Northeast Fisheries Science Center Reference Document 06-03
Summer Abundance Estimates
of Cetaceans
in US North Atlantic
Navy Operating Areas
by Debra L. Palka
National Marine Fisheries Serv.,
166 Water St.,
Woods
Hole, MA 02543
Print
publication date March 2006;
web version posted March 30, 2006
Citation: Palka DL. 2006. Summer abundance estimates of cetaceans in US North Atlantic Navy Operating
Areas. US Dep Commer, Northeast Fish Sci Cent Ref Doc. 06-03; 41 p.
Download complete PDF/print version
ABSTRACT:
The
US Fleet Forces Command, Department of the Navy, contracted the consulting
firm Geo-Marine, Inc. (GMI) to generate technical reports that provide
marine mammal and sea turtle density estimates for Navy operating areas.
Some of the needed density estimates are for areas off the northeast
US coast, an area that has been surveyed by marine mammal abundance surveys
conducted by the Northeast Fisheries Science Center. GMI requested my
aid in preparing summer density estimates for the northeast operating
areas (NE OPAREA) using data collected from 1998, 1999, 2002, and 2004.
The Gulf of Maine Central and Offshore NE OPAREAs had the highest numbers
of cetaceans, although the NE OPAREAs with the highest densities (abundance
divided by area) were the Gulf of Maine North and Scotian NE OPAREAs
(both in Canadian waters). Within US waters, the stratum
with the highest density was the Gulf of Maine Central, followed by the
Shelf Central, Shelf West, and Georges Bank Central strata. The
strata with the lowest densities and lowest species diversity were the
Mid-Atlantic and Georges Bank West strata. The 2004 estimates appear
to be more representative of a springtime distribution or the transition
between spring and summer distributions, while the 2002 and earlier estimates
appear to be more representative of mid summer distributions.
INTRODUCTION
The US Fleet Forces Command, Department of the Navy, contracted
the environmental consulting firm Geo-Marine, Inc. (GMI) to generate
technical reports that provide marine mammal and sea turtle density
estimates for Navy operating areas (OPAREAs). These density estimates
will be used for the purposes of Navy environmental planning and compliance
and will serve as the basis for future documentation under
federal reporting requirements.
Some of the needed density estimates
are for OPAREAs off the northeast US coast (NE OPAREAs), an area that
has been surveyed by marine mammal abundance surveys conducted by the
Northeast Fisheries Science Center (NEFSC). GMI requested my aid in
providing survey data and in preparing summer density estimates for
the NE OPAREA region. In response to this request, I re-analyzed data
that were previously collected to estimate abundance of cetaceans detected
within and beyond the NE OPAREAs (Figure
1; Table 1). The shipboard and aerial line
transect data used in this analysis were collected during the summers
of 1998 (Palka 2005a), 1999 (Palka 2000), 2002 (Palka 2005b; in review),
and 2004 (Palka in review).
METHODS
Field
methods for shipboard surveys
Shipboard
data included in this analysis came from the NEFSC 1998, 1999, and 2004
abundance surveys (Figures 2 to 4). The 1998
and 2004 shipboard surveys (Table 2) covered similar
areas: an area bounded to the south at the 37°N latitudinal line
(off Chesapeake Bay, Virginia), to the north by Georges Bank (41°N),
to the west at 74°W, and
to the east at the US-Canadian EEZ line at 65° 30’W. This
covered waters between approximately the 100 m and 4000 m isobaths. The
original study area was divided into two strata defined by bio-geographic
habitats: a shelf edge stratum, and an offshore stratum that was offshore
of the shelf and included the Gulf Stream. The shelf bio-geographic
stratum is the sum of the following NE OPAREAs: Shelf West, Shelf Central,
and Shelf East. The offshore bio-geographic stratum and the offshore
NE OPAREA are similar. Saw-toothed transects were placed to cross the
bathymetry gradient and were started at a random point within each stratum.
The 1999 shipboard survey (Table 2) covered shallow waters of the northern
Gulf of Maine (to approximately the 100m depth contour), western Scotian
Shelf and lower Bay of Fundy (Figure 3). The coastal sections of
the Gulf of Maine Central NE OPAREA stratum was surveyed in 1999 by a
ship, while the offshore section was surveyed by a plane (Figure
2; see
more details about the aerial survey in the next section).
On all of the shipboard surveys, two visual observer teams on independent
platforms simultaneously collected data. Data from both teams were
needed to estimate g(0), the probability of detecting a group
on the track line. Each team consisted of three observers on duty
and one observer at rest. Each platform had three observation stations. Observers
changed stations every 30 minutes. Observers searched during daylight
hours (usually 6 am to 6 pm with one hour off for lunch), when weather
permitted (i.e., when Beaufort sea state conditions were below five,
and when there was at least 3.7 km of visibility). Observers searched
the area between 90° on both sides of the transect line, and from
the ship to the horizon.
Because the ships and target species differed between
the three shipboard surveys, the locations of the platforms and searching
tools also differed (Table 3). This was done
to ensure as many animal groups as possible were detected. In the
lower density pelagic surveys (1998 and 2004), high-powered binoculars
were used by two of the three observers on both teams, while the third
on-effort observer searched using naked eye and also recorded the data
from all the observation stations on that team. In
the higher density coastal survey (1999), all observers on both teams
used naked eye and recorded their own sightings.
On all three shipboard surveys, data collected included information
on sightings, effort, and environmental factors. For each cetacean
group detected, sightings data included time, ship’s latitude and
longitude, bearing between the transect line and line of sight to the
location of the group, radial distance between the ship and the center
of the group, species composition, group size (best high and low estimate),
swim direction (0° indicates swimming parallel to the track line
in the direction the ship was traveling, 90° indicates swimming perpendicular
to the track line and towards the right, etc.), behavior (swimming, charging,
milling, etc.), and cue (factor that attracted the observer to the group:
body, splash, blow, etc.). When binoculars were used, bearings
were measured using angle rings around the tripod-mounted binoculars
and radial distances were measured using reticles in the eyepiece of
the binoculars. When
the naked eye was used, bearings were measured using calibrated polaruses
that were mounted in front of each observer, and radial distances were
estimated visually. All observers were trained and tested to ensure
accurate radial distances. The “best” estimate for
group size was used in the abundance estimates because this value was
the result of assessing the group size as often as possible as the group
passed by the ship. Species were identified to the lowest taxonomic level
possible. When not possible to reliably distinguish an animal to the
species level, species groupings were used, such as "pilot whale" spp.,
which could be either a short-finned (Globicephala macrorhynchus)
or long-finned (G. melas) pilot whale. Another example is
"unidentified dolphin," which could be any dolphin species. Groups identified
to a level with the word “unidentified” were included in
abundance estimates that were separate from abundance estimates derived
from groups identified to a specific species. Therefore, all abundance
estimates of a specific species are negatively biased because an unknown
proportion of groups of that species were detected but were included
in the unidentified abundance estimate.
When high-powered binoculars were used (1998 and 2004), it was not always
possible to
confirm the species identification or group size. For many of the unidentified
groups within
about 5.5 km (3 nautical miles) of the ship, the ship went off-effort
and approached the group to
a distance from which it was possible to confirm the identification and
group size. When a group
was approached, both teams were off-effort, so any additional sightings
were not recorded. On-effort
sightings were resumed when the ship was back on the original track line.
When naked
eye was used, the ship did not go off-effort to identify species.
At the beginning of each track line segment (called a leg) and when
conditions changed, effort and environmental data were collected. These
data included: time, observer at each observation station, ship’s
position (latitude and longitude), ship’s speed and course, wind
speed and direction, water depth, surface temperature, air temperature,
swell height and direction (relative to the ship’s track line),
Beaufort sea state (0 to 4.9 in 0.1 increments), direction of sun (relative
to the ship’s track line), magnitude of glare (none, slight, moderate,
and excessive), and distance with clear visibility.
Field methods for aerial surveys
Aerial data included in this analysis came from the NEFSC 1998, 1999,
2002, and 2004 summer abundance surveys (Figures 2 to
5). All of these
aerial surveys were conducted on the NOAA DeHavilland Twin Otter DHC-6,
Series 300 aircraft (Table 2). The portion of the study area covered
by all the aerial surveys extended from waters south of Rhode Island,
northward through the Gulf of Maine to the lower Bay of Fundy and to
Scotian waters south of Nova Scotia. The 1998 and 2004 aerial surveys
also covered shelf waters along the Mid-Atlantic states of New York to
Virginia. The original aerial survey study areas were divided into
bio-geographic habitat strata: a southern region below Long Island, NY
(Mid-Atlantic NE OPAREA), a central region consisting of Georges Bank
(NE OPAREAs Georges East, Georges Central, and Georges West), and a northern
region consisting of the Gulf of Maine, lower Bay of Fundy, and southern
Scotian shelf (NE OPAREAs Gulf of Maine (GOM) south, GOM central, GOM
north, and Scotian).
During all surveys, track lines were flown 182 m (600 feet) above the
water surface, at about 200 km/hr (110 knots), when Beaufort sea state
conditions were below four, and when there was at least 3.7 km (2 nmi)
of visibility. During all surveys, there were two pilots and five
scientists onboard. Three scientists were observers searching for animals
using the naked eye; the fourth scientist was at rest; and the fifth
scientist recorded the data. The recorder worked at this position
for the entire survey. The other four scientists rotated between
the three observation stations and the rest station. Rotations
occurred at the end of track lines or about every 30-40 minutes. Two
observers, located behind the pilots, looked through side-viewing large
bubble windows, where one observer was on each side of the plane. The
third observer was at the back of the plane lying on the ground to look
through a belly window. The belly window observer was limited to
approximately a 28° view on both sides of the track line. The
bubble window observers concentrated searching from straight down (0°)
up to about 45° from the track line; the area from 45° to the horizon
(90°) was also searched, though less frequently. Handheld binoculars
were available to confirm species identifications and group sizes, if
desired.
During all surveys, when an animal group was observed the following
data were collected: time group passed perpendicular to the window; species
identification; group size; angle of declination from the track line
(measured by inclinometers or marks on the windows); cue (animal, splash,
blow, footprint, birds, vessel/gear, windrows, or other); swim direction
(0E indicates swimming parallel to the track line in the direction the
plane was flying, 90E indicates swimming perpendicular to the track line
and towards the right, etc.); if the animal appeared to react to the
plane (yes or no); if the animal was diving (yes or no), and; comments,
if any.
At the beginning of each leg and when conditions changed, the following
data were collected: initials of persons in the two pilot seats and three
observation stations; Beaufort sea state (0 to 3.9 in 0.1 increments);
water color (deep blue, blue, greenish blue, green, light green, yellowish
green, yellow green, green yellow, greenish yellow, or yellow); percentage
of cloud cover (0-100%); angle glare started and ended at (0-359°,
where 0° was the track line in the direction of flight and 90° was
directly abeam to the right side of the track line, etc.); magnitude
of glare (none, slight, moderate, and excessive); and subjective overall
quality for each observer (excellent, good, moderate, fair, and poor). Data
collected in poor conditions were not used in the abundance estimate.
To estimate g(0), the Hiby circle-back data
collection method (Hiby 1999) was used for harbor porpoise sightings
only during the 1998 survey, and for all species after that. The aerial
Hiby circle-back method is comparable to the two-team shipboard method. Both
methods result in data used to estimate g(0). The circle-back
method modified standard single-plane line-transect methods by circling
back and re-surveying a portion of the track line (Figure
6). The portions of track lines that
were re-surveyed were called “trailing” legs. The portions
of the track lines that initiated a circle were called “leading” legs,
and the portions of the track lines that were between the end of a trailing
leg and the beginning of the subsequent leading leg were called “single-plane” legs. As
in the case of two teams on a ship, g(0) can be estimated using
the aerial data collected during the leading and trailing legs, as they
are comparable to data collected by two teams. That is, data collected
on trailing legs corresponded to data from a second team, data collected
on leading legs corresponded to data from a primary team when a second
team was on-effort, and data collected on single-plane legs corresponded
to data collected by the primary team when the second team was off-effort.
The criterion that started a circle was a small group (
5
animals) of cetaceans or turtles that was the only sighting of the same
species within a 30 second time period. The circle-back procedure was
as follows (Figure 4):
- Time and location of an initial
sighting when it passed abeam of the plane was recorded and started
a 30-second timer (Point 1 in Figure 6),
- During
the 30 seconds, additional sightings were recorded. If more than
one additional sighting of the same species that triggered the circle
was recorded during this 30 seconds, then the circle-back procedure
was aborted, because the density may be too high to accurately determine
if a group of animals was the same group on both the leading and trailing
legs of the track line.
- At
the end of the 30 seconds, if the criterion in number 2 was passed,
the plane started to circle back and the observers went off-effort. The
time leaving the track line was recorded, which also started another
timer for 120 seconds (Point 2 in Figure 6).
- During
this 120 seconds the plane circled back 180° and traveled parallel
to the original track line about 1.5 km (0.8 nmi) away, in the opposite
direction, and on either side of the original track line.
- At
the end of the 120 seconds, the plane started to fly back to the
track line (Point 3 in Figure 6).
- When
the plane intercepted the original track line, the time was recorded,
observers went back on-effort, they started searching again,
and a 5-minute timer was started (Point 4 in Figure
6).
- All
sightings were then recorded.
- The
circle-back procedure was not initiated again until a sighting was
made after the 5-minute timer expired (Point 5 in Figure
6). This
was to ensure forward progress on the track line.
Shipboard analytical methods
In the original analyses for 1998, 1999, and 2004 shipboard data, abundance
estimates were calculated for large bio-geographic habitat strata (Palka
2000; 2005a; in review). The 1998 and 2004 data, collected while
surveying with high-powered binoculars, were investigated to determine
if animals responded to the ship. To estimate the abundance for
those species that demonstrated responsive movements, the Palka-Hammond
analytical method (Palka & Hammond 2001) was used. To estimate
the abundance of all other species, the direct-duplicate method (Palka
1995) was used. Covariates were investigated to determine if any
can improve the detection function of the 1998 (Palka 2005) and 2004
data (Palka in review).
To estimate abundance within the smaller NE OPAREA strata, the survey
track line and sighting data were first divided into the NE OPAREA strata. Track
line lengths, sighting rates and average group sizes within each NE OPAREA
stratum were then calculated using only the data with a NE OPAREA. Using
the direct-duplicate method (Palka 1995), the abundance (Nil)
for species l (within species group j) from NE OPAREA stratum i was
then estimated as the product of the density (Dil)
and area (Ai) of stratum i: Nil = Dil • Ai. Density
(Dil), was calculated as:
![](e1.gif)
(1)
where
Dupper |
= density, assuming g(0) = 1, using only the upper team’s
data in Eq. 2; |
Dlower |
= density, assuming g(0) = 1, using only the lower team’s
data in Eq. 2; |
Ddup |
= density, assuming g(0) = 1, using only duplicate sighting’s
data in Eq. 2. |
and
(2)
where
n |
= number of groups detected; |
E(s) |
= expected group size; |
L |
= length of transect line while on-effort; |
ESHW |
= Effective Strip Half Width;
= inverse of the sighting
probability density at zero perpendicular distance using data with
a perpendicular distance of less than or equal to w; |
w |
= maximum perpendicular distance used in analysis; |
k |
= team: upper=upper team, lower=lower team, dup =
duplicate sightings; |
j |
= species group; |
l |
= species; |
i |
= stratum. |
Duplicate sightings were defined as groups seen by both the upper
and lower teams, though not necessarily at exactly the same time. During
the analysis phase, the duplicate sightings were determined by a computer
program that compared the position of sightings detected by each team. Timing,
swim direction, and species identification were taken into account
when comparing the position of a sighting from one team to the predicted
position of previous sightings from the other team.
Species groups (j) were defined as an individual
species when there were a sufficient number of sightings for an individual
species. This occurred for offshore bottlenose dolphins, common
dolphins, Risso’s dolphins, white-sided dolphins, harbor porpoises,
humpback whales (during 1999 only), minke whales, right whales, and
sperm whales (Table 4). A species group
was defined as several species pooling together when it was not possible
to distinguish the species while in the field and/or there were an
insufficient number of sightings per individual species, and the species
within a species group had similar behaviors, and so approximately
equal chances of being detected. This occurred for pilot whales
(pooled short-finned and long-finned pilot whales); cryptic whales
(pooled beaked whales and Kogia spp.); and pelagic dolphins (pooled
spotted, spinner, and striped dolphins). During 1998 and 1999,
"large whales" was defined as pooling fin whales, sei whales, and animals
identified as either fin or sei whales. During 2004, "large whales"
was defined as pooling humpback whales, fin whales, sei whales, animals
identified as either a fin or sei whale, and animals identified as
an unknown large whale. Pilot
whales and beaked whales were pooled because it was not always possible
to positively identify the species. The other species groups
were formed because of insufficient sample sizes of each individual
species.
During 1998 and 2004, because binoculars were used, the angle and
radial distances could have been rounded when recorded (Palka in review). If
present, to correct for rounding error, recorded values were smeared
using Method 2 of Buckland and Anganuzzi (1988) before further analyses
were conducted.
The ESHW for each species group l and
team k (ESHWlk)
was estimated in the initial analyses using data pooled over all bio-geographic
habitat strata (Table 4). The 1998 and 2004
estimates of ESHW were
corrected for heterogeneities by incorporating significant covariates
into the detection function using the computer package DISTANCE 4 (Buckland et
al. 2001). The 1999 data have not yet been investigated to determine
if covariates improve the ESHW estimates. Model and covariate
selection was based on minimum Akaike Information Criterion (AIC).
The following animal-related covariates were investigated: group size,
group behavior (swimming, porpoising, and charging) and initial cue
(body, splash, and blow). The following survey-related covariates
were investigated: observer experience level (highest sighting rate,
intermediate sighting rate, lower sighting rate), Beaufort sea state
(0 to 4.9 in 0.1 increments), and wind speed. The following covariates
that could be either animal-related or survey-related were also investigated:
sea surface water temperature (SST), bottom depth, and bottom slope. In
addition, for the 2004 data, the time period the data were collected
-- time period 1 (23 June to 12 July 2004) versus time period 2 (16
July to 4 August 2004) -- was also included as a covariate to investigate
if the different sets of observers had an effect. A complete
description of the covariates is in Appendix 1 of
Palka (in review). Potential
detection function models without covariates included the uniform with
cosine adjustments, half-normal with polynomial or cosine adjustments,
and hazard-rate with polynomial or cosine adjustments. Potential
detection function models with covariates included the hazard rate
with polynomial or cosine adjustments and half-normal with polynomial
or cosine adjustments.
Estimates of g(0) for each species group and team was determined
in the initial analyses using data pooled over all bio-geographic habitat
strata (Table 5). The 1998 and 2004 g(0) estimates
included effects of covariates, when significant.
In cases of no duplicate sightings for a species group
within a NE OPAREA, it was not possible to use Eq. 1. Instead,
if within a NE OPAREA there were data from only one team, the abundance
estimate for that NE OPAREA was the product of the abundance estimated
from the data of the only team available and the species group-team-specific
estimate of g(0) as determined in the original analysis. If
within a NE OPAREA there were data from both teams, but no duplicates,
then the abundance estimate was the sum of the upper and lower team-g(0) corrected
abundance estimates.
It was assumed the best species abundance estimates were from the
larger bio-geographic habitat strata analysis and not the smaller NE
OPAREA strata analysis. Because the NE OPAREA strata were subsets
of the bio-geographic habitat strata, it was possible to correct the
NE OPAREA stratum-specific abundance estimates so that the sum of the
abundance from all the NE OPAREA strata equaled the sum from the applicable
bio-geographic habitat strata. That is, the best abundance within NE
OPAREA stratum i for species l (BNil) was
estimated as a proportion of the best abundance estimate derived from
the bio-geographic habitat strata (Nbiogeo):
(3)
where Nil was estimated using Eqs. 1 and 2 and Nj.biogeo was
estimated in the original analysis (Appendix I).
Coefficient of variations (CV) of the abundance estimates were
determined using bootstrap re-sampling techniques (Efron and Tibshirani
1993). Portions of the track line within each NE OPAREA were
re-sampled with replacement, so that the track line length within a
NE OPAREA from a bootstrap iteration was approximated equal to the
actual track line length within that NE OPAREA. The re-sampled
portions of the track line were defined as “legs” of effort
in which each leg was about 9.3 km (5 nmi) long, and where all
conditions (weather and position of observers) were similar. For
each of the 1000 bootstrap iterations, the abundance estimate of each
species within each stratum (
)
was estimated using the above equations. The CV of an abundance
estimate within a stratum was:
(4)
Aerial analytical methods
Abundance estimates from the 1998, 1999, 2002, and 2004 aerial
surveys were originally calculated using larger bio-geographic habitat
strata (Palka 2000; 2005b; in prep). To estimate abundance within
the smaller NE OPAREA strata, the survey data were first divided into
the NE OPAREA strata, then track line lengths, sighting rates, and
average group sizes within each NE OPAREA stratum were calculated.
Abundance of a species was calculated in a three-step procedure. First,
abundance uncorrected for g(0) was estimated for each year using
data collected during that year on the single-plane and leading (SL)
legs (i.e., corresponding to a conventional single plane survey). Second,
using only the 2002 and 2004 data, an estimate of g(0)leading was
derived from the data pooled over years collected by the “two
teams”; that is, from the leading and trailing legs. Finally,
to obtain an abundance estimate corrected for g(0) for all
years, g(0)leading obtained in step 2 was applied
to the abundance estimate derived from the SL legs, obtained in step
1. That is,
the same estimate of g(0) was applied to each year’s data.
Because the criteria used to start a circle was the detection of
a small group of animals (
5
animals), the estimate of g(0) was
only applicable to groups of animals with
5
animals. Consequently,
it was assumed the estimate of g(0) for group sizes of over
five was one.
In summary, abundance from year y in stratum i of
species l that belongs to species group j (Nily)
was estimated as:
(5)
where
nsmall.SL |
= number of groups 5 seen on the single
and leading (SL) legs; |
nlarge.SL |
= number of groups > 5 seen on the single and leading legs; |
E(s) small.SL |
= expected group size of groups 5
seen on the single and leading legs; |
E(s) large.SL |
= expected group size of groups > 5 seen on the single and leading
legs; |
ESHWj.SL |
= Effective half strip width of species group j using data
from the single-plane and leading legs;
= inverse of the sighting probability density at zero perpendicular
distance using data with a perpendicular distance of less than or equal
to w; |
w |
= maximum perpendicular distance used in analysis; |
LSL |
= length of transect line while on-effort on the single and leading
legs; |
Ai |
= area of stratum i |
i |
= stratum; |
j |
= species group of which species l belongs to; |
l |
= species; |
y |
= year: 1998, 1999, 2002 or 2004. |
and g(0) for all years, for species l that were in
groups of size 5 or less when detected during the leading legs was
estimated using data only from 2002 and 2004:
(6)
where
nsmall.dup |
= number of groups 5 seen on both
the leading and trailing legs; |
nsmall.trailing |
= number of groups 5
seen on the trailing legs; |
ESHWj.trailing |
= Effective half strip width of species group j using data
from the trailing legs; |
ESHWj.dup |
= Effective half strip width of species group j using data
from the duplicate sightings seen during the leading and trailing
legs |
|
|
Ideally, the estimates of E(s), ESHW,
and g(0) would
be estimated separately for each species. However, sample sizes
were small, especially for those relatively rare species. Thus,
estimates of g(0) and the ESHW were derived for groups
of species, sometimes over years. (Table 6). Species
groups were defined to meet the following criteria: include all species
detected, have a sufficiently large sample size, and have similarities
in the physical and behavioral attributes that affect the detectability
of these animals. Three species groups were defined. One
group consisted of only harbor porpoises. A second group was
small cetaceans: common dolphins, bottlenose dolphins, white-sided
dolphins, Risso’s dolphins, pilot whales, and unidentified dolphins. The
third group was large cetaceans: minke whales, fin whales, sei whales,
right whales, humpback whales, beaked whales, and unidentified whales.
Using the computer package DISTANCE (version 4), the various ESHWs
were estimated from a detection model of unbinned perpendicular distances. The
perpendicular distances were right truncated, when appropriate. For
the 2002 and 2004 data, the detection models accounted for heterogeneities
by including significant covariates, where a significant covariate
was a covariate that contributed to a significantly improved fit as
defined by the AIC criterion. Choices of covariates included
group size, initial cue (body of animal, splash, or blow), percent
cloud cover (0 to 100), Beaufort sea state (0 to 3.9 in 0.1 increments),
average subjective quality of the sighting conditions (excellent=1,
good=2, moderate=3, fair=4, poor=5, in 0.1 increments), water color
(deep blue, blue, greenish blue, green, light green, yellowish green,
yellow green, green yellow, greenish yellow or yellow) and species.
Potential models without covariates included the uniform with cosine
adjustments, half-normal with polynomial or cosine adjustments, and
hazard-rate with polynomial or cosine adjustments. Potential
models with covariates included the hazard rate with polynomial or
cosine adjustments and half-normal model with polynomial or cosine
adjustments.
It was assumed the best species abundance estimates were from the
larger bio-geographic habitat strata analysis and not the smaller NE
OPAREA strata analysis. Because the NE OPAREA strata were subsets
of the biogeographic habitat strata, it was possible to correct the
NE OPAREA stratum-specific abundance estimates so that the sum of the
abundance from all the NE OPAREA strata equaled the sum from the applicable
bio-geographic habitat strata. That is, the best abundance within NE
OPAREA stratum i for species l (BNil) was
estimated as a proportion of the best abundance estimate derived from
the bio-geographic habitat strata (Nbiogeo), as defined in Eq.
3.
The CVs of the abundance estimates were estimated using
the delta method (Buckland et al. 2001). Bootstrapping,
such as was done for the shipboard data, would have been preferred,
however; because of the complications of having leading and trailing
legs that have to be paired together, re-sampling the track lines was
difficult. Thus, the CV of the small and large abundance estimates
within NE OPAREA stratum i for species l that was within
species group j was estimated as:
![](e7a.gif)
(7)
where
(8)
smi equals the size of group m in stratum i,
and nvi equals the number of observations of species l within
stratum i, and
(9)
where there are p legs (track lines with no changes)
within stratum i, n =
nm,
T =
tm,, tm was
the length of the mth track line, and nm was
the number of groups detected on the mth track line.
RESULTS
Shipboard
surveys
The 1998 shipboard survey covered 4,270 km in the three Shelf strata
and the Offshore stratum (Table 1). The 1999 shipboard survey covered
2,382 km in the Gulf of Maine North, Gulf of Maine Central, and Scotian
strata. The 2004 shipboard survey covered 3,991 km of track lines in
the three Shelf strata and the Offshore stratum.
As determined in the original analyses, two species demonstrated responsive
movement. During the 2004 survey, Risso’s dolphins avoided
the ship. During the 1998 and 2004 surveys, pilot whales spp. were
attracted to the ship.
Estimates of ESHW for each species group for the upper team,
lower team, and duplicate sightings, as derived in the original analysis,
were generally in the 1500 to 3000 m range for the surveys using high-powered
binoculars (1998 and 2004; Table 4) and in the 200 to 1500 m range for
the 1999 survey where observers searched with naked eye. At least
one covariate was
found to be significant for at least one of the years for the detection
function of all species
investigated (Table
4). Group
size, Beaufort sea state (or wind speed), and cue were the most commonly
significant covariates.
As derived in the original analysis for the upper and lower teams, estimates
of g(0) for harbor porpoises and beaked whales were the
lowest (about 0.25), while some of the dolphins were the highest (about
0.8) (Table 5). Estimates of g(0) when
searching with the naked eye (during 1999) were, in general,
lower than estimates of g(0) when
using high-powered binoculars (during 1998 and 2004).
Aerial surveys
The 1998 aerial survey in the Mid-Atlantic stratum covered 1,734 km
of track lines. The 1999 aerial survey covered 3,741 km in the
Gulf of Maine Central, Gulf of Maine South, Georges East, Georges Central,
and Scotian stratum (Table 1). The 2002 aerial
survey covered 7,487 km in three Gulf of Maine strata, three Georges
Bank strata, two Shelf strata, and the Scotian stratum. The 2004 aerial
survey covered 3,991 km of track lines in the three Gulf of Maine, three
Georges Bank, three Shelf, and Offshore strata (Table
1).
From the pooled 2002 and 2004 aerial data, the original
estimates of the ESHW and g(0)leading were the
lowest for harbor porpoises, higher for small cetaceans, and highest
for the large whales (Table 6). Cue was a
significant covariate for the model of the detection function for large
whales, as was size for harbor porpoises. There were no significant
covariates for small cetaceans (Table
6).
Joint aerial and shipboard abundance estimates
Combining the 1998 and 1999 aerial and shipboard surveys provides one
set of abundance estimates for all species located within all of the
strata for the months of July and August. Combining the 2004 shipboard
and aerial surveys provides another set of abundance estimates for all
species that were located within all strata, but during the months of
June and July.
The
total abundance over all strata and all species covered during 1998/99
was nearly the same as during 2004: 279,583 versus 256,737, respectively
(Table 7). However, the distribution of animals
between the two years differed. During 1998/99 the most populated
strata (with over 50,000 animals) were the Offshore, Gulf of Maine Central,
and Scotian strata (Tables 8-11). During 2002,
although the survey only covered the northern strata, the Gulf of Maine
Central stratum was the only stratum with over 50,000 animals (Table
12). During 2004, only the
Offshore stratum had over 50,000 animals (Tables 13-14).
The Gulf of Maine Central and Offshore strata had the highest numbers
of cetaceans (Table 7), although the strata
with the highest densities (abundance divided by area) were the Gulf
of Maine North and Scotian strata (both mostly in Canadian waters). Within
US waters, the stratum with the highest density was the Gulf of Maine
Central, followed by the Shelf Central, Shelf West, and Georges Bank
Central strata (Table
7). The strata with the lowest densities and lowest species
diversity were the Mid-Atlantic and the western part of Georges Bank.
DISCUSSION
The
2002 aerial survey was not able to complete the planned track lines in
the GOMN stratum north of Grand Manan Island, Nova Scotia, Canada. In
the summer, many harbor porpoises and right whales, along with fewer
animals of other species such as fin whales, humpback whales, and minke
whales, usually inhabited the GOMN stratum. Thus, the 2002 estimates
for the GOMN are biased low.
The 2002 aerial survey was only conducted in the Gulf of Maine and Georges
Bank regions. Thus, the lack of estimates for the Shelf, Offshore,
and Mid-Atlantic strata for 2002 are an indication of no survey effort,
not an indication of depleted numbers of animals.
As noted above, the
strata with the lowest densities and lowest species diversity were
the Mid-Atlantic and the western part of Georges Bank. However, the survey
effort in these two strata was the lowest, after the Offshore stratum
(Table 1). Thus, to be confident
with this generalization, more future survey effort is needed in the
Mid-Atlantic and Georges Bank West strata.
The 2004 aerial survey was conducted from 12 June to 12 July, which
was several weeks earlier than the 2002 and other past surveys. It
is generally known that cetaceans that inhabit the Gulf of Maine during
the summer (e.g., harbor porpoises, white-sided dolphins, humpback whales,
minke whales, and pilot whales) enter the Gulf of Maine in early
summer and appear to peak in abundance during August. Comparing
the 2002 to 2004 estimates illustrate this movement into the Gulf of
Maine. That is, for the southern strata (GOMS, GeorgesW, and GeorgesC),
the 2004 estimate was larger than the 2002 estimate, and for the more
northern stratum (GOMC) it was the opposite: the 2002 estimate was larger
than the 2004 estimate. In addition, species thought to be more numerous
in springtime US waters, like sei whales and common dolphins, were more
numerous in the 2004 survey as compared to the 2002 survey. Thus, the
2004 distibutions and estimates appear to be more representative the
springtime distribution or the transition period between spring and summer,
while the 2002 and earlier distributions and abundance estimates appear
to be more representative of the summertime.
REFERENCES
Buckland,
S.T., Anderson, D.R., Burham, K.P., Laake, J.L., Borchers, D.L., and
Thomas, L. 2001. Introduction to Distance Sampling. Oxford University
Press, Oxford.
Buckland, S.T. and Anganuzzi, A.A. 1988. Comparison of smearing methods
in the analysis of Minke sightings data from IWC/IDCR Antarctic cruises. Rep.
Int. Whal. Commn. 38:257-63.
Efron, B. and Tibshirani, R.J. 1993. An Introduction to the Bootstrap.
Chapman and Hall. New York. 436 pp.
Hiby, L. 1999. The objective identification of duplicate sightins in
aerial survey for porpoise. In Marine Mammal Survey and Assessment
Methods (eds. G.W. Garner, S.C. Amstrup, J.L. Laake, B.F.J. Manly,
L.L. McDonald, and D.G. Robertson). Balkema, Rotterdam, pp. 179-189.
Palka, D. 1995. Abundance estimate of the Gulf of Maine harbor porpoise.
In Marine Mammal Survey and Assessment Methods (eds. G.W. Garner,
S.C. Amstrup, J.L. Laake, B.F.J. Manly, L.L. McDonald, and D.G. Robertson). Balkema,
Rotterdam, pp. 27-50.
Palka, D. 2000. Abundance of the Gulf of Maine/Bay of Fundy harbor porpoise
based on shipboard and aerial surveys during 1999. Northeast Fisheries
Science Center Reference Document 00-07. Available from: NEFSC,
166 Water St., Woods Hole, MA 02543.
Palka, D. 2005a. Shipboard surveys in the Northwest Atlantic: estimation
of g(0). In: Proceedings of the workshop on Estimation of g(0) in line-transect
surveys of cetaceans. (eds. F.Thomsen, F. Ugarte, and P.G.H. Evans).
pp. 32-37.
Palka, D. 2005b. Aerial surveys in the Northwest Atlantic: estimation
of g(0). In: Proceedings of the workshop on Estimation of g(0) in line-transect
surveys of cetaceans. (eds. F.Thomsen, F. Ugarte, and P.G.H. Evans).
pp. 12-17.
Palka, D. (In review). Evaluating assumptions underlying a line transect
analysis of a 2004 Northwest Atlantic shipboard abundance survey for
cetaceans. Submitted to Marine Mammal Science.
Palka, D. (in prep). Estimate of abundance from aerial surveys
during 2002 and 2004 when using the circle-back method.
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.