1

Centre for Marine Science and Technology

Habitat preferences and distribution of

killer whales (Orcinus orca) in the Bremer

Sub-Basin, Australia

Prepared for

The National Environmental Science Programme

Prepared by:

Chandra Salgado-Kent, Iain Parnum, Rebecca Wellard, Christine Erbe, Leila Fouda

PROJECT: CMST 1505

REPORT: 2017-15

October 2017

2

Contents

1. Introduction ..................................................................................................................................... 4

2. Methodology .................................................................................................................................... 6

2.1. Study Site and Survey Information ...................................................................................... 6

2.1.1. Boat Surveys ................................................................................................................. 6

2.1.2. Aerial Surveys ............................................................................................................... 7

2.1.3. Satellite Remote Sensing Data ..................................................................................... 8

2.2. Data Preparation .................................................................................................................. 8

2.3. Statistical Analyses ............................................................................................................... 9

3. Results ........................................................................................................................................ 11

3.1. Encounter Histories ............................................................................................................ 11

3.2. Environmental Variables .................................................................................................... 18

3.3. Distribution of Marine Mammals relative to Environmental Parameters ........................ 24

4. Discussion ....................................................................................................................................... 34

Acknowledgements ............................................................................................................................ 35

Data Sources ...................................................................................................................................... 35

References.......................................................................................................................................... 35

3

Abstract

This report summarises visual sightings of marine mammals (primarily killer whales) in the Bremer

Sub-Basin between February and April in the years 2015-2017. Observations were made from

boats in 2015-2017 and by airplane in 2017. Environmental parameters (bathymetry, slope,

distance from shore, sea surface temperature, Chlorophyll-A) were compiled for the days of

surveying in 2015-2017 and overlain with survey track lines and locations of marine mammals.

Generalized Estimating Equations (GEE) were used to investigate the influence of environmental

parameters on killer whale groups sighted for the two years of greatest effort: 2015 and 2016.

Using data collected opportunistically from a tour vessel during this period, a preliminary

assessment of the potential distribution of killer whales in the broader Bremer Sub-Basin region

was undertaken using the best fit GEE for predictions. Killer whale occurrence peaked in areas

with highest productivity which occurred at the edge of the canyon where bathymetry dropped

the fastest. Locations where other orca ‘hotspots’ may be expected were predicted in two other

canyon systems, including Hood Canyon and Pallinup Canyon; where productivity associated with

orcas was comparable to those in the Bremer Sub-Basin. The predicted hotspot in the areas

covered by relatively high survey effort were comparable with empirical observations. Estimates

of model prediction error and confidence intervals indicated that predictions had relatively wide

confidence intervals. The results presented in this report should be viewed as a preliminary

assessment to guide future research, as the model was based on highly biased (unsystematic)

surveys in one location. Robust, systematic line-transect surveys would allow for a more powerful

assessment of killer whale distribution, and reduce biases inherent in opportunistic data sets.

Systematic surveys are recommended to confirm predicted distributions reported here, as well as

to investigate killer whale and other megafauna distributions outside the months of February to

April.

4

1. Introduction

The killer whale (Orcinus orca; Linnaeus, 1758) is a cosmopolitan species, occurring throughout the

world’s oceans from warm, equatorial to cold, polar waters (Dahlheim and Heyning, 1999). While

the species is globally distributed, it is currently classified on the IUCN Red List as data deficient

with high uncertainty around abundance estimates (Taylor et al., 2013).

Killer whales occur within distinct subpopulations and populations around the world ranging from

below 100 to over several thousand animals (e.g., Berzin and Vladimirov, 1989; Branch and

Butterworth, 2001; Ford et al., 2000; Guerrero-Ruiz et al., 1998; Gunnlaugsson and Sigurjónsson,

1990; Hammond, 1984; Iñíguez, 2001; Kasamatsu and Joyce, 1995; Keith et al., 2001; Lopez and

Lopez, 1985; Mironova et al., 2002; Miyashita, 1993; Ohsumi, 1981; Øien, 1990; Sigurjónsson et

al., 1989; Visser, 1999; Wade and Gerrodette, 1993; Zerbini et al., 2007). These animals appear in

a diversity of habitats ranging from the surf zones, to river mouths, enclosed seas and open

oceans. Overall, killer whales are reported as being most common within continental margins

(Forney and Wade, 2006). Killer whale distribution is thought to be mainly linked to high

productivity, although they can occur in less productive waters (Forney and Wade, 2006). Thus,

the reported increase in killer whale density from low to high latitudes has been attributed to

greater productivity and prey abundance at higher latitudes (Dahlheim and Heyning, 1999; Forney

and Wade, 2006). In some locations where their food source is reliable year-round, resident

populations occur. However, there are many locations where killer whales occur seasonally

(Wellard et al., 2016), which is likely linked to seasonal prey availability. Based on killer whale

habitat and ecological specialisations (including prey targeted), and regionally-linked

distinguishable morphological features, colouration, acoustics and genetics, distinct killer whale

‘morphotypes’ have been described for populations that have been relatively well studied. While

morphotypes currently are considered to all fall within one species (Orcinus orca), separate

species or subspecies have been suggested for some morphological forms (Barrett-Lennard, 2000;

Berzin and Vladimirov, 1983; Hoelzel et al., 1998; Hoelzel and Dover, 1991; Mikhalev et al., 1981;

Pitman and Ensor, 2003; Pitman et al., 2007; Stevens et al., 1989).

In the North Pacific, killer whale morphotypes are described as “resident”, “Bigg’s” (or “transient”)

and “offshore” morphotypes (Ford, 2002). While killer whales around the world consume a wide

range of prey (e.g., Dahlheim and Heyning, 1999; Pitman and Ensor, 2003; Similä et al., 1996;

Visser, 1999), including cephalopods, fish (such as salmon, herring, bluefin tuna and other

schooling fish as well as sharks, stingrays and toothfish), seabirds (including penguins), turtles and

marine mammals (including dugongs, seals, sea lions, porpoises, dolphins, beaked whales, minke

whales and other baleen whales, particularly their calves), resident killer whales in the North

Pacific occurring in inshore waters year-round specialise in preying on fish (particularly Chinook

salmon (Oncorhynchus tshawytscha); Ford and Ellis, 2006). Bigg’s killer whales in this region are

reported to almost exclusively prey on marine mammals (Ford and Ellis, 1999; Ford et al., 1998).

5

Offshore killer whales have a larger geographic range than residents and transients and primarily

feed on fish (Herman et al., 2005).

In the North Atlantic, three killer whale populations have been identified: ‘A’, ‘B’, and ‘C’ (e.g.,

Matkin et al., 1997). A and B populations feed on fish and marine mammals, while C specialises on

fish (Cosentino, 2015; Foote et al., 2009; Vongraven and Bisther, 2014).

In Antarctic and sub-Antarctic waters, morphotypes have been classified as ‘A’, ‘B’, ‘C’ and ‘D’

(which are different from A, B and C in the North Atlantic). Morphotypes A and B have been

reported to feed on marine mammals, whereas Types C and D have been reported to prey on fish

(Durban et al., 2017; Pitman and Durban, 2010, 2012; Pitman et al., 2011; Pitman and Ensor, 2003;

Tixier et al., 2010).

While these distinct morphotypes vary widely in their ecological specialisations, habitat types and

regions they occupy, each population’s distribution also varies on a local scale (e.g., Pitman and

Ensor, 2003; Secchi et al., 2001). For instance, within Antarctic waters killer whale densities are

known to be higher near the ice edge (Berzin and Vladimirov, 1983). The spatial distribution and

abundance of killer whales locally is likely linked directly with the distribution of food resources.

In Australian waters, little is known regarding killer whale distribution regionally or locally, or their

ecological specialisations. Killer whale sightings have been reported in all state and territory

waters, with greater numbers of sightings off Australia’s southern coasts (Morrice, 2004; Mustoe,

2008; Pitman et al., 2015).

In the Bremer Sub-Basin off Western Australia relatively high numbers of killer whale sightings are

made consistently in late austral summer and early austral autumn (Wellard et al., 2015; Wellard

et al., 2016). Killer whales have been reported to forage in the Bremer Sub-Basin, including a few

documented beaked whale predation events (Wellard et al., 2016). Killer whale occurrence in the

Bremer Sub-Basin is most likely linked to high seasonal productivity and prey abundance.

However, the physical and environmental features that are responsible for driving productivity,

prey abundance and, ultimately, killer whale habitat use are unknown. This study investigates the

influence of environmental and physical features that may drive habitat use of killer whales during

their seasonal occupancy of the Bremer Sub-Basin. Specifically, the distribution and relative

abundance of killer whales observed from (i) boats between February and April in 2015, 2016 and

2017 and (ii) an airplane in 2017 are determined. For the years 2015 and 2016, environmental and

physical features influencing habitat use of killer whales within the Sub-Basin were identified

based on modeling outputs.

The results will provide insight in the use of other similar canyon systems in Australian waters and

ultimately help identify other locations likely used by killer whales that have not yet been

explored.

6

2. Methodology

2.1. Study Site and Survey Information

The Bremer Sub-Basin is located off the southwest continental shelf of Australia, approximately

180 km southeast of the city of Albany. The Sub-Basin contains numerous submarine canyons

extending from the shelf edge at approximately 100 m to water depths of 4500 m in the abyssal

plain (Exon et al., 2005). The submarine canyon system is biologically significant for a range of

species, including a number of toothed whales such as killer whales (Department of Sustainability,

2012).

Figure 1: Study site within the Bremer Sub-Basin, with overlaid vessel tracks (2015-2017; light red lines) and aerial

survey tracks (2017; light blue lines).

2.1.1. Boat Surveys

Opportunistic, non-systematic vessel surveys were conducted in an area approximately 4000 km2

within the canyon system (centred at 34°44.30'S latitude and 119°35.55'E longitude) between

February and April 2015 and 2016. These vessel-based surveys were undertaken from commercial

ecoutourism vessels Cetacean Explorer and Dhu Force (operated by Naturaliste in 2015 and 2016)

while these vessels were carrying out normal ecotourism manouvres taking tourists to see killer

whales. Between March and April 2017 non-systematic vessel surveys were also conducted in the

same vicinity, onboard a dedicated research vessel Big Dreams chartered by MIRG Australia.

Vessels departed from Bremer Bay, southern Western Australia, and headed offshore,

approximately 50 km south-east of Bremer Bay. Surveys were conducted on every good (up to

7

Beaufort 5) weather day (generally 6 days per week) over a period of 6 to 11 daylight hours and

consisted of passing and closing modes. At the beginning of each survey the date, observers and

start time were recorded and the track logged throughout the entire survey with a Holux wireless

GPS receiver. During passing mode, observations were undertaken by a minimum of two

individuals scanning the area forward to abeam the vessel on port and starboard sides up to the

horizon continually while on survey. Upon sighting a group of killer whales, the GPS position of the

vessel, group size and behavioural state were noted. The vessel then switched to closing mode and

the group was slowly approached to within 50 m. The GPS position of the group, group size

(minimum, maximum and best guess), group composition (numbers of adults and calves) and

predominant behavioural state were recorded and photographs taken. Behavioural states and

photographs were collected as part of a separate study and are not described here (see Wellard

and Erbe, 2017). Environmental conditions were recorded at the beginning of the survey and as

conditions changed. Conditions included Beaufort (sea state), cloud cover (okta scale), swell height

(estimated), glare (intensity from 0 to 3), visibility (up to 10 km) and wind direction and velocity

(estimated). Once all information had been collected, the vessel resumed closing mode and the

process repeated. If the same group of killer whales was encountered more than once within a

day, the observation was noted as a repeat sighting. If the group could not be confirmed as a

resight, it was considered a new group. At the end of the survey, the end time was recorded. All

survey information was recorded on a digital voice recorder. Voice recorded information was

transcribed at the end of each day and entered into the detailed field log. Photos were

downloaded to a laptop computer and backed up on an external hard drive.

2.1.2. Aerial Surveys

Aerial surveys were conducted in March 2017 with a total of 6 flight survey days. A twin-engine

over-head wing Cessna 337 aircraft with bubble windows was chartered from Norwest Air Work

Pty Ltd to survey transect lines and departed Albany airport each day. Pre-determined transects

were flown at an altitude of 300 m and a speed of 120 knots.

Personnel for each survey included one pilot and two observers. The pilot was not responsible for

spotting and was separated acoustically from the two observers during transects. The pilot’s only

role in data collection was to report the horizontal drift of the aircraft from its flight path at the

end of each transect. Two trained observers on each side of the aircraft scanned the transects and

recorded their sightings onto separate tracks on their Olympus digital recorders via headset. The

two observers worked independently and did not communicate whilst on transect.

At the beginning of each survey the date, observers and start time were recorded, as were wind

speed (knots) and wind direction. The track was logged throughout the entire survey with a Holux

wireless GPS receiver. Additional environmental conditions were recorded from each observer for

each side of the aircraft at the beginning and end of each transect, and updated immediately if any

of these conditions changed. Various weather covariates recorded included glare intensity (from 0

to 3), glare type (reflective or glitter), turbidity, Beaufort (sea state), visibility (up to 10 km) and

8

cloud cover (okta scale). All conditions were recorded onto digital voice recorder separately from

each observer accounting for different sides of aircraft.

Upon sighting marine fauna, the observer recorded time (WST), species, species confidence

(definite, probable, likely), group size, group composition, swim direction (relative to the plane)

and sighting cue. Although due to speed of plane, not all group size, composition and direction of

travel could be noted. Observers used inclinometers (Suunto PM-5/360PC) and compass boards to

record relative vertical and horizontal positions of each sighting. All sighting details were recorded

on an Olympus digital recorder for post-flight data entry by the team leader.

The surveys were conducted in passing mode with killer whales as the main focus. Due to this

focus and opportunity to gather as much information possible on this species, when a killer whale

was sighted, passing mode was aborted and the plane circled back to obtain and validate as much

detailed information as possible for this sighting. When requesting a “circle-back”, the observer

blew their whistle to alert the pilot and direct the plane to the sighting. Once observers were

confident all information necessary had been recorded, the plane travelled back to the waypoint

where the transect line had been aborted and recommenced on transect in passing mode and

continued with pre-determined transects.

The sampling consisted of two transects. Transect A consisted of a set of diagonal transects that

allowed large-scale coverage inside and adjacent to the Commonwealth Marine Reserve, whilst

Transect B consisted of a series of denser vertical transects across the known aggregation area.

2.1.3. Satellite Remote Sensing Data

Further data on environmental conditions within the study area were obtained for the survey

periods in 2015-2017, including daily (night time) sea surface temperature (SST) at a 0.01 degree

(approximately 1.1 km) resolution derived from the Advanced Very High Resolution Radiometer

(AVHRR) satellite data (IMOS 2015-2017a), daily Chlorophyll-A (Chl-A) at 0.01 degree (1.1 km)

resolution derived from Moderate Resolution Imaging Spectroradiometer (MODIS) satellite data

using the OC3 model (IMOS 2015-2017b), and depth from the 250 m grid created by Geoscience

Australia (2009). The downloaded SST were adjusted for sensor-specific error statistics (SSES) bias

and only data with a quality level 5 were used and converted from degrees Kelvin to Centigrade.

Chl-A data with a quality level below 3 were discarded. The following statistics were calculated

over the SST and Chl-A data for the survey days: minimum, maximum, mean, std, median, 25th

quantile and 75th quantile.

2.2. Data Preparation

To model the occurrence of killer whales over the spatial extent of the study area, the study area

was gridded into two grid resolutions: 250 m (i.e. to match the depth resolution) and 1.1 km (i.e.

to match the SST and Chl-A resolutions). Grid cells were allocated an ID and the latitude and

longitude of each grid cell’s centre was extracted. Thus, environmental conditions were allocated

9

to each grid cell with known centre point position. For the 250 m resolution data set, the statistics

calculated for the SST and Chl-A were linearly interpolated in MATLAB (The MathWorks Inc.) using

the inbuilt interp2.m function. For the 1.1 km dataset the depth was resampled from the 250 m

grid using the same function. In addition, the distance in km of the centre point of each grid to the

nearest coastline (defined as 0 m elevation in the GA 250 m bathymetric and topographic grid)

was calculated. The slope of the seafloor (in degrees) was calculated from the 250 m depth grid.

Data preparation and plotting was undertaken using MATLAB.

2.3. Statistical Analyses

The spatial occurrence of killer whale groups over time was modelled using Generalized Estimation

Equations (GEEs) with a log-link function (for Poisson distributed count data) using data gridded on

the 1.1 km resolution. GEEs account for residual autocorrelation in data (Zeger & Liang, 1986; Thall

& Stephen, 1990; Zorn, 2001; Zuur et al., 2009; Staley, 2013). Models using the CReSS-GEE (Scott-

Hayward et al., 2013a) framework in R (R Core Team, 2013) were used to allow for smooth terms

to be included. RStudio, Version 0.98.501 (© 2009-2013 RStudio, Inc.), using MRSea (Scott-

Hayward et al., 2013b) and geepack (Yan, 2002; Yan & Fine, 2004), was used to fit models.

To build the model, first independent variables were explored to identify and remove variables

with insufficient spread over the study area and period. In addition, data were truncated to only

include observation that were deeper than -250 m and shallower than -1600 m bathymetry since

no killer whale sightings were made in depths outside of this range. Thus, zero inflation in the

model was significantly reduced. Model predictors considered included grid cell centre depth,

distance from nearest land, seabed slope, mean SST, median SST, maximum SST, minimum SST,

SST standard deviation, 25th quantile SST, 75th quantile SST, mean Chl-A concentration, median

Chl-A, maximum Chl-A, minimum Chl-A, Chl-A standard deviation, 25th quantile Chl-A and 75th

quantile Chl-A. Because there was autocorrelation, a spatial blocking structure was applied as a

vector of integers numbered as sequential ~3 km2 blocks (the correlation over grid cells (clusters)

declined to approximately zero after 3 km). Grid cell was used as the ID.

Regression analyses were run in order to investigate the presence of highly correlated variables,

and Variance Inflation Factors (VIFs) of > 3 considered collinear (Zuur et al., 2009) and removed

from analyses. The function “runSALSA1D” in the MRSea package was used to select smoothness

of the covariates. An AR-1 correlation structure was initially selected (Zuur et al., 2009), however

independence, unstructured, and exchangeable correlation structures were also tested in models,

and the best fit was selected using Quasi-likelihood information criterion (QIC; Pan, 2001).

A full model was first fitted including all biologically relevant variables and interactions (that

allowed model convergence and were not collinear). Several different combinations of ‘Full

Models’ were required to test the effects of all covariates, because a number of covariates were

collinear. Explanatory terms that were not significant were eliminated one by one, and the

resulting submodel refitted. Each submodel was checked and validated using observed versus

10

fitted values to assess model fit and fitted values versus scaled Pearson’s residuals were plotted to

assess the mean-variance relationship. The model that reduced the QICu by > 2 values with the

fewest terms was selected as the best fit (Pan, 2001; Hardin & Hilbe, 2002; Cui, 2007; Hudecová &

Pešta, 2013). ANOVA was applied to cross-compare models with terms removed. All significant

terms were retained in the final model.

The final best model was then used to predict the expected distribution of groups of orcas in a

~130 km longitudinal range along the shelf edge and canyon systems in the region. Methods by

Scott-Hayward et al. (2013b) were applied. Model validation was undertaken using k-fold cross-

validation (Boyce et al., 2002) using the package boot (Davison & Hinkley, 1997; Canty & Ripley,

2017) in R (R Core Team, 2013) and k=10. In addition, bootstrap confidence intervals were

estimated for the model coefficients (using 1000 iterations) and for predicted responses (using

methods in Scott-Hayward et al., 2013b).

11

3. Results

3.1. Encounter Histories

A total of 594 hours was spent on 77 survey days over the course of 3 years. Twenty-eight surveys

were conducted in 2015, 36 in 2016 and 13 in 2017. A total of 304 cetacean encounters were

recorded. Of these, 177 were encounters with killer whales. These data are summarised in Table 1.

Table 1: Summary of visual observations.

2015 2016

2017 -

plane

2017 -

boat

2017

total TOTAL

Killer whale encounters 71 92 7 7 14 177

Other marine mammal encounters 11 31 54 31 85 127

Days surveyed 28 36 6 7 13 77

Total hours surveyed 237.8 259.2 24 72.8 96.8 593.8

Average killer whale encounters per

day 2.4 2.6 1.2 0.9 1.0 -

Average other encounters per day 0.4 0.9 9.0 3.9 6.1

Average survey hours per day 8.2 7.2 4.0 9.1 6.8 -

Average hours between killer whale

encounters 3.3 2.8 3.4 10.4 6.8 -

Average hours between other

encounters 21.6 8.4 0.4 2.4 1.1

Survey start/1st sighting 9-Feb 19-Feb 16-Mar 25-Mar - -

Survey end/last sighting 24-Mar 17-Apr 24-Mar 4-Apr - -

Figure 2 shows a map of the cumulative search effort for the visual surveys by boat and plane.

Figure 3 and Figure 4 show the survey tracks (boat and plane respectively) with the sites of killer

12

whale and other marine mammal encounters overlain. Figure 5 and Figure 6 show the same data,

but have the other marine mammals broken down by species.

Figure 2: Search effort in 1.1 km grids; (top) in cumulative hours of vessel surveys 2015-2017, (bottom) in minutes of

plane survey 2017.

13

Figure 3: Survey tracks and marine mammal encounters from vessel surveys 2015 to 2017; killer whale encounters as

pink dots and other marine mammal encounters as white dots. The lower panel is zoomed-in version of the upper

panel.

14

Figure 4: Tracks and marine mammal encounters from the plane surveys in 2017; killer whale encounters as pink dots

and other marine mammal encounters as white dots.

15

Figure 5: Tracks and marine mammal encounters from vessel surveys 2015 to 2017 by species. The lower panel is a

zoomed-in version of the upper panel.

16

Figure 6: Tracks and marine mammal encounters from the plane surveys in 2017.

17

Figure 7: The summation of the minimum group size divided by search effort in hours for vessel surveys 2015 to 2017

at 1.1 km grid resolution, for: (top) killer whales, (bottom) other marine mammals.

18

3.2. Environmental Variables

The following figures show the range of environmental parameters encountered at the locations

of surveying and sighting marine mammals. Specifically, marine mammal encounters are plotted

on top of distance from shore (Figure 8 and Figure 9), bathymetry (Figure 10 and Figure 11),

bathymetric slope (Figure 12 and Figure 13), mean SST (Figure 14) and mean Chl-A (Figure 15).

Figure 8: Distance from shore (km) with killer whale (pink) and other marine mammal (white) encounters from vessel

surveys 2015-2017. The lower panel zooms into the region of most marine mammal sightings.

19

Figure 9: Distance from shore (km) with killer whale (pink) and other marine mammal (white) encounters from plane

surveys in 2017.

20

Figure 10: Depth (m) with killer whale (pink) and other marine mammal (white) encounters from vessel surveys

between 2015-2017. The lower panel zooms into the region of most marine mammal sightings.

21

Figure 11: Depth (m) with killer whale (pink) and other marine mammal (white) encounters from plane surveys in

2017.

22

Figure 12: Slope of the seafloor (degrees) with killer whale (pink) and other marine mammal (white) encounters from

vessel surveys between 2015-2017. The lower panel zooms into the region of most marine mammal sightings.

23

Figure 13: Slope of the seafloor (degrees) with killer whale (pink) and other marine mammal (white) encounters from

plane surveys in 2017.

Figure 14: Mean SST on days of boat surveys 2015 to 2017. Killer whale encounters as pink dots and other marine

mammal encounters white dots.

24

Figure 15: Mean Chl-A (mg/m3) on days of boat surveys 2015 to 2017. Killer whale encounters as pink dots and other

marine mammal encounters white dots.

3.3. Distribution of Marine Mammals relative to Environmental Parameters

For modelling the number of killer whale groups encountered in relation to environmental

parameters, sighting data from a total of 497 h over 64 days of effort during 2015 and 2016 were

available. Over the 64 survey days, a total of 170 orca groups were encountered in 10,820 linear

km travelled.

Vessel tracks crossed a total of 1,130 1-km2 grid cells in the study area, from the point of

departure to the Bremer Sub-Basin. The effort varied among 1-km2 grid cells. Most grid cells had

less than 1 h and 10 km of effort (Figure 16). The mean number of orca groups per grid cell

observed was 0.15 and ranged between 0 and 17.

25

Figure 16: Distribution of hours and km of effort per 1 km2 grid cell in the Bremer Sub-Basin study area over the 2015

and 2016 survey seasons, overlaid with orca group sightings (red dots).

The mean depth over the grid cells traversed was 649 m and ranged between 0 and 2479 m

(Figure 17). The mean slope over the grid cells was 5.2° and ranged from 0 to 34.5°. The mean

average SST was 20.1°C and ranged over only 3 degrees from 19.8° C to 20.9° C. The mean average

Chl-A concentration was 0.16 mg/m3 and ranged from 0.11 to 0.51 mg/m3. The maximum and

minimum values for SST and Chl-A varied more widely with highest maximum SST at 22.6°C and

lowest minimum of 17.3°C, and the highest maximum Chl-A at 3.06 mg/m3 and lowest minimum of

0 mg/m3.

26

Figure 17: Distribution of environmental conditions per 1 km2 grid cell in the Bremer Sub-Basin study area taken over

the entire 2015 and 2016 survey seasons.

The statistics (mean, minimum, maximum, standard deviation, and 25th and 75th quantiles) had

limited variability over the study area, with the exception of some gradients associated with

changes in bathymetry between the mainland coast and the Bremer Sub-Basin (Figure 18).

The density of killer whale groups sighted was greatest approximately 38 km from the coast,

where the seafloor was between 800 and 1000 m deep with a slope between 5 and 8 degrees

(Figure 19). Peak numbers of killer whale groups sighted occurred in grids with environmental

conditions of an overall average temperature (across years) of approximately 20.1°C, with

minimum and maximum ranging from 19.4 to 22.1 °C. Peak numbers of killer whale groups were

observed in average Chl-A concentrations of approximately 0.15 mg/m3 with minimum and

maximum values for peak numbers of killer whales of 0.05 and 0.5 mg/m3.

However, peak numbers in killer whale groups sighted may have been influenced by high search

effort within grid cells with specific environmental conditions (Figure 20). In addition, many grid

cells had no killer whale groups sighted, despite having environmental conditions within the range

where peak number were also sighted (Figure 20).

27

Figure 18: Distribution of environmental conditions over 1 km2 grid cells surveyed in the Bremer Sub-Basin study area over the 2015 and 2016 survey period, overlaid with killer

whale group sightings (red dots). Grey region corresponds to areas where environmental conditions were absent and could not be plotted.

28

Figure 19: Number of killer whale groups sighted over the different environmental conditions in each 1 km2 grid cell surveyed in the Bremer Sub-Basin study area over the 2015

and 2016 survey period.

29

Figure 20: Number of killer whale groups sighted (red dots) in relation to environmental conditions over 1 km2 grid cells surveyed in the Bremer Sub-Basin study area over the 2015

and 2016 survey period. Grey region corresponds to km of search effort (in 10 km increments). Blue dots correspond to environmental conditions within grid cells having 0 killer

whale groups sighted.

30

To reduce the influence of values within the data that were associated with very little effort

in the model, the data were first trimmed so that grids with fewer than 10 km effort were

removed. Any observations with extreme values of environmental variable separated by

large gaps from the rest of the measures were also removed. In addition, preliminary

analyses indicated that depth and nearest distance to the coast were collinear. Distance was

removed from the model terms as there was clear evidence from the mapping exercise that

killer whales occurred approximately 38 km from the coast and beyond (near the shelf edge;

Figure 21).

Figure 21: Distribution of killer whale (pink dots) and other marine mammal (white dots) encounters as a

function of distance from shore, shown with the average depth, and survey hours for vessel surveys (2015-

2017) and plane surveys (2017).

In addition, SST measures were collinear, as were Chl-A measures. Thus, the means over the

2015 and 2016 survey seasons for SST and Chl-A were selected as priority conditions for

modelling as these represented central tendency, but also reflected smaller and greater

values (which could influence the presence of killer whale groups; Figure 22).

31

Figure 22: Number of killer whale groups sighted (red dots) in relation to modelled environmental conditions

over 1 km2 grid cells surveyed in the Bremer Sub-Basin study area over the 2015 and 2016 survey period. Grey

region corresponds to km of search effort (in 10 km increments). Blue dots correspond to environmental

conditions within grid cells having 0 killer whale groups sighted.

The subset of data used for modelling included a total of 360 h over a total of 4,284 linear km

of effort in the survey area during 2015 and 2016. Effort (km) was included in the model as

an offset. The model with the best fit (the final model) included Chl-A Max only, as a significant

explanatory term (Table 2). The model scale parameter was 0.60.

Table 2: Generalized Estimating Equations coefficients, parameter estimates, standard error, Wald

statistic,p-values and significance level (* ≤ 0.05) for the final model for predicting the number of killer

whale groups encountered.

Coefficients Estimate Std Error Wald Pr (>W)

Intercept -4.220 0.276 234.4 <2e-16 *** Chl-A Max 1.892 0.466 16.5 4.8e-05 ***

The final model indicated that Chl-A Max had a relatively large influence on the number of

killer whale groups sighted. While most observed measurements of Chl-A were below 0.6, as

the Chl-A increased so did the number of killer whale groups observed (Figure 23).

32

Figure 23 Partial residual plot produced from the best Generalized Estimating Equation model for predicting the

number of killer whale groups sighted over 1.1 km2 grid cells surveyed in the Bremer Sub-Basin study area over

the 2015 and 2016 survey periods. Error bars represent 95% confidence intervals.

Using the above model, the expected relative distribution of groups of orcas in a ~130 km

longitudinal range along the shelf edge and associated canyon systems in the region was

assessed (Figure 24). The numbers of groups predicted (corrected for km transited by the

survey vessel) were comparable. The mean relative number of observed killer whales sighted

per grid cell (1.1 km2) in the Bremer Sub-Basin (empirical data) was 0.015 (min=0, max=0.995,

SE=0.004, 95% CI=0.008), and the mean predicted relative number of killer whales per grid

cell (1.1 km2) across the broader region from Bremer Canyon to Pallinup Canyon was 0.012

(min=0, max=0.14, SE=0.0006, 95% CI=0.001). In the broader region, hotspots were identified

over three canyon systems including the Bremer Sub-Basin, Hood Canyon, and Pallinup

Canyon (Figure 24). The predicted hotspot in the areas covered by relatively high survey effort

were consistent with empirical observations in the region.

The GEE model prediction error (k-folds cross-validation mean squared error) was 0.728 (with

corrected bias of 0.723). Multiple runs of the k-folds cross-validation resulted in consistent

mean squared errors. The root mean squared error (RMSE) was 0.853 (corrected for

bias=0.850). Bootstrap 95% Confidence Intervals for the estimated coefficient of ChlA as the

predictor were 1.643 (lower) and 2.204 (upper). Spatial bootstrap estimates of 95%

confidence intervals of predicted relative density of whales indicated that the predictions had

wide confidence intervals (see Figure 24 for spatial maps of 95% confidence intervals).

33

Figure 24 Predicted (a), predicted lower (b) and upper (c) 95% Confidence Intervals, and

observed (d) relative densities of orca groups (corrected for km transited by the vessel).

(b) Predicted Lower 95% CI

Pallinup Canyon

Pallinup Canyon

(c) Predicted Upper 95% CI

Pallinup Canyon

(a) Predicted

(d) Observed

Pallinup Canyon

34

4. Discussion

Between January and April killer whales can be encountered within the Bremer Sub-Basin with

high probability (91.5%; Wellard et al., 2016). Thus, opportunistic boat-based observations have

been spatially concentrated near in the Bremer Sub-Basin in past years. The environmental

variables vary little over the extend of this ‘hotspot’.

Nonetheless, the occurrence of groups sighted between 2015-2017 was associated with

productivity. Higher productivity, measured in the form of Chlorophyll A, was mostly at the edge of

canyon systems, which may be due to seasonal upwelling. Cold Antarctic water rises to the surface

at the canyon edge, and brings with it nutrients and consequently higher order predators (e.g. Allen

and Durrieu de Madron, 2009; Middleton and Bye, 2007; Pattiaratchi and Woo, 2009; Waite et al.,

2007). A predictive model identified three other locations in the region that had productivity levels

expected to be associated with orca presence. These locations were associated with deep canyon

systems located at the edge of the continental shelf. The predictive model resulted in comparable

relative densities and hotspots of killer whales as those observed in the Bremer Sub-Basin. Estimates

of model prediction error and confidence intervals indicated that predictions had relatively wide

confidence intervals.

In conclusion, this study used data collected opportunistically from a tour vessel to produce a

preliminary assessment of the potential distribution of killer whales in the broader Bremer Sub-

Basin region. The predictions rely heavily on observations from a very limited region, as surveys

were mostly focused in an area where killer whales had previously been sighted reliably. As a result,

the spatial extent of the study was highly biased, and mainly dependent on observations from one

canyon area. These biases led to heavy culling of the data to exclude environmental condition ranges

that had insufficient observations to allow model stability. This led to limitations in the area in which

predictions could be made. In addition, predictions of killer whale occurrence in the broader region

of this study assume that conditions not measured in this study (prey availability, predator

prevalence, etc.) are comparable to the region surveyed by the tour vessel. Systematic line-transect

surveys (in space and time) are recommended to test and validate predicted killer whale

distributions with empirical measurements. In addition, such surveys would allow modelling of

other species of marine mammals (and megafauna in general). While the aerial surveys in 2017 were

an attempt at surveying a larger area systematically, these were too few to be useful in statistical

modelling. Line-transect surveys over a larger canyon region and at times outside of February-April

are strongly recommended.

35

Acknowledgements

Cetacean Explorer and Dhu Force were operated by Naturaliste Charters in 2015 and 2016. Big

Dreams was chartered by MIRG Australia for the 2017 season. We are further grateful for all

assistance given with the 2017 aerial surveys by Northwest Air Work pilot Conor and marine

wildlife consultant Verity Steptoe. The 2015-2017 visual data were collected under the Curtin

University Animal Ethics Committee Approval No. AEC_2013_27 and under the Australian

Government Department of the Environment Research Permits Nos. 2014/0008 and 2017/0001.

The 2015 data were collected under a contract with the SeaWorld Busch Gardens Conservation

Fund and Curtin University (recipients Christine Erbe and Leila Fouda). The 2016 and 2017 boat-

based data were collected with support from the Holsworth Wildlife Research Endowment and

The Australian Geographic Society (recipient Rebecca Wellard). This report was written with

funding from the National Environmental Science Programme.

Data Sources

Geoscience Australia. "Australian bathymetry and topography grid, June 2009." Canberra:

Geoscience Australia, http://pid.geoscience.gov.au/dataset/67703.

IMOS (2015-2017a). IMOS - SRS Satellite - SST L3S - 01 day composite - night time,

https://portal.aodn.org.au/, Accessed: 11th August 2017.

IMOS (2015-2017b). IMOS - SRS - MODIS - 01 day - Chlorophyll-a concentration (OC3 model),

https://portal.aodn.org.au/, Accessed: 11th August 2017.

References

Allen, S., and Durrieu de Madron, X. (2009). A review of the role of submarine canyons in deep-

ocean exchange with the shelf. Ocean Science, 5(4), 607-620.

Barrett-Lennard, L. G. (2000). Population structure and mating patterns of killer whales (Orcinus

orca) as revealed by DNA analysis. University of British Columbia.

Berzin, A., and Vladimirov, V. (1983). A new species of killer whale (Cetacea, Delphinidae) from the

Antarctic waters. Zoologichesky Zhurnal, 62(2), 287-295.

Berzin, A., and Vladimirov, V. (1989). Recent distribution and abundance of cetaceans in the Sea of

Okhotsk. Biologiya Morya-Marine Biology, (2), 15-23.

Boyce, M. A., Vernier, P. R., Nielsen, S. E., and Schmiegelow, F.K.A. (2002). Evaluating resource

selection functions. Ecological Modelling, 157, 281–300.

Branch, T., and Butterworth, D. (2001). Estimates of abundance south of 60^oS for cetacean

species sighted frequently on the 1978/79 to 1997/98 IWC/IDCR-SOWER sighting surveys.

Journal of Cetacean Research and Management, 3(3), 251-270.

Canty, A, and Ripley, B. (2017). boot: Bootstrap R (S-Plus) Functions. R package version 1.3-20.

36

Cosentino, A. M. (2015). First record of Norwegian killer whales attacking and feeding on a

harbour porpoise. Marine Biodiversity Records, 8.

Cui, J. (2007). QIC program and model selection in GEE analyses. The State Journal, 7(2), 209-220.

Dahlheim, M. E., and Heyning, J. E. (1999). Killer whale Orcinus orca (Linnaeus, 1758). In S. H.

Ridgway & R. Harrison (Eds.), Handbook of Marine Mammals (Vol. 6, pp. 281-322), doi:

Davison, A. C., and Hinkley, D. V. (1997) Bootstrap Methods and Their Applications. Cambridge

University Press, Cambridge. ISBN 0-521-57391-2.

Department of Sustainability, Environment, Water, Population and Communities. (2012). Marine

bioregional plan for the South-west Marine Region.

Durban, J., Fearnbach, H., Burrows, D., Ylitalo, G., and Pitman, R. (2017). Morphological and

ecological evidence for two sympatric forms of Type B killer whale around the Antarctic

Peninsula. Polar Biology, 40(1), 231-236.

Exon, N., Hill, P., Mitchell, C., and Post, A. (2005). Nature and origin of the submarine Albany

canyons off southwest Australia. Australian Journal of Earth Sciences, 52(1), 101-115.

Foote, A. D., Newton, J., Piertney, S. B., Willerslev, E., and Gilbert, M. T. P. (2009). Ecological,

morphological and genetic divergence of sympatric North Atlantic killer whale populations.

Molecular Ecology, 18(24), 5207-5217.

Ford, J. K. (2002). Killer whale Orcinus orca. In W. F. Perrin, B. Würsig & J. G. M. Thewissen (Eds.),

Encyclopedia of Marine Mammals (pp. 669-676): Academic Press, doi:

Ford, J. K., and Ellis, G. M. (1999). Transients: mammal-hunting killer whales of British Columbia,

Washington, and southeastern Alaska: UBC Press.

Ford, J. K., and Ellis, G. M. (2006). Selective foraging by fish-eating killer whales Orcinus orca in

British Columbia. Marine Ecology Progress Series, 316, 185-199.

Ford, J. K., Ellis, G. M., and Balcomb, K. C. (2000). Killer whales: the natural history and genealogy

of Orcinus orca in British Columbia and Washington: UBC press.

Ford, J. K., Ellis, G. M., Barrett-Lennard, L. G., Morton, A. B., Palm, R. S., and Balcomb III, K. C.

(1998). Dietary specialization in two sympatric populations of killer whales (Orcinus orca) in

coastal British Columbia and adjacent waters. Canadian Journal of Zoology, 76(8), 1456-

1471.

Forney, K. A., and Wade, P. R. (2006). Worldwide distribution and abundance of killer whales.

Whales, whaling and ocean ecosystems, 145-162.

Guerrero-Ruiz, M., Gendron, D., and Urbán, J. (1998). Distribution, movements and communities

of killer whales (Orcinus orca) in the Gulf of California, Mexico. Report of the International

Whaling Commission, 48, 537-543.

Gunnlaugsson, T., and Sigurjónsson, J. (1990). NASS-87: Estimation of whale abundance based on

observations made onboard Icelandic and Faroese survey vessels. Reports of the

International Whaling Commission, 40, 571-580.

Hammond, P. S. (1984). Abundance of killer whales in Antarctic Areas II, III, IV and V. Reports of the

International Whaling Commission, 34, 543-548.

37

Hardin, J. W., and Hilbe, J. M. (2002). Generalized Estimating Equations. Boca Raton, FL: Chapman

and Hall/CRC Press.

Herman, D., Burrows, D., Wade, P., Durban, J., Matkin, C., LeDuc, R., Barrett-Lennard, L., and

Krahn, M. M. (2005). Feeding ecology of eastern North Pacific killer whales Orcinus orca

from fatty acid, stable isotope, and organochlorine analyses of blubber biopsies. Marine

Ecology Progress Series, 302, 275-291.

Hoelzel, A., Dahlheim, M., and Stern, S. (1998). Low genetic variation among killer whales (Orcinus

orca) in the eastern North Pacific and genetic differentiation between foraging specialists.

Journal of Heredity, 89(2), 121-128.

Hoelzel, A. R., and Dover, G. A. (1991). Genetic differentiation between sympatric killer whale

populations. Heredity, 66(2), 191-195.

Hudecová, S., and Pešta, M. (2013). Modeling dependencies in claims reserving with GEE.

Insurance: Mathematics and Economics, 53(3), 786-794.

Iñíguez, M. A. (2001). Seasonal distribution of killer whales (Orcinus arca) in Northern Patagonia,

Argentina. Aquatic Mammals, 27, 154-161.

Kasamatsu, F., and Joyce, G. G. (1995). Current status of odontocetes in the Antarctic. Antarctic

Science, 7(4), 365-379.

Keith, M., Bester, M., Bartlett, P., and Baker, D. (2001). Killer whales (Orcinus orca) at Marion

Island, Southern Ocean. African Zoology, 36(2), 163-175.

Lopez, J. C., and Lopez, D. (1985). Killer whales (Orcinus orca) of Patagonia, and their behavior of

intentional stranding while hunting nearshore. Journal of Mammalogy, 66(1), 181-183.

Matkin, C. O., Matkin, D. R., Ellis, G. M., Sau-litis, E., and McSweeney, D. (1997). Movements of

resident killer whales in southeastern Alaska and Prince William Sound, Alaska. Marine

Mammal Science, 13(3), 469-475.

Middleton, J. F., and Bye, J. A. (2007). A review of the shelf-slope circulation along Australia’s

southern shelves: Cape Leeuwin to Portland. Progress in Oceanography, 75(1), 1-41.

Mikhalev, Y. A., Ivashin, M., Savusin, V., and Zelenaya, F. (1981). The distribution and biology of

killer whales in the Southern Hemisphere. Report of the International Whaling Commission,

31, 551-566.

Mironova, A., Burdin, A., Hoyt, E., Jikiya, E., Nikulin, V., Pavlov, N., Sato, H., Tarasyan, K., Filatova,

O., and Vertyankin, V. (2002). O. orca abundance, distribution, seasonal presence,

predation and strandings in the waters around Kamchatka and the Kommander Islands: An

assessment based on reported sightings 1992-2000. Paper presented at the Proceedings of

the 4th International Orca Symposium and Workshop, Chize, France, pp106-108.

Miyashita, T. (1993). Abundance of dolphins stocks in the western North Pacific taken by the

Japanese drive fishery. Reports of the International Whaling Commission, 43, 417-437.

Morrice, M. G. (2004). Killer whales (Orcinus orca) in Australian territorial waters. Technical Paper.

Deakin University, Victoria. Australia.

Mustoe, S. (2008). Killer Whale Orcinus' Orca'Sightings in Coastal Victoria. Victorian Naturalist,

The, 125(3), 76.

38

Ohsumi, S. (1981). Distribution and abundance of killer whales in the Southern Hemisphere.

Reports of the International Whaling Commission, 31, 1-9.

Øien, N. (1990). Sightings surveys in the northeast Atlantic in July 1988: distribution and

abundance of cetaceans. Reports of the International Whaling Commission, 40, 499-511.

Pan, W. (2001). Akaike’s information criterion in Generalized Estimation Equations. Biometrics, 57,

120-125.

Pattiaratchi, C., and Woo, M. (2009). The mean state of the Leeuwin current system between

North West Cape and Cape Leeuwin. Journal of the Royal Society of Western Australia,

92(2), 221-241.

Pitman, R. L., and Durban, J. W. (2010). Killer whale predation on penguins in Antarctica. Polar

Biology, 33(11), 1589-1594.

Pitman, R. L., and Durban, J. W. (2012). Cooperative hunting behavior, prey selectivity and prey

handling by pack ice killer whales (Orcinus orca), type B, in Antarctic Peninsula waters.

Marine Mammal Science, 28(1), 16-36.

Pitman, R. L., Durban, J. W., Greenfelder, M., Guinet, C., Jorgensen, M., Olson, P. A., Plana, J.,

Tixier, P., and Towers, J. R. (2011). Observations of a distinctive morphotype of killer whale

(Orcinus orca), type D, from subantarctic waters. Polar Biology, 34(2), 303-306.

Pitman, R. L., and Ensor, P. (2003). Three forms of killer whales (Orcinus orca) in Antarctic waters.

Journal of Cetacean Research and Management, 5(2), 131-140.

Pitman, R. L., Perryman, W. L., LeRoi, D., and Eilers, E. (2007). A dwarf form of killer whale in

Antarctica. Journal of Mammalogy, 88(1), 43-48.

Pitman, R. L., Totterdell, J. A., Fearnbach, H., Ballance, L. T., Durban, J. W., and Kemps, H. (2015).

Whale killers: Prevalence and ecological implications of killer whale predation on

humpback whale calves off Western Australia. Marine Mammal Science, 31(2), 629-657.

Scott-Hayward, L., Mackenzie, M. L., Donovan, C. R., Walker, C. G., & Ashe, E. (2013a). Complex

Region Spatial Smoother (CReSS). Journal of Computational and Graphical Statistics, 23(2),

340-360. http://dx.doi. org/10.1080/10618600.2012.762920

Scott-Hayward, L., Oedekoven, X. S., Mackenzie, M. L., Walker, C. G., & Rexstad, E. (2013b). User

guide for the MRSea package: Statistical modelling of bird and cetacean distributions in

offshore renewables development areas (Contract with Marine Scotland: SB9

[CR/2012/05]). St Andrews, Fife, UK: University of St. Andrews. Retrieved 17 September

2015 from http:// creem2.st-and.ac.uk/software.aspx.

Secchi, E. R., Dalla Rosa, L., Kinas, P. G., Santos, M. C. d. O., Zerbini, A. N., Bassoi, M., and Moreno,

I. B. (2001). Encounter rates of whales around the Antarctic Peninsula with special

reference to humpback whales, Megaptera novaeangliae, in the Gerlache Strait: 1997/98

to 1999/2000.

Sigurjónsson, J., Gunnlaugsson, T., and Payne, M. (1989). NASS-87: Shipboard sightings surveys in

Icelandic and adjacent waters June-July 1987. Report of the International Whaling

Commission, 39, 395-409.

39

Similä, T., Holst, J. C., and Christensen, I. (1996). Occurrence and diet of killer whales in northern

Norway: seasonal patterns relative to the distribution and abundance of Norwegian spring-

spawning herring. Canadian Journal of Fisheries and Aquatic Sciences, 53(4), 769-779.

Staley, T. (2013) The case for Generalized Estimating Equations in state-level analysis. Working

draft prepared for the State Politics and Policy Conference, Iowa City, IA.

Stevens, T. A., Duffield, D. A., Asper, E. D., Hewlett, K. G., Bolz, A., Gage, L. J., and Bossart, G. D.

(1989). Preliminary findings of restriction fragment differences in mitochondrial DNA

among killer whales (Orcinus orca). Canadian Journal of Zoology, 67(10), 2592-2595.

Taylor, B. L., Baird, R., Barlow, J., Dawson, S. M., Ford, J., Mead, J. G., Notarbartolo di Sciara, G.,

Wade, P., and Pitman, R. L. (2013). Orcinus orca. The IUCN Red List of Threatened Species

e.T15421A44220470, doi: 10.2305/IUCN.UK.2013-1.RLTS.T15421A44220470.en.

Tixier, P., Gasco, N., Duhamel, G., Viviant, M., Authier, M., and Guinet, C. (2010). Interactions of

Patagonian toothfish fisheries with killer and sperm whales in the Crozet islands Exclusive

Economic Zone: an assessment of depredation levels and insights on possible mitigation

strategies. CCAMLR Science, 17, 179-195.

Thall, P. F. and Stephen C. V. (1990). Some covariance models for longitudinal count data with

overdispersion. Biometrics, 46, 657-671.

Visser, I. N. (1999). Antarctic orca in New Zealand waters?

Vongraven, D., and Bisther, A. (2014). Prey switching by killer whales in the north-east Atlantic:

observational evidence and experimental insights. Journal of the Marine Biological

Association of the United Kingdom, 94(6), 1357-1365.

Wade, P. R., and Gerrodette, T. (1993). Estimates of cetacean abundance and distribution in the

eastern tropical Pacific. Report of the International Whaling Commission, 43, 477-493.

Waite, A., Thompson, P., Pesant, S., Feng, M., Beckley, L., Domingues, C., Gaughan, D., Hanson, C.,

Holl, C., and Koslow, T. (2007). The Leeuwin Current and its eddies: An introductory

overview. Deep Sea Research Part II: Topical Studies in Oceanography, 54(8), 789-796.

Wellard, R., and Erbe, C. (2017). Killer whales of the Bremer Sub-Basin: A photo-ID catalogue.

Perth, WA, Australia: Centre for Marine Science and Technology, Curtin University.

Wellard, R., Erbe, C., Fouda, L., and Blewitt, M. (2015). Vocalisations of killer whales (Orcinus orca)

in the Bremer Canyon, Western Australia. PLoS ONE, 10(9), e0136535, doi:

10.1371/journal.pone.0136535.

Wellard, R., Lightbody, K., Fouda, L., Blewitt, M., Riggs, D., and Erbe, C. (2016). Killer whale

(Orcinus orca) predation on beaked whales (Mesoplodon spp.) in the Bremer Sub-Basin,

Western Australia. PLOS ONE, 11(12), e0166670, doi: 10.1371/journal.pone.0166670.

Yan, J. (2002). Geepack: Yet another package for Generalized Estimating Equations. R-News, 2(3),

12-14

Yan, J. and Fine, J. P. (2004). Estimating equations for association structures. Statistics in Medicine,

23(6), 859-880.

Zeger, S. L. and Liang, K. (1986). Longitudinal data analysis for discrete and continuous outcomes.

Biometrics, 42, 121-130.

40

Zerbini, A. N., Waite, J. M., Durban, J. W., LeDuc, R., Dahlheim, M. E., and Wade, P. R. (2007).

Estimating abundance of killer whales in the nearshore waters of the Gulf of Alaska and

Aleutian Islands using line-transect sampling. Marine Biology, 150(5), 1033-1045.

Zorn, C. (2001). Generalized Estimating Equation models for correlated data: A review with

applications. American Journal of Political Science, 45(2), 470-490.

Zuur, A. F., Ieno, E. N., Walker, N. J., Saveliev, A. A., and Smith, G. M. (2009). Mixed effects models

and extension in Ecology with R. New York: Springer Science+Business Media, LLC. 574 pp.