University of Southern California Dissertations and Theses

Using volunteered geographic information to model blue whale foraging habitat, Southern California Bight

(USC Thesis Other)

Using volunteered geographic information to model blue whale foraging habitat, Southern California Bight

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

USING VOLUNTEERED GEOGRAPHIC INFORMATION
TO MODEL BLUE WHALE FORAGING HABITAT,
SOUTHERN CALIFORNIA BIGHT

Matthew W. Bissell

A Thesis Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(GEOGRAPHIC INFORMATION SCIENCE AND TECHNOLOGY)

December 2013
! ""!
Acknowledgements

Several accolades are in order to acknowledge the people who made this thesis possible.
Thank you to Travis R. Longcore, my thesis chair, who was always available and flowing
with knowledge. And to the committee: Yao-Yi Chiang, Meredith Franklin, and Flora
Paganelli. Thank you all.

Thank you to the captains and crews aboard the five whale-watching vessels who took
the time to record whale observations during their many voyages, and for allowing me
access to their datasets: Mike Bursk, Peter Heistand, Larry Hartmann, Kera Mathes, Carl
Mayhugh, Cici Sayer, and Melissa Galieti. And thank you Greg Campbell, for allowing
access to the CalCOFI cetacean dataset.

And to Mary Elizabeth Portwood: the best friend, research assistant, secretary, cook,
companion, and partner in crime one could wish for. Thank you and I love you.

Thank you Mom, thank you Dad, for everything.

! """!
Table of Contents
Acknowledgements ii
List of Figures and Tables iv
Abstract v
Chapter 1: Introduction 1
Chapter 2: Literature Review 5
Chapter 3: Study Area and Data 14
Chapter 4: Methodology 21
Chapter 5: Results 28
Chapter 6: Discussion and Conclusion 40
References 46

! "#!
List of Figures and Tables
Figures

1. Temporal coverage of individual datasets 17
2. Map of whale observations 23
3. Spatial patterns of blue whale observations 29
4. Radar plots depicting seasonal variability 31
5. Maxent model: science-quality 33
6. Maxent model: whale-watch 34
7. Maxent model: whale-watch 35
8. Graph of model performance (AUC) 36
9. Graph of model performance (rate of omission) 37
10. Model comparison 38
11. Blue whale tracking 41
12. Blue whale habitat and commercial shipping routes 44

Tables

1. Description of whale presence datasets 17
2. Environmental variables 20
3. Seasonal variability analysis results 30
4. Model contribution of environmental variables 32

! "!
Abstract

Using Volunteered Geographic Information (VGI) to model blue whale
(Balaenoptera musculus) foraging habitat, this thesis assesses the utility of citizen science
in cetacean research and marine spatial management. A unique and new data source on
whale locations, observation data collected voluntarily by whale-watching vessels, was
procured, compiled, and digitized. The utility of this newfound dataset was investigated
through its use in probabilistic habitat suitability analyses and description of species
phenology. A statistical analysis of whale observations was used to quantify seasonal
variability of three common baleen whale species within the study area. Among these,
blue whales exhibit the highest degree of seasonal variability with a mean seasonal
abundance occurring in late July. Maximum entropy modeling was used to illustrate
potential blue whale foraging areas based on three environmental variables: bathymetry,
sea surface temperature, and chlorophyll-a concentrations. Spatial patterns of whale
observations recorded by whale watchers and scientists indicate a strong habitat
preference of steep bathymetric features in and around the 300-m isobath. Models using
whale-presence data collected by whale-watchers were compared to similar models using
science-quality whale observation data. Differences between these models are minimal
and the results of the comparison support the usefulness of citizen science in cetacean
research.
!
! "!
Chapter 1: Introduction
Volunteered Geographic Information (VGI) involves human volunteers, as
sensors of the environment, contributing in whole or in part to the creation, collection,
and/or dissemination of geographic information (Goodchild, 2008). This thesis composes
and analyzes the utility of a volunteered geographic dataset describing whale presence
locations in the coastal waters of Southern California. This new dataset was created by
combining the geographic information stored in the logbooks of five whale-watching
vessels operating daily and simultaneously from four ports within the Southern California
Bight. The high temporal (daily) resolution and vast geographic coverage of this dataset
is characteristic of VGI. These highly opportunistic observations are used here to create
probabilistic habitat suitability models for blue whales within the Southern California
Continental Borderland. Habitat modeling is an essential component of conservation
biology and key to understanding how we can better share the marine environment with
vulnerable species. The potential for future use of this new whale observation dataset in
scientific research and marine management is the focus of this thesis.

Hunted to near extinction during the nineteenth and twentieth centuries, many
baleen whale species exist at a fraction of their pre-whaling population numbers (Mate
and Calambokidis, 1999). Since the international ban on commercial whaling in 1965,
many of these whale species have been slowly recovering from their once intensive
exploitation (Calambokidis and Barlow, 2004). Nevertheless, baleen whales are still a
vulnerable group and are faced with continuing threats from human activity. These
include water contamination, high levels of anthropogenic sound, entanglement with
! "!
commercial fishing gear, and collisions with ships (Laist et al., 2001). These physical
threats have physical locations, and if we intend to protect these threatened species, it is
necessary to identify the geographic areas where whales might be affected by such
activity. The understanding of when and where a population is most likely to occur is a
primary goal in conservation biology (Phillips et al., 2006). In addition, knowing how a
species interacts with its environment can help to forecast future habitat selection by that
species as the environment changes.

Habitat modeling is a useful tool that can provide important information
pertaining to a species’ potential range and distribution. Taking what is known about an
organism’s presence and habitat requirements, statistical models can predict the
probability that other areas will also offer suitable habitat for that organism (Phillips et
al., 2006). Much work has been done to study whale habitat and many variations of whale
habitat models have been developed to give us a better understanding of how these
animals interact with their environment (Munger et al., 2009; Burtenshaw et al., 2004;
Moore et al., 2002; Fiedler et al., 1998; Croll et al., 1998). Recently, several studies have
combined this information with data on shipping routes and other human-induced
hazards, identifying regions with high risk of human-induced whale mortality (Redfern et
al., 2013; Pittman and Costa, 2010). This knowledge can be used to inform marine
managers when designing and implementing maritime regulations or to recommend best
practices to boaters operating their vessels within known whale habitat. The predictive
capabilities of these habitat models are commensurate with the quality and resolution of
the data put into the model. Currently, most whale habitat models utilize whale presence
! "!
data obtained from scientific surveys with regularized spatial sampling techniques
performed by trained biologists. These surveys are often laborious, time consuming, and
expensive, and as a result are conducted infrequently. For example, a commonly cited
dataset for whale observations in California is from the California Cooperative Oceanic
Fisheries Investigation (CalCOFI), with surveys being conducted quarterly (i.e., four
times per year). While this data is of high quality on many levels, it is of low temporal
resolution. Models using this data can still be effective, but they are often limited by
small sample sizes resulting from scarcity of surveys and lack of data during winter
months when weather conditions deteriorate (Munger et al., 2009). One solution to this
limitation is the use of citizen science in the data collection process. A sub-category of
VGI, citizen science is capable of amassing large amounts of data covering vast
geographic areas over short periods of time (Conrad and Hilchey, 2011).

This thesis utilizes a newfound volunteered geographic dataset provided by the
eco-tourism industry of Southern California. Assembled from five whale-watching
vessels operating daily and simultaneously from four ports within the Southern California
Bight, the data is of remarkably high temporal resolution and was collected at no cost.
While subject to several biases, which will be discussed in subsequent chapters, this new
dataset demonstrates the potential for increased citizen science in marine mammal
research.

! "!
Objectives
The objectives of this thesis are: (1) to demonstrate the spatial patterns of opportunistic
observations of baleen whales within the Southern California Bight; (2) to provide
descriptive statistics of the seasonal variability of whale sightings within the Southern
California Bight; (3) to use these opportunistic observations to produce probabilistic
habitat suitability analyses describing the potential spatial and temporal extent of whale
presence within the larger Southern California Continental Borderland; (4) to compare
these results with models using science-quality data from expeditions with more regular
spatial sampling schemes but much lower temporal resolution; and (5) to thereby assess
the utility of opportunistic observations in scientific research.
! "!
Chapter 2: Literature Review
Whales have a long documented history in the Southern California Bight. Much
of the early documentation of whales in this area came from industrial whaling outposts
operating along the coast during the nineteenth and twentieth centuries; six such outposts
were known to exist between Point Conception and San Diego (Starks, 1922). These
operations often maintained records of their takes of whales and sales of their products,
which contributed to early estimates of whale populations in the area. After nearly 100
years of intense exploitation, the hunting of whales off the California coast was brought
to an end in 1965 (Calambokidis et al., 2009). Following a half-century of slow
repopulation, these whales are again a valuable resource for the local economies. This
time, however, it is not for their oil and meat. Whale-watching boats take droves of
passengers into the waters of Southern California to view these animals nearly every day
of the year. And much like the whaling industry before them, these eco-touring vessels
hold the potential to be very rich sources of whale presence data.

An estimated 30 species of cetaceans reside in the eastern Pacific Ocean (Balance
et al., 2006). A number of these whale species can be observed in the waters off Southern
California; CalCOFI biologists identified 15 different whale species in this area during
their quarterly survey cruises in 2009–2010. This number was also observed during the
2010–2011 cruises, and 14 different species were recorded in the 2011–2012 surveys
(Campbell et al., 2010, 2011, 2012). Of these observed species, fin (Balaenoptera
physalus, blue (Balaenoptera musculus), gray (Eschrichtius robustus), and humpback
(Megaptera novaeangliae) whales were the most common of the baleen whales.
! "!
This thesis focuses on the blue whale for four reasons. (1) Blue whale occurrences
in the VGI dataset exhibit the most consistent seasonal variability with over 90 percent of
their occurrences taking place in July, August, and September. This phenomenon makes
for convenient monthly comparisons between years. (2) Blue whales are actively feeding
in the study area during these months and their presence or absence is largely affected by
the availability of their food source (Croll et al., 1998; Fiedler et al., 1998). This creates a
starting point for model development; e.g., which environmental variables have an effect
on prey production? (3) Blue whales feed almost exclusively on krill, a planktonic
crustacean whose presence and abundance is closely linked to the local environmental
conditions (Croll et al., 1998; Fiedler et al., 1998). Other species of baleen whales known
to forage in the study area have a much more varied diet consisting of zooplankton and
several species of small fishes, making their food source more difficult to model. (4)
Records of blue whales in the ships’ logbooks are often more complete than records of
other species. This is likely due to the intrinsic value of the blue whale to the whale-
watching industry. A captain can more easily please a group of passengers by presenting
them with a whale of many superlatives than with smaller, less impressive animals. As a
result, a blue whale observation will almost always be recorded with precise coordinates
of latitude and longitude. The locations of other whale species (fin, gray, minke, etc.)
recorded in the logbooks were sometimes noted as “out front,” “near red buoy,” or “off
blue house.” These types of vague descriptors were seldom used for describing a blue
whale’s location.

! "!
The blue whale is said to be the largest animal known to have lived on this planet;
they are capable of reaching lengths of nearly 100 feet, and weighing up to 120 tons
(Hass, 2011). Harvested to near extinction in the nineteenth and twentieth centuries, the
global population is estimated between ten and fifteen thousand whales: merely 8 percent
of their pre-commercial whaling numbers (Mate and Calambokidis, 1999). The National
Oceanic and Atmospheric Administration (NOAA) lists the blue whale as a depleted
species (population below optimum sustainable levels) under the Marine Mammal
Protection Act of 1972, and in danger of extinction under the Endangered Species Act of
1973. Blue whales are also listed as endangered species by the International Union for
Conservation of Nature (IUCN) and the Convention on International Trade in
Endangered Species (CITES). The remaining blue whales are found throughout the
world’s oceans and reside in distinct populations that seldom mix (Burtenshaw et al.,
2004). As members of a suborder of cetaceans called Mysticeti, their mouths are
equipped with baleen plates designed for capturing small planktonic prey. They feed
almost exclusively on krill, a small crustacean found in all of the world’s oceans (Croll et
al., 1998; Fiedler et al., 1998). Very mobile animals, they partake in extensive annual
migrations from summer foraging grounds to areas of breeding and calving during the
winter (Pittman and Costa, 2010). While we still do not fully understand the complete
annual migrations of these animals (i.e., where they go for breeding and calving), they do
show a high level of fidelity to their summer foraging grounds (Pittman and Costa, 2010;
Mate and Calambokidis, 1999). The coastal waters off California serve as a foraging area
for possibly the largest remnant population of blue whales in the world (Mate and
Calambokidis, 1999). Recent estimates of this population suggest between 2,000 and
! "!
3,000 individuals (Calambokidis and Barlow, 2004). Current threats to this species
include anthropogenic sound production, water contamination, entanglement with
commercial fishing gear, collisions with ships, and illegal whaling (Laist et al., 2001).

Several methods have been used to model whale habitat. Most variations use a
combination of environmental variables as indicators of suitable habitat or proxies for
potential food availability. Bathymetry, sea surface temperature (SST), and chlorophyll-a
are common physical variables used in these models (e.g., Redfern et al., 2010; Munger
et al., 2009; Balance et al., 2006), as described in the following sections.

Bathymetry
Bathymetry is the measurement of the ocean’s depth and describes the underwater
features that make up the sea floor. Bathymetry has a profound impact on the abundance
and diversity of organisms living in the water column above (Pittman and Costa, 2010).
Pittman and Costa (2010) discuss the high predictive powers of bathymetry alone in
modeling whale abundance and distribution, noting that edge habitats (e.g., continental
slopes) are strongly linked to high concentrations of prey. Seafloor features have a
significant influence on the vertical and horizontal movement of water and the resulting
eddies can serve to collect and maintain large concentrations of krill (Croll et al., 1998).
As a result, steep bathymetric features are necessary for blue whales to exploit their tiny
prey (Fiedler et al., 1998). In the northwest Pacific, seamounts, slopes, and other
prominent bathymetric features were identified as focal points for blue whales throughout
the year (Moore et al., 2002). Pitman and Costa (2010) describe the 100-m isobath line as
! "!
a “cetacean superhighway” for whales along the southern gulf of Maine and argue that
bathymetry data should be a prime candidate when choosing environmental variables to
model whale habitat. Burtenshaw and others (2004) identified bathymetry as an import
variable that could have added predictive power to their model (they did not include
bathymetry).

Sea Surface Temperature
Sea Surface Temperature (SST) is a measure of the temperature of the uppermost
layer of the ocean to about one meter (Campbell and Wynne, 2011). It can be measured
in situ via boats, buoys, and underwater autonomous gliders, or remotely via airborne and
space-borne sensors. SST is a fundamental component of marine ecology; oceanic
temperatures can define marine habitats and detect biological hotspots (Etnoyer et al.,
2006). A study of blue whale distributions off Southern California found SST to be an
important variable influencing the presence or absence of whales (Munger et al., 2009).
The authors found blue whales to be associated with colder SST when compared to
random locations in the study area. This was thought to be a result of oceanic processes
that, in addition to bringing cold water to the surface, foster prey production,
accumulation, and retention (Munger et al., 2009). Also in waters off Southern
California, two reports published in 1998 identify low relative water temperatures as an
indicator of potential blue whale habitat (Croll et al., 1998; Fiedler et al., 1998). In each
of these studies the majority of blue whale observations were made in cold, well-mixed
water that had been upwelled north of the sighting location and advected south via the
California Current System. Similarly, a study conducted in the Northwest Pacific found
! "#!
blue whales to be associated with colder than area-average SST (Moore et al., 2002).
These whale clusters were also near SST fronts with sharp gradients. A study in 2007
found blue whales to be more closely correlated with SST fronts than any other whale
species in their study (Doniol-Valcroze et al., 2007). This relationship between blue
whale habitat selection and SST is also documented in the Great Australian Bight where
the species has been linked to SST of about 1 degree Celsius cooler than average SST in
the study region (Gill et al., 2011).

Chlorophyll-a
Chlorophyll-a is the measure of primary productivity in the ocean’s upper layer.
Plantlike plankton occupying the sunlit portion of the world’s ocean use chlorophyll-a
and other pigments to perform photosynthesis. These colorful pigments can be detected
via satellite remote sensing and the varying concentrations of chlorophyll-a on the
ocean’s surface can be discerned. Chlorophyll-a has been shown to be an important
environmental variable when modeling the habitat of blue whales. Because this pigment
is an indicator of primary productivity (the food source of zooplankton) it can be used as
a proxy for blue whale prey production. Several studies have associated blue whale
presence with high levels of chlorophyll-a. In 2002, Moore and others noted that blue
whales in the northwest Pacific were associated with high concentrations of chlorophyll-a
in the spring; this strong association was not observed later in the foraging season. The
authors hypothesized this was due to the voracious primary consumption of
phytoplankton by the zooplankton, coupled with the reduced input of nutrients on the
back end of the upwelling season (Moore et al., 2002). In a separate study of blue whales
! ""!
in the northeast Pacific, their abundance was also associated with high levels of
chlorophyll-a (Burtenshaw et al., 2004). This study identifies a time lag between peak
chlorophyll-a concentrations and whale presence on the order of several months
(Burtenshaw et al., 2004).

Citizen science
Citizen science is the involvement of volunteers in some or all aspects of
scientific research (Conrad and Hilchey, 2010). There are several benefits of citizen
science: large data sets can be compiled very quickly and inexpensively (Trumbull et al.,
2000) and processes can be observed over large geographic areas (Dickenson et al.,
2010). The results produced by citizen science not only provide decision makers with
vital information on important matters, but the entire process increases public awareness
and public involvement in these same issues (Bonney et al., 2009; Goffredo et al., 2010).

Using nonprofessional scientists as volunteer sensors of the environment is
nothing new. One enduring citizen science campaign, the Christmas Bird Count, began in
1900 as a means to discourage over-hunting and promote ecological awareness (Bianchi,
1999). The 27 volunteer birders that took part in the inaugural bird count has grown to
involve over 50,000 citizens worldwide volunteering to collect bird observation data each
year. Continuing today, it is producing an ever-expanding online data set providing
critical information on distributions, ranges, and migration patterns of avian species
(Audubon, 2013). An effort of this scale, both spatially and temporally, would be nearly
impossible without the involvement of citizen scientists. This successful use of
! "#!
volunteered data in scientific research can also be seen in the realm of marine science and
marine management. In an effort to survey the underwater ecosystems of coastal Italy,
researchers utilized the effort and enthusiasm of 3,825 volunteer SCUBA divers. In just
four years the project was able to amass nearly 19,000 biological surveys of Italian
marine ecosystems. Collectively these divers contributed over 13,000 hours of
underwater data collection at no cost to the research organization. A later assessment of
this data concluded that the quality and accuracy of the volunteered data was equal to
data collected by trained divers on precise science-based transects. A subsequent and
independent study performed by Italy’s Ministry of the Environment validated the
ecological findings of the VGI dataset (Goffredo et al., 2010). Similarly, a study on the
east coast of the United States involved nearly 1,000 citizen scientists to monitor 750
kilometers of coastline. Throughout this vast geographic area, volunteers surveyed the
inter-tidal ecosystems in search of invasive species of crabs. These volunteers were found
to have a high level of accuracy when identifying species and the volunteered data
detected a range expansion of one species of invasive crab (Delaney et al., 2008).

The development and evolution of several technologies have increased the ability
of citizens to participate in science and, more specifically, geographic research
(Goodchild, 2007). The World Wide Web has essentially connected the world, allowing
large amounts of data to be easily shared, compiled, and analyzed (Goodchild, 2007).
Global Positioning Systems (GPS) allow for locations of objects to be easily and
accurately measured and recorded by non-trained individuals; this exercise is an essential
component of geography. Digital cameras (many enabled with GPS technology) allow
! "#!
the average person far more access than ever before to photography, further enabling the
citizen scientist (Goodchild, 2007). Mobile web devices and the ever-expanding
capabilities of smart phones also increase the ability of the general public to participate in
scientific research (Haklay, 2013).

With more citizens now capable of collecting accurate geographic data, the quality and
size of VGI and citizen science-based datasets have increased. And because humans are a
rather ubiquitous species, these datasets are often of very high spatial and temporal
resolution. Temporal resolution refers to the frequency of data collection with respect to
time: Data collected daily gives the dataset a higher degree of temporal resolution than a
dataset whose measurements are made weekly, monthly, annually, etc. Temporal
resolution is an important factor in habitat modeling, and fine-scale (daily) temporal
resolution may be crucial for detecting temporal trends (Kearney et al., 2011).!
! "#!
Chapter 3: Study Area and Data
Study Area
The study area is described in two parts: the geographic area from which the
whale observation data was collected, and the geographic extent of the extrapolated
probabilistic habitat suitability models. The whale observation data was collected from
within the Southern California Bight. This area is defined as the coastal waters (from the
shoreline to the continental shelf) from Point Conception in the north to the U.S.-
Mexican border in the south. It is defined by a wider than average continental shelf, with
complex bathymetry consisting of many basins and ridges. Oceanic circulation within the
Southern California Bight is also unique; the northbound Davidson Current brings warm
water up along the coast, while the California Current carries cold nutrient-rich water
southward and further offshore. These opposing oceanic currents create a biological
transition zone that supports nearly 500 species of fish and more than 5,000 species of
invertebrates (Southern California Coastal Water Research Project, 2013).

The probabilistic habitat suitability models were applied to the Southern
California Bight and extrapolated south to Vizcaino Bay, Mexico, about halfway down
the Baja California peninsula. This larger area known as the Southern California
Continental Borderland is a natural extension of the Southern California Bight. The
continued biologic, oceanographic, and bathymetric complexity of this area creates a
geographic unit ideal for studying the behavioral ecology of marine mammals
(Henderson, 2010). The width of the continental shelf narrows quickly just north and
south of this region.
! "#!
Data
Whale presence data can be acquired via three general methods: visual
observation, acoustic detection, and radio telemetry (tagging). Most scientific research on
live whales uses at least one of these techniques. Each of these methods has its benefits as
well as its shortfalls. This thesis focuses on the use of visual observations in whale
research. Specifically, a comparison between whale observation data collected by
scientists on scientific surveys and whale observation data collected voluntarily on
commercial whale-watching tours will be made.

Science-Quality Whale Presence Data
A commonly cited data source for blue whale presence (visual observations) in
the Southern California Bight is the CalCOFI dataset. CalCOFI is a partnership formed in
1949 between the Department of Fish and Wildlife, NOAA Fisheries Service, and
Scripps Institute of Oceanography. The organization conducts quarterly cruises along a
series of transect lines extending perpendicular from the central and Southern California
coastlines. Originally commissioned to study the collapse of the sardine fishery in the
1940s, its mission has evolved and in 2004 the cruises began recording baleen whale
observations (among many other data). While this is an impressively longstanding study
with remarkable consistency, the temporal resolution of the data is low (quarterly).
Throughout a nine-year period, between 2004 and 2012, the CalCOFI cruises observed
blue whales (at least one) 121 times, accounting for 212 blue whale records (Campbell et
al., 2012). This dataset is referred to as the “science-quality” whale observation data used
! "#!
in this study. It includes 29 blue whale observations accounting for 64 blue whale
sightings within the study area and time frame.
Non-systematic research cruises (i.e., opportunistic, non-transect) typically
experience a higher number of whale sightings. Without the gridded constraints of a
transect line, this more dynamic approach can purposefully place the observer in areas of
high whale density. Using this method, professional scientists sighted 2,403 blue whales
in the waters off the coasts of Washington, Oregon, and California in a seven-year period
between 1991 and 1997 (Calambokidas and Barlow, 2004). While this opportunistic
approach might be more efficient at finding and documenting whales, it is also much
more susceptible to recounting the same whale multiple times. Testament to this, using
photo-identification techniques, the authors of this study determined that only 908 of the
2,403 blue whales were unique individuals (Calambokidas and Barlow, 2004).

Whale-watch observation data
Perhaps the most opportunistic of all whale observations are the ones made
aboard commercial whale-watching vessels. An increasingly popular eco-tourism
activity, whale-watching tours can be found in nearly every harbor in Southern
California. Operating daily and simultaneously throughout the Southern California Bight,
these vessels are potentially a rich source of whale presence data with very high temporal
resolution. The whale-watch data used in this study was collected from five whale-
watching vessels operating in four ports within the Southern California Bight: two vessels
in San Diego, and one each in Dana Point, Newport Beach, and San Pedro. Together, in a
! "#!
five-year period, these vessels have logged over 875 blue whale observations accounting
for more than 2,250 individual whale counts (Figure 1 and Table 1).

Figure 1: Gantt chart illustrating start dates and end dates of data collection by various
sources. Green indicates daily temporal resolution of whale-watch datasets, red indicates
quarterly temporal resolution of the science-quality CalCOFI dataset.

Table 1: Metadata describing the five individual datasets used in this study. The VGI
dataset is comprised of whale observations from whale-watching vessels operating out of
San Pedro, Newport Beach, Dana Point, and San Diego. The science-quality dataset is
collected and maintained by CalCOFI.
Dataset Start date End date
Duration
(years)
Temporal
resolution
Blue whales
counted
San Pedro 1/2/08 9/29/12 4.74 Daily 1412
Newport Beach 5/14/11 11/14/12 1.51 Daily 306
Dana Point 5/11/11 12/10/12 1.59 Daily 315
San Diego 1/29/12 12/1/12 0.84 Daily 217
CalCOFI 7/28/04 11/5/12 8.27 Quarterly 64

! "#!
Environmental Variables
Bathymetry, sea surface temperature, and chlorophyll-a are used here as
indicators of potential blue whale habitat (Table 2). These oceanographic parameters
were chosen due to their prevalent use in blue whale habitat modeling per the scientific
literature.

Bathymetry
Bathymetry is the measure of the ocean’s depth. This variable is highly correlated
with the fine-scale presence and distribution of blue whales in the literature. The
bathymetric data used in this study, ETOPO1, is a product of NOAA’s National
Geophysical Data Center. This global relief model is a compilation of numerous global
and regional datasets. Used here in raster format, the data has a spatial resolution of
0.016667 arc-degrees (1.85 km
2
) and a vertical accuracy of about 10 meters.

Sea Surface Temperature
Sea Surface Temperature is the measured temperature of the uppermost layer of
the ocean to about one meter (Campbell and Wynne, 2011). Temperature is an important
environmental variable that can influence the presence and distribution of many marine
species. While blue whales are capable of residing in a wide range of temperatures, the
literature suggests that SST can be used as an indicator of prey production, thus blue
whale occurrence. The SST data used here is captured by the Moderate Resolution
Imaging Spectroradiometer (MODIS), a 36-band spectroradiometer measuring visible
and infrared radiation, onboard NASA’s Aqua satellite. Specifics of the MODIS
! "#!
instrument can be found at: http://aqua.nasa.gov/about/instrument_modis.php. The Aqua
satellite scans the earth’s surface every one to two days recording SST with a spatial
resolution of 0.0125 arc-degrees (1.47 km
2
). The Goddard Ocean Biology Processing
Group processes this raw data using multi-sensor level-1 to level-2 software. More on
this process can be accessed via the SeaWiFS data processing website at:
http://oceancolor.gsfc.nasa.gov/DOCS/SW_proc.html. Processed SST values are
validated with in-situ SST buoy data. This dataset covers the eastern pacific (155º W to
105º W Longitude, 22º N to 51º N Latitude) and maintains a nominal accuracy of ± 1
degree Celsius. Summary data of one-month averages are used in this study.

Chlorophyll-a
Chlorophyll-a is a specific pigment essential to photosynthesis. Found in the cells
of many photosynthesizing marine organisms, this pigment can be detected via remote
sensing satellites and is a strong indicator of primary productivity. The chlorophyll-a
dataset used here was obtained from MODIS aboard NASA’s Aqua satellite.
Concentrations of chlorophyll-a are recorded in milligrams per meter
3
with a spatial
resolution of 0.0.05 arc-degrees (5.55 km
2
). Raw data is processed at the Goddard Space
Flight Center using SeaWIFS Data Analysis System software (NOAA Coast Watch,
2013). With near-global spatial coverage (180º W to 180º E longitude, and 75º N to 75º S
Latitude) this dataset maintains a nominal accuracy of 40 percent. It is important to note
the significant discrepancies that exist between datasets derived from different remote
sensing apparatus as well as with high-quality in-situ measurements (NOAA Coast
Watch, 2013). Summary data of one-month averages are used in this study.
! "#!
Table 2: Descriptions of environmental variable datasets. Each dataset was downloaded
from NOAA’s Coast Watch Program,
(http://coastwatch.pfeg.noaa.gov/coastwatch/CWBrowser.jsp).

Environmental
variable
Source /
instrumentation
Spatial
resolution
Temporal
Resolution
Unit of
measurement
Bathymetry ETOPO1 1.85 km
2
Static Meter (m)
SST Aqua MODIS 1.47 km
2
1-month average Degrees Celsius
Chlorophyll-a Aqua MODIS 5.55 km
2
1-month average Milligrams/meter
3

!
! "#!
Chapter 4: Methodology
Science-quality whale observation data
The science-quality whale observation data used in this study was collected as
part of the CalCOFI project. Four times per year research vessels follow a series of
transect lines within the CalCOFI study area. Trained observers scan the ocean, during
daylight hours and when weather conditions permit, using 7x power binoculars. When a
whale is spotted, observers use 18x power binoculars to aid in the identification of the
species. Each whale sighting is logged and includes distance and bearing from ship,
species identification, group size, group composition, and the animal’s behavior
(Campbell et al., 2009). Greg Campbell of Scripps Institute of Oceanography currently
maintains CalCOFI cetacean data. At the time of writing this thesis the CalCOFI cetacean
data is undergoing comprehensive quality control, and is not yet publicly available.
Campbell generously, and personally, provided the CalCOFI blue whale data used here.

Whale-watch observation data set
Each commercial vessel operating more than three miles offshore is required by
the United States Coast Guard to maintain a logbook documenting all maintenance of
safety equipment and the dates emergency drills are performed. In addition, the logbooks
often list dates of re-fueling, oil changes, and miscellaneous other maintenance
operations, number of passengers, length of voyage, and the coordinates of ash dispersal
during burials at sea, among other data. Apart from this information, vessels involved in
whale-watching tours often record the locations of whale observations. Their motivation
to record a whale’s location is simple; if you know where a whale was, you have a better
! ""!
chance of finding it on a future trip. Many of these vessels are running two or three trips
daily, and if an opportunity arises to revisit the same whale later in the day they will.

The methods of whale detection and documentation employed by whale-watching
vessels vary. In general, the first boat out in the morning will set a course to an area
frequented by wildlife the day before. While motoring to this location, the captain and
possibly one or two deckhands will scan the area for signs of whale activity. These
observers are usually equipped with binoculars of varying optical magnification.
Generally, the first and or closest whale to be spotted will be visited and should be
recorded. Because the purpose of the trip is to let the passengers experience a whale,
rather than count every whale in the area, the vessel will likely stay with one or two
whales for a long period of time before continuing the search or returning to port. It is
usually the whale’s last known location that is recorded in the logbooks. A whale’s
location is used to revisit the same whale on a later excursion or to share with other
whale-watching vessels in the area; the captains of these vessels maintain open dialog via
VHF radio.

Currently, not every commercial whale-watching vessel maintains a log of whale
sightings. Furthermore, of the vessels that do record whale sightings, not all contain
geographic references. After inquiring at each whale-watching operation in the study
area, five vessels were found to maintain georeferenced whale records. Each of these
vessels was visited and the logbooks were recorded using the camera-video function on
an iPhone (the captains all preferred that the logbooks not physically leave the vessels).
! "#!
Later, the video recordings were played back, and the whale records were viewed and
manually transfer into an Excel spreadsheet. To control and assure the quality of this
data, and to minimize error in the transcription process, the following measures were
taken. If the species and/or coordinates of any logbook entry were undecipherable by two
persons, they were not included in the digital dataset. If an entry was legible by only one
of two persons, a third party was used to confirm or repudiate the record. Only when the
unbiased third party confirmed was the entry included. For every 100 entries in Excel, 15
were chosen at random and crosschecked with the original source document. Entries were
imported into ArcMap 10.1 (Figure 2).

Figure 2: Two maps comparing the total number of blue whale observations contained in
each dataset. Left: Science-quality dataset collected over 9 years from 2004-2012 yields
26 blue whale observations (red dots) representing 64 blue whale counts. Right: Whale-
watch data collected over 5 years from 2008-2012 yields 847 whale observations (green
dots) representing 2,250 blue whale counts.
! "#!
Seasonal Variability

The seasonal variability of blue, fin, and minke whales was quantified using
circular statistics. One year is divided into 360 degrees with months replacing the
corresponding angles on a circular plot: 0 degrees is replaced with January, 30 degrees is
replaced with February, etc. Averaged monthly whale abundance values are recorded as
radii. These radar plots demonstrate mean whale abundances for each month at each
location throughout the year. The following equation was used to calculate the mean
angle and mean radius for each species at each location (Zar, 1996):

Where:
f
i
is one month’s whale abundance in each month
a
i
is each months corresponding angle
a is the mean angle
r is the radius of the mean vector

The mean angle (a) represents the time of year (month) when a particular whale
species, on average, can be observed. The radius of the mean vector (r) is a measure of
the amount of seasonal variability displayed by a species. This value ranges from a
minimum of zero to a maximum of one. When r equals zero, there is said to be no
detectable seasonal variability. A value of one indicates a significant change in species
occurrences between seasons.
! "#!
Habitat Suitability Models
Maximum entropy (Maxent) is a sophisticated approach to modeling a species’
geographic distribution (Phillips et al., 2004). Using a general-purpose, machine-learning
method, Maxent can generate probabilistic habitat suitability analyses describing the
spatial and temporal extent of a given species. Using a set of data points marking where a
species has been observed, and the environmental conditions associated with each at the
time the observations were made, Maxent will estimate the environmental requirements
for that species. The information is then used to estimate the range and distribution of this
species in non-sampled regions where the environmental conditions are known. It is
assumed that the localities of the sample points are collected without concern or influence
of the environmental variables used in the model.

Maxent is free software that can be downloaded from the Internet
(http://www.cs.princeton.edu/~schapire/Maxent/). It requires all species location data
points to be comma-separated values (CSV) in the form of species, longitude, and
latitude. This task was performed using Microsoft Excel and the data points were
uploaded into the Maxent software using the browser function on the Maxent interface.
Similarly, a directory containing the environmental variable files to be used in the model
(bathymetry, SST, chlorophyll-a) was uploaded. The files in this directory must all be in
ASCII format and contain the same geographic reference system, geographic extent, and
grid cell size (bathymetry and chlorophyll-a datasets are resampled to conform to the SST
grid cell size of 1.47 km
2
). Each of these formatting requirements was executed using
ArcMap 10.1. First, a map was composed including a raster file of each environmental
! "#!
variable and a shapefile of the study area. In the Spatial Analyst toolbox, the Extraction
by Mask tool was used to extract the study area from each raster file. The dialog box for
this operation allows the user to set the parameters for each output file; the geographic
extent, geographic reference system, and grid cell size were set to the same values for
each environmental variable. The conversion tool in the spatial analyst toolbox was used
to convert the extracted raster layers to ASCII files (a Maxent requirement). Once the
species locations and environmental variables were uploaded into the Maxent software,
the model could be performed. Before running a model, several parameters were adjusted
in the Maxent settings field. The number of samples to be set aside for testing was set to
25 percent, allowing the performance of the resulting model to be tested using a random
selection of 25 percent of the species location points. The number of model replications
was set to 15. This tells the software to independently create 15 versions of the model and
to average the results. The resulting model is theoretically more robust than any single
replication. This process was repeated for each month of whale observation data.

Samples With Data
A second approach to running a Maxent model is to provide each species location
data point with an individual set of environmental variable values corresponding to the
time and place the observation was made. This method is referred to as the samples with
data (SWD) format. It can be advantageous when dealing with sample points collected
during different time periods and thus different environmental conditions. And because a
Maxent model’s performance increases with an increase in training data (Phillips and
Dudik, 2008), this approach can lead to higher performance by increasing the sample
! "#!
size. SWD was chosen due to the modest sample size and long periods of time between
sample points in the CalCOFI dataset.

The following procedure was used to create a file of sample points with
environmental data to be used in SWD format. A series of maps was created containing
sample points collected within the same time period (month) and the corresponding
environmental variables for that time period. This was accomplished in ArcMap. In the
Spatial Analyst toolbox the Extract Multiple Values to Points tool was used to affix the
environmental variable values to each corresponding whale observation point location.
This procedure was repeated for each month’s set of whale observations. The amended
attribute tables for the point locations were copied and pasted into Excel and saved as a
CSV file. These files were uploaded into Maxent.

Model Comparison
Model outputs were compared and contrasted using map algebra, a way to
analyze multiple maps using algebraic expressions. Used here to view the differences and
similarities between two Maxent model outputs, the analysis was accomplished using
ArcMaps’s map algebra function in the Spatial Analyst Toolbox. Using the raster
calculator within the map algebra toolset, cell values from a model using volunteered
whale-watch data were subtracted from the corresponding cell values of a model using
science-quality whale presence data. The map algebra output file demonstrates areas
where the two models agree and disagree (Figure 9).

!"#
Chapter 5: Results
Volunteered Whale Observation Dataset
The volunteered dataset composed here of the observational records of five whale-watching
vessels included over 3,000 logbook pages describing five years of whale-watching activity
within the study area. Held in these pages were 2,250 blue whale observations, 1,218 fin whale
observations, and 172 minke whale observations. Also found in these logbooks were accounts of
rare whales seldom seen on infrequent scientific surveys, including sperm whales (Physeter
macrocephalus), orca whales (Orcinus orca), sei whales (Balaenoptera borealis), and bairds
beaked whales (Berardius bairdii). Figure 3 demonstrates the spatial patterns of opportunistic
observations of blue whales.

Seasonal variability
Each whale species observed and recorded by the whale-watching vessels displayed unique
seasonal variability. Table 3 shows the statistical results of the seasonal variability analysis. The
radius of the mean vector (r) is a measure of the amount of seasonal variability displayed by a
species. This value ranges from a minimum of zero to a maximum of one. When r equals zero
there is said to be no detectable seasonal variability. A value of one indicates a significant
change in species occurrences between seasons. Among the whales in the volunteered dataset the
blue whale exhibited the strongest and most consistent seasonal variability (r=0.83). The mean
angle (a) represents the time of year (month) when a particular whale species, on average, can be
observed. The cumulative mean angle for blue whales (averaged over four locations) is 206º,
suggesting that blue whales are most likely to be observed in late July. Radar plots show each
month’s mean whale abundance for blue, fin, and minke whales at each location (Figure 4).
!"#
Spatial patterns of opportunistic observations of blue whales

Figure 3: Spatial patterns of blue whale observations (white dots) and associated bathymetry.
The average depth of blue whale observations was 297 meters (162 fathoms). Within the study
area, this depth is associated with steep bathymetric features thought to be responsible for blue
whale prey production, accumulation, and retention (Croll et al., 1999).

NOAA/NGDC
NOAA/NGDC
0 2 4
Nautical Miles
0 10 20
Nautical Miles
±
!"#
Table 3: Statistical results of the seasonal variability analysis and the contributing data for the
ensuing radar plots. Mean angle (a) represents the time of year (month) when a particular whale
species, on average, can be observed (January=0 º, April=90 º, July=180 º, etc.). Radius of the
mean vector (r) is a measure of the amount of seasonal variability displayed by a species
(0=low, 1=high).

Species Location Number
of whales
(n)
Mean
angle
(a)
Mean
vector
(r)
Cumulative
mean angle
Cumulative
mean
vector
Blue San Pedro 1412 214 0.79 206 º 0.83
Newport Beach 306 204 0.82
Dana Point 315 216 0.83
San Diego 217 188 0.89
Fin San Pedro 854 359 0.35 254 º 0.43
Newport Beach 111 260 0.42
Dana Point 96 268 0.56
San Diego 157 130 0.40
Minke San Pedro 120 228 0.29 170 º 0.57
Newport Beach 8 248 0.88
Dana Point 24 142 0.78
San Diego 20 60 0.34

!"#
Seasonal Variability

San Pedro:

Newport Beach:

Dana Point:

San Diego:

Blue Whale Fin Whale Minke Whale

Figure 4: Radar plots showing seasonal variability of three species of baleen whale observed
within the study area. Among these, blue whales exhibit the highest degree of seasonal
variability. Rings represent numbers of whales; radii represent seasonal occurrence.
!"#
Model Results
Maxent models demonstrate potential blue whale foraging habitat within the Southern
California Continental Borderland (Figures 5, 6, and 7). Warmer colors (yellows and reds)
indicate higher probability of suitable habitat, while cooler colors (blues and greens) indicate
lower probability of suitable habitat. Because bathymetry was given such a high predictive value
by the computer-learning models, seasonal variations due to temperature and chlorophyll-a are
minimal. Because of this, the aforementioned phenology of each species is highly important
when analyzing these models. Blue whales are primarily observed in the study area from July
through September and models projected onto other months will not be as accurate. Models
produced using the whale-watch data did not differ qualitatively from models using the science-
quality data. Both models rank the influence of environmental variables in identical order with
similar model contribution values assigned to each variable (Table 4). The extrapolated
predictions of each model are also very similar. The whale-watch data produced a habitat
suitability model with more definitive predictions, while the science-quality data produced a
more generalized model. This is most apparent throughout the southern portion of the models in
Vizcaino Bay, and is likely a result of the differences in model training sample sizes.

Table 4: Each environmental variable’s relative contribution to the Maxent model given as a
percent. The rank of contribution among environmental variables was consistent among models.

Whale Presence Data Bathymetry Sea Surface Temperature Chlorophyll-a
Science-quality (n=30) 51.9 29.6 18.5
Whale-watch (n=30) 60.6 26.9 12.5
Whale-watch (n=250) 49.7 34.1 16.1

!!"
Maxent model: science-quality dataset (n=30)

Figure 5: Probabilistic blue whale habitat suitability model using 30 samples of blue whale
locations collected during CalCOFI cruises between 2004 and 2012. The model is projected onto
the environmental conditions of August 2011. Warm colors indicate regions with a high
probability of suitable habitat; cool colors indicate regions with lower probability of suitable
habitat (AUC=0.945).

Sources: Esri, USGS, NOAA
Sources: Esri, USGS, NOAA
±
High
Low
0 100 200 50
Nautical Miles
Probability of Suitable
Blue Whale Habitat
!"#
Maxent model: whale-watch dataset (n=30)

Figure 6: Probabilistic blue whale habitat suitability model using 30 random samples of blue
whale locations collected by whale-watching vessels between 2008 and 2012. The model is
projected onto the environmental conditions of August 2011. Warm colors indicate regions with
a high probability of suitable habitat; cool colors indicate regions with lower probability of
suitable habitat (AUC=0.964).

Sources: Esri, USGS, NOAA
Sources: Esri, USGS, NOAA
±
High
Low
0 100 200 50
Nautical Miles
Probability of Suitable
Blue Whale Habitat
!"#
Maxent model: whale-watch (n=250)

Figure 7: Probabilistic blue whale habitat suitability model using 250 locations of blue whales
collected by whale-watching vessels between 2008 and 2012. The model is projected onto the
environmental conditions of August 2011. Warm colors indicate regions with a high probability
of suitable habitat; cool colors indicate regions with lower probability of suitable habitat
(AUC= 0.953).

Sources: Esri, USGS, NOAA
Sources: Esri, USGS, NOAA
±
High
Low
0 100 200 50
Nautical Miles
Probability of Suitable
Blue Whale Habitat
!"#
The receiver operating characteristic (ROC) curve is a graphical representation
comparing the fraction of true positives versus the fraction of false positives committed by the
model during test runs (Figure 8). Because presence-only data is used, the rate of commission
(false positives) is unable to be calculated. Maxent replaces this statistic with the fraction of the
total study area predicted present (Phillips, 2008). The area under this curve (AUC) is the
measure of a model’s performance. An AUC of 1 indicates a perfect model where every test
sample is accurately described. Alternately, an AUC value of 0.5 describes a random model with
a predictive average of 50 percent. This means half of the predictions are erroneous and half are
accurate (a performance that can be achieved by flipping a coin). The AUC for both the science-
quality and whale-watch-generated models indicate high levels of performance (0.945 and 0.953,
respectively). The increase in performance with the whale-watch data is most likely the result of
a larger dataset.

Figure 8: Average model sensitivity vs. specificity. The red line illustrates the mean area under
the curve (AUC). The blue buffer shows the mean standard deviation and the black line
represents random prediction. The two charts compare model performance between science-
quality (AUC=0.945) and whale-watch (AUC=0.953) datasets.

!"#
The calculated omission rate for whale-watch data was very similar to the predicted
omission rate, another indication of high performance (Figure 9). The omission rate for the
science-survey data did not conform to the line of predicted rate of omission, indicating a less
robust model. The probabilistic habitat suitability models show an increase in predictive
performance, with an increase in sample size. As a result, the large dataset provided by the
whale-watch vessels out-performed the science-quality data with a smaller sample size.

Figure 9: Average omission and predicted area. The black line (largely hidden behind yellow)
represents the model’s predicted omission. The green line depicts the mean omission on test data
and the orange buffer illustrates the mean standard deviation.

Model Comparison
An algebraic comparison (whale-watch model output subtracted from science-quality
model output) shows significant agreement between the two models (Figure 10). Red shows
areas with a high probability of suitable blue whale habitat as predicted by the science-quality
data and less suitable by the whale-watch data. Green, on the contrary, indicates areas predicted
as highly suitable by the whale-watch data and less suitable by the science-quality data. Yellow
shows areas where the two models are in agreement. The main difference between the models is
the nearness to shore of predicted blue whale foraging habitat. Models produced using whale-
! "#!
Model Comparison: whale-watch model subtracted from science-quality model

Figure 10: Using map algebra, cell values from the whale-watch model (Figure 6)
subtracted from corresponding cell values of the science-quality model (Figure 5). Red
indicates areas predicted highly suitable by the science-quality data and less suitable by
the whale-watch data. Green indicates areas predicted highly suitable by the whale-
watch data and less suitable by the science-quality data. Yellow shows areas where the
two models are in agreement.
Sources: Esri, USGS, NOAA
Sources: Esri, USGS, NOAA
±
Positive
Models agree
Negative
0 100 200 50
Nautical Miles
Whale-watch values subtracted
from Science-quality values
! "#!
watch data are more likely to imply near-shore foraging areas than are models using
science-quality whale observation data. And, vice-versa, the models using science-quality
whale data are more likely to suggest offshore suitable habitat. This result may be a
consequence of the nature and behavior of the whale-watching industry: Their restricted
temporal and spatial ranges for individual trips limit their offshore observations. The
science-quality whale data are supplied by cruises operating further offshore than the
typical whale-watching vessel.
!
! "#!
Chapter 6: Discussion and Conclusion

Volunteered whale-watch data proved to be highly effective for the creation of
probabilistic habitat suitability models. The large and inexpensive dataset created here of
whales occurring over a vast geographic area is characteristic of VGI. And the high
temporal resolution of this data can give insight into phenomena that cannot be detected
from infrequent scientific surveys. Because of Maxent’s tolerance to presence-only data it
is well equipped to utilize this highly opportunistic dataset. And the large sample size
provided the software with ample training information to learn and test the blue whale’s
habitat requirements.

The use of Maxent in whale habitat modeling can be further explored by the use
of additional environmental variables. With the apparent correlation between blue whale
occurrences and steep bathymetric features, slope and aspect values may be of significant
importance. These variables can be derived from the same bathymetric dataset and can
help explain in more detail the importance underwater features on blue whale prey
production and retention. Additionally, the use of oceanic current data may help describe
the spatial separation between these features and the areas where whales are sighted. By
incorporating these and other environmental variables the predictive power of the model
may be enhanced.!

Consistent in each model is an area of highly probable blue whale foraging habitat
in Vizcaino Bay, Mexico. This area, located in the middle of the Baja peninsula, occurs
outside of the CalCOFI study area and out of range for Southern California whale-
! "#!
watching vessels. As a result, the whale observation datasets used in this thesis could not
confirm this prediction. Model confirmation comes from a study conducted in 1995
where five blue whales were tagged in the Santa Barbara Channel, California. During the
time the tags remained attached to the whale’s bodies they allowed researchers to track
the paths of these animals (Figure 11). Of the five whales tagged, four made southerly
routes to Vizcaino Bay (Mate and Calambokidis, 1999). Furthermore, three of the four
southbound whales followed a path of suitable habitat as predicted by the Maxent model.

Figure 11: Tracks of five blue whales tagged in the Santa Barbara Channel in 1994 and
1995 are displayed in various colors. Four of the five whales traveled south making
temporary stops in Vizcaino Bay (Mate and Calambokidis, 1999). Vizcaino Bay is
identified as highly probable blue whale foraging habitat by the Maxent models.

! "#!
While whale-watch data is successfully used here in modeling blue whale
foraging habitat, several issues concerning the quality of this data limit its use in other
scientific pursuits. For this type of data to be used in a more encompassing scientific
capacity, the following issues must be addressed. The nature and overall intent of a
whale-watching cruise is fundamentally different from a true research cruise. Whale-
watching vessels are limited by the short duration of their search; a normal whale-
watching excursion ranges from two to five hours in length. As a result, they tend to
cover the same geographic area trip after trip. Whale-watching vessels may run two or
more trips per day, and to save fuel they will often only visit whales closest to the harbor-
relying heavily on locations of previous whale sightings. This can lead to recounting the
same whale as well as omitting whales that are outside of this limited area. Once a whale
is sited, whale-watching vessels tend to stay with the whale for a prolonged period of
time instead of continuing to search for different whales in the area. This cessation of
search will result in lower whale counts for the larger geographic area.

Because the nature and behavior of a whale-watching tour boat strongly
influences the pattern of its whale observations, the spatial patterns of opportunistic
observations of whales do not reflect the true spatial arrangement of the population. Maps
produced using whale-watch observation data are indicative of where whales are seen on
eco-tours rather than how the area’s whale population is distributed. The opportunistic
observations do, however, reveal an interesting pattern; the vast majority of the
observations were made on or near the 300m-isobath, a geographic feature locally
! "#!
associated with high bathymetric relief. This phenomenon is also detected in the science-
quality dataset and supported in the literature.

By accounting for the observational effort exhibited by a whale-watching vessel,
many of the aforementioned concerns may be mitigated. This can be accomplished by
logging the hours of active whale searching in addition to total hours of excursion. In
doing so, researchers can better understand the density or scarcity of whales within the
area. In addition, recording GPS tracks of each excursion can help researchers more
accurately define the observational area of effort covered by a whale-watching vessel.

Whale-watch data is also limited in other aspects of marine mammal research.
Without a means of identifying individual whales (i.e., photographic identification),
opportunistic observations are not conducive to studies of population numbers or
dynamics. One solution to this problem is for whale-watching vessels to photograph each
whale they encounter with a GPS-enabled digital camera. These geo-tagged photographs
can then be analyzed by scientists (citizen or professional) and individual whales within
the population can be identified. This would allow whale-watch data to be used for
additional scientific purposes other than habitat suitability analyses.

! ""!
Commercial shipping and blue whale foraging habitat
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
Figure 12: Probability of blue whale habitat versus areas of high shipping activity. Maps
such as these can be used to inform marine spatial planners when designing shipping
lanes, marine protected areas, and other marine management areas. Note the area of
highly suitable blue whale habitat where shipping lanes cross.

Sources: Esri, USGS, NOAA
±
Shipping
0 20 40 10
Nautical Miles
Probability of suitable blue whale habitat
with areas of dense shipping traffic
High
Low
! "#!
The goal of this and similar studies is to alert and bring awareness to boat
operators when in the vicinity of whales. Maps incorporating marine traffic and potential
whale habitat can be used to inform the maritime community of this spatial conflict
(Figure 12). An important aspect of using volunteered geographic information in this
pursuit is the direct involvement of the target audience. By involving this demographic in
the scientific process they are more likely to be interested in the results and take active
roles in the solutions.

Citizen science and volunteered geographic information can be of great service to
the future of marine mammal research and marine spatial planning. Currently, the
availability of these datasets is limited by the current method of recording data by hand
and storing data in logbooks dispersed among many vessels. Future work in this area
should include the unification and digitization of whale observations made by whale-
watching vessels and citizen scientists in general. This would allow for the acquisition of
near real-time whale location data by scientists and marine managers. This streamlining
of data acquisition can be accomplished via the creation of a main database capable of
receiving data entries from various mobile devices (phones, tablets, etc.). Users of these
devices could upload geo-tagged photos of whales in real-time. By increasing the
availability of this data and by decreasing the time between data collection and
utilization, marine managers will be better equipped to deal with the dynamic spatial
conflict between humans and whales.

! "#!
References

Ballance, L., R. Pitman, and P. Fiedler. 2006. Oceanographic Influences on Seabirds
and Cetaceans of the Eastern Tropical Pacific: A review. Progress in
Oceanography 69:360–390.
Bianchi, J. 1999. Christmas Bird Count. New York State Conservationist 54(3):6.
Bonney, R., C. Cooper, J. Dickinson, S. Kelling, T. Phillips, K. Rosenberg, and J. Shirk.
2009. Citizen Science: A Developing Tool for Expanding Science Knowledge and
Scientific Literacy. BioScience 59(11):977–984.
Burtenshaw, J., E. Oleson, J. Hilderbrand, M. McDonald, R. Andrew, B. Howe, and J.
Mercer. 2004. Acoustic and Satellite Remote Sensing of Blue Whale Seasonality
and Habitat in the Northeast Pacific. Deep Sea Research Part II 51:967–986.
Calambokidis, J., and J. Barlow. 2004. Abundance of Blue and Humpback Whales in
the Eastern North Pacific Estimated by Capture-Recapture and Line-Transect
Methods. Marine Mammal Science 20(1):63–85.
Calambokidis, J., J. Barlow, J. Ford, T. Chandler, and A. Douglas. 2009. Insights Into
The Population Structure of Blue Whales in the Eastern North Pacific From
Recent Sightings and Photographic Identification. Marine Mammal Science
25(4):816–832.
Campbell, G., K. Merkens, and J. Hilderbrand. 2010. California Cooperative Oceanic
Fisheries Investigation (CalCOFI) Cruises: 2009-2010. San Diego, CA. USA:
Marine Physical Laboratory, Scripps Institution of Oceanography.
Campbell, G., L. Roche, H. Basset, A. Simonis, D. Camacho, K. Whitaker, and J.
Hilderbrand. 2011. California Cooperative Oceanic Fisheries Investigation
(CalCOFI) Cruises: 2010–2011. San Diego, CA. USA: Marine Physical
Laboratory, Scripps Institution of Oceanography.
Campbell, G., L. Roche, K. Whitaker, and J. Hilderbrand. 2012. California Cooperative
Oceanic Fisheries Investigation (CalCOFI) Cruises: 2011–2012. San Diego, CA.
USA: Marine Physical Laboratory, Scripps Institution of Oceanography.
Campbell, James B. and Randolph H. Wynne. Introduction to Remote Sensing (5th
Edition). (Guildford Press, New York, 2011).
Conrad, C. C., and K. G. Hilchey. 2011. A Review of Citizen Science and Community-
Based Environmental Monitoring: Issues and Opportunities. Environmental
Monitoring Assessment 176:273–291.
Croll, D., B. Tershy, R. Hewitt, D. Demer, P. Fiedler, S. Smith, W. Armstrong, J. Popp,
T. Kiekher, R. Lopez, J. Urban, and D. Gendron. 1998. An Integrated Approach
to the Foraging Ecology of Marine Birds and Mammals. Deep Sea Research II
45:1353–1371.
Delaney, D., C. Sperling, C. Adams, and B. Leung. 2008. Marine Invasive Species:
Validation of Citizen Science and Implications For National Monitoring
Networks. Biological Invasions 10:117–128.
Dickinson, J., B. Zuckerberg, and D. Bonter. 2010. Citizen Science as an Ecological
Research Tool: Challenges and Benefits. Annual Review of Ecology, Evolution,
and Systematics 41:149–172.

! "#!
Doniol-Valcroze, T., D. Berteaux, P. Larouche, and R. Sears. 2007. Influence of
Thermal Fronts on Habitat Selection by Four Rorqual Whale Species in The Gulf
of St. Lawrence. Marine Ecology Progress Series 335:207-216.
Etnoyer, P., D. Canny, B. Mate, L. Morgan, J. Ortega-Ortiz, and W. Nichols. 2006. Sea-
Surface Temperature Gradients Across Blue Whale and Sea Turtle Foraging
Trajectories off the Baja California Peninsula, Mexico. Deep Sea Research Part II
53:340–358.
Fiedler, P., S. Reilly, R. Hewitt, D. Demer, V. Philbrick, S. Smith, W. Armstrong, D.
Croll, B. Tershy, and B. Mate. 1998. Blue Whale Habitat and Prey in the
California Channel Islands. Deep Sea Research II 45:1781–1801.
Gill, P., M. Morrice, B. Page, R. Pirzl, A. Levings, and M. Coyne. 2011. Blue Whale
Habitat Selection and Within-Season Distribution in a Regional Upwelling
System off Southern Australia. Marine Ecology Progress Series 421:243–263.
Goffredo, S., F. Pensa, P. Neri, A. Orlandi, M. Gagliardi, A. Velardi, C. Piccinetti, and F.
Zaccanti. 2010. Unite Research With What Citizens Do For Fun: "Recreational
Monitorring" of Marine Biodiversity. Ecological Applications 20(8):2170–2187.
Goodchild, M. F. 2008. Commentary: Whither VGI? GeoJournal 72:239–244.
Goodchild. M. F. 2007. Citizens as Sensors: The World of Volunteered Geography.
GeoJournal 69:211–221.
Haklay, M. 2013. Citizen Science and Volunteered Geographic Information: Overview
and Typology of Participation. In Crowdsourcing Geographic Knowledge:
Volunteered Geographic105 Information (VGI) in Theory and Practice, ed. D.
Sui, 105–122.
Hass, T. C. 2011. Improving Natural Resource Management : Ecological and Political
Models. Hoboken, NJ, USA: Wiley.
Henderson, E. E. 2010. Cetaceans in the Southern California Bight: Behavioral,
Aucustical, and Spatio-temporal Modeling. Ph.D. dissertation, University of
California, San Diego.
International Union for Conservation of Nature. IUCN Red List of Threatened
Species. In International Union for Conservation of Nature [database online].
Cambridge, United Kingdom, 2008. Available from
http://www.iucnredlist.org/details/2477/0 (last accessed June/15 2013).
Kearney, M., A. Matzelle, and B. Helmuth. 2012. Biomechanics Meets the Ecological
Niche: The Importance of Temporal Data Resolution. Experimental Biology
15:922–933.
Laist, D., A. Knowlton, J. Mead, A. Collet, and M. Podesta. 2001. Collisions Between
Ships and Whales. Marine Mammal Science 17:35–75.
Mate, B., and J. Calambokidis. 1999. Movements of North Pacific Blue Whales During
the Feeding Season Off Southern California and Their Southern Fall Migration.
Marine Mammal Science 15(4):1246–1257.
Moore, S., W. Watkins, M. Daher, J. Davies, and M. E. Dahlheim. 2002. Blue Whale
Habitat Associations in the Northwest Pacific: Analysis of Remotely-Sensed Data
Using a Geographic Information System. Oceanography 15(3):20–25.

! "#!
Munger, L., D. Camacho, A. Havron, G. Campbell, J. Calambokidis, A. Douglas, and J.
Hilderbrand. 2009. Baleen Whale Distribution Relative to Surface Temperature
and Zooplankton Abundance off Southern California. San Diego, California:
CalCOFI, Report Number, 50.
National Audubon Society Web Site. In National Audubon Society [database online].
New York, NY. USA, 2013 Available from
http://birds.audubon.org/christmas-bird-count (last accessed June 15, 2013).
National Oceanic and Atmospheric Administration. Coast Watch Website. Washington,
DC. USA, 2013 Available from
http://coastwatch.pfeg.noaa.gov/coastwatch/CWBrowser.jsp (last accessed
August 21, 2013).
Phillips, S. J., R. P. Anderson, and R. E. Schapire. 2006. Maximum Entropy Modeling of
Species Geographic Distributions. Ecological Modeling 190:231–259.
Phillips, S. J., and M. Dudik. 2008. Modeling of Species Distributions With Maxent:
New Extensions and a Comprehensive Evaluation. Ecography 31:161–175.
Phillips, S. J., M. Dudik, and R. E. Schapire. 2004. A Maximum Entropy Approach to
Species Distribution Modeling. Proceedings of the 21st International Conference
on Machine Learning 655–655–662.
Pittman, S., and B. Costa. 2009. Linking Cetaceans to Their Environment: Spatial
Data Acquisition, Digital Processing and Predictive Modeling for Marine Spatial
Planning in the Northwest Atlantic. In Spatial Complexity, Informatics, and
Wildlife Conservation, ed. Huettmann & Cushman, 387–408.
Redfern, J., M. McKenna, T. Moore, J. Calambokidis, M. Deangelis, E. Becker, J.
Barlow, K. Forney, P. Fiedler, and S. Chivers. 2013. Assessing the Risk of Ships
Striking Large Whales in Marine Spatial Planning. Conservation Biology 27:292–
302.
Southern California Coastal Water Research Project Website. Costa Mesa, CA. USA.
Available from
http://www.sccwrp.org/researchareas/RegionalMonitoring/BightRegionalMonitori
ng.aspx. (Last accessed August 21, 2013).
Starks, E. C. 1922. A History of California Shore Whaling. San Diego: State of
California, Department of Fish and Game, Report Number, Fish Bulletin No. 6.
Trumbull, D., R. Bonney, D. Bascom, and A. Cabral. 2000. Thinking Scientifically
During Participation in a Citizen-Science Project. Science Education 84:265–275.
Zar, Jerrold H. Biostatistical Analysis (4th Edition). (Prentice Hall, NJ. 1999).

Abstract (if available)

Abstract Using volunteered geographic information (VGI) to model blue whale (Balaenoptera musculus) foraging habitat, this thesis assesses the utility of citizen science in cetacean research and marine spatial management. A unique and new data source on whale locations, observation data collected voluntarily by whale-watching vessels, was procured, compiled, and digitized. The utility of this newfound dataset was investigated through its use in probabilistic habitat suitability analyses and description of species phenology. A statistical analysis of whale observations was used to quantify seasonal variability of three common baleen whale species within the study area. Among these, blue whales exhibit the highest degree of seasonal variability with a mean seasonal abundance occurring in late July. Maximum entropy modeling was used to illustrate potential blue whale foraging areas based on three environmental variables: bathymetry, sea surface temperature, and chlorophyll-a concentrations. Spatial patterns of whale observations recorded by whale watchers and scientists indicate a strong habitat preference of steep bathymetric features in and around the 300-m isobath. Models using whale-presence data collected by whale-watchers were compared to similar models using science-quality whale observation data. Differences between these models are minimal and the results of the comparison support the usefulness of citizen science in cetacean research.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Habitat suitability modeling of Mexican spotted owl (Strix occidentalis lucida) in Gila National Forest, New Mexico

PDF

Predictive habitat distribution modeling of sperm whale (Physeter macrocephalus) within the central Gulf of Alaska utilizing passive acoustic monitoring

PDF

A Maxent-based model for identifying local-scale tree species richness patch boundaries in the Lake Tahoe Basin of California and Nevada

PDF

3D visualization models as a tool for reconstructing the historical landscape of the Ballona Creek watershed

PDF

Using Maxent modeling to predict habitat of mountain pine beetle in response to climate change

PDF

Predicting Hydromantes shastae occurrences in Shasta County, California

PDF

An analysis of the North Rainier Elk Herd area, Washington: change detection and habitat modeling with remote sensing and GIS

PDF

Building better species distribution models with machine learning: assessing the role of covariate scale and tuning in Maxent models

PDF

Modeling burn probability: a Maxent approach to estimating California's wildfire potential

PDF

A comparison of GLM, GAM, and GWR modeling of fish distribution and abundance in Lake Ontario

PDF

Generating bicyclist counts using volunteered and professional geographic information through a mobile application

PDF

Deriving traverse paths for scientific fieldwork with multicriteria evaluation and path modeling in a geographic information system

PDF

Validation of volunteered geographic information quality components for incidents of law enforcement use of force

PDF

Spatial distribution of the endangered Pacific pocket mouse (Perognathus ssp. pacificus) within coastal sage scrub habitat at Dana Point Headlands Conservation Area

PDF

Assessing the transferability of a species distribution model for predicting the distribution of invasive cogongrass in Alabama

PDF

Testing LANDIS-II to stochastically model spatially abstract vegetation trends in the contiguous United States

PDF

Social media to locate urban displacement: assessing the risk of displacement using volunteered geographic information in the city of Los Angeles

PDF

Modeling open space acquisition in Boulder, Colorado

PDF

Guiding business oriented volunteered geographic information through geotigger services: a case study of CrossFit affiliates

PDF

Creating Hot Streets: developing an automated approach using ModelBuilder

Asset Metadata

Creator Bissell, Matthew W. (author)

Core Title Using volunteered geographic information to model blue whale foraging habitat, Southern California Bight

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 10/01/2013

Defense Date 06/16/2013

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag blue whales,citizen science,habitat modeling,Maxent,maximum entropy modeling,OAI-PMH Harvest,volunteered geographic information

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Longcore, Travis R. (committee chair), Chiang, Yao-Yi (committee member), Franklin, Meredith (committee member), Paganelli, Flora (committee member)

Creator Email bissell.matt@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-331686

Unique identifier UC11297477

Identifier etd-BissellMat-2057.pdf (filename),usctheses-c3-331686 (legacy record id)

Legacy Identifier etd-BissellMat-2057.pdf

Dmrecord 331686

Document Type Thesis

Format application/pdf (imt)

Rights Bissell, Matthew W.

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA