View
214
Download
0
Category
Preview:
Citation preview
Patterns of Natal Recruitment in the Atlantic Puffin (Fratercula arctica)
by
S. Erin Whidden
BSc. Biology, University of New Brunswick, 2005
A Thesis Submitted in Partial Fulfillment
of the Requirements for the Degree of
Master of Science
in the Graduate Academic Unit of Biology
Supervisor: Antony Diamond, Ph.D., Biology Department and Forestry and
Environmental Management
Examining Board: Dr. Les Cwynar, Ph.D. Biology Department, Chair
Dr. Stephen Heard, Ph.D., Biology Department
Dr. Heather Major, Ph.D., Biology Department at UNBSJ
This thesis is accepted by the
Dean of Graduate Studies
THE UNIVERSITY OF NEW BRUNSWICK
January, 2016
©S. Erin Whidden, 2016
ii
ABSTRACT
Seabirds exhibit considerable variation in demographic parameters within a typical K-
selected framework. Identification of the most sensitive of these parameters for
population change is crucial for successful management and conservation. When adult
survival is high and stable in seabird populations, bottom-up forces prevail and
reproductive success and subsequent recruitment of juveniles will govern population
dynamics. Recruitment is difficult to study, requiring large samples from throughout a
metapopulation. The Atlantic puffin (Fratercula arctica) colony on Machias Seal Island
is the largest in the Gulf of Maine population, and is ideally suited to explore recruitment
in this species since the entire metapopulation is monitored. I first assessed trends in
hatch-year parameters since the 1980s and studied whether hatch year conditions for
Atlantic puffins affected the probability of natal recruitment. Despite significant declines
in hatch-year parameters (wing growth, fledging body condition, quality food delivered
to chicks, fledge date), none had a statistically significant effect on the probability of
natal recruitment. The number of islands visited between fledging and recruiting was the
only parameter with a statistically significant effect on natal recruitment. Puffins are less
likely to recruit to their natal island the more islands they visit between fledging and
recruitment age. I conclude that conditions experienced between fledge and recruitment
may be more important in the process of natal recruitment than conditions experienced by
chicks on their natal colony. These findings highlight the importance of multi-year, multi-
site metapopulation-level studies when tracking changes in seabird dynamics.
iii
DEDICATION
To Iris. May you always be inspired to learn.
iv
ACKNOWLEDGEMENTS
I would like to thank my supervisor, Dr. Tony Diamond. Your seemingly infinite
patience and thoughtful advice helped me to regain direction and push forward so many
times. Thank you for this opportunity. Thanks also to my committee members, Drs.
Linley Jesson and Graham Forbes for your encouragement and constructive criticism.
The finished product is vastly improved thanks to you.
Thanks to my lab mates for your support and for many insightful conversations, ideas and
questions. And to a collaborator, Dr. Andre Breton, whose hours of hard work sadly did
not make it into the final draft. I am a better scientist for your efforts.
Finally, I would like to thank my family. Mum, Dad, Megan, Patrick, and Shannon. Your
love, support (in all ways possible), humour, and encouragement held me together
through all I faced as a student and in life. Thanks for listening, even when you had no
idea what I was going on about.
And to the puffins - thanks for always inspiring my curiosity.
v
Table of Contents
ABSTRACT ........................................................................................................................ ii
DEDICATION ................................................................................................................... iii
ACKNOWLEDGEMENTS ............................................................................................... iv
Table of Contents ................................................................................................................ v
List of Tables .................................................................................................................... vii
List of Figures .................................................................................................................. viii
Chapter 1 : Introduction ...................................................................................................... 1
Background ..................................................................................................................... 1
Hypotheses and Predictions ........................................................................................ 4
General Methods ............................................................................................................. 8
Species Account .......................................................................................................... 8
Study Area ................................................................................................................. 10
Objective and Chapter Summary ...............................................................................11
Chapter 2 : Trends in Chick Development on Machias Seal Island between 1980 and
2014................................................................................................................................... 13
Introduction ................................................................................................................... 13
Methods......................................................................................................................... 14
Field Protocol ............................................................................................................ 14
Statistical Analyses ................................................................................................... 16
Results ........................................................................................................................... 17
Chick Development .................................................................................................. 17
Food .......................................................................................................................... 18
vi
Breeding Success ...................................................................................................... 18
Timing of Fledge ....................................................................................................... 18
Discussion ..................................................................................................................... 19
Chapter 3 : Effects of hatch year conditions on natal recruitment in Atlantic Puffins
breeding on Machias Seal Island ...................................................................................... 30 Introduction ................................................................................................................... 30
Methods......................................................................................................................... 33
Data ........................................................................................................................... 33
Statistical Analyses ................................................................................................... 35
Results ........................................................................................................................... 37
Discussion ..................................................................................................................... 38
Chapter 4 : Synthesis ........................................................................................................ 55
Bibliography ..................................................................................................................... 59
Curriculum Vitae
vii
List of Tables
Table 2-1. Yearly sample sizes for parameters spanning from 1980 to 2014. .................. 23
Table 3-1. List and descriptions of parameters included in models. ............................... 42
Table 3-2. Count of how many birds were detected (survived) for each level of number of
islands visited. The proportion of birds that survived increases as more islands are
visited. ............................................................................................................................... 41
Table 3-3. Yearly means for parameters included in models. .......................................... 43
Table 3-4. Summary of results from generalised linear models with binomial response =
recruited to MSI (natal recruitment). Models were compared using Akaike's Information
Criterion (AICc). Number of parameters in the model (K), AICc, delta AICc (difference
between AICc of current and top model), AICc weight (a measure of support for each
model; values sum to 1) and cumulative AICc weight are reported. ................................ 44
Table 3-5. Model averaged estimates of parameter effects (with shrinkage). Weighted
averages of parameters present in top models (ΔAICc < 4). ........................................... 45
Table 3-6. Summary of results from generalised linear models with binomial response =
survived to recruitment age (1), was not detected at recruitment age (0). Models were
compared using Akaike's Information Criterion (AICc). AICc, delta AICc (difference
between AICc of current and top models), AICc weight (a measure of support for each
model; values sum to 1). K = number of parameters. ...................................................... 46
Table 3-7. Estimates of parameter effects (with shrinkage) for the only supported survival
model (ΔAICc < 4)............................................................................................................ 47
viii
List of Figures
Figure 1-1. Map showing locations of puffin breeding islands in the Bay of Fundy/Gulf
of Maine metapopulation. Islands in the inset represent monitored islands, while Bird
Island (BI), Pearl Island (PI), and the Tusket and Mud Island groups (TMG) represent
puffin breeding islands that are not monitored for banded birds. The size of each dot
roughly represents the relative sizes of the breeding populations on each island. ........... 12
Figure 2-1. Trends in Atlantic Puffin (Fratercula arctica) chick growth rates on Machias
Seal Island New Brunswick from 1997 to 2014. a. Mass growth rate (g/day) measured
during the linear growth phase (10-30 days old), showed no significant trend over time. b.
Wing Chord growth rate (mm/day), also measured during the linear growth phase ,
showed a slight negative trend over time. Shown with 95% confidence limits. .............. 24
Figure 2-2. Trends in Atlantic Puffin (Fratercula arctica) fledging condition on Machias
Seal Island, New Brunswck from 1980-2013. a. Declining trends in mass (blue
diamonds) and wing chord (red squares) over time. Shown with 95% confidence
intervals, linear formula and significance value; b. Declining trend in condition index
over time. Shown with 95% confidence limits. ................................................................ 25
Figure 2-3. Different food types (g per hour) fed to Atlantic Puffin (Fratercula arctica)
chicks on Machias Seal Island, New Brunswick from 1995 to 2014. Prey categorised as
"good" food (metamorphosed fish) and "poor" food (non-fish) from 1995 to 2014.
Butterfish have been included in "poor" food because its shape makes it difficult or
impossible for puffin chicks to ingest. .............................................................................. 26
ix
Figure 2-4. Breeding success of Atlantic Puffins (Fratercula arctica) on Machias Seal
Island, New Brunswick from 1995 to 2014. Breeding success ranged from 89% in 2012
to 25% in 2013. Shown with 95% confidence limits. ....................................................... 27
Figure 2-5. Trends in Atlantic Puffin (Fratercula arctica) fledge dates on Machias Seal
Island, New Brunswck from 1980-2013. a. Increasing trend in fledge date over time; b.
No trend in the difference in fledge date from the mean within a year over time. Shown
with 95% confidence intervals. ......................................................................................... 28
Figure 2-6. Correlation matrix of parameters. Pearson correlation coefficients calculated
for the relationships between each parameter. Values above 0.50 indicate strong
collinearity. ....................................................................................................................... 29
Figure 3-1. The cumulative percentage of puffins tagged as chicks that recruited per year
from ages 5 - 7 for each year cohort form 1997 to 2006. Recruitment by cohort ranges
from 0.12 of the 2004 cohort, to 0.46 of the 2001 cohort. Lines are terminated at age 7 so
that early cohorts (which have had more time to recruit) are not over-represented; overall,
96% of birds that ever recruit, have done so by age 7 (Figure 2). ................................... 48
Figure 3-2. Resightings of banded puffins showing the proportion of all resighted birds
that were seen at a given age. Most (96%) puffins that are resighted are seen by age 7. 49
Figure 3-3. Relationships between the probability of natal recruitment to Machias Seal
Island and variables present in top models from Table 3-3 (ΔAICc < 4). Shading
represents 95% confidence limits. "Rugs" on the top and bottom of each graph represent
the actual data where point on the top rug are observations with positive residuals and
those on the bottom are observations with negative residuals. Jitter was applied to
continuous variables (i,e, all but IV) to better visualise density of the data points. ......... 51
x
Figure 3-4. Model averaged effect sizes (95% confidence intervals) of predictors for natal
recruitment models. Parameter from models in Table 3-3 with ΔAICc < 4 were used for
averaging based on AICc weights..................................................................................... 52
Figure 3-5. Relationships between the probability of survival and variables present in top
models from Table 3-5 (ΔAICc < 4). Shading represents 95% confidence limits. "Rugs"
on the top and bottom of each graph represent the actual data where point on the top rug
are observations with positive residuals and those on the bottom are observations with
negative residuals. Jitter was applied to better visualise density of the data points. ........ 53
Figure 3-6. Model averaged effect sizes (95% confidence intervals) of predictors for
survival models. Parameter from models in Table 3-5 with ΔAICc < 4 were used for
averaging based on AICc weights..................................................................................... 54
Figure 3-7. Correlation matrix for predictors. Correlation coefficients for predictors
included in the model set. Size and colour of circles indicates strength of the relationship
and corresponds to the numbers on the bottom left portion of the figure. ........................ 50
1
Chapter 1 : Introduction
Background
Seabirds are widely acknowledged as potential indicators of the health of the
marine environment (Cairns 1987; Bost and Le Maho 1993; Barrett 2002; Diamond and
Devlin 2003; Boyd et al. 2006; Frederiksen et al. 2007; Iverson et al. 2007; Parsons et
al. 2008; Einoder 2009) depending entirely on the ocean for food and coming to land
only to raise their offspring. They are abundant throughout the globe in diverse marine
environments. Despite their abundance, Paleczny et al. (2015) report that seabirds have
declined by nearly 70% from 1950 to 2010. Croxall et al. 2012 report that seabirds as a
group are declining faster than any other group of birds and are under threat from human-
induced changes in their environment due mainly to the negative impacts of commercial
fisheries and pollution. Several recent studies have also shown that seabirds are affected
by climate (Springer et al. 2007, Frederiksen et. al 2013). In the North Atlantic in
particular, Frederiksen et al. (2013) report a strong positive correlation between the
presence of suitable habitat for a low trophic-level copepod (Calanus finmarchicus) and
long-term breeding success of several seabird species feeding at a higher trophic level.
This suggests a bottom-up effect on seabirds of a changing marine climate.
The life history characteristics of seabirds make the study of them particularly
useful as the proverbial "canary in the coal mine" of changing marine ecosystems.
Consistent with a K-selected strategy, seabirds are long-lived, show delayed reproductive
maturity, and have low annual fecundity (Weimerskirch 2002). However, there is
considerable variation within this framework (Russell 1999). Life expectancy ranges
2
from roughly five years to 35. Young seabirds typically delay breeding for between two
and five years and when they breed lay one to two eggs annually, or less often
(Weimerskirch 2002). These characteristics combined with the fact that seabirds spend
the vast majority of their time at sea, where they are very difficult to monitor can make
them a challenge to study, require data collection over long time scales with large sample
sizes. Such studies are few and far between due to these logistical, financial and
dedicative requirements.
Variation in population size is of particular interest in conservation ecology
because population change is crucial for successful management and conservation,
particularly when combined with knowledge of the effects and effect sizes affecting these
demographic processes (Croxall and Rothery 1990; Skalski et al. 2005). While, in
general, changes in adult survival have the greatest impact on population growth rate in
long-lived species such as seabirds (Russell 1999; Saether and Bakke 2000; Cooch et al.
2001), adult survival tends to be insensitive to environmental change relative to
parameters such as reproduction, juvenile survival, and natal recruitment (Lebreton and
Clobert 1991). Natal recruitment is defined as the return of an individual to its place of
birth to breed. If we understand how each of the latter population parameters contributes
to population change, and the major biological and environmental effects modulating
these parameters, we will be in a better position to isolate the principal factor(s)
responsible for population changes (Perrins 1991; Russell 1999).
Hanski and Gilpin (1997) define a metapopulation as a set of local populations in
an area of suitable habitat that are separated from other occupied patches by
fragmentation, or patches of unsuitable habitat. Local populations within a
3
metapopulation are connected through infrequent dispersal, and are subject to extinction
and reoccupation. During the breeding season seabird populations generally function on a
metapopulation scale, existing in patchy environments often consisting of discrete island
populations, or breeding colonies, connected through natal and/or breeding dispersal.
Although survival probabilities and reproductive parameters have been well studied in
some metapopulations, natal dispersal and recruitment represent a major gap in our
knowledge of metapopulation dynamics in seabirds, and are important parameters for
population change.
Researchers in the Atlantic Laboratory for Avian Research (ALAR) at the
University of New Brunswick have been monitoring seabird populations on Machias Seal
Island (MSI), a migratory bird sanctuary, since 1995. MSI is home to the largest breeding
population of Atlantic puffins (Fratercula arctica, hereafter ‘puffin’) south of
Newfoundland; over 6000 pairs reproduce on the island annually (Diamond, unpublished
data). The next largest breeding population in the Gulf of Maine is the restored colony on
Seal Island National Wildlife Refuge, Penobscot Bay (Kress and Jackson 2015); in recent
years, approximately 500 pairs have nested on the island (Harris and Wanless 2011).
The Machias Seal Island puffin population has remained relatively stable since
2000 but changes seem likely because the ecosystem is changing (Croxall et al. 2012).
The fragility of the system was noted when the common and Arctic tern (Sterna hirundo
and S. paradisaea, respectively) breeding colonies on MSI experienced a complete
collapse, even though they had appeared to be the most stable seabird populations in the
area (Diamond 2009, Gaston, et al. 2009). The Arctic tern colony in particular was the
largest in North America (Devlin 2006).
4
Given the recent decline and complete breeding failure of Arctic and common
terns, and our lack of understanding of the cause or causes, it seems reasonable to
conclude that the largest population of puffins in the Gulf of Maine, and the most
important for recruitment to other colonies in the metapopulation (Breton et al. 2005),
may also be susceptible to the same forces responsible for the breeding failure and
decline of the MSI terns (Diamond and Devlin 2003). Because adult survival is high and
stable for puffins in the Gulf of Maine (Breton et al. 2005) and recruitment has been
identified for this metapopulation as the likely factor responsible for differing colony
sizes (Breton 2005), it is important to consider recruitment as a key factor affecting
population dynamics (Harris et al. 2005).
Hypotheses and Predictions
Juveniles breeding for the first time must decide where to do so based on
information gathered prior to breeding. Some information about their natal colony is
undoubtedly obtained during their hatch year. In a review of research on recruitment in
seabirds, Becker and Bradley (2007) report that fledging mass and fledging date affected
recruitment probability in common terns (Sterna hirundo), Cory's shearwater
(Calonectris diomedea), and sooty terns (Sterna fuscata). Although several studies have
reported no effect of conditions on the breeding grounds on recruitment probability in
auks (Lloyd 1979, Hedgren 1981, Harris et al. 1992), these studies were performed on
species that continue to provision and care for their young well after fledging, such as
common murres and razorbills. Any effect that conditions at the natal colony have on
5
recruitment would be more easily detected if not confounded by the extension of those
effects beyond the colony.
Hypothesis 1: Juveniles are more likely to return to MSI to breed if they left the
island as fledgers in good condition.
Chick growth and ultimately size at fledging are affected by a number of factors
including food availability (Baillie and Jones 2004), food quality (Harris and Hislop
1978), and adult foraging ranges (Anker-Nilssen and Lorentsen 1990). In this way, chick
size at fledging reflects the foraging conditions at the time of breeding. Perrins et al.
(1973) showed that mass at fledging affected survival of immature manx shearwaters
(Puffinus puffinus). The same was not found to be true for puffins on the Isle of May
(Harris and Rothery 1985), however this study was relatively short (i.e., 5 years) and
sample sizes were small (fewer than 100 chicks banded per year). In addition, conditions
were near peak for the Isle of May colony at the time of the study, so selection for
dispersal was probably weak.
During chick growth, energy is not distributed equally in the body. Oyan and
Anker-Nilssen (1996) showed that growth of body mass in puffins was most affected by
quantity of food, because energy was preferentially assigned to maintain growth of the
wings and skull. In order to account for this discrepancy, I used an index of body
condition (mass divided by wing chord; Lewis et al. 2006, Cohen et al. 2014) that
represents condition of a chick at fledging better than mass or wing chord alone. If the
value is high (chick has a large mass in relation to its wing chord), then it is likely that
foraging conditions are good, while a low value (chick mass is low in relation to wing
6
chord) indicates that development of the wing has taken priority over body tissue
development and foraging conditions are likely not as good.
Prediction 1: Index of body condition at fledge is positively related to probability of natal
recruitment.
Hypothesis 2: Juveniles that fledge MSI earlier are more likely to return to MSI to
breed.
Presumably, puffins time their arrival at the breeding colony to correspond with a
reliable source of food for themselves and their young (Lack 1968). Changes in the
marine environment can affect timing of breeding in puffins (Durant et al. 2004, Stenseth
et al. 2002). Wanless et al. (2009) showed that over a 33-year period (1970s to 2000s)
puffins at the Isle of May, Scotland, have begun breeding later and later in the season,
due in part to changes in the ocean climate. In Norway, Durant et al. (2006) found that
the nestling period was affected by sea surface temperature and salinity in March, a
reflection of the feeding conditions later in the season when adults are feeding young.
When conditions were favourable (i.e., sea surface temperature and salinity were
favourable for development of prey) the nestling period for puffins was shorter and
chicks fledged earlier. Additionally, Ydenberg et al. (1995) showed that rhinoceros auklet
(Cerorhinca monocerata, a close relative of puffins) chicks that hatched later fledged
smaller. Overall, seabirds are thought to be more successful the earlier they nest, though
whether this results in better recruitment needs more investigation.
7
Prediction 2: Fledge date and the probability of natal recruitment will covary inversely.
Hypothesis 3: Natal recruitment of young Atlantic puffins to the breeding colony on
MSI is affected by the availability of high quality food during the chick growth
period
Yearling Atlantic herring (Clupea harengus) have historically been the primary
food for puffins on MSI (Breton and Diamond 2014). However, in recent years
availability of herring on MSI has been highly variable ranging from 80% to nearly 0%
of chick diet (A. W. Diamond, unpublished data). In years when herring were presumably
unavailable, adults brought alternative food such as hake (Urophycis spp.), sandlance
(Ammodytes sp.), butterfish (Peprilus triacanthus), larval fish, and euphausiids (krill) to
their chicks. While sandlance, herring and hake are considered relatively nutritious food,
butterfish, although also high in fat, are difficult for chicks to swallow due to their shape
(S. M. Kress, personal comm.), and larval fish and krill are low in calories (A.W.
Diamond, unpublished data).
Prediction 3: Amount of quality food in the diet of chicks is positively related to
probability of natal recruitment.
8
Hypothesis 4: Natal recruitment of young Atlantic puffins to the breeding colony on
MSI is affected by the average breeding success on MSI during their hatch-year.
Young puffins may use information obtained during their hatch-year about the
overall success of their breeding colony when deciding where to settle and breed
(Danchin et al. 1998).
Prediction 4: Mean breeding success in an individual's hatch-year is positively related to
probability of natal recruitment.
To assess these non-exclusive hypotheses, I used a model selection framework in
which the global model was an additive model, and all other models were versions of the
global model with one or more predictors excluded. Biologically relevant interactions
were included. Additionally, in order to provide an alternative to the hypothesis that
conditions during the hatch year drive natal recruitment in puffins, I included in my
analysis a parameter (number of islands visited) relating not to the hatch year, but to the
pre-breeding or "prospecting" period.
General Methods
Species Account
Species account taken, unless stated otherwise, from Lowther et al. 2002.
The puffin is a member of the Alcid family of seabirds (Order Charadriiformes).
They are stout birds: 310-550 g, 26-29 cm in length as adults and showing the typical
contrasting black upper and white lower colouration of diving seabirds. They have a
large, brightly coloured bill that grows larger and more colourful with age until breeding
9
age is reached at 5 years old (Harris and Wanless 2011). Sexes are visually identical,
although males may be slightly larger and heavier than females (Friars and Diamond
2011, Harris and Wanless 2011). Puffins can live upwards of 30 years. They live and
breed throughout the North Atlantic, spending the summer breeding season on small
offshore islands from 43o - 80
o north latitude and wintering at sea.
During the summer breeding season, mated pairs occupy an earthen or rock
burrow, often returning to the same burrow year after year (Hudson 1985). Females lay a
single egg and both parents incubate the egg for 36-45 days. Chicks are born semi-
precocial (mobile and downy, but remaining in the burrow to be fed) and are fed by both
parents until fledging (leaving the nest) at 38-44 days old. Chicks fledge at night and go
directly to the sea; at this point they are completely independent of their parents. Most
young remain at sea for their first year, returning to breeding colonies (but not breeding)
in their 2nd
, 3rd
, and 4th
years. Immature individuals may visit several breeding colonies
before settling in their 5th
year (Harris 1983, Breton et al. 2005). Emigration after this
time is rare (Harris and Wanless 1991; Breton et al. 2005).
Atlantic puffin global population estimates range from 3.8 – 8.212-14 million.
The vast majority breed in the eastern North Atlantic, specifically in Iceland, Great
Britain, and Norway (Harris and Wanless 2011). Most puffins in the western Atlantic
breed off the coast of Newfoundland and Labrador. However, there is a growing
metapopulation of puffins on managed islands in the Bay of Fundy/Gulf of Maine
(BOF/GOM) region, where Machias Seal Island, the source of data for this thesis, is
located. This region represents the southern-most population of puffins in the world.
10
Study Area
Individuals were captured and biological covariates recorded on MSI in the Bay
of Fundy. MSI (44o30' N, 67
o06' W; Fig. 3) is managed by the Canadian Wildlife Service
(CWS) as a Migratory Bird Sanctuary, and monitored cooperatively by CWS and ALAR.
It is located 19 km SE of Cutler, Maine and 19 km SW of Grand Manan, New
Brunswick, at the convergence of the Gulf of Maine and the Bay of Fundy. The small,
treeless, 9.5 ha island boasts a breeding population of puffins over 6000 pairs (Diamond,
unpublished data) - far fewer than the millions found in the colonies of Newfoundland
and the eastern Atlantic (Harris and Wanless 2011), but a significant number for the
region. MSI is composed mostly of exposed bedrock and boulders with grasses and other
non-woody plants covering the interior of the island.
Other species breeding on the island are razorbill (Alca torda), common murre
(Uria aalge), Leach's storm petrel (Oceanodroma leucorhoa), spotted sandpiper (Actitis
macularius), common eider (Somateria mollissima), savannah sparrow (Passerculus
sandwichensis), great black-backed gull (Larus marinus), and herring gull (L.
argentatus). Prior to 2006, a large breeding population of common and Arctic terns
(Sterna hirundo and S. paradisaea, respectively) was also present on the island.
In a study of recruitment using only data from Machias Seal Island, recruitment of
MSI-born puffins to other colonies, a process known as natal dispersal, would be
problematic as I would be unable to distinguish between death and dispersal. Fortunately,
five of the eight breeding colonies in the Gulf of Maine puffin metapopulation, and the
vast majority (~98%, Harris and Wanless 2011) of the breeding population, are
monitored by other researchers throughout each breeding season, providing data for
11
estimating natal dispersal. Specifically, resighting of banded puffins was conducted each
year not only on MSI, but also on Petit Manan, Seal Island, Matinicus Rock, and Eastern
Egg Rock (Figure 1-1). Resighting protocols are similar across all islands.
Objective and Chapter Summary
The main objective of my study was to determine whether factors experienced by
puffins during their hatch year on Machias Seal Island affect natal recruitment. In order
to determine which parameters to include in the analysis (detailed in chapter three), I first
explored trends and inter-annual variability in hatch-year conditions over a considerable
time (up to 33 years depending on the parameter) for parameters of interest in chapter
two. In chapter three I modeled recruitment as a function of several hatch-year and one
sub-adult parameter using a generalised linear model approach with model selection
based on AIC. The results provide valuable information about the juvenile phase of
puffin life-history in the Gulf of Maine/Bay of Fundy metapopulation, as reviewed in
Chapter 4.
12
Figure 1-1. Map showing locations of puffin breeding islands in the Bay of
Fundy/Gulf of Maine metapopulation. Islands in the inset represent monitored islands,
while Bird Island (BI), Pearl Island (PI), and the Tusket and Mud Island groups (TMG)
represent puffin breeding islands that are not monitored for banded birds. The size of
each dot roughly represents the relative sizes of the breeding populations on each island.
13
Chapter 2 : Trends in Chick Development on Machias Seal Island
between 1980 and 2014
Introduction
Systematic monitoring of seabirds breeding on MSI since the 1980's provides rare
insight into changes in populations over a considerable time-scale. Many questions about
the mechanisms driving observed changes arise from a simple view of population
characteristics over time. For puffins in particular, MSI is by far the largest breeding
colony in the Bay of Fundy/Gulf of Maine metapopulation, playing host to ~6000
breeding pairs of puffins during the summer season (A.W. Diamond, personal
communication). In comparison, (Matinicus) Seal Island (the next largest colony in the
metapopulation) hosts only an estimated 600 breeding pairs (Harris and Wanless 2011).
Thus, the importance of changes in the population ecology of MSI puffins, as a source of
recruits for the remaining islands in the metapopulation (MacArthur and Wilson 1967,
Clobert et al. 2004), cannot be underestimated.
Here I describe changes over 20 years in population parameters related to chick
development (hatch date, growth rate of chicks, mass at fledge, wing chord at fledge,
fledge date, fledge date relative to others within a year, fledge date relative to others
among years, breeding success) and some post-fledge parameters (number of islands
within the metapopulation visited before recruitment age, and recruitment/dispersal). In
particular, I am interested in detecting meaningful declining trends, if any, in these
parameters over time in order to determine whether puffins on MSI may be headed
towards a similar fate to the MSI terns. As the largest (by far) breeding colony in the
14
metapopulation, and the source for the majority of dispersed birds in the region, failure of
the MSI colony could be catastrophic for the metapopulation as a whole. In particular,
MSI is important as a source of "rescue" numbers to other colonies in the region that may
at times be unable to maintain their own numbers.
Methods
Field Protocol
Growth Rates and Breeding Success
To calculate daily growth rates for puffin chicks we monitored active burrows and
measured mass and wing chord twice or thrice during the linear growth period (age 10-30
days old). We determined breeding success for the same monitored burrows (n=26-84
burrows, depending on year) as the number of young fledged per egg laid. Burrows were
not included in growth or breeding success calculations if the fate of the egg or chick was
unknown. Chicks reaching 35d and not subsequently found dead were treated as fledged.
Food
The type and amount of prey fed to chicks was determined by conducting 3-hr
“feeding watches” throughout the chick-rearing period. Items carried in from the sea by
adults were identified by researchers concealed by blinds. Amount (by mass) was
determined based on the length of each item in relation to the length of the bill,
multiplied by mass per length as determined by length/mass regressions from collected
samples.
15
Food items were categorised as "good" or "poor" based on fat content of collected
samples. With one exception (Atlantic butterfish, Peprilus triacanthus), samples could be
easily categorised as good or poor based on their fat content. Good food included
metamorphosed, or non-larval (usually one year old or 1-group), fish species (herring,
hake, and sandlance), while poor or "junk" food included invertebrates and larval fish, as
their fat content is low in comparison to other food items delivered. I made an exception
to this categorisation in the case of butterfish (sp.), which have a high fat content, but
their laterally flattened shape makes them difficult or impossible for puffin chicks to
swallow (S. M. Kress, pers. comm.), I included them in the poor category.
Fledge Parameters and Resighting
Approximately 200-400 puffin chicks per year (approximately 5-10% of fledgers
based on an average breeding success of 0.60 chicks fledged per egg laid, and a
population of 6,000 breeding pairs) were captured at night on MSI as they were fledging
from their nest burrow. Chicks are easy to capture in this way because they are (for as yet
unknown reasons) attracted to the lighthouse at the centre of the island, and as such can
be easily seen and captured in numbers on the surrounding lawn. For each individual, we
recorded the date of capture (fledge date), mass (g), and wing chord (mm) and fitted two
uniquely engraved metal bands - one a USGS band (Bird Banding Laboratory, BBL), the
other a metal field-readable band (MFR). Numbers on these bands were used to identify
individuals in subsequent years.
16
I calculated an index of fledging body condition (CI) as fledge mass(g) / fledge
wing chord (mm). Chicks with a relatively low CI were light with long wings at fledging,
whereas those with relatively high CI were heavier with shorter wings at fledging.
Although we have data spanning 1980 to 2014, we do not have all data for all
years. No data were recorded in 1981-1983. In addition, data for food fed to chicks were
available only for chicks born between 1995 and 2014, as these data were not collected
prior to 1995. Parameters of interest are presented in Table 2-1 with their corresponding
sample sizes.
Statistical Analyses
I performed simple linear regressions for continuous response variables and
logistic regressions for binomial response variables modeled as a function of year in R
version 3.1.1 (R Core Team, 2014) to examine whether fledge parameters declined over
time. In an effort to include all data for all years, separate regressions were performed
rather than a multiple regression where some data would be lost. For simplicity, linear
relationships were assumed for all regressions except breeding success (which is a
binomial parameter) in which a logistic relationship was used. The assumption of
linearity was tested by examining residuals versus predicted values plots in which the
points should be symmetrically distributed around a horizontal line. I created a
correlation matrix with corresponding correlation coefficients (values above 0.50 signify
strong collinearity) to assess the possibility of correlation between predictors. Results of
the correlation matrix are presented with correlation coefficients and regressions are
presented with their slope, correlation coefficient (or 1-residual deviance/null deviance;
17
an analogous estimate to r2 for logistic regression) and significance value (H0: slope = 0;
alpha = 0.05).
Because field protocols differed before and after 1995, I looked at trends from
both 1980 on and 1995 on to determine if there would be a difference in results. In
particular, I was concerned that earlier dates of departure of researchers from MSI prior
to 1995 might skew the data toward earlier fledge dates and lower fledge mass and wing
chord values. However, neither r2
values nor significance (and therefore interpretation)
differed between the two datasets and so the full dataset from 1980 and after is presented
here.
Results
Chick Development
There was no significant trend in chick mass growth rates (mass; Figure 2-1a.;
slope = 0.00, r2 = 0.00, p=0.94), while rate of growth in the length of the wing (wing
chord) declined slightly from 1997 to 2014 (Figure 2-1b.; slope = -0.03, r2=0.03,
p<0.001). Both mass at fledge and wing chord at fledge declined slightly over time
(Figure 2-2a.; slope = -1.11, r2=0.07, p<0.001, Figure 2-2a.; slope = -0.24, r
2=0.08,
p<0.001 respectively). Furthermore, when we combined the two parameters as a single
index of fledging condition (mass/wing chord), the same slight negative trend over time
was seen (Figure 2-2b.; slope = -0.005, r2=0.02, p<0.001). However, more interesting is
the high degree of variation from year to year, particularly in wing chord (from a high of
152mm in 1984 to a low of 129mm in 2013) and condition index (high of 2.23 in 2007
and low of 1.73 in 2013).
18
Food
Prior to 2004, the majority of food delivered to puffin chicks on MSI could
consistently be categorised as good quality based on high fat content. This proportion (by
mass calculated using length estimates and known mass per length values) varies from
59% in 2004 to nearly 100% in many other years. However, there has been a significant
decline in the amount (in grams per hour) of good food, and in fact total amount of food,
fed to MSI puffin chicks (Figure 2-3; slope = -2.68, r2
= 0.04, p < 0.001). This decline in
good food was not accompanied by an increase in poor ("junk") food fed to chicks
(Figure 2-3; slope = 0.07, r2=0.00, p=0.31). Notable however, is that very little junk food
(<4% of total food) was detected prior to 2002, while values in subsequent years were as
high as 41% of total food.
Breeding Success
During 1995 - 2014 breeding success, a combined measure of hatching success
and fledging success, measured as the number of chicks fledged per egg laid, ranged from
an overall high of 0.89 in 2012 to an overall low of 0.25 the very next year, 2013. Overall
there was a very slight but statistically significant positive trend over time (Figure 2-42-4;
slope = 0.005, p=0.05).
Timing of Fledge
During 1981 - 2013 chick fledging dates ranged from 6 July to 24 August, with
the mean fledge date occurring nearly a month later in 2013 (16 August; 15 Aug - 17
Aug, 95% CI) than in 1981 (26 July; 23 Jul - 29 Jul, 95% CI). Chicks are fledging later
over time (Figure 2-5; slope = 0.40, r2
= 0.15, p < 0.001).
19
Results of the correlation matrix are shown in Figure 2-6. Condition index and
fledge date had a Pearson correlation coefficient above 0.50 (r = 0.60). All other values
were 0.50 or less.
Discussion
Parameters relating to fledging conditions for puffin chicks on MSI show a high
degree of variation from year to year with a generally, but weakly, declining trend over
time. In a study on tufted puffins (Fratercula cirrhata) breeding on Triangle Island, B.C.
(another species for which independence begins upon fledging) Morrison et al. (2009)
found that survival of juveniles depended strongly on wing length at fledge and weakly
on mass and date at fledge. The same may be true for puffins on this side of the continent
where negative trends in several fledgling characteristics (fledging mass, wing chord,
fledge date, quality food) may be concerning. After nine years, researchers are still trying
to understand the reasons for the failure of the tern colonies on MSI. No signs of decline
were detected prior to that failure, and such changes may be more subtle than one would
expect given the consequences.
Food availability during the summer breeding season is affected by oceanic
conditions in the spring when fish species rely on the phytoplankton bloom for
development. Durant et al. (2006) reported that high salinity and low sea surface
temperatures in March in the Barents Sea resulted in a longer fledgling period for puffins
several months later. Major changes in food brought back to the colony by adults and fed
to several species of seabird chicks on MSI and other breeding islands in the western
Atlantic have been documented in recent years (Gaston et al. 2009, Breton & Diamond
20
2014, A.W. Diamond unpublished data). Similar changes have been documented across
the breeding range of puffins (Martin 1989, Barrett 2002, Baillie and Jones 2003, 2004,
Regher and Rodway 1999). Prior to 2001, <1 year old Atlantic herring (Clupea
harengus), a high-fat and energy-rich fish, were the major food source for most species
on MSI. Since that time, sandlance (Ammodytes sp.), hake (Merluccius sp.), larval fish,
and euphausiid shrimp (Meganyctiphanes norvegicus) have become more common in the
diet than herring. Abandonment of the tern colonies on MSI coincided with a switch from
herring as the dominant food source to larval fish and euphausiids, followed by declining
reproductive success, an increase in predation by gulls and finally complete abandonment
in the span of 2 years (Gaston et al. 2009, Diamond 2009). We are seeing the same shift
away from herring as the main prey fed to puffin chicks (Breton and Diamond 2014). In
addition, adults do not seem to be compensating for lack of quality by increasing their
rate or quantity of feeding. In fact, the opposite trend appears in the data with a decline in
the total amount of food brought in. On a positive note however, we do not see the same
decline in breeding success for puffins as preceded the tern collapse. In addition, Baillie
and Jones (2003) reported that puffins breeding on islands in Newfoundland and
Labrador with low abundance of their main food source (capelin Mallotus villosus) were
able to maintain breeding parameters similar to those experienced by breeders on islands
where Capelin were abundant by switching to other prey species. They concluded that
puffins in that region were resilient to large-scale changes in prey abundance.
In colonial seabirds, reproductive synchrony (how temporally closely
reproduction occurs between individuals) is crucial to reproductive success (Wittenberger
and Hunt 1985). In particular, timing of fledging is important where chicks that fledge at
21
an optimal time will have access to abundant seasonal food resources and better oceanic
weather conditions. If chicks fledge sooner or later than usual, match-up with beneficial
oceanographic conditions may not be optimal. In good years, this is likely a bigger
problem for chicks traveling with adults than for species independent at fledging (such as
puffins) who are capable of traveling greater distances for food. But in poor years when
fledgers may leave the colony in less than optimal condition, survival, challenging for
juveniles during the best of times, could be compromised.
Morrison et al. (2009) report that survival of juvenile tufted puffins was related to
wing length at fledge and to a lesser degree mass and date at fledge. A delay of nearly
one month, as has been documented here for puffins on MSI, could drastically affect their
ability to find food and survive at sea, especially in years when chicks may be fledging in
poor condition. Wanless et al. (2009) report a similar trend towards later fledge dates of
puffins breeding in the North Sea. Barrett and Rikardsen (1992) and Diamond and
Nettleship (1989) showed that later fledging dates in the Barents Sea and Northwestern
Atlantic respectively were a result of prolonged food stress.
In general, although some puffin pre-fledge characteristics on MSI remain stable
(mass growth rate, breeding success, synchrony within years), others show negative
trends (feeding and body condition) and may warrant concern. In particular changes in
food and apparent inability of adults to compensate by increasing feeding rate may be
causing puffins to fledge later and in poorer condition than is optimal. If this were the
case, we would expect to see an accompanying decline in growth rates, contrary to my
results. We may have missed seeing these effects by measuring growth rate only during
the linear growth phase (between 10 and 30 days post hatch); much growth happens after
22
this age, and in particular our method would not detect whether or not chick mass reaches
a peak and then declines before fledging, as well-fed puffin chicks often do (A.W.
Diamond, pers. comm.). If these trends continue, and a decline in breeding success
develops, puffins on MSI may be heading for the same fate as their tern neighbours.
Although the overall trends in parameters relating to pre-fledging and fledging
conditions are significant, the time trend accounts for very little of the variation in these
parameters (r2 = 2 -8%). Valuable insight would undoubtedly be gained by determining
what other forces contribute to this high degree of variation. Possibilities for further
research include the effects of oceanic conditions prior to the breeding season, and
parental quality.
23
Table 2-1. Yearly sample sizes for parameters spanning from 1980 to 2014.
Year
Mass at Fledge,
Wing Chord at
Fledge, Fledge
Date
Mass Growth
Rate, Wing Chord
Growth Rate
Breeding
Success
Feeding
Watches
1980 71
1981 25
1982 57
1983 68
1984 69
1985
1986
1987
1988 117
1989 134
1990 192
1991 152
1992 102
1993 17
1994 124
1995 267 26 25
1996 274 54 21
1997 264 16 64 10
1998 264 17 54 11
1999 267 18 84 14
2000 202 23 69 27
2001 150 18 75 22
2002 490 36 81 23
2003 394 40 74 21
2004 344 43 59 28
2005 158 34 49 49
2006 235 21 72 62
2007 255 36 71 28
2008 305 62 74 37
2009 86 64 82 40
2010 136 41 57 44
2011 169 53 63 52
2012 202 33 53 47
2013 103 39 68 53
2014 41 50 42
24
Figure 2-1. Trends in Atlantic puffin (Fratercula arctica) chick growth rates on
Machias Seal Island New Brunswick from 1997 to 2014. a. Mass growth rate (g/day)
measured during the linear growth phase (10-30 days old), showed no significant trend
over time. MGR = -0.005*YR +16.34. p = 0.94, R2 = -0.002. b. Wing Chord growth rate
(mm/day), measured during the linear growth phase, showed a slight negative trend over
time. WGR = -0.03*YR + 59.79. p < 0.001, R2 = 0.03. Data shown as means with 95%
confidence limits.
25
Figure 2-2. Trends in Atlantic puffin (Fratercula arctica) fledging condition on
Machias Seal Island, New Brunswick from 1980-2013. a. Declining trend in mass
(blue diamonds; M = -1.11*YR + 2515.92, p < 0.001, R2 = 0.07) and wing chord (red
squares; W = -0.244*YR + 630.91, p < 0.001, R2 = 0.08) over time; b. declining trend in
condition index (mass/wing chord) over time. CI = -0.005*YR + 11.42, p < 0.001, R2 =
0.02. Shown with 95% confidence intervals.
26
Figure 2-3. Food fed to Atlantic puffin (Fratercula arctica) chicks on Machias Seal
Island, New Brunswick from 1995 to 2014. Prey categorised as "good" food
(metamorphosed fish; GF = -7.03*YR + 14233.20, p < 0.001, R2 = 0.04) and "poor" food
(non-fish; PF = 0.21*YR - 29.18, p = 0.31, R2 = 0.000) from 1995 to 2014. Butterfish
have been included in "poor" food because its shape makes it difficult or impossible for
puffin chicks to ingest.
27
Figure 2-4. Breeding success of Atlantic puffins (Fratercula arctica) on Machias Seal
Island, New Brunswick from 1995 to 2014. Significant positive trend in breeding
success over time. Breeding success ranged from 89% in 2012 to 25% in 2013. Data
shown as means with 95% confidence limits (Wilson Score Interval). BS = 0.005*YR -
9.30, p = 0.04, R2 = 0.002.
28
Figure 2-5. Trends in Atlantic puffin (Fratercula arctica) fledge dates on Machias
Seal Island, New Brunswick from 1980-2013. Increasing trend in fledge date over
time; shown with 95% confidence intervals. FD = 0.40*YR - 698.53, p < 0.001, R2 =
0.13.
29
Figure 2-6. Correlation matrix of parameters. Pearson correlation coefficients
calculated for relationships between parameters (FD = fledge date (days since 1 May), CI
= condition index (mass/wing chord), BS = breeding success (chicks fledged per egg),
MGR = mass growth (grams per day), WGR = wing chord growth (mm per day), GF =
good food (grams per hour), PF = poor food (grams per hour)). Values above 0.50
indicate strong collinearity. Size of the circle and colour indicate degree of correlation.
30
Chapter 3 : Effects of hatch year conditions on natal recruitment in
Atlantic puffins breeding on Machias Seal Island
Introduction
Recruitment of individuals into the breeding population is a complex process in
seabirds. For these long-lived species where reproductive maturity is often delayed this
process can occur over several years. Danchin et al. (1991) described recruitment in long-
lived birds as a 3-stage process whereby young that have fledged the nest must 1) survive
to recruitment age; 2) choose a breeding location; and 3) become a breeder. It is often
difficult or impossible to tease apart these three stages, as most studies focus on a single
breeding colony and therefore cannot distinguish between natal dispersal and mortality.
Through a partnership between ALAR and several other organizations monitoring
seabirds in the Gulf of Maine, resighting data are available from all relevant breeding
colonies in the metapopulation (islands with more than a few pairs of breeding puffins).
With these data, it is possible to target stage 2 only in order to determine factors
influencing whether an immature puffin returns to the same colony (natal recruitment), or
settles in a different colony (natal dispersal) for breeding.
Recruitment generally involves dispersal from the natal site to varying extents
(Duncan and Monaghan 1977). Relatively few studies (e.g., Pradel et al. 1997, Crespin et
al. 2006, Votier et al. 2008, Aubry et al. 2009) have attempted to determine what factors,
be they environmental, ecological, behavioural, or physiological, may be driving natal
recruitment in seabirds. The paucity of studies is due, in part, to the difficulty of
obtaining enough data over sufficient time to measure the process. In chapter one I
31
outlined the importance of natal recruitment in understanding seabird population
dynamics, especially when adult survival is stable and thus relatively insensitive to
changes in environmental conditions (Lebreton and Pradel 2002). In populations
conforming to a metapopulation framework the processes of recruitment and dispersal
among patches can lead to local extinctions and "rescue effects" (in which dispersal to
small patches can "rescue" that patch from extinction; Hanski & Gilpin 1997). If we
understand the major biological and environmental effects modulating these processes,
we will be in a better position to isolate the principal factor(s) responsible for such
population changes (Perrins 1991, Russell 1999).
Although puffins have been studied extensively, especially in the European parts
of their breeding range, little is known about the process of recruitment. Both Harris et al.
(2005) and Breton et al. (2005) found that adult survival was similar among colonies of
puffins showing marked differences in patterns of population change. Both concluded
that recruitment must be underlying the observed patterns. Of the few estimates of natal
recruitment that have been published (Harris and Wanless 1991; Finney et al. 2003;
Breton et al. 2006a; Sandvik et al. 2008) most report considerable inter-annual variability
(e.g., ranging from 6.4% to 30.1% of the fledging cohort on the Isle of May, Scotland;
(Harris & Wanless 1991; but see Sandvik et al. 2008), but few of these studies attempted
to identify effects responsible for the observed dynamics. Notable exceptions include
Finney et al. (2003) who showed that 21% of variation in puffin recruitment was
explained by nesting density of coexisting gulls, and Breton et al. (2006a) who showed
that colony size affected natal dispersal (and conversely natal recruitment). Data from
MSI also show a substantial inter-annual variability in recruitment (Figure 3-1) Only
32
resightings of birds up to age seven were included here and in subsequent analyses (i.e.
resightings of birds age 8 and older are omitted) because 96% of all birds resighted have
been seen by age seven (Figure 3-2).
Puffins on MSI present an ideal scenario for studying the process of natal
recruitment in seabirds. Each breeding pair raises a single chick, reducing the need for
determining compensatory or confounding effects of siblings. After 38-44 days (Lowther
et al. 2002) the chick, at about two-thirds of adult mass but completely independent of its
parents, moves from the nest burrow to the sea (a process known as "fledging") where it
will remain for 1 to 3 years before returning to land for the first time. Several hundred
chicks are captured and banded each year on MSI as they transition from their burrows to
the sea. Returning birds are identified in subsequent years by their unique band numbers.
Consequently, 73% of the >8,000 puffins banded on MSI since 1995 are of known age.
Ideally, recruitment and survival should be estimated simultaneously because the
process of recruitment is dependent upon survival. However, data for this project were
too sparse for the rigorous and complex statistical analyses required to do so. The process
of survival is unavoidably linked to recruitment, thus I controlled for survival in my
analysis by assigning a binomial response that included only those individuals known to
have survived to recruitment age. In an attempt to tease apart this issue, I analysed a
similar set of models but with return rate as the binomial response. Return rate is the
product of true survival and recapture rate and as such will underestimate true survival.
For convenience, I refer to return rate as survival. I did not include number of islands
visited as a predictor in the survival models because the number of islands visited
33
increases with survival because those that do not survive cannot be observed on
additional islands. Most juvenile mortality occurs in the first two years (Harris and
Wanless 2011) and as such, birds with survival equal to zero (were not detected at
recruitment age) will almost always have number of islands visited equal to one, and
sometimes two (Table 3-1). I would expect that parameters found to be good predictors
for the recruitment models would also be good predictors in the survival models.
However, a parameter with a stronger effect in the survival models may contribute more
to the survival process than to the recruitment process.
Methods
See chapter one for description of study area and species account.
Data
Fledge Parameters and Resighting
Puffin chicks were captured at night as they were fledging. For each individual (n
= 3,142 fledglings from 1997 to 2013), we recorded the date of capture (fledge date,
"FD"; recorded as days since 1 May), mass (g), and wing chord (mm). I calculated an
index of fledging body condition, "CI", as fledge mass (g) / fledge wing chord (mm).
Chicks with a relatively low CI were light with long wings at fledging, whereas those
with relatively high CI were heavier with shorter wings at fledging. Although data exist
over a longer time period for fledge parameters, in order to minimise resighting-related
heterogeneity in the response variable due to varying resighting rates of different band
34
types (Breton et al. 2006b), I chose to use data only from individuals banded with metal
field readable bands (1997-2006).
Bands were resighted each year on all five puffin breeding islands in the
BOF/GOM metapopulation using spotting scopes and binoculars between mid-May and
mid-August by researchers concealed in blinds. Resighting was conducted consistently
throughout the breeding season in order to ensure all age-classes were represented.
An individual was considered "recruited" when it was observed on one of the
study islands during the breeding season between the ages of 5 and 7. Breton et al. 2006a
reported the probability of moving between islands to be around 2% for 5 year old
puffins in the Gulf of Maine, getting less likely with age. Birds observed on MSI were
assigned a "1" and those observed elsewhere were assigned a "0", producing the binomial
response variable used for analyses (natal recruitment vs natal dispersal; Table 3-2).
Individuals that were not observed between ages five and seven were not included in
natal recruitment analyses. This likely provides an underestimate of true recruitment as it
does not account for resighting rates that are less than one, nor does it include birds
recruiting at older ages (but see Figure 3-2). Resighting rates for birds between the ages
of three and five on MSI and on other islands are approximately 0.40 and 0.73,
respectively (20 year average from 1983 to 2003; Breton 2006a). MSI has lower
resighting rates due to the larger number of birds compared with other islands in the
metapopulation. The number of islands visited by an individual from age one to
recruitment age, "IV", was recorded based on resightings from all islands.
Breeding success, "BS", was determined for monitored burrows on MSI as the
mean number of young fledged per egg laid per year. Burrows were not included if the
35
fate of the egg or chick was unknown. Chicks reaching 35d and not subsequently found
dead were treated as fledged.
Food
The type and amount of prey fed to chicks was determined by conducting three-
hour “feeding watches” throughout the chick-rearing period (n = 568 feeding watches
from 1997 to 2013). Three sample plots were monitored each year and items carried in
from the sea by adults were identified by researchers concealed by blinds. Hourly amount
(by mass) for each three-hour stint was determined for common prey types based on the
length of each item in relation to the length of the bill multiplied by mass per length.
Mass per length was determined by length/mass regressions from collected samples
(A.W. Diamond, unpublished data).
As a clear division can be made, food items were categorised as "good" or "poor"
based on fat content of collected samples. My measure of good food, "GF", included
metamorphosed (or non-larval) fish species (herring, hake, and sandlance), while poor
food (or "junk-food", Romano et al. 2006) included invertebrates, butterfish, and larval
fish.
Statistical Analyses
All analyses were conducted in R v. 3.1.2 (R Core Team 2014). Response
variables natal recruitment and survival were recorded as binomial variables (1 =
recruited to MSI, 0 = dispersed, N=898; and 1 = detected at recruitment age, 0 = was not
detected at recruitment age, N=3,143, respectively) and was modeled using a generalised
linear model (family = binomial, link function = logit) approach. I tested fit of the global
36
(fully additive) model with the logit, probit, and cloglog link functions. Fit was best using
the logit link function and so this was applied to all nested models. Response was
modeled as a function of time (year), five parameters relating to conditions during an
individual's hatch year (body condition (unique to individual), fledge date (unique to
individual), mean breeding success, mean proportion of good food delivered by adults; all
continuous variables), and one parameter relating to behaviour between hatch year and
recruitment (number of islands visited; a categorical variable unique to individual; Table
3-1). In addition, three biologically relevant interaction terms were included: condition
index × proportion of × good food delivered, condition index × fledge date, and fledge
date × proportion of × good food delivered.
Ideally, I would have also included several parameters relating to the juvenile
period (the time between fledging and recruitment). However, detailed data were not
available on conditions at colonies other than MSI during the study period.
Parameters were centered and scaled around a mean of zero and standard
deviation of one. Competing models were compared using Akaike's Information Criterion
(AIC). Weighted averages (weighted by AICc weights) were calculated for parameter
estimates using models with ΔAICc less than four, and are reported with their standard
error and significance (H0: slope = 0, alpha = 0.05). I employed the 'shrinkage' method of
calculating model averaged estimates (Burnham and Anderson 2002) in order to reduce
model selection bias. The shrinkage correction provides more conservative variance
estimates by including a zero parameter estimate in models where the parameter being
estimated is not present (as opposed to dropping the model) resulting in more
conservative variance estimates.
37
In order to avoid difficulties in interpreting models, I tested for collinearity of
predictors using a correlation matrix. Correlation coefficients were 0.50 or less for all
predictors indicating no strong collinearity. In addition, changing the order of predictors
in the models had no effect on results.
I assessed model fit by visual inspection of modeled relationships created with R
package visreg (Breheny and Burchett 2013). Visually, I assessed model fit as "good" if
there was no obvious pattern in the residuals, no clear changes in variance or obvious
outliers.
Results
I detected no strong collinearity between predictors (all correlation coefficients ≤
0.50; Figure 3-3). Visual inspection of regression plots on the logit scale showed
reasonable model fit, with some evidence of skewness for good food and number of
islands visited. However, the skewness was mild and logistic regression is robust to
variation in distribution of the predictors.
Of the 28% of banded individuals that were detected at recruitment age, 86%
(84.4% - 88.9%, 95% CI) recruited to MSI, their natal colony. Values ranged from 75%
of the 2001 cohort to 94% of the 1998 cohort (Table 3-3). There was not strong support
for any single model in the analysis, with many models having essentially equal support.
The top supported models (ΔAICc < 4) collectively included all predictors (Table 3-4).
Number of islands visited appeared in all top models.
The number of islands visited had a negative effect on the probability of natal
recruitment (Table 3-5). A juvenile puffin that visits two islands before recruitment age is
38
89% (83% - 92%, 95% CI) less likely to recruit to MSI than one that visited only one
island. The relationship between recruitment and number of islands visited at the five
island level (n = 2) was not significant with confidence intervals incorporating the
possibility of no effect (Figure 3-4, Figure 3-5). However, removing these data points did
not change the results qualitatively.
The single top model (ΔAICc < 4) in the survival model set included the
predictors good food and year (Table 3-6). For each unit increase in good food, puffins
are 48% (32% - 66%, 95% CI; Table 3-7, Figure 3-6) more likely to be detected at
recruitment age while for each unit increase in year, puffins are 18% (7% - 31%, 95% CI;
Table 3-7, Figure 3-6) more likely be detected. These relationships are depicted visually
in Figure 3-6.
Discussion
In contrast to my predictions but similar to other auks (Lloyd 1979, Hedgren
1981, Harris et al. 1992), I found that hatch year conditions for chicks on the natal colony
did not have a strong effect on the probability of natal recruitment in puffins. While hatch
year parameters were present in models with ΔAICc < 4, support for the effects was not
strong, as the effect sizes bounded zero (on the log-odds scale) for all parameters.
The number of islands visited (a parameter relating to the juvenile period) was the
only parameter with a statistically significant effect on natal recruitment in my models.
Intuitively, this is not a surprising result; the more islands a bird visits, the more likely
they are to settle on one of them, rather than their natal island. What is surprising is that
39
this has not yet been reported for puffins. Likely, this is due to a lack of resighting effort
and subsequent data.
One possible reason for the support of hatch year parameters in the models is that,
instead of affecting recruitment rates, they may have an effect on the number of islands
visited. However, in a preliminary look at simple correlations between predictors I
detected no strong correlations between islands visited and any other predictor. A path
analysis could provide better insight into these relationships and allow for a deeper
understanding of how each contribute to the probability of natal recruitment. Another
plausible explanation is that these predictors are actually affecting the survival process,
rather than the natal recruitment process. The survival models provided some evidence to
suggest that the effect of food may be attributable to the carryover affect of survival in
the recruitment models.
Based on these results I suggest that either (a) destination colony may be more
important than colony of origin in the natal recruitment/dispersal process for puffins;
and/or (b) conditions experienced by puffins during the juvenile phase may be more
important than those experienced during hatch year in the natal recruitment process.
Although several studies have reported that the location an individual breeds depends on
factors relating to the destination colony such as colony size (Breton et al. 2006a,
Coulson and Coulson 2008) and density (Gauthier et al. 2010), to my knowledge no
previous study has reported an effect of the extent of prospecting (number of islands
visited) on natal recruitment probability in seabirds. This novel finding is the direct result
of ongoing examination of long-term data and collaborations between researchers on all
islands in the region's puffin metapopulation. Without resighting taking place on all
40
possible breeding islands and data sharing, these results would not have been possible.
This is an important finding, as it suggests that conditions on potential breeding islands
during the prospecting period, rather than conditions on the natal colony during the hatch
year, play an important role in the process of natal recruitment and dispersal for puffins.
A worthwhile future endeavor would be to study the effect of conditions at potential
breeding islands during the prospecting period on natal recruitment probability. Such a
study should yield interesting results and is within the scope of current research being
conducted on the islands by ALAR and others.
My findings highlight the importance of long-term studies of seabirds at the
metapopulation level. Studies of seabirds that focus on a single colony treated as a closed
population ignore two major characteristics of metapopulation dynamics: movement and
dispersal. Puffins in the GOM/BOF are a good example of a metapopulation with discrete
local breeding populations (colonies) where movement between colonies is common and
mitigates the effects of stochasticity at the local level. Modeling population parameters
using data from a single colony in isolation would surely be inaccurate.
41
Table 3-1. Count of number of birds that were detected (survived) by number of
islands visited. The proportion of birds that survived increases as more islands are
visited.
Number of
Islands VisitedSurvived Did Not Survive
1 685 2087
2 151 136
3 53 20
4 7 1
5 2 0
42
Table 3-2. List and description of parameters included in statistical models
describing natal recruitment at Machias Seal Island during 1997 – 2014.
Parameter Code Units Description
Natal Recruitment RE0 = Dispersed
1 = RecruitedRecruited to/Dispersed from MSI
Return Rate (Survival) SU0 = Did not Survive
1 = SurvivedSurvived/Did Not Survive to Recruitment Age
Condition Index CI g/mm fledge mass (g) / fledge wing chord (mm)
Fledge Date FD days fledge date as the number of days since 1 May
Breeding Success BSchicks fledged per
egg laidmean breeding success in an individual's hatch year
Good Food GF g/hourmean hourly amount of herring, hake, and sandlance delivered
to sample plots in an individual's hatch year
Islands Visited IV number of Islandsnumber of islands visited by an individual between fledging and
recruitment
Year YR Year an individual's hatch year
43
Table 3-3. Yearly means from 1997 - 2006 for parameters included in natal
recruitment models.
YearNatal
Recruitment
Condition
Index
Fledge
Date
Number of
Islands
Visited
Breeding
Success
Proportion of
Good Food
1997 0.89 2.14 89.06 1.16 0.61 1.00
1998 0.94 2.17 87.69 1.08 0.65 0.99
1999 0.89 2.11 90.87 1.38 0.60 1.00
2000 0.79 1.91 98.28 1.40 0.48 0.97
2001 0.75 2.16 93.25 1.46 0.71 1.00
2002 0.92 2.02 92.21 1.23 0.59 0.88
2003 0.89 2.15 94.01 1.42 0.67 0.97
2004 0.87 1.87 105.48 1.30 0.76 0.60
2005 0.82 1.97 103.37 1.33 0.82 0.61
2006 0.84 2.07 96.59 1.30 0.35 0.85
44
Table 3-4. Summary of results from generalised linear models with binomial
response recruited to MSI (natal recruitment). Models were compared using Akaike's
Information Criterion (AICc). Number of parameters in the model (K), AICc, ΔAICc
(difference between AICc of current and top model), AICc weight (a measure of support
for each model; values sum to 1) and cumulative AICc weight are reported. Parameters as
in Table 3-1. AICc was 598.27 for the top model.
Model K ΔAICc AICc Wt. Model K ΔAICc AICc Wt.
IV 5 0.00 0.17 IV_GFCI_FD 9 6.48 0.01IV_FDGF 8 1.03 0.10 IV_BS_GFCI 9 6.63 0.01
IV_BS_FDGF 9 1.24 0.09 Null 1 109.97 0.00FD_IV 6 1.30 0.09 FD 2 110.03 0.00IV_GF 6 1.72 0.07 YR 2 110.72 0.00CI_IV 6 1.76 0.07 FDGF 4 110.75 0.00IV_BS 6 1.99 0.06 CIFD 4 111.57 0.00IV_YR 6 2.00 0.06 CI 2 111.66 0.00
IV_FDGF_CI 9 2.84 0.04 FD_YR 3 111.70 0.00IV_YR_FD 7 3.30 0.03 GF 2 111.84 0.00IV_GF_FD 7 3.31 0.03 FD_GF 3 111.88 0.00IV_GF_CI 7 3.58 0.03 BS 2 111.97 0.00IV_GF_BS 7 3.61 0.03 FD_BS 3 112.04 0.00IV_GF_YR 7 3.67 0.03 CI_FD 3 112.04 0.00IV_CIFD 8 3.70 0.03 GF_YR 3 112.45 0.00
CI_YR 3 112.54 0.00
BS_YR 3 112.69 0.00
IV_GFCI 8 4.62 0.02 CI_GF 3 113.61 0.00
IV_GF_BS_FD 8 5.22 0.01 CI_BS 3 113.65 0.00
IV_BS_CIFD 9 5.56 0.01 BS_GF 3 113.86 0.00
IV_CIFD_GF 9 5.73 0.01 CIGF 4 114.77 0.00
CI_FD_IV_BS_GF_YR
_CIFD_CIGF_FDGF13 3.99 0.02
45
Table 3-5. Model averaged estimates of parameter effects (with shrinkage).
Weighted averages of odds ratios for parameters present in top natal recruitment models
(ΔAICc < 4). IV3, IV4 and IV5 = three, four and five islands visited respectively, SYR =
year banded (scaled and centred), SGF = good food (grams per hour; scaled and centred),
SFD = fledge date (days since 1 May; scaled and centred).
Parameter Estimate LCL UCL Pr(>|z|)
(Intercept) 14.84 10.44 21.09 0.00
IV2 0.11 0.07 0.17 0.00
IV3 0.07 0.03 0.16 0.00
IV4 0.04 0.00 0.45 0.01
IV5 0.07 0.00 1.11 0.06
SFD 0.95 0.77 1.17 0.61
SGF 0.94 0.65 1.33 0.71
SFD:SGF 1.09 0.79 1.49 0.60
SBS 1.02 0.88 1.18 0.78
SCI 1.01 0.90 1.14 0.83
SYR 1.00 0.89 1.11 0.98
SCI:SFD 1.01 0.94 1.08 0.85
SCI:SGF 1.01 0.91 1.12 0.89
46
Table 3-6. Summary of results from generalised linear models with binomial
response detected at recruitment age (1), or not detected at recruitment age (0); a
measure of apparent survival. Models were compared using Akaike's Information
Criterion (AICc). AICc, ΔAICc (difference between AICc of current and top models),
AICc weight (a measure of support for each model; values sum to 1). K = number of
parameters.
Model K ΔAICc AICc Wt.
GF_YR 3 0.00 0.87CI_FD_BS_GF_YR_CIFD_CIGF_FDGF 9 6.19 0.04
CIFD_GF 5 7.17 0.02GF 2 7.92 0.02
CI_GF 3 9.24 0.01FD_GF 3 9.30 0.01CIGF 4 9.32 0.01
BS_GF 3 9.57 0.01GFCI_FD 5 10.04 0.01
FDGF 4 10.79 0.00FDGF_CI 5 10.86 0.00BS_GFCI 5 11.16 0.00BS_FDGF 5 12.71 0.00BS_CIFD 5 28.11 0.00
CIFD 4 31.97 0.00
CI_BS 3 35.04 0.00
FD_BS 3 36.04 0.00
BS_YR 3 36.93 0.00
BS 2 39.09 0.00
FD 2 40.71 0.00
CI_FD 3 40.88 0.00
CI_YR 3 41.37 0.00
FD_YR 3 41.86 0.00
CI 2 42.10 0.00
YR 2 44.29 0.00
Null 1 46.44 0.00
47
Table 3-7. Estimates of parameter effects (with shrinkage) for the only supported
survival model (ΔAICc < 4). Estimates are odd ratios. SYR = year banded (scaled and
centred), SGF = good food (grams per hour; scaled and centred).
Parameter Estimate LCL UCL Pr(>|z|)
(Intercept) 0.39 0.36 0.42 2.00E-16
SGF 1.48 1.32 1.66 4.03E-11
SYR 1.18 1.07 1.31 1.59E-03
48
Figure 3-1. The cumulative percentage of puffins tagged as chicks that recruited to
Machias Seal Island per year from ages 5 - 7 for each year cohort form 1997 to 2006.
Recruitment by cohort ranges from 0.12 of the 2004 cohort, to 0.46 of the 2001 cohort. N
= 3,143. Lines are terminated at age 7 so that early cohorts (which have had more time to
recruit) are not over-represented; overall, 96% of birds that recruit, have done so by age 7
(Figure 2). Note that these proportions incorporate individuals that were not seen at
recruitment age as well as those that dispersed to other islands, and as such the values are
lower than "natal recruitment" in the modeling results where only individuals that were
known to have recruited or dispersed are included.
49
Figure 3-2. Resightings of banded puffins showing the proportion of all resighted
birds that were observed at a given age. Most (96%) puffins that are resighted are seen
by age 7. N = 898.
50
Figure 3-3. Correlation matrix for predictors. Correlation coefficients for predictors
included in the model set. Size and colour of circles indicates strength and direction of
the relationship, respectively and corresponds to the numbers on the bottom left portion
of the figure. Parameters as in Table 3-2.
51
Figure 3-4. Relationships between the probability of natal recruitment to Machias
Seal Island and the number of islands visisted. Shading represents 95% confidence
limits. "Rugs" on the top and bottom represent the actual data where points on the top rug
are observations with positive residuals and those on the bottom are observations with
negative residuals.
52
Figure 3-5. Model averaged effect sizes (with 95% confidence intervals) of
predictors for natal recruitment models. Parameter from models in Table 3-4 with
ΔAICc < 4 were used for averaging based on AICc weights. Effect size was calculated
for each parameter with all others held at their mean. IV2, IV3, IV4, and IV5 = two,
three, four and five islands visited respectively, SFD = fledge date (days since 1 May;
scaled and centred), SGF = good food (grams per hour; scaled and centred), SFD:SGF =
interaction between fledge date and good food, SBS = breeding success (chicks fledged
per egg laid; scaled and centred), SCI = condition index (mass at fledge/wing chord at
fledge; scaled and centred), SYR = year banded (scaled and centred), SCI:SFD =
interaction between condition index and fledge date.
53
Figure 3-6. Relationship between the probability of survival and variables present in
top models from Table 3-5 (ΔAICc < 4). Shading represents 95% confidence limits.
"Rugs" on the top and bottom of each graph represent the actual data where points on the
top rug are observations with positive residuals and those on the bottom are observations
with negative residuals. Jitter was applied to better visualise density of the data points.
54
Figure 3-7. Model averaged effect sizes (with 95% confidence intervals) of
predictors for survival models. Parameter from models in Table 3-5 with ΔAICc < 4
were used for averaging based on AICc weights. Effect size was calculated for each
parameter with all others held at their mean. SGF = good food (grams per hour; scaled
and centred), SYR = year banded (scaled and centred).
55
Chapter 4 : Synthesis
My thesis explores the significance of declines in fledging parameters of Atlantic
puffins on MSI over the past three decades. Three parameters, in particular, have all
shown weakly declining trends with time over this period: physical condition (mass/wing
chord); quality of food; and timing of fledging. Similar trends (for condition, food, and
timing of breeding) that were observed in puffins breeding on the Isle of May in Scotland
during a comparable time period were accompanied by a decline in breeding success
(Wanless et al. 2009, Harris and Wanless 2011). I expected to see a similar decline in
breeding success (number of young fledged per egg hatched) in puffins on MSI.
Surprisingly, I found a slight increase over time. However, differences in breeding
success from year to year are much greater than any long-term trend. Puffins in
Newfoundland and Labrador have demonstrated resilience to considerable changes in
food conditions (Baillie 2001), and the same may be true of puffins in the Gulf of Maine.
Young puffins appear to be able to fledge the nest even in years when conditions are not
ideal; one mechanism identified is the preferential allocation of growth to aspects of
development that are the most important for survival, such as head and wing development
(Oyan and Anker-Nilssen 1996, Morrison et al. 2009). A striking decline in breeding
success was observed, however, during a year (2013) when body condition was the worst
(lowest mean mass/wing chord) in the 33 years of observation and the fledge date was the
latest (15 days later than the mean) over that same 33 year period. This is consistent with
a limit to the ability of puffins to overcome poor conditions; the existence of such a limit
is supported by patterns of declining puffin populations that are currently being observed
at the Isle of May and elsewhere in Europe (BirdLife International 2015).
56
The declining trends in fledging condition might be expected to produce
corresponding negative patterns in natal recruitment, especially for a population existing
on the far southern boundary of the species' range. However, I observed only weak (i.e.,
not statistically significant) effects of hatch-year parameters on the probability of natal
recruitment five years later. The degree to which an individual prospected for breeding
sites during the sub-adult life stage (ages 1-4) was the only parameter included in my
models showing a detectable effect on probability of natal recruitment. To my
knowledge, this is a novel result, and provides rare insight that may not have been
apparent had I not been able to control for survival in my analysis. This leads me to
speculate that conditions on destination colonies during the prospecting period may be
more important than conditions at the natal colony when young puffins are deciding
where to breed. Further insight into the natal recruitment process may be gained by
investigating the effect of conditions on destination islands during the prospecting period
on the probability of natal recruitment.
After observing the weak effects of fledging parameters that appeared in the top
models for the recruitment analysis, I suspected that these parameters may contribute
more to the survival process than to natal recruitment. Although I controlled for survival
in this analysis by excluding those that did not survive to breeding age, the process of
survival is still inherent in recruitment. To explore this hypothesis, I conducted an
analysis with survival to breeding age as the response variable. Results of that analysis
confirmed my suspicion, at least partially, with food and year having significant effects
on survival to breeding age. Interpretation of these results is complicated by difficulty in
controlling for recruitment in a survival analysis and, to a lesser degree, the reverse,
57
controlling for survival in a recruitment analysis. The latter difficulty was controlled by
excluding individuals that did not survive to recruitment age. However, contrary to a
study on tufted puffins (Morrison et al. 2009) reporting better survival for birds that
fledged heavier and with longer wings I found no effect of fledge condition on survival.
Natal philopatry of puffins on MSI is very strong with an average of 86% of
survivors from each cohort (1997-2006) returning to MSI (their natal island) to breed.
This is high in comparison to other islands in the metapopulation (ranging from 42% -
75%, Breton et al. 2006a), and other reports for puffins (6.5% - 30.1%, Harris and
Wanless 2011). The proportion that settle on MSI declines with increased prospecting
during the juvenile stage. The reason that some young prospect more than others is
unclear and does not appear to be linked to conditions during their hatch year.
Ideally, natal recruitment in any study would be determined by using data directly
related to breeding status (such as presence of a brood patch, capture within a breeding
burrow, etc.). In the absence of such data, movement patterns of juveniles can be used to
confirm that puffins on MSI, similar to puffins elsewhere (Petersen 1976, Breton et al.
2006a, Harris and Wanless 2011), begin breeding around the age of five years (Breton et
al. 2005). This approach to assigning breeding status is useful when data directly relating
to breeding status are unavailable, and should be considered by those without access to
such data.
In science we seek to gain an understanding of a particular system by observing
patterns and then determining the underlying processes responsible for the observed
patterns. Certainly, the scale at which any process should be studied is an important
consideration. Organisms respond to heterogeneity at multiple spatial and temporal
58
levels, and deciding at what scale to measure this heterogeneity is essential to obtaining
reliable insight into a particular process (Fahrig 1992, Levin 1992). For puffins (and
seabirds in general), the importance of studying dispersal and recruitment at the
metapopulation scale cannot be overemphasized. In terms of the spatial scale considered
for such studies, data from as many locations as possible within the metapopulation are
essential in obtaining reliable recruitment rates, as death and dispersal cannot be
separated in the absence of such data. This process takes many years, especially in long-
lived species, so data must also be collected over a large time scale. Such large-scale data
sets are understandably rare. Data from the Gulf of Maine puffin metapopulation are a
fortunate exception and have allowed for investigations into many aspects of puffin
ecology, including recruitment and dispersal, at an appropriate scale. My study in
particular has provided valuable insight into the nature of natal recruitment in puffins and
should ideally lead to further research into conditions faced during the juvenile period
between fledging and recruitment.
59
Bibliography
Anker-Nilsson, T. and Lorentsen, S.-H. 1990. Distribution of Puffins Fratercula arctica feeding
off Rost, northern Norway, during the breeding season, in relation to chick growth, prey
and oceanographical parameters. Polar Research 8: 67-76
Aubry, L. M., Cam, E., Monnat, J.-Y. 2009. Habitat selection, age-specific recruitment and
reproductive success in a long-lived seabird. In: Modeling demographic processes in
marked populations. Thomson, D.L., Cooch, E.G. & Conroy, M.J. 2009. Environmental
and Ecological Statistics Series, Volume 3. Springer, New York.
Baillie, S. M. 2001. Atlantic puffin response to changes in Capelin abundance in Newfoundland
and Laborador: inter-colony and inter-decade comparison. MSc Thesis. Memorial
University.
Baillie, S.M., Jones, I.L. 2003. Atlantic Puffin (Fratercula arctica) chick diet and reproductive
performance at colonies with high and low Capelin (Mallotus villosus) abundance.
Canadian Journal of Zoology 81: 1598-1607.
Baillie, S.M., and Jones, I.L. 2004. Responses of Atlantic Puffins to a decline in Capelin
abundance at the Gannet Islands, Labrador. Waterbirds 27(1): 102–111.
Barrett, R.T. 2002. Atlantic Puffin Fratercula arctica and Common Guillemot Uria aalge chick
diet and growth as indicators of fish stocks in the Barents Sea. Marine Ecology Progress
Series 230: 275-287.
60
Barrrett, R.T., Rikardsen, F. 1992. Chick growth, fledging periods and adult mass loss of
Atlantic Puffins Fratercula arctica during years of prolonged food stress. Colonial
Waterbirds 15(1): 24-32.
Becker, P. H., & Bradley, J. S. 2007. The role of intrinsic factors for the recruitment process in
long-lived birds. Journal of Ornithology, 148(S2), 377–384.
Bost, C. A., and LeMaho, Y. 1993. Seabirds as bio-indicators of changing marine ecosystems:
new perspectives. Acta Oecologica 14: 463–470.
Boyd, I. L., Wanless, S. and Camphuysen, K., eds. 2006. Top predators in marine ecosystems:
their role in monitoring and management. Cambridge, UK: Cambridge University Press.
Breheny, P. and Burchett, W. 2013. Visualizing regression models using visreg.
http://myweb.uiowa.edu/pbreheny/publications/visreg.pdf
Breton, A. R. 2005. Demographic parameter estimates and their biological implications for a
seabird metapopulation. Ph.D. Thesis. University of New Brunswick.
Breton, A.R. and Diamond, A.W. 2014. Annual survival of adult Atlantic Puffins Fratercula
arctica is positively correlated with Herring Clupea harengus availability. Ibis 156: 35-
47.
Breton, A. R., Diamond A.W., and Kress S.W. 2005. Adult survival estimates from two Atlantic
puffin (Fratercula arctica) colonies in the Gulf of Maine. Auk 122:773–782.
61
Breton, A.R., Diamond, A.W., and Kress, S.W. 2006a. Encounter, survival and movement
probabilities from an Atlantic puffin (Fratercula arctica) metapopulation. Ecological
Monographs 76: 133–149.
-------------------, ----------------, and --------------- 2006b.Surface wear of incoloy and darvic bands
on Atlantic Puffin adults and chicks. Journal of Field Ornithology 77(2): 111-119.
Burnham, K.P. and Anderson, D.R. 2002. Model selection and multimodel inference: A practical
information theoretic approach, 2nd ed. Springer, New York.
Cairns, D.K. 1987. Seabirds as indicators of marine food supplies. Biological Oceanography
5(4): 261-271.
Clobert, J., Ims, R.A., Rousset, F. 2004. Causes, mechanisms and consequences of dispersal. In:
Ecology, genetics, and evolution of metapopulations. Hanski, I., Gaggiotti, O.E. (eds.).
Pp 307-335. Elsevier Academic Press, Massachusetts.
Cooch, E. G., Rockwell, R. F. and Brault, S. 2001 Retrospective analysis of demographic
responses to environmental change: a lesser snow goose example. Ecological
Monographs 71: 377-400.
Cohen, L. A., Pichegru, L., Grémillet, D., Coetzee, J., Upfold, L., & Ryan, P. G. 2014. Changes
in prey availability impact the foraging behaviour and fitness of Cape gannets over a
decade. Marine Ecology Progress Series (505), 281–293.
62
Coulson, J. C., & Coulson, B. A. 2008. Measuring immigration and philopatry in seabirds;
recruitment to Black-legged Kittiwake colonies. Ibis 150, 288–299.
Crespin, L., Harris, M. P., Lebreton, J.-D., Frederiksen, M. &Wanless, S. 2006. Recruitment to a
seabird population depends on environmental factors and on population size. Journal of
Animal Ecology 75: 228–238.
Croxall, J.P., Butchart, S.M., Lascelles, B., Stattersfield, A.J., Sullivan, B., Symes, A., and
Taylor, P. 2012. Seabird conservation status, threats and priority actions: a global
assessment. Bird Conservation International 22: 1-34.
Croxall, J. P. and Rothery, P. 1990. Population regulation of seabirds and implications of their
demography for conservation. In: Bird population studies, relevance to conservation and
management (ed. C. M. Perrins, J.-D. Lebreton & G. J. M. Hirons), pp. 272-296. Oxford
University Press.
Danchin, E., Cadiou, B., Monnat, J.-Y., & Rodriguez Estrella, R. 1991. Recruitment in long-
lived birds: conceptual framework and behavioural mechanisms. In: B. D. Bell (Ed.),
Acta XX Congressus Internationalis Ornithologici (pp. 1641–1656). Wellington, New
Zealand: Congress Trust Board.
Danchin, E., Boulinier, T., and Massot, M. 1998. Conspecific reproductive success and breeding
habitat selection: Implications for the study of coloniality. Ecology. 79(7): 2415-2428.
Devlin, C.M. 2006. Birds crossing borders: A population study of Arctic terns (Sterna
paradisaea). Ph.D. Thesis. University of New Brunswick.
63
Diamond, A.W. 2009. Vicissitudes of the Machias Seal Island tern colony. NB Naturalist 36
(3):116-117.
Diamond, A.W. and Devlin, C.M. 2003. Seabirds as indicators of changes in marine ecosystems:
ecological monitoring on Machias Seal Island. Environmental Monitoring and
Assessment 88: 153-175.
Diamond, A.W., Nettleship, D.N. 1989. Experimental studies of growth in Atlantic Puffins. In:
Population Biology and Conservation of Marine Birds. Abstracts of workshop. Gaston,
A. J., Montevecchi, W. A., Elliot, R. D. (eds.). Pp 15-16. Can. Wildl. Serv. and
Newfoundland Inst. Cold Ocean Sci., Memorial University.
Duncan, W. N. M., & Monaghan, P. 1977. Infidelity to the natal colony by breeding herring
gulls. Ringing & Migration 1(3)
Durant, J.M., Anker-Nilssen, T., Stenseth, N.C. 2006. Ocean climate prior to breeding affects the
duration of the nestling period in the Atlantic Puffin. Biology Letters 2(4): 628-631.
Durant, J.M., Anker-Nilssen, T., Hjermann, D.O., and Stenseth, N.C. 2004. Regime shifts in the
breeding of an Atlantic puffin population. Ecology Letters 7: 388-394.
Einoder, L.D.. 2009. A review of the use of seabirds as indicators in fisheries and ecosystem
management. Fisheries Research 95: 6-13.
Fahrig, L. 1992. Relative importance of spatial and temporal scales in a patchy environment.
Theoretical Population Biology 41(3), 300–314.
64
Finney, S.K., Harris, M.P., Keller, L.F., Elston, D.A., Monaghan, P., and Wanless, S. 2003.
Reducing the density of breeding gulls influences the pattern of recruitment of immature
Atlantic puffins Fratercula arctica to a breeding colony. Journal of Applied Ecology 40:
545-552.
Frederiksen, M., Mavor, R.A., Wanless, S. 2007. Seabirds as environmental indicators: the
advantages of combining data sets. Marine Ecology Progress Series 352: 205-211.
Frederiksen, M., Anker-Nilssen, T., Beaugrand, G., Wanless, S. 2013. Climate, copepods and
seabirds in the boreal Northeast Atlantic - current state and future outlook. Global
Change Biology 19(2): 364-372.
Friars, K.A. and Diamond, A.W. 2011. Predicting the Sex of Atlantic Puffins, Fratercula
arctica, by Discriminant Analysis. Waterbirds 34(3): 304-311.
Gaston, A.J., Bertram, D.F., Boyne, A.W., Chardine, J.W., Davoren, G., Diamond, A.W., Hedd,
A., Montevecchi, W.A., Hipfner, J.M., Lemon, M.J.F., Mallory, M.L., Rail, J.-F., and
Robertson, G.F. 2009. Changes in Canadian seabird populations and ecology since 1970
in relation to changes in oceanography and food webs. Environmental Review 17: 267-
286.
Gauthier, G., Milot, E., & Weimerskirch, H. 2010. Small-scale dispersal and survival in a long-
lived seabird, the wandering albatross. The Journal of animal ecology, 79(4), 879–87.
Greenwood, P. J., & Harvey, P. H. 1982. The natal and breeding dispersal of birds. Annual
Review of Ecology and Systematics 13: 1–21.
65
Hanski, I., & Gilpin, M. E. (Eds.). 1997. Metapopulation biology: ecology, genetics and
evolution. San Diego, California: Academic Press.
Harris, M.P. 1983. Biology and survival of the immature Puffin Fratercula arctica. Ibis 125: 56-
73.
Harris, M.P., Anker-Nilssen, T., McCleery, R.H., Erikstad, K.E., Shaw, D.N., and Grosbois, V.
2005. Effect of wintering area and climate on the survival of adult Atlantic puffins
Fratercula arctica in the eastern Atlantic. Marine Ecology Progress Series 297: 283-296.
Harris, M. P., Halley, D. J., & Wanless, S. 1992. The post-fledging survival of young Guillemots
Uria aalge in relation to hatching date and growth. Ibis 134: 335–339.
Harris M.P., Hislop J.R.G.1978. The food of young puffins (Fratercula arctica). Journal of
Zoology 185: 213–239.
Harris, M.P. and Rothery, P. 1985. The post-fledging survival of young puffins Fratercula
arctica in relation to hatching date and growth. Ibis 127: 243-250.
Harris, M.P. and Wanless, S. 1991. Population studies and conservation of puffins (Fratercula
arctica). In: C. M. Perrins, J.-D. Lebreton & G. J. M. Hirons (Eds.) Bird population
studies, relevance to conservation and management. pp. 230-248. Oxford University
Press.
Harris, M.P. and Wanless, S. 2011. The Puffin. London. Poyser.
66
Hedgren, S. 1981. Effects of fledging weight and time of fledging on survival of Guillemot Uria
aalge chicks. Ornis Scandinavica 12(1), 51–54.
Hudson, P.J., 1985. Population parameters for the Atlantic Alcidae. In: D.N. Nettleship and T.R.
Birkhead (Eds.) The Atlantic Alcidae. Academic, London, 233-261.
Iverson, S.J., Springer, A.M., Kitaysky, A.S. 2007. Seabirds as indicators of food web structure
and ecosystem variability: qualitative and quantitative diet analyses using fatty acids.
Marine Ecology Progress Series 352: 235-244.
Lack, D. 1968.Ecological Adaptations for Breeding in Birds. Methuen, London.
Lebreton, J.-D. and Clobert, J. 1991 Bird population dynamics, management, and conservation:
the role of mathematical modelling. In: C. M. Perrins, J.-D. Lebreton and G. J. M. Hirons
(Eds.) Bird population studies, relevance to conservation and management. pp. 105-125.
Oxford University Press.
Lebreton, J.-D., & Pradel, R. 2002. Multistate recapture models: modelling incomplete
individual histories. Journal of Applied Statistics 29, 353–369.
Levin, S. A. 1992. The problem of pattern and scale in ecology : The Robert H . MacArthur
Award Lecture. Ecology, 73(6), 1943–1967.
Lewis, S., Grémillet, D., Daunt, F., Ryan, P. G., Crawford, R. J. M., & Wanless, S. 2006. Using
behavioural and state variables to identify proximate causes of population change in a
seabird. Oecologia, 147(4), 606–14.
67
Lloyd, C. S. 1979. Factors affecting breeding of razorbills Alca torda on Skokholm. Ibis 121:
165–176.
Lowther, P.E., Diamond, A.W., Kress, S.W., Robertson, G.J., and Russell, K. 2002. Atlantic
Puffin (Fratercula arctica), The Birds of North America Online (A. Poole, Ed.). Ithaca:
Cornell Lab of Ornithology.
MacArthur, R.H. and Wilson, E.O. 1967. The Theory of Island Biogeography. Princeton
University Press, Princeton, N.J.
Martin, A.R. 1989. The diet of Atlantic Puffin Fratercula arctica and Northern Gannet Sula
Bassana chicks at a Shetland colony during a period of changing prey availability. Bird
Study 36: 170-180.
McFadden, D. Quantitative methods for analysing travel behaviour of individuals: some recent
developments. In D. Hensher and P. Stopher (Eds.), Behavioural Travel Modeling. (pp.
279–318). London: Groom Helm.
Morrison, K.W., Hipfner, J.M., Gjerdrum, C., Green, D.J. 2009. Wing length and mass at
fledging predict local juvenile survival and age at first return in tufted puffins. The
Condor 111(3): 433-441.
Oyan, H.S., Anker-Nilssen, T. 1996. Allocation of growth in food stressed Atlantic Puffin
chicks. The Auk 113(4): 830-841.
68
Parsons, M., Mitchell, I., Butler, A., Ratcliffe, N., Frederiksen, M., Foster, S. and Reid, J. B.
2008. Seabirds as indicators of the marine environment. ICES J. Mar. Sci. 65: 1520–
1526.
Perrins, C.M. 1991. Constraints on the demographic parameters of bird populations. In: C. M.
Perrins, J.-D. Lebreton & G. J. M. Hirons (Eds.) Bird population studies, relevance to
conservation and management. pp. 190-206. Oxford University Press.
Perrins, C. M., M. D. Harris, and C. K. Britton.1973. Survival of manx shearwaters, Puffinus
puffinus. Ibis 115: 535-548.
Petersen, A. 1976. Age of first breeding in Puffin, Fratercula arctica. Astarte, 9: 43–50.
Paleczny, M., Hammill, E., Karpouzi, V., and Pauly, D. 2015. Population trend of the world’s
monitored seabirds, 1950-2010. PLoS ONE 10:e0129342.
doi:10.1371/journal.pone.0129342
Pradel, R., Johnson, A.R., Viallefont, A., Nager, R.G., and Cezilly, F. 1997. Local recruitment in
the greater flamingo: a new approach using capture-mark-recapture data. Ecology 78(5):
1431-1445.
R Core Team. 2014. R: A language and environment for statistical computing. R Foundation for
Statistical Computing, Vienna, Austria. URL http://www.R-project.org/.
69
Regehr, H. M. and Rodway, M. S. 1999. Seabird breeding performance during two years of
delayed Capelin arrival in the Northwest Atlantic: a multi-species comparison.
Waterbirds 22(1): 60-67.
Romano, M. D., Piatt, J. F., and Roby, D. D. 2006. Testing the junk-food hypothesis on marine
birds: effects of prey type on growth and development. Waterbirds 29(4): 407-414.
Russell, R. W. 1999. Comparative demography and life history tactics of seabirds: implications
for conservation and marine monitoring. American Fisheries Society Symposium 23: 51-
76.
Saether, B. E. and Bakke, 0. 2000. Avian life history variation and contribution of demographic
traits to the population growth rate. Ecology 81: 642-653.
Sandvik, H., Erikstad, K.E., Fauchald, P., and Tveraa, T. 2008. High survival of immatures in a
long-lived seabird: insights from a long-term study of the Atlantic puffin (Fratercula
arctica). Auk 125(3): 723-730.
Skalski, J.R., Ryding, K.E., and Millspaugh, J.J. 2005. Wildlife demography: analysis of sex,
age, and count data. Burlington, MA. Elsevier Academic Press.
Springer, A.M., Byrd, G.V., Iverson, S.J. 2007. Hot oceanography: planktivorous seabirds reveal
ecosystem responses to warming of the Bering Sea. Marine Ecology Progress Series 352:
289-297.
70
Stenseth, N.C., Mysterud, A., Ottersen, G., Hurrell, J.W.,Chan, K.-S., and Lima, M. 2002.
Ecological effects of climate fluctuations. Science 297: 1292–1296.
Votier, S. C., Birkhead, T. R., Oro, D., Trinder, M., Grantham, M. J., Clark, J. A., … Hatchwell,
B. J. 2008. Recruitment and survival of immature seabirds in relation to oil spills and
climate variability, 974–983.
Wanless, S., Frederiksen, M., Walton, J., & Harris, M. P. 2009. Long-term changes in breeding
phenology at two seabird colonies in the western North Sea. Ibis 151(2), 274–285.
Waugh, S.M., Barbraud,C., Adams,L., Freeman, A.N.D., Wilson, K.-J., Wood, G., Landers,T.J.
and G. B. Baker. 2015. Modeling the demography and population dynamics of a
subtropical seabird, and the influence of environmental factors. Condor 117:147-164.
Weimerskirch, H. 2002. Seabird demography and its relationship with the marine environment.
In: Schreiber, E.A. (Ed.) 2001. Biology of marine birds. CRC Marine Biology Series, 1:
pp. 115-135
Wittenberger, J.F. Hunt, G.L. 1985. The adaptive significance of coloniality in birds. Pages 1-78
In: Farner, D.S. and King J.R., (Eds.) Volume VIII In: Avian Biology. Academic Press,
Orlando Florida.
Ydenberg, R.C., Clark, C.W., and Harfenist, A. 1995. Intraspecific fledging mass variation in the
alcidae, with special reference to the seasonal fledging mass decline. American Naturalist
145(3): 412-433.
Curriculum Vitae
Candidate’s full name: Susan Erin Whidden
Universities attended (with dates and degrees obtained):
University of New Brunswick, 2005 B.Sc. Biology
University of New Brunswick, 2016 M.Sc. Biology
Publications:
Whidden, S. E., C. T. Williams, A. R. Breton and C. L. Buck. 2007. Effects of
transmitters on the reproductive success of Tufted Puffins. Journal of Field Ornithology,
78(2), 206-212.
Conference Presentations and Posters:
Whidden, S. Erin, and A. W. Diamond. Patterns of Natal Recruitment in the Atlantic
Puffin (Fratercula arctica). Presentation given to the Gulf of Maine Seabird Working
Group (GOMSWG) spring meeting, 2012. Bangor, Maine.
Whidden, S. Erin, Cory T. Williams, Andre R. Breton and C. Loren Buck. Effects of
Radiotransmitters on Tufted Puffin (Fratercula cirrhata) Reproductive Success. Poster
Presented at Pacific Seabird Group annual meeting, 2006. Girdwood, Alaska.
Whidden, S. Erin, Cory T. Williams, Andre R. Breton and C. Loren Buck. Effects of
Radiotransmitters on Tufted Puffin (Fratercula cirrhata) Reproductive Success. Poster
Presented at Seabird Group annual meeting 2006. Aberdeen, Scotland.
Recommended