams et al.,
2002b) and disruption of feeding activity (Williams et al.,
2006) and emission of outboard motor exhaust. All marine
mammals in BC, with their acoustic sensitivity and high
trophic position, are vulnerable to impacts of intense
anthropogenic noise and toxicity of fat-soluble
contaminants. In recent years, there has been considerable
discussion about lifting existing moratoria on offshore oil
and gas exploration and extraction off the north and central
coasts of BC, which has created a heightened sense of
urgency to collect baseline data on marine mammal
distribution and abundance (Royal Society of Canada,
2004).
The following is a summary of frequently seen cetacean
and pinniped species in BC’s inshore waters in summer.
With few exceptions, our knowledge of populations reflects
our pattern of use of that species. Exploited populations
(either for hunting, live capture, culling, or non-
consumptive uses such as whale watching), have received
much greater scientific attention than unexploited ones. The
species’ status in BC refers to that determined by the
Committee on the Status of Endangered Wildlife in Canada
(COSEWIC). COSEWIC uses a variety of information
sources to assess species’ extinction risk (ranging from
Extinct, Extirpated, Endangered, Threatened and Special
Concern to Data Deficient or Not at Risk) and to report its
recommendation to the Canadian government and the
____________________________________________________________________________________
* Raincoast Conservation Society, PO Box 2429, Sidney, BC, Canada, V8L 3Y3 and Sea Mammal Research Unit, Gatty Marine
Laboratory, University of St Andrews, St Andrews, Fife KY16 8LB Scotland.
+ Research Unit for Wildlife Population Assessment, Centre for Research into Ecological and Environmental Modelling, University of St
Andrews, KY16 9LZ Scotland.
public. At that point, species that have been designated by
COSEWIC may or may not qualify for legal protection and
recovery efforts under Canada’s Species at Risk Act
(SARA).
Cetaceans
The harbour porpoise (Phocoena phocoena) is listed as a
species of Special Concern in Canada’s Pacific waters; a
designation indicating that it is considered a species to
watch although not in obvious danger of extinction in the
near term (COSEWIC, 2003a). Anthropogenic activity,
pollution and bycatch have been flagged as conservation
threats (Baird, 2003a). Rates of bycatch were estimated by
Hall et al. (2002) and studies on harbour porpoise habitat
usage are taking place off southern Vancouver Island (Hall,
2004).
Dall’s porpoise (Phocoenoides dalli) is thought to be Not
at Risk in BC (Jefferson, 1990). It is widely distributed, and
commonly seen in deep coastal waters.
The Pacific white-sided dolphin (Lagenorhynchus
obliquidens) is thought to be Not at Risk in BC (Stacey and
Baird, 1991). This species returned to inshore waters in BC
relatively recently after a decades-long absence (Heise,
1997; Morton, 2000), and has been called the most abundant
cetacean in the region (Heise, 1997). Interactions with
fisheries are rare locally (Hall et al., 2002) and in other areas
of the North Pacific (Anon., 2000).
The common minke whale (Balaenoptera acutorostrata)
status is currently under review by COSEWIC. The species
has never been a target of coastal or pelagic whalers in
western Canadian waters. There is some evidence that
individual common minke whales may be resident to
inshore coastal waters of Washington State (Dorsey et al.,
1990), but the species is relatively poorly studied in BC
waters.
The humpback whale (Megaptera novaeangliae) is listed
as Threatened in Canada’s Pacific waters (COSEWIC,
2003b). They were reduced to a fraction of pre-exploitation
numbers by commercial whaling (Baird, 2003b), but there is
strong evidence to suggest that the North Pacific population
is recovering (Calambokidis et al., 1997). Photo-
identification on animals that use BC waters is extensive,
collaborative and ongoing (Cetacean Research Program
1
,
Fisheries and Oceans Canada (DFO)). While the primary
focus of DFO’s Cetacean Research Program is on humpback
and killer whales, sightings of all cetacean species are
recorded during the non-randomised surveys that they have
conducted since 2002.
The fin whale (Balaenoptera physalus) is listed as
Threatened in Canada’s Pacific region (COSEWIC, 2005).
The species was heavily exploited by commercial whaling,
with evidence suggesting that the population was hunted to
near commercial extinction by the 1960s (Gregr et al., 2000;
Gregr and Trites, 2001). Ship strikes and fishing gear
entanglement are potential threats to fin whale recovery in
BC. Photo-identification studies are beginning on this
species in this region (coordinated by DFO).
Two fish-eating (i.e. northern and southern ‘resident’)
populations of killer whales inhabit the coastal waters of
BC, as do a mammal-hunting ‘transient’ population and a
recently discovered and poorly studied ‘offshore’ population
(Ford et al., 2000; Ford et al., 1998). Not only is abundance
known for the fish-eating killer whales, but also it is known
with an unusually high degree of confidence (Ford et al.,
2000; Olesiuk et al., 1990). Absolute abundance of
mammal-hunting killer whales is more difficult to estimate
than that of fish-eating killer whales, because the strong
differences in social structure make it difficult to choose
appropriate capture-recapture statistical models for transient
killer whale photo-identification data. Northern resident and
transient populations are considered Threatened in BC
waters, while the southern resident population is listed as
Endangered. The offshore population is considered to be of
Special Concern. Conservation threats to the species in BC
waters include: small population size due to a previous live-
capture fishery for display (Bigg and Wolman, 1975;
Williams and Lusseau, 2006); anthropogenic noise and
repeated disturbance (Williams et al., 2002b); contaminants
(Ross et al., 2000); and prey availability (Baird, 2001b).
Sei (Balaenoptera borealis), blue (Balaenoptera
musculus) and North Pacific right whales (Eubalaena
glacialis) are all listed as Endangered in BC, due primarily
to historic overexploitation (Gregr et al., 2000) which
resulted in the small current population size. The gray whale
(Eschrichtius robustus) is listed as a species of Special
Concern in BC (COSEWIC, 2004), although the population
has recovered since its heavy exploitation in the 19
th
century. The deep-diving and relatively poorly studied
beaked whales are rarely reported in BC waters, although
they may be more common than scarce sightings would
suggest (Willis and Baird, 1998).
Pinnipeds
The harbour seal (Phoca vitulina) is considered to be Not at
Risk in western Canada, due to its large and increasing
population size (Baird, 2001a; Olesiuk, 1999). Conservation
concerns include prey availability and illegal and unreported
shooting or mortality incidental to fish farming operations
(Baird, 2001a). Their widespread distribution makes them
less vulnerable to oil spills than those species that haul out
in few locations, although their tolerance of urbanised
habitat lends them susceptible to bioaccumulation of
contaminants. The DFO conducts regular counts of
pinnipeds at haul-out sites and corrects for animals likely to
be at sea, so trend data are available, particularly in southern
BC, however density of harbour seals on the north and
central BC coasts is less well studied, and at-sea distribution
is poorly studied in BC waters generally (Olesiuk, 1999).
The Steller sea lion (Eumetopias jubatus) is listed as a
species of Special Concern in western Canada (COSEWIC,
2003c). While the species is locally abundant and the
population growing, the breeding population in BC waters is
composed of only three known breeding sites (COSEWIC,
2003c), which makes them inherently vulnerable to
catastrophic events such as oil spills. At-sea distribution of
Steller sea lions is not well studied in BC waters.
The northern elephant seal (Mirounga angustirostris) is
considered to be Not at Risk in BC waters. The population
was hunted to near extinction in the 19
th
century, but the
surviving population has increased exponentially since then.
At-sea distribution of elephant seals is not well studied in
BC waters.
Systematic sightings survey of Canada’s Inside Passage
We conducted a line transect survey in the inshore western
Canadian waters between the BC-Washington and the BC-
Alaska borders. The primary objective of the survey was to
generate design-unbiased estimates of abundance of marine
mammal species in BC coastal waters during the summer
months. A related goal was to provide estimates of at-sea
distribution, to begin to understand which areas of the coast
16 WILLIAMS & THOMAS
:
MARINE MAMMALS IN COASTAL WATERS OF BRITISH COLUMBIA
1
Cetacean Research Program: http://www.pac.dfo-mpo.gc.ca/sci/
sa/cetacean/default_e.htm
JNL339 015-028:Layout 1 29/10/07 07:16 Page 16
may represent the most important habitat to each species.
This paper reports results from our systematic survey of
marine mammals of the Inside Passage, which were
completed in the summers of 2004 and 2005.
METHODS
Survey design
The survey design is described in detail in Thomas et al.
(2007). A stratified survey design was used where the study
area was divided into four strata, within which a sample of
equal-spaced zig-zag or parallel transect lines was placed
with a random start point to ensure equal coverage
probability within strata. Area names are provided in the
map shown in Fig. 1.
(1) Queen Charlotte Basin – this roughly convex region has
been proposed for offshore oil exploration and
extraction. For our purposes, it extends to a maximum
of 82 n.miles offshore in the west to an eastern boundary
line drawn down the edges of the outer coastal islands,
east of which we considered part of a mainland inlet
stratum (see Stratum 4). The southern boundary was the
narrow neck of Johnstone Strait.
(2) Strait of Georgia and Juan de Fuca Strait – Canadian
waters off southern Vancouver Island to the BC
mainland shore.
(3) Johnstone Strait and Discovery Pass – narrow
passageway between northeastern Vancouver Island and
mainland BC.
(4) Mainland inlets – this collection of fjords, passages,
straits and inlets was grouped using a Geographical
Information System (GIS) into 33 irregular-shaped
bodies of water that could be surveyed in 1-3 days.
From these ‘Primary Sampling Units’ (PSUs), a sample
of five was selected using a systematic random design,
with probability of sampling proportional to area.
Within each of these, a systematic parallel line design
was used to generate transect lines. This provided a 10-
day, cluster sample of the mainland inlets, which was
designed to provide a reasonable starting point to
represent the mainland inlet stratum.
Stratum 1 (Queen Charlotte Basin), was surveyed twice, in
the summers of 2004 and 2005. The 2005 survey design for
Stratum 1 was similar to 2004, but used a new, random
starting point. The effort (and consequently coverage
probability) was similar in both years.
Field methods to measure animal density
Visual search effort
Data were collected aboard the motor-sailing vessel
Achiever (a 21m steel-hulled sailboat) in 2004 and the Gwaii
Haanas (a 20m aluminium power boat) in 2005. An
aluminium platform was built for Achiever to increase the
eye height of observers well above the ship’s boom, and
both vessels steamed at approximately 8 knots (15km h
–1
)
during searching effort. Data were collected from the
highest accessible point (the primary observer platform) on
the ships used in this study, such that eye height was
approximately 5m in both years. The team consisted of six
people. Three people served on the primary observer team,
namely a port and starboard observer and a data recorder. In
addition, one observer operated the computer while two
team members were on rest periods.
The primary observer team searched ahead of the ship,
that is, a sector from the trackline to 90° abeam the ship,
while concentrating primarily on the trackline. Each
observer used 7350 or 8350 binoculars to search a sector
spanning from 30° on one side of the trackline to 90° on the
other side. The data recorder recorded whenever a sighting
was made, and assisted the observer with species
identification or group size estimation when needed.
A Global Positioning System (GPS) was connected to a
computer running Logger software (Logger 2000,
International Fund for Animal Welfare). This collected
positional information every 10s, which was used for
calculating length of trackline covered, as well as ship’s
course and speed. The computer operator entered
information on sighting conditions every 15min, or as
conditions changed. The computer operator also noted the
position of each team member at the beginning of every
hour. Observer rotation occurred every hour. Information
collected on factors that could affect sighting conditions
included sea-state, cloud cover and precipitation and a
subjective sightability code.
As well as cetaceans, pinniped sightings were also
recorded, both in the water and hauled out. In searching for
hauled out pinnipeds, the shoreline was scanned at the end
of each transect carefully, concentrating particularly on the
region where the transect line met the shore. For the
purposes of analysis, animals seen hauled out on the shore
past the end of a transect line were treated as if they were
seen right at the end of the line, that is, just within the study
area.
Sightings
Whenever a sighting was made of a marine mammal, it was
assigned a sighting number and reported to the data recorder
on the bridge via two-way radio. An angle board mounted
on the deck railing was used to measure radial angle to the
school, and a measurement was made of the range to the first
sighting using 7350 reticle binoculars or a graduated
perpendicular sighting gauge. Distance to pinnipeds was
recorded when possible using a Bushnell Yardage Pro laser
rangefinder. If a visual estimate had to be made, then those
radial distance estimates were corrected subsequently using
observer-specific distance estimation experiments. During
distance estimation experiments, observers recorded their
visual estimates of distance to 20 continuously visible
J. CETACEAN RES. MANAGE. 9(1):15–28, 2007
17
Fig. 1. Realised search effort: tracklines followed during the 2004 and
2005 summer field seasons. Stratum 1 (Queen Charlotte Basin) was
re-surveyed in 2005, with the same amount of trackline effort
allocated in 2004 but with a new random start point.
JNL339 015-028:Layout 1 29/10/07 07:16 Page 17
targets (rocks, floating logs, etc.) to which a data recorder
measured range using a laser rangefinder. A linear
regression model with error proportional to true distance
was fitted to the data. Visual estimates from the survey were
subsequently corrected by dividing estimated distance by
the estimated slope through the origin. No attempt was made
to assess whether range or angle measurements (i.e. those
ranges measured using binocular reticles, rangefinders or a
sighting gauge) were biased; only visual estimates of range
were corrected using these calibration experiments. The
computer operator and data recorder noted ship location and
the time of the sighting, and binoculars were used to confirm
species and school size. Additional information was
recorded on cue type (inter alia body, blow, seabird
activity), the animal’s behaviour (swimming normally
[travel/forage], avoid, approach, feeding, breaching, other),
and its heading relative to the ship (profile, head-on, tail-on
or other/unsure).
When time permitted, a decision would be made to
‘close’ on certain sightings to confirm species identification
(usually for balaenopterids); to allow collection of
identification photographs of humpback and killer whales;
or to obtain more accurate estimates of school size. When
the ship left the trackline, search effort was terminated.
Search effort was resumed once the ship reached cruising
speed and rejoined the original trackline. We also
occasionally stayed ‘on-effort’ during transit legs to increase
the number of sightings available for fitting the detection
function. These sightings are subsequently referred to as
transit-leg sightings and were not included in density
estimates.
Data analysis
To analyse the data, the methods described by Buckland et
al. (2001) were followed, which are referred to as
‘conventional distance sampling’ (CDS) in this paper. Note
that these methods assume that probability of detection of
animals at zero distance from the trackline is one (the so-
called g(0)=1 assumption), an issue we return to later. The
analysis can be split into three parts: (i) fitting a detection
function g(x), where xis perpendicular distance, to observed
distances of sightings from the transect to estimate average
probability of detection, p; (ii) using observed school sizes
to estimate mean school size in the population, E(s); (iii)
estimation of animal density, D, using the formula
(1)
where nis the number of schools seen within w;wis the
truncation distance; and Lis the total length of the transects
searched on effort. We deal with each of these parts below.
Data for each species were treated separately. Estimation
of detection probability and mean school size was
performed using the free software Distance 5.0 Beta 5
(Thomas et al., 2005). Distance was not used to estimate
overall density or abundance due to the non-standard
stratification used, so this was done using free statistical
software R(R development core team 2005), version 2.2.0.
Distance projects containing the data and analyses, and R
code are available on request from the first author.
Estimation of detection probability
Models were fitted to the observed distribution of distances
based on the key function and series expansion formulation
of Buckland et al. (2001). The uniform, half-normal and
hazard-rate key functions were used, together with
polynomial or cosine series expansion terms as required.
The model that minimised the Akaike Information Criterion
(AIC) was selected, unless the behavioural observations
indicated a problem with avoidance or attraction, in which
case the model that we felt best reflected the relationship
between probability of detection and distance for that
species was subjectively chosen. In practice the lowest AIC
model was not chosen in only one case: Pacific white-sided
dolphin (see Results). The absolute fit of models was judged
using diagnostic plots and the Kolmogorov-Smirnov
goodness-of-fit test (Buckland et al., 2004).
The analysis of each species’ data was begun by
considering the need for truncation of the largest distances,
since no truncation of sightings was performed in the field.
To do this, detection functions were fitted to data with 0, 5
and 10% of the most distant sightings removed and the least
amount of truncation necessary was chosen to achieve the
same number of series expansion terms as were fit at 10%
truncation (fitting fewer series terms will tend to give better
precision) while keeping g(w)>0.10. The truncation distance
was then rounded to the nearest 100m. In general, the
amount of truncation used had very little effect on the
results.
As the survey design involved stratification, the
possibility of fitting a separate detection function in each
stratum for species was investigated where sample size was
sufficient (>60 observations per stratum). For each of these
species, the AIC for a detection functions fit was compared
to all data, with the sum of AICs for detection functions fit
to data for each stratum separately. The stratified detection
functions were used if they had a lower AIC than the pooled
function.
Data collected for one pinniped species, the harbour seal
were also analysed. It was anticipated that there would be a
qualitatively different detection function for seals seen in the
water compared with those hauled out, so fitting detection
functions stratified by in/out of water was attempted.
Estimation of mean school size
The default method in Distance was used to obtain an
unbiased estimate of mean school size, as follows. The
natural logarithm of school size, ln(s), was regressed on the
estimated probability of detection at the distance the school
was seen. The predicted value of ln(s) at zero distance
(where detection probability is 1) was then back-
transformed to provide the required estimate.
For harbour seals, it was expected that the school size
would be different for observations in the water and seals
that were hauled out, so the two groups were analysed
separately.
Density, abundance and variance estimation
For each species, density, abundance and associated
measures of uncertainty were estimated within each stratum
and then combined to produce results for the whole study
area. For harbour seals, these statistics were estimated
separately within each stratum for seals on land versus those
in the water. The methods used are an application of those
given by Buckland et al. (2001, section 3.6), but are slightly
more complicated due to the stratification used in the survey
design and the cluster sample in stratum 4.
For stratum 1, the estimate of mean density for both years
was calculated as the effort-weighted mean of the year-
specific estimates. For stratum 4, the estimate of mean
density was an unweighted mean of the estimates in each
PSU, rather than an effort-weighted mean (see Discussion
18 WILLIAMS & THOMAS
:
MARINE MAMMALS IN COASTAL WATERS OF BRITISH COLUMBIA
JNL339 015-028:Layout 1 29/10/07 07:16 Page 18
for an alternative). Overall mean density for the study area
was calculated as the area-weighted average of the stratum
estimates.
Variances were calculated using the delta method, and
log-normal, t-based, two-sided 95% confidence limits for
the estimates of density and abundance were obtained using
equations 3.72-3.76 of Buckland et al. (2001).
RESULTS
Realised survey effort
On-effort transects covered in 2004 and 2005 are shown in
Fig. 1. Table 1 includes area of each stratum and realised
survey effort (trackline length). Some small segments of
trackline were unsurveyed due to poor weather conditions,
while others were excluded because they proved in the field
to be non-navigable. Nearly 100% of planned survey effort
was realised in strata 1 and 3, and 92% of stratum 4. The US
waters south of Vancouver Island (stratum 2), were
eliminated for logistical reasons, which resulted in only 64%
of those planned tracklines being surveyed. Consequently,
the US waters in stratum 2 were removed from the survey
area at the abundance estimation step.
Cetacean sightings
The following section summarises the number and
behaviour of animals sighted during our surveys and some
technical issues relating to the selected detection function
for each species. Table 2 lists the truncation distance,
number of observations (before and after truncation), fitted
detection function model, p-value from Kolmogorov-
Smirnov (K-S) goodness-of-fit (GOF) test, and estimated
mean detection probability of observed schools (p
ˆ) for each
species analysed. Fig. 2 shows the selected detection
function for each species. In no cases did detection functions
fitted separately to each stratum have a lower AIC than those
fit to all data pooled, so the pooled functions were used.
Table 3 shows the mean school size for each species, both
from the observed data and the size-bias regression, as well
as summary statistics on observed group sizes. Table 4
summarises our species-specific estimates of density and
abundance, with corresponding 95% confidence intervals
(CIs) and percentage coefficient of variation (%CV), by
stratum and combined. Sightings distribution maps for 12 of
the most frequently seen species are shown in Fig. 3.
Harbour porpoise
A total of 68 harbour porpoise schools was sighted while on-
effort (Table 2). Harbour porpoise were seen throughout the
study area; most commonly in the southern straits, but also
frequently in mainland inlets and Queen Charlotte Basin
(Fig. 3). Notes on animal behaviour collected at the time of
first sighting do not indicate any severe problem with
responsive movement. Most observations were scored as
travel/forage (64/68=94%), with the remainder as follows:
avoid (3/68=4.4%); and feeding (1/68=1.5%). Data on body
aspect relative to the ship showed no obvious signs of
responsive movement. Most of the 68 sightings were
observed in profile (31 heading left, and 31 heading right),
with nearly equal numbers of sightings observed head-on
(3) and tail-on (2), and only one sighting of unknown
aspect.
The selected truncation distance was 500m, and at this
distance the lowest-AIC detection function model was a
hazard rate with no adjustment terms (Fig. 2). A half-normal
model would have provided a wider shoulder in the
estimated detection function, however the AIC for that
model was 3.28 higher than the hazard rate. Nevertheless,
the half-normal and the hazard rate models resulted in quite
similar values for p
ˆ(approximately 25% higher for half-
normal: 0.27 vs 0.21).
An outstanding issue for the harbour porpoise sighting
data is an apparent spike in the detection function at zero
distance (Fig. 2), which might suggest attractive movement
or that trackline detection for this species was less than unity
(i.e. g(0)<1). However, note that little other evidence of
attractive movement by harbour porpoise was seen in the
data.
Dall’s porpoise
A total of 112 schools of Dall’s porpoises was recorded
during the surveys. Dall’s porpoises were seen most
commonly in the offshore waters of Queen Charlotte Basin,
occasionally in the southern straits, and relatively
infrequently in mainland inlets (Fig. 3). Of the 112
sightings, most observations were recorded at the time of
first sighting as exhibiting normal, slow-rolling
(travel/forage) behaviour (108/112=96.4%), with other
behaviours scored as follows: approach (2/112=1.8%);
avoid (1/112=0.9%); and feeding (1/112=0.9%). Data on
body aspect revealed no obvious evidence of responsive
movement. Most of the 112 sightings were observed in
profile (48 heading left and 48 heading right), but more
observations were scored as head-on sightings (8) than tail-
on (5); although 3 were of unknown aspect.
At 700m truncation, the best model was half-normal with
no adjustments (Fig. 2), but there was some ambiguity in
this case about which detection function best described the
observed sightings data. The half-normal, which was used,
and the uniform with one cosine adjustment models gave
near-identical results in terms of p
ˆand AIC. There was some
support from the data (DAIC=1.33) for choosing a hazard-
rate model, which would have fitted the apparent spike near
zero and resulted in an approximately 50% lower estimate of
p(0.35 vs 0.54) and consequently higher estimate of animal
density.
While failure of the g(0)=1 assumption likely introduced
some negative bias in our estimate, obvious cases of
Dall’s porpoise being attracted to or avoiding our survey
vessel were rare (3/112 sightings), which suggests that any
bias associated with responsive movement was probably
low.
J. CETACEAN RES. MANAGE. 9(1):15–28, 2007
19
JNL339 015-028:Layout 1 29/10/07 07:16 Page 19
Pacific white-sided dolphin
A total of 117 schools of Pacific white-sided dolphins were
seen during the survey. Dolphins were most frequently seen
in Queen Charlotte Basin, but also in Johnstone Strait and
occasionally in the southern straits (Fig. 3). Responsive
movement was a bigger problem for this species than with
any other. While most observations were recorded at the
time of first sighting as exhibiting normal, ‘slow-roll’
(travel/forage) surfacing behaviour (84/117=71.8%), the
second-most frequently recorded behaviour was ‘approach’
(16/117=13.7%). Remaining behaviours recorded were as
follows: feeding (9/117=7.7%); breaching (6/117=5.1%);
avoid (1/117=0.9%); and other/unsure (1/117=0.9%).
Information on animal heading relative to the ship showed a
similar pattern; while most of the sightings were of animals
in profile (25 heading left and 30 heading right), there were
ten times as many sightings scored as head-on (20) than tail-
on (2) and 40 sightings were of uncertain aspect. Not
surprisingly, given these observations, the perpendicular
distance data exhibit a spike at low distances (note the
higher than expected probability of detections at and near 0
perpendicular distance, Fig. 2) that is consistent with
attractive movement.
A truncation distance of 700m was chosen, and selected
the half-normal detection function with no adjustments.
Note that this was not the model with the lowest AIC (which
was the hazard rate model), but the half-normal model was
used to avoid fitting the spike at zero distance, which was
believed to be an artefact of responsive movement towards
the boat (see Discussion). The half-normal was chosen
because it was felt that the detection function should be
qualitatively similar to that for Dall’s porpoise (Fig. 2), so
the same shape (half-normal) was forced, with the data
being used to select the scale (parameter of the half-normal).
The estimated pturned out to be very close to that of Dall’s
porpoise (Table 2).
Humpback whale
A total of 76 humpback whale schools was seen on-effort
during the study. Humpback whales were seen most
frequently in Queen Charlotte Basin and the mainland inlets
of the north and central coasts (Fig. 3). Of the 76 sightings,
most observations were recorded at the time of first sighting
as exhibiting normal, travel/forage behaviour
(66/76=86.8%), with other behaviours scored as follows:
breaching (4/76=5.3%); feeding (4/76=5.3%); and
other/unsure (2/76=2.6%). No observations were scored as
representing avoidance or attractive behaviour, although
more animals were observed head-on than tail-on (9 vs 6).
The model that fitted these data best was a half-normal
detection function with one cosine adjustment term, using a
2,000m truncation distance (Fig. 2).
Fin whale
A total of 35 fin whale schools was recorded during the
study (including three transect-leg sightings used for
estimating effective strip width and mean school size, but
not for estimating density). All fin whale sightings were
made in the Queen Charlotte Basin or adjacent north-coast
mainland inlets (Fig. 3). Of the 35 sightings, most
observations were recorded at the time of first sighting as
exhibiting normal, travel/forage behaviour (30/35=85.7%),
20 WILLIAMS & THOMAS
:
MARINE MAMMALS IN COASTAL WATERS OF BRITISH COLUMBIA
JNL339 015-028:Layout 1 29/10/07 07:16 Page 20
J. CETACEAN RES. MANAGE. 9(1):15–28, 2007
21
Cont.
JNL339 015-028:Layout 1 29/10/07 07:16 Page 21
22 WILLIAMS & THOMAS
:
MARINE MAMMALS IN COASTAL WATERS OF BRITISH COLUMBIA
Fig. 2. Histograms of observed distances and fitted detection functions for the species analysed. In constructing the histograms, 1.5AU
nequally spaced
intervals were used (where nis the number of detections), except that the first two intervals were further sub-divided into two to make the pattern
of detections close to the transect line more clear.
JNL339 015-028:Layout 1 29/10/07 07:16 Page 22
with other behaviours scored as follows: feeding
(2/35=5.7%); and other/unsure (3/35=8.6%). No
observations were scored as representing avoidance or
attractive behaviour. However, while 4 observations were
made of animals in tail-on aspect, only 1 head-on sighting
was made, as were 8 sightings of unsure aspect.
Despite the small sample size, the perpendicular distance
data were reasonably well behaved; there was evidence of a
shoulder in the detection function (Fig. 2), and after
truncation at 2,000m, the data were described well by a
uniform model with one cosine adjustment term.
Killer whale
Only 18 killer whale schools were recorded during the
survey (15 ‘residents’ and 3 ‘transients’). Four of these
sightings were recorded during transit-leg surveys and used
for estimating effective strip width and mean school size,
but not for estimating density. Sightings were most common
in Queen Charlotte Basin and Johnstone Strait, and in
adjacent north- and central-coast mainland inlets (Fig. 3).
No southern resident killer whales were observed while on
effort in the southern straits. Most observations of killer
whales were recorded as exhibiting normal, travel/forage
behaviour (16/18=88.9%), with the remaining two scored as
other/unsure (11.1%). No observations were scored as
representing avoidance or attractive behaviour. Information
on animal heading also failed to show evidence of
responsive movement. Equal numbers of head-on and tail-
on sightings (2 each) were recorded.
A truncation distance of 1,500m was chosen, in that it
only required dropping two observations, but provided a
reasonable fit to the data (Fig. 2) using a half-normal
function.
J. CETACEAN RES. MANAGE. 9(1):15–28, 2007
23
Fig. 3. Maps showing distribution of on-effort sightings, uncorrected for unequal survey effort between strata, for 12 marine
mammal species in coastal waters of BC in summer months of 2004 and 2005.
JNL339 015-028:Layout 1 29/10/07 07:16 Page 23
Common minke whale
Only 14 common minke whale schools were recorded
during the survey. Sightings were distributed more or less
uniformly from north to south, but all were seen somewhat
offshore; no sightings of minke whales were made in
mainland inlets (Fig. 3). Of the 14 sightings, all
(14/14=100%) were recorded as displaying normal,
travel/forage behaviour at the time of first sighting; no
evidence of avoidance or attractive behaviour was observed.
Information on animal heading relative to the ship indicated
that one animals was observed head-on, and two tail-on.
A variety of detection functions were fitted to the data and
it was found that, with truncation at 300m (that is, truncating
one observation and leaving 13), AIC was lowest for a
uniform detection function with no adjustments (DAIC over
uniform + 1 cosine adjustment term was 2.2; and DAIC over
half-normal was 1.64). This model corresponds to a strip
transect out to 300m (Fig. 2). Note that the K-S p-value was
p=0.8903 and N
ˆ=388 (CV 26.8%, 95%CI 222-680). For
comparison, an analysis with 150m truncation was done, as
the assumption of a strip transect should be even better
justified then. While the number of observations was
reduced to 8, the results changed little (N
ˆ=475, CV37.3%,
95% CI 221-1,020). So, it can be concluded that there were
fewer than 1,000 minke whales in the area during the study,
and most likely around 400 minke whales.
Pinniped sightings
Harbour seal
Harbour seals were by far the most frequently sighted
marine mammal species during the study (Table 2; 350
sightings [104 hauled out, 246 in water]). Sightings were
most frequent in southern straits and mainland inlets, but
nevertheless quite common in Queen Charlotte Basin. There
was strong support from the data (as indicated by AIC) for
stratification of detection function estimation by whether
seals were observed hauled out, or in water. For each
stratum, 500m truncation was used, so that estimated g(w)
was >0.1. For both strata, a half-normal detection function
was selected (Table 2, Fig. 2); however two cosine
adjustment terms were preferred for the in-water stratum.
School size was estimated separately by stratum as well
(Table 3).
Density and abundance were estimated for each part of
the population separately and combined by summing
density across the parts (see Methods). The abundance
estimate for harbour seals is an underestimate due to
incomplete trackline detection. Trackline detection of
hauled-out seals is also uncertain, but g(0) is likely to be
especially low for those in water. However, as an index of
relative abundance, two results emerge. First, the point
estimates indicate that at on average, at least two-thirds of
the seals were in the water (perhaps much more than this, if
g(0) for animals in the water is particularly low), while one-
third were hauled out (Table 4). Secondly, in terms of spatial
variability, while the majority of harbour seals were found in
the southern straits and in the mainland inlets, a substantial
number of animals were in the Queen Charlotte Basin as
well (Table 4).
Northern elephant seal and Steller sea lion
Too few sightings were made of elephant seals or Steller sea
lions to fit a detection function. Locations of sightings are
shown in Fig. 3. All sightings of northern elephant seals
(7/7) were made in Queen Charlotte Basin, as were the
majority (17/21) of Steller sea lions (Fig. 3). Remaining
sightings of Steller sea lions (4/21) were made in mainland
inlets (Stratum 4). All sightings of elephant seals were of
animals at sea (that is, no sightings were made of hauled-out
elephant seals), which is unsurprising given that the survey
was conducted in a feeding area. Similarly, most Steller sea
lion sightings (16/21) were of animals in the water, with the
remaining 5/21 sightings hauled out on land.
DISCUSSION
Preliminary estimates of abundance and distribution
This systematic survey of Inside Passage waters achieved its
primary objectives and preliminary (see the section on
reliability below) estimates of abundance for six cetacean
species in the inshore coastal waters of Canada’s Pacific
region are reported. This information is needed for
informing a variety of conservation and management issues,
were Canada to assess the sustainability of observed levels
of bycatch of small cetaceans in commercial fisheries (Hall
et al., 2002), or to pursue proposals to incorporate predator
needs when setting fish quotas through ecosystem-based
fisheries management (Larkin, 1996). In addition, the
distribution data provide a systematic snapshot of marine
mammal distribution, which it is hoped will be of use when
reviewing permit applications to conduct seismic surveys in
the Queen Charlotte Basin region, and if seismic surveys do
proceed, for mitigating the impacts of these and other
intense anthropogenic noise sources on acoustically
sensitive animals. It should be noted that the highest
numbers of common minke, fin and humpback whales were
observed in this stratum (Stratum 1), with fin whales found
here exclusively. It should be noted that although the term
‘population’ is used for convenience in this discussion, this
does not imply that the animals found in the surveyed areas
necessarily represent discrete biological populations.
Rather, the numbers presented represent the estimated
numbers of animals found in the surveyed waters at the time
of the surveys.
The best estimates of abundance throughout the study
area, and the stratum-specific estimates, are presented in
Table 4. While small sample size makes the northern
resident killer whale estimate tenuous, it is one worth
noting, because this is the only finding that can be
corroborated against the true number. The best estimate
(161) is close to the true population size of 219 animals
(2004 census, Cetacean Research Program, Pacific
Biological Station, DFO Canada), and the 95% confidence
interval (45-574) comfortably includes this number. While
the common minke whale abundance estimate is similarly
tentative, we are confident in our finding that minke whales
were relatively rare. Future research on common minke
whales should be encouraged to assess the relatedness of
BC’s inshore minke whales to stocks in Canadian offshore
or adjacent US waters.
Also relevant to this study were those species that were
not encountered at all – no blue, sperm or right whales or
northern fur seals were seen, for example. Blue and sperm
whales were certainly caught in these waters historically
(Gregr et al., 2000), although the study area was admittedly
on the periphery of the preferred habitat for those two
species (Gregr and Trites, 2001). Right whales do appear in
historic catch data for the BC coast, but had already been
largely depleted before record keeping began in earnest
around 1908 (Gregr et al., 2000). Only one sighting of a sei
whale was made during the survey (Fig. 3), although sei
whales were frequently caught in Queen Charlotte Sound
24 WILLIAMS & THOMAS
:
MARINE MAMMALS IN COASTAL WATERS OF BRITISH COLUMBIA
JNL339 015-028:Layout 1 29/10/07 07:16 Page 24
J. CETACEAN RES. MANAGE. 9(1):15–28, 2007
25
and southern Hecate Strait (our Stratum 1) and this area was
predicted to represent a good habitat for this species in the
region (Gregr and Trites, 2001). No beaked whales were
seen during the study, despite their tendency to be reported
in stranding records relatively frequently along the Queen
Charlotte Islands, which are sparsely populated and do not
have a dedicated strandings network (Willis and Baird,
1998). Gray whales were also scarce in these inshore waters,
but are known to be common on the west coasts of
Vancouver Island and Queen Charlotte Islands. It is hoped to
expand the study area to survey the outer coastal waters to
complete the assessment of cetacean abundance at least to
the edge of the continental shelf. There is at present an
international collaboration
2
to use photo-identification to
estimate current abundance of humpback whales in the
North Pacific basin (SPLASH), and it is hoped that the
emerging analyses of that project will give a new basis for
comparison within the study region of overlap.
Reliability of abundance estimates
The accuracy of the abundance estimates hinges on three
primary assumptions: that detection on the trackline was
certain; that no responsive movement occurred prior to
detection; and that distances and angles were measured
without error (Buckland et al., 2001, section 2.1).
Measurement error in radial ranges was unlikely to have
introduced major bias, in that most ranges were measured
rather than estimated visually, and estimates were calibrated
post-hoc using calibration experiments to remove
systematic bias. The two remaining assumptions are
problematic for some species. Summarised below are the
major conclusions regarding responsive movement and
uncertain trackline detection from the behavioural and
animal heading data, the shape of the fitted detection
function, and comparisons with other studies. After
summarising these outstanding technical issues, the utility
of these preliminary results are discussed along with
planned future work for conservation and management.
Responsive movement and uncertain trackline detection
are outstanding issues for three species: harbour and Dall’s
porpoises and Pacific white-sided dolphins. For harbour
porpoises, the shape of the detection function illustrates
outstanding technical issues for this species. If the observed
spike in detections near zero distance represented attractive
movement, then the selected hazard-rate detection function
would result in an overestimate of abundance. However, (1)
no evidence was seen of attractive movement in the study;
(2) harbour porpoise are generally thought to avoid ships
(Palka and Hammond, 2001); and (3) only 4% of the
sightings suggested avoidance behaviour. It may be that the
observed spike simply reflects the true detection process:
that a very narrow effective strip width was covered for this
species, possibly because observers were searching with
low-power binoculars from a relatively low platform (~5m).
A second alternative is that the detection function lacked a
wide shoulder because g(0) was substantially less than 1.
Efforts are underway to increase sample size of double-
platform trials to estimate g(0) for this platform. However,
until g(0) can be estimated directly for the survey, it may be
useful to consider a range of likely values. Barlow (1995)
estimated g(0) due to perception bias for cryptic species
(harbour and Dall’s porpoises and pygmy sperm whale) to
be 0.78 for a team of three people. Barlow (1995) suggested
that this may be an overestimate of g(0) for the harbour
porpoise, because it is pooled with the more detectable
Dall’s porpoise. Barlow et al. (1997) reviewed g(0)
estimates reported from shipboard surveys for harbour
porpoises in US waters that ranged from 0.4 to 0.78,
depending, inter alia, on the number of observers and
sighting conditions. Palka (2000) conducted ship-board
sightings surveys for harbour porpoises in the Gulf of Maine
and Bay of Fundy that produced estimates of g(0) ranging
from 0.25 to 0.74, depending on platform height (9 or 14m
above the sea surface), number of observers and stratum.
Clearly, our estimate of harbour porpoise abundance is
negatively biased to some degree.
For the Dall’s porpoise, responsive movement has been
found to be a large problem in some studies (Buckland and
Turnock, 1992), and not in others (Barlow, 1995). The
behavioural data presented here suggest that Dall’s porpoise
were detected before most animals started to respond to the
ship, which may reflect the tendency for observers to search
as instructed, namely ahead of the ship using binoculars, or
suggest that the small, relatively quiet research vessel
elicited weaker avoidance responses from Dall’s porpoise
than has been reported from larger ships. The assumption
that g(0)=1 no doubt did introduce some negative bias into
the abundance estimate, but trackline detection probability
may be higher for Dall’s porpoise than for the more cryptic
harbour porpoise. Other studies have reported estimates of
g(0) for Dall’s porpoise of 0.78 (Barlow, 1995) and 0.6
(Buckland and Turnock, 1992).
For the Pacific white-sided dolphin, the observed spike in
detections near zero distance (Fig. 2) was thought to have
arisen due to attractive movement, as this species is known
to bow-ride; the behavioural and animal heading data
indicated a responsive movement problem. An attempt to
correct for this was done by fitting a half-normal detection
function to the data (that is, by not fitting the spike). As a
result a much lower estimated pwas obtained than the data
would otherwise indicate (because AIC favoured the hazard
rate model). However, if the spike occurred due to a sharp
decline in detection probability with increasing distance,
rather than responsive movement, then the decision not to fit
the spike means that abundance would be underestimated.
Alternatively, if animals were attracted in from far outside
the surveyed strip, then overestimation may have occurred,
as the encounter rate will be higher than it would be if there
were no responsive movement. These issues need to be
explored further with future double-platform data collection
in which one team searches farther ahead of the ship than the
primary platform was able to do, to assess the point at which
responsive movement occurs, and to correct for it (Dawson
et al., 2004; Palka and Hammond, 2001). In contrast to the
responsive movement problem, it is not expected that the
g(0)=1 assumption introduced major bias to the dolphin
abundance estimate. Barlow (1995) reported an estimate of
g(0) for large delphinids of 0.736 for small groups (1-20
animals) and 1.0 for large (>20) groups; the mean observed
school size was approximately 19 animals (Table 3). The
best estimate of abundance for this species is the highest
estimated for any cetacean species in the area, and we
concur with Heise (1997) that Pacific white-sided dolphins
are the most abundant cetacean in BC coastal waters in
summer months.
In contrast to the small cetacean species, the abundance
estimates for whale species seem robust to the responsive
movement and g(0)=1 assumptions. The humpback whale
abundance estimate should be reasonably robust. The
detection function possessed a shoulder and responsive
2
http://hawaiihumpbackwhale.noaa.gov/special_offerings/sp_off/
splash.html
JNL339 015-028:Layout 1 29/10/07 07:16 Page 25
movement was not found to be a problem. Negative bias due
to g(0)<1 was unlikely to have been large. Estimates of g(0)
for large whales due to perception bias have ranged from
0.9-1.0 (Barlow, 1995; Williams et al., 2006) under calm
conditions. Detection probability for fin whales should be
near 1 (Barlow, 1995; Williams et al., 2006). For fin whales,
the primary issue here is simply one of small sample size.
The abundance estimate for killer whales is very
preliminary, but in light of the small sample size, the
detection function fitted the data surprisingly well (Fig. 2).
Ideally, abundance would have been estimated with more
than 18 sightings, but at this rate, one would need two more
seasons to have a large enough sample size to begin to
assess model fit. However, this relatively imprecise estimate
(161) is of the right order of magnitude (true population size
in 2004 was 219, Cetacean Research Program, Pacific
Biological Station, DFO), and the confidence interval (45-
574) comfortably spans the true population size. For such a
small and highly clustered population though, one would not
choose distance sampling a priori as the most efficient
method to estimate population size. Since identification
photographs for each killer whale encounter were collected
off-effort, it is possible to compare distance sampling to
capture-recapture estimates of abundance for this species, in
addition to the true, known population size.
For common minke whales, the preliminary abundance
estimate is admittedly tenuous given the small sample size
and should be interpreted with caution. But at this stage,
increasing the precision of the estimate for this species is of
greater concern than for reducing bias. A strip transect was
fitted, however responsive movement could have caused
animals to enter or leave the strip prior to detection. While
responsive movement is not thought to be a problem, a
larger sample size will be required to address this.
Meanwhile, strip transects of 300 and 150m gave roughly
the same point estimates of abundance (388 and 475,
respectively), with different precision (as one would expect
when sample size varies: 95% CIs 222-680 and 221-1,020,
respectively). Detection probability was certainly less than 1
for minke whales. Perception bias may not have introduced
much negative bias in the abundance estimate. Under
excellent sighting conditions, Williams et al. (2006)
estimated g(0) for Antarctic minke whales to be
approximately 0.9 due to perception bias and Barlow (1995)
has reported an estimate of g(0) due to perception bias for
small whales of 0.840. However, attempts to address
availability bias have produced substantially lower
estimates of g(0) for minke whales. Skaug and Schweder
(1999) have estimated that 56 to 68% of minke whales on
the trackline may be missed by observers during North
Atlantic surveys.
The pinniped abundance estimates will require increased
sample size in the case of Steller sea lions and elephant seals
and estimates of g(0) for all three species.
Nevertheless, these preliminary estimates are a good
starting point for discussion, which in turn will help to
prioritise future research activities. If it should turn out that
precise, accurate estimates of absolute abundance are
needed for management purposes – if, for example, bycatch
of harbour porpoise were found to be near some threshold
like Potential Biological Removal (Wade, 1998) that might
trigger some management action based on the abundance
estimate reported (e.g. Hammond et al., 2002) – then it
would be worth investing more time and money to address
responsive movement and g(0). A similar small-boat survey
was conducted recently for coastal dolphins in New Zealand
(Dawson et al., 2004), where simultaneous helicopter and
boat surveys were used to calculate a correction factor for
responsive movement and uncertain trackline detection. A
similar approach could be used in BC in future. Apilot study
was initiated in 2005 to conduct double-platform trials to
begin to assess how much bias the g(0)=1 assumption might
introduce, so that the abundance estimates reported here can
be adjusted accordingly. As expected, this attempt to isolate
independent platforms on a small boat was problematic and
consequently the sample size is currently too small to permit
statistical analyses; however this did work better on the
powerboat (2005) than on the sailboat (2004). Double-
platform data will continue to be collected as opportunities,
platform and funding permit, and we hope to report on this
in future. In the meantime, we plan to explore the sensitivity
of PBR calculations to varying levels of bias and precision
in our porpoise abundance estimates in a simulation
framework.
Another consideration affecting the reliability of the
abundance estimates is the need for a large enough sample
size of observations for fitting the detection function. This
may be a particularly important issue for the estimates of
abundance for small cetaceans. The decision to use AIC to
guide the choice of detection function (unless there was
good reason not to do so) seems sensible, but requires closer
inspection. For Dall’s porpoise, the two detection functions
with lowest AIC (half-normal and uniform) gave similar
results, but one with only marginally higher AIC (hazard
rate) would have resulted in a much higher estimate of
animal density. For Pacific white-sided dolphins, the
behavioural data gave strong support for not choosing the
detection function with the lowest AIC, that is, for not fitting
the spike near zero distance, but it is unable to determine
whether the analyses addressed responsive movement issues
completely. For that, work on better double-platform data
collection or the use of higher power binoculars are
required. For harbour porpoise, the detection function
lacked a shoulder, but the next-best detection function
produced a much worse AIC and gave similar results in any
case.
Future work
The next steps are to improve the precision and accuracy of
the abundance estimates through more data collection
(funding permitting) and additional analyses and
simulations, and to begin to apply the distribution data to
define areas of important habitat for at-risk species. In terms
of additional fieldwork, one obvious need is to increase
sample sizes of observations for common minke, fin and
killer whales. Secondly, it is hoped that the 2005 pilot study
can be expanded to collect double-platform data, to be used
to address outstanding issues of g(0)<1 and responsive
movement.
Over time, one possibility might be to expand the
analyses to use multiple covariate distance sampling
methods (MCDS); (Marques and Buckland, 2003; Marques,
2001), in which models allow factors (such as group size or
sighting conditions) to alter the scale of the detection
function without affecting its shape. MCDS offers a
parsimonious intermediate between full stratum-level
stratification of the detection function and full pooling. Note
that covariates were tried and found that no MCDS model
was selected over CDS, with two exceptions. The only
analyses where AIC favoured the use of MCDS were for
Dall’s porpoise, for which it made little difference to the
estimates, and for harbour seals, where stratification was
preferred because it was suspected that the different sighting
26 WILLIAMS & THOMAS
:
MARINE MAMMALS IN COASTAL WATERS OF BRITISH COLUMBIA
JNL339 015-028:Layout 1 29/10/07 07:16 Page 26
processes for seals on land versus in water could have
resulted in different shapes for the detection functions. More
specifically, the Dall’s porpoise MCDS analysis suggested
that the subjective code describing sightability conditions
might have improved the detection function fit, but the
stratum-specific estimates of density using sightability were
very similar to the pooled estimate, so it would not have
make any difference to the results. Overall, we conclude that
if a bias exists due to pooling, we expect it to be small
because there was little variability in detectability
between strata. Year was specifically looked at as a
covariate in stratum 1 and no support for including it was
found, even for harbour seals, for which there was a large
sample size.
A more salient application of MCDS methods might be to
improve the estimates for species with very small sample
sizes, by using MCDS to model detection function for
multiple species simultaneously. It is possible to combine
species with detection functions thought to have similar
shapes, however the initial investigation showed that single-
species models were preferred. As our sample sizes and
statistical power increase over time, it is suspected that
MCDS methods may produce slightly better inferences at
the stratum level for some species.
A final general comment involves the desire for a
‘sufficient’ number of transects or primary sampling units
(PSUs) per stratum. The number of replicate transects was
sufficient for strata 1-3, but there were only 5 PSUs in
stratum 4, so the variance estimates there were high (Table
4) and may be unreliable. However, surveying the mainland
inlets was a secondary goal, and the overall global variance
estimates are quite reasonable (Table 4).
Two additional lines of work are being investigated in
terms of reanalysis of existing data. The first is to explore
applications of improved design-based variance estimators
being developed by Buckland and colleagues (R. Fewster
and S. Buckland, pers. comm.) for systematic (c.f. random)
sampling designs. The other area is to apply spatial habitat
models of encounter rate, both to reduce variance and also
to uncover potential habitat associations (e.g. Hedley et al.,
1999; Williams et al., 2006). The long-term goal is to build
habitat models that help to understand species-specific
factors determining marine mammal distribution in our
study area. Ancillary data (namely water temperature and
salinity to a depth of 150m, and zooplankton samples) were
collected simultaneously on these surveys, which will be
used to model factors that influence cetacean and pinniped
distribution and density. Such results could be used to
inform marine planning processes and to identify candidate
areas for protection. Over time, it is hoped that the replicate
surveys will allow predictive models of animal distribution
to be constructed so that together, estimates of interannual
variability and model uncertainty can inform a quantitative
risk assessment framework for exposure of marine
mammals to anthropogenic activity.
In spite of the work remaining to be done to improve the
estimates from this small-boat survey, this report provides
the first comprehensive line transect survey of marine
mammals in these waters. Such surveys, undertaken with a
limited budget, can still be valuable when carefully designed
(Thomas et al., 2007), planned and implemented, as was the
case in this low-cost university-NGO partnership. We hope
that these results form a baseline against which population
trends may be measured in future, and that in the meantime,
these preliminary results can be of use in regional, national
and international efforts to study, conserve and manage
marine mammal populations.
ACKNOWLEDGEMENTS
We thank many people and agencies that provided Raincoast
Conservation Society and the authors with logistical or
financial support (including Bullitt Foundation, Vancouver
Foundation, Marisla Foundation, McLean Foundation, The
Russell Family Foundation, Endswell, Mountain Equipment
Co-op, Patagonia, The Jane Marcher Foundation, Willow
Grove Foundation, Canadian Whale Institute, the Hanson
and Clark families, Dave German (Fathom Expeditions),
BaseCamp Adventure Outfitters, and anonymous donors).
Brian Falconer, Erin Nyhan, Stephen Anstee, Heidi
Krajewsky, Misty MacDuffee, Olive Andrews, Chad
Malloff, Nick Engelmann, Anneli Englund, Sonja Heinrich,
Hayley Shephard, George Hudson, Stephanie Fernandez,
Erin Ashe and Matt Farley assisted with fieldwork,
preparation and data collection. Jonathan Gordon provided
equipment, as did our colleagues at the Institute of Ocean
Sciences, Universities of British Columbia and Victoria and
Alert Bay Marine Laboratory (namely John Nelson, Bon
van Hardenberg, Eddy Carmack, Grant Ingram, Mike Berry,
Jamie Pepper and Nick Engelmann). Jay Barlow, Arne
Bjørge, Steve Dawson, Greg Donovan, Phil Hammond and
Paul Wade provided useful feedback on our survey design
and field methods. We acknowledge the developers of the
following free software programs: Logger 2000 developed
by the International Fund for Animal Welfare (IFAW);
Distance (www.ruwpa.st-and.ac.uk/distance) and R(www.r-
project.org). We thank Steve Buckland, John Calambokidis,
Tiago Marques, Arliss Winship, two anonymous reviewers
and members of the Scientific Committee of the
International Whaling Commission for feedback on an
earlier draft of this manuscript. All data presented here were
collected under a Marine Mammal/Species at Risk Act
Scientific License issued to RW by Fisheries and Oceans
Canada, thereby conforming to Canadian legal requirements
including those relating to conservation and animal
welfare.
REFERENCES
Anon. 2000. Pacific white-sided dolphin (L agenorhynchus
obliquidens): North Pacific stock. Stock Assessment Reports (SAR).
National Marine Fisheries Service, NOAA. 4pp. [Available from:
www.noaa.gov].
Baird, R.W. 2001a. Status of harbour seals, Phoca vitulina, in Canada.
Can. Field-Nat. 115(4): 663-75.
Baird, R.W. 2001b. Status of killer whales, Orcinus orca, in Canada.
Can. Field-Nat. 115: 676-701.
Baird, R.W. 2003a. Update COSEWIC status report on the harbour
porpoise Phocoena phocoena (Pacific Ocean population) in Canada,
in COSEWIC assessment and update status report on the harbour
porpoise Phocoena phocoena (Pacific Ocean population) in Canada.
Committee on the Status of Endangered Wildlife in Canada, Ottawa,
Canada. 22pp.
Baird, R.W. 2003b. Update COSEWIC status report on the humpback
whale Megaptera novaeangliae in Canada, in COSEWIC assessment
and update status report on the humpback whale Megaptera
novaeangliae in Canada. Committee on the Status of Endangered
Wildlife in Canada, Ottawa, Canada. 25pp.
Barlow, J. 1995. The abundance of cetaceans in California coastal
waters: I. Ship surveys in the summer/fall 1991. Fish. Bull. 93: 1-14.
Barlow, J., Gerrodette, T. and Silber, G. 1997. First estimates of vaquita
abundance. Mar. Mammal Sci. 13(1): 44-58.
Bigg, M.A., Olesiuk, P.F., Ellis, G.M., Ford, J.K.B. and Balcomb, K.C.
1990. Social organization and genealogy of resident killer whales
(Orcinus orca) in the coastal waters of British Columbia and
Washington State. Rep. int. Whal. Commn (special issue) 12: 383-
405.
Bigg, M.A. and Wolman, A.A. 1975. Live-capture killer whale (Orcinus
orca) fishery, British Columbia and Washington, 1962-73. J. Fish.
Res. Bd Can. 32(7): 1,213-1,223.
Buckland, S.T., Anderson, D.R., Burnham, K.P., Laake, J.L., Borchers,
D.L. and Thomas, L. 2001. Introduction to Distance Sampling:
J. CETACEAN RES. MANAGE. 9(1):15–28, 2007
27
JNL339 015-028:Layout 1 29/10/07 07:16 Page 27
Estimating Abundance of Biological Populations. Oxford University
Press, Oxford, UK. vi+xv+432pp.
Buckland, S.T., Anderson, D.R., Burnham, K.P., Laake, J.L., Borchers,
D.L. and Thomas, L. 2004. Advanced Distance Sampling: Estimating
Abundance of Biological Populations. Oxford University Press,
Oxford, UK. 416pp.
Buckland, S.T. and Turnock, B.J. 1992. A robust line transect method.
Biometrics 48: 901-09.
Calambokidis, J., Steiger, G.H., Straley, T.J., Quinn, L.M., Herman,
L.M., Cerchio, S., Salden, D.R., Yamaguchi, M., Sato, F., Urbán R,
J., Jacobsen, J., von Ziegesar, O., Balcomb, K.C., Gabriele, C.M.,
Dahlheim, M.E., Higashi, N., Uchida, S., Ford, J.K.B., Miyamura, Y.,
Ladrón de Guevara P, P., Mizroch, S., Schlender, A. and Rasmussen,
K. 1997. Abundance and population structure of humpback whales in
the North Pacific basin. Contract report submitted to NMFS. 72pp.
COSEWIC. 2003a. COSEWIC assessment and update status report on
the harbour porpoise Phocoena phocoena (Pacific Ocean population)
in Canada. Committee on the Status of Endangered Wildlife in
Canada, Ottawa, Canada. 22pp. www.sararegistry.gc.ca.
status/status_e.cfm.
COSEWIC. 2003b. COSEWIC assessment and update status report on
the humpback whale Megaptera novaeangliae in Canada. Committee
on the Status of Endangered Wildlife in Canada, Ottawa, Canada.
25pp.
COSEWIC. 2003c. COSEWIC assessment and update status report on
the Steller sea lion Eumetopias jubatus in Canada. Committee on the
Status of Endangered Wildlife in Canada, Ottawa, Canada. 47pp.
COSEWIC. 2004. COSEWIC assessment and update status report on
the gray whale (Eastern North Pacific population) Eschrichtius
robustus in Canada. Committee on the Status of Endangered Wildlife
in Canada, Ottawa, Canada. 31pp. www.sararegistry.gc.ca/
status/status_e.cfm.
COSEWIC. 2005. COSEWIC assessment and update status report on
the fin whale Balaenoptera physalus in Canada. Committee on the
Status of Endangered Wildlife in Canada, Ottawa, Canada. 37pp.
Dawson, S.M., Slooten, E., Du Fresne, S.P., Wade, P.R. and Clement, D.
2004. Small-boat surveys for coastal dolphins: line-transect surveys
of Hector’s dolphins. Fish. Bull. 201: 441-51.
Dorsey, E.M., Stern, J.S., Hoelzel, A.R. and Jacobsen, J. 1990. Minke
whales (Balaenoptera acutorostrata) from the west coast of North
America: individual recognition and small-scale site fidelity. Rep. int.
Whal. Commn (special issue) 12: 357-68.
Ford, J.K.B., Ellis, G.M. and Balcomb, K.C. 2000. Killer Whales: The
Natural History and Genealogy of Orcinus orca in British Columbia
and Washington. 2nd ed. UBC Press, Vancouver. 102pp.
Ford, J.K.B., Ellis, G.M., Barrett-Lennard, L.G. and Morton, A.B. 1998.
Dietary specialization in two sympatric populations of killer whales
(Orcinus orca) in coastal British Columbia and adjacent waters. Can.
J. Zool. 76: 1456-71.
Gregr, E.J., Nichol, L., Ford, J.K.B., Ellis, G. and Trites, A. 2000.
Migration and population structure of northeast Pacific whales of
coastal British Columbia: an analysis of commercial whaling records
from 1908-1967. Mar. Mammal Sci. 16(4): 699-727.
Gregr, E.J. and Trites, A.W. 2001. Predictions of critical habitat for five
whale species in the waters of coastal British Columbia. Can. J. Fish.
Aquat. Sci. 58: 1269-85.
Hall, A., Ellis, G. and Trites, A.W. 2002. Harbour porpoise interactions
with the 2001 selective salmon fisheries in southern British Columbia
and license holder reported small cetacean bycatch. Selective Salmon
Fisheries Science Program Report, Fisheries and Oceans Canada.
51pp.
Hall, A.M. 2004. Seasonal abundance, distribution and prey species of
harbour porpoise (Phocoena phocoena) in southern Vancouver Island
waters, University of British Columbia, Department of Zoology,
Vancouver. 109pp.
Hammond, P.S., Berggren, P., Benke, H., Borchers, D.L., Collet, A.,
Heide-Jørgensen, M.P., Heimlich, S., Hiby, A.R., Leopold, M.F. and
Øien, N. 2002. Abundance of harbour porpoise and other cetaceans in
the North Sea and adjacent waters. J. Appl. Ecol. 39(2): 361-76.
Hedley, S., Buckland, S.T. and Borchers, D.L. 1999. Spatial modelling
from line transect data. J. Cetacean Res. Manage. 1(3): 255-64.
Heise, K. 1997. Life history and population parameters of Pacific white-
sided dolphins (Lagenorhynchus obliquidens). Rep. int. Whal.
Commn 47: 817-25.
Jefferson, T.A. 1990. Status of Dall’s porpoise, Phocoenoides dalli, in
Canada. Can. Field-Nat. 104(1): 112-16.
Larkin, P.A. 1996. Concepts and issues in marine ecosystem
management. Rev. Fish Biol. Fish. 6: 139-64.
Marques, F.F. and Buckland, S.T. 2003. Incorporating covariates into
standard line transect analyses. Biometrics 59: 924-35.
Marques, F.F.C. 2001. Estimating wildlife distribution and abundance
from line transect surveys conducted from platforms of opportunity,
School of Mathematics and Statistics. PhD thesis, University of St
Andrews, UK. 57pp.
Morton, A. 2000. Occurrence, photo-identification and prey of Pacific
white-sided dolphins (Lagenorhyncus obliquidens) in the Broughton
Archipelago, Canada 1984-1998. Mar. Mammal Sci. 16(1): 80-93.
Olesiuk, P., Bigg, M.A. and Ellis, G.M. 1990. Life history and
population dynamics of resident killer whales (Orcinus orca) in the
coastal waters of British Columbia and Washington State. Rep. int.
Whal. Commn (special issue) 12: 209-43.
Olesiuk, P.F. 1999. An assessment of the status of harbour seals (Phoca
vitulina) in British Columbia. Canadian Stock Assessment Secretariat
Research Document 99/33. ISSN 1480-4483, Ottawa, 71pp.
Palka, D. 2000. Abundance of the Gulf of Maine/Bay of Fundy harbor
porpoise based on shipboard and aerial surveys during 1999.
NOAA/NMFS/NEFSC Ref. Doc. 00-07 (unpublished). 29pp.
[Available from: National Marine Fisheries Service, 166 Water
Street, Woods Hole, MA 02543-1026].
Palka, D.L. and Hammond, P.S. 2001. Accounting for responsive
movement in line transect estimates of abundance. Can. J. Fish.
Aquat. Sci. 58: 777-87.
Ross, P.S., Ellis, G.M., Ikonomou, M.G., Barrett-Lennard, L.G. and
Addison, R.F. 2000. High PCB concentrations in free-ranging Pacific
killer whales, Orcinus orca: effects of age, sex and dietary
preference. Mar. Poll. Bull. 40(6): 504-15.
Royal Society of Canada. 2004. Report of the expert panel on science
issues related to oil and gas activities, offshore British Columbia: an
expert panel prepared by the Royal Society of Canada at the request
of Natural Resources Canada, Ottawa, Canada. 178pp.
Skaug, H.J. and Schweder, T. 1999. Hazard models for line transect
surveys with independent observers. Biometrics 55(1): 29-36.
Stacey, P.J. and Baird, R.W. 1991. Status of the Pacific white-sided
dolphin, Lagenorhynchus obliquidens, in Canada. Can. Field-Nat.
105(2): 219-32.
Thomas, L., Laake, J.L., Strindberg, S., Marques, F.F.C., Buckland,
S.T., Borchers, D.L., Anderson, D.R., Burnham, K.P., Hedley, S.L.,
Pollard, J.H., Bishop, J.R.B. and Marques, T.A. 2005. Distance 5.0.
Beta 5. Research Unit for Wildlife PopulationAssessment, University
of St Andrews, UK. [Available from: http ://www.ruwpa.st-
and.ac.uk/distance/].
Thomas, L., Williams, R. and Sandilands, D. 2007. Designing line
transect surveys for complex survey regions. Journal of Cetacean
Research and Management 9(1): [This volume].
Wade, P.R. 1998. Calculating limits to the allowable human-caused
mortality of cetaceans and pinnipeds. Mar. Mammal Sci. 14(1): 1-37.
Williams, R., Bain, D.E., Ford, J.K.B. and Trites, A.W. 2002a.
Behavioural responses of male killer whales to a ‘leapfrogging’
vessel. J. Cetacean Res. Manage. 4(3): 305-10.
Williams, R.M., Trites, A.W. and Bain, D.E. 2002b. Behavioural
responses of killer whales to whale-watching traffic: opportunistic
observations and experimental approaches. J. Zool. (Lond.) 256: 255-
70
Williams, R., Hedley, S.L. and Hammond, P.S. 2006. Modeling
distribution and abundance of Antarctic baleen whales using ships of
opportunity. Ecology and Society 11(1): [online:
http://www.ecologyandsociety.org/vol11/iss1/art1/].
Williams, R. and Lusseau, D. 2006. A killer whale social network is
vulnerable to targeted removals. Biology Letters 2: 497-500.
Williams, R., Lusseau, D. and Hammond, P.S. 2006. Estimating relative
energetic costs of human disturbance to killer (Orcinus orca). Biol.
Conserv. 133(3): 301-11.
Willis, P.M. and Baird, R.W. 1998. Sightings and strandings of beaked
whales on the west coast of Canada. Aquat. Mamm. 24(1): 21-25.
Date received: June 2006.
Date accepted: January 2007.
28 WILLIAMS & THOMAS
:
MARINE MAMMALS IN COASTAL WATERS OF BRITISH COLUMBIA
JNL339 015-028:Layout 1 29/10/07 07:16 Page 28
