foraging behavior of brown bears on kodiak island, alaska · foraging behavior of brown bears on...
TRANSCRIPT
i
FORAGING BEHAVIOR OF BROWN BEARS
ON KODIAK ISLAND, ALASKA
By
WILLIAM WELLING DEACY
A Dissertation
presented in partial fulfillment of the requirements
for the degree of
Doctor of Philosophy
in Systems Ecology
College of Humanities and Sciences
University of Montana
Missoula, MT
Spring 2016
Approved by:
Scott Whittenburg, Dean of The Graduate School
Graduate School
Dr. Jack Stanford, Chair
Flathead Lake Biological Station, Division of Biological Sciences
Dr. Jonny Armstrong
Department of Fisheries and Wildlife, Oregon State University
Dr. Lisa Eby
College of Forestry and Conservation
Dr. Bonnie Ellis
Flathead Lake Biological Station, Division of Biological Sciences
Dr. Chris Servheen
College of Forestry and Conservation
ii
© COPYRIGHT
by
Full Legal Name
Year
All Rights Reserved
iii
Deacy, William, Ph.D., Spring 2016 Systems Ecology
Foraging behavior of brown bears (Ursus arctos middendorffi) on Kodiak Island, Alaska
Chairperson: Dr. Jack Stanford
ABSTRACT
A key challenge for ecologists is understanding how organisms achieve a positive live history
energy balance in spite of resources which vary in abundance across space and through time.
Recently, two foraging ecology themes have emerged which contribute to our understanding of
this topic. First, resource waves describe how animals can use spatial variation in resource
phenology to extend access to foods. Several publications have highlighted animals using
resource waves caused by elevational or latitudinal gradients, however, none have demonstrated
animals tracking more complex resource waves. Second, the macronutrient optimization
hypothesis (MOH) provides a more nuanced model animal diet selection; rather than simply
maximizing energy intake, the MOH says animals also attempt to minimize digestive costs by
consuming diets with specific mixtures of macronutrients (protein, carbohydrates, and fat). In
this dissertation, I used the foraging behavior of Kodiak brown bears in southwest Kodiak Island,
Alaska to contribute to these two foraging ecology themes: resource waves and macronutrient
optimization. The body of the dissertation consists of four chapters, detailed below.
First, to understand how bears respond to sockeye salmon spawning in tributaries, I
developed a monitoring method that did not disturb foraging bears, was inexpensive, and could
be deployed in remote locations. The system used time-lapse photography and video to observe
passing salmon accurately, but at a fraction of the equipment costs and footage review time
required by conventional methods. I used these systems to monitor 9-11 streams from 2013-
2015. A manuscript detailing this method is currently in review at PeerJ.
In southwest Kodiak Island, sockeye salmon spawning phenology varies among different
spawning locations, creating a resource wave. While spawning at each of these rivers, lake
beaches, and streams may only last for 30-40 days, salmon are spawning somewhere in the study
area for over three months. I used data from GPS collared bears to determine the extent to which
bears used phenological variation in spawning to extend their access to salmon. Bears used an
average of 3 different streams, rivers, and lakes to access salmon, and they visited these sites in
the order predicted by spawn timing. More importantly, the number of spawning sites used was
positively correlated with salmon feeding duration, suggesting phenological variation allowed
bears to increase their access to salmon, a resource linked to bear fitness. These findings were
reported in a paper entitled “Kodiak brown bears surf the salmon red wave: direct evidence from
GPS collared individuals” published in Ecology in May, 2016.
In 2014 and 2015, I observed periods where few bears seemed to be foraging on salmon
despite strong salmon returns. The explanation from local Kodiak naturalists was bears were
abandoning salmon to eat seasonally abundant red elderberries (Sambucus racemosa). Although
this behavior seemed maladaptive from an energetics perspective, the macronutrient optimization
hypothesis (MOH) predicts more efficient weight gain by bears foraging on elderberries
compared to salmon. I used three years of bear distribution data and natural variation in
elderberry phenology to test whether bears foraged according to the MOH in the wild.
iv
Elderberry phenology was relatively early in 2014 and 2015, overlapping the second half of the
salmon runs, whereas the elderberry crop and salmon were discrete in time in 2013. In both
2014 and 2015, bear detection along streams dropped considerably when elderberries became
ripe, while in 2013 bear activity was synchronous with salmon abundance. During the lull in
bear activity on streams, collared bears were using elderberry habitat. Together, these data
suggest wild bears facing real-world foraging constraints forage according to the MOH.
Although bears preferred berries to salmon, salmon were available for much longer, and likely
contribute more to bear annual energy budgets.
In addition to creating a salmon monitoring method that expands the breadth of sites where
salmon monitoring is feasible, I contributed to the foraging ecology literature by testing two
aspects of foraging theory. I used the movements of brown bears to determine whether a mobile
consumers can track a complex resource wave caused by variation in salmon run phenology, and
I used natural variation in red elderberry phenology to test whether wild bears forage according
to the macronutrient optimization hypothesis, foraging on red elderberry when abundant salmon
are available.
ACKNOWLEDGEMENTS
This study was only possible because of the support of many people. I first want to thank my
graduate advisor, Dr. Jack Stanford, who steadfastly supported me throughout the last four and a
half years. Jack helped me see the larger context of my work, and gave me the freedom to
follow my own intellectual path. I am truly grateful to be among his last graduate students. I
would also like to thank my graduate committee, Jonny Armstrong, Lisa Eby, Bonnie Ellis, and
Chris Servheen; their hard work greatly improved this work. I would also like to thank the staff
at the Flathead Lake Biological Station; they provided logistical support which greatly enhanced
my productivity.
This work was funded by the Jessie M. Bierman Professorship, and the USFWS Inventory
and Monitoring, Youth Initiative, and Refuge Programs. I am indebted to the managers,
biologists, pilots, and staff at the Kodiak National Wildlife Refuge, whose assistance made this
project possible. Lecita Monzon, Cinda Childers, and Gerri Castonguay all provided crucial
logistical support, patiently working with our crew to solve the crisis du jour. Many pilots, but
especially Kevin VanHatten and Kurt Rees, flew hundreds of hours to collect data, move us
around the island, and keep us provisioned. Their safe flying despite the foul weather in Kodiak
was impressive. My good friend and predecessor, Mat Sorum, first experimented with video
monitoring of salmon, greatly accelerating the development of the work in chapter 1. Finally, I
want to thank Bill Pyle and McCrea Cobb for providing valuable edits to my proposal.
This study would not have been possible without the hard work of many volunteers. Despite
poor weather and tough work, these folks volunteered their time: Tip Leacock, Barbara Svoboda,
Alex May, Bill Dunker, Tim Melham, Tyler Tran, Marie Jamison, Louisa Pless, Francesca
v
Cannizzo, Isaac Kelsey, Mark Melham, Jane Murawski, Prescott Weldon, Shelby Flemming,
Andy Orlando, and Kristina Hsu.
None of this work would have happened without my friend and mentor, Bill Leacock. He
has been with me every step of the way, from project conception to our last day of field work.
He cares about bears more than anyone I know, and took the time to teach me about their
ecology and how to be safe near bears without scaring them away. I also learned a lot from him
about being a fun, caring supervisor. I will sorely miss the late night laughs and stories.
My greatest thanks goes to my fiancé, Caroline Cheung, who has supported me every day of
the last four and a half years. She put her plans on hold so that I could pursue my passion. She
made fieldwork feel like vacation, adding joy to even mundane tasks. More than anything, I am
going to miss watching bears with her. I would also like to thank my family. My parents have
always let me pursue my passions, and instilled in me a sense of wonder which continues to lead
me towards adventure.
Finally, this work is dedicated to the memory of Tip Leacock. I will always cherish the
memory of her chasing down salmon in O’Malley Creek. I feel fortunate to have spent time with
her in the field, where her calm presence made Camp Island feel like home.
vi
TABLE OF CONTENTS
ABSTRACT ................................................................................................................................... iii
ACKNOWLEDGEMENTS ........................................................................................................... iv
LIST OF FIGURES ..................................................................................................................... viii
LIST OF TABLES ......................................................................................................................... ix
INTRODUCTION .......................................................................................................................... 1
Background ................................................................................................................................. 1
The Problem ............................................................................................................................ 1
Salmon and Bears ................................................................................................................... 3
Foraging Ecology ................................................................................................................... 5
Overview of Chapters ................................................................................................................. 7
Study Area .................................................................................................................................. 9
Broader Impacts ........................................................................................................................ 11
Literature Cited ......................................................................................................................... 11
CHAPTER 1 ................................................................................................................................. 14
Abstract ..................................................................................................................................... 14
Introduction ............................................................................................................................... 15
Methods..................................................................................................................................... 18
Approach ............................................................................................................................... 18
Study Streams ........................................................................................................................ 18
Time lapse camera system .................................................................................................... 19
Modelling salmon escapement (abundance) ......................................................................... 20
Modelling number of living salmon in streams..................................................................... 22
Results ....................................................................................................................................... 24
Salmon Escapement .............................................................................................................. 24
Modelling number of living salmon in streams..................................................................... 25
Discussion ................................................................................................................................. 26
Literature Cited ......................................................................................................................... 31
Chapter 1 Tables ....................................................................................................................... 34
Chapter 1 Figures ...................................................................................................................... 36
vii
CHAPTER 2 ................................................................................................................................. 41
Abstract ..................................................................................................................................... 41
Introduction ............................................................................................................................... 42
Methods..................................................................................................................................... 44
Study site ............................................................................................................................... 45
Variation in timing of salmon availability for bears ............................................................ 46
Bear movements in relation to salmon abundance ............................................................... 46
Movements of collared bears ................................................................................................ 48
Results ....................................................................................................................................... 49
Discussion ................................................................................................................................. 50
Acknowledgements ................................................................................................................... 54
Literature Cited ......................................................................................................................... 54
Chapter 2 Figures ...................................................................................................................... 58
Appendix A ............................................................................................................................... 62
CHAPTER 3 ................................................................................................................................. 63
Abstract ..................................................................................................................................... 63
Introduction ............................................................................................................................... 63
Methods..................................................................................................................................... 68
Study Area ............................................................................................................................. 68
Salmon Abundance................................................................................................................ 69
Elderberry phenology ........................................................................................................... 69
Bear habitat use .................................................................................................................... 70
Results ....................................................................................................................................... 72
Elderberry Phenology ........................................................................................................... 72
Salmon Abundance................................................................................................................ 73
Bear habitat use .................................................................................................................... 73
Discussion ................................................................................................................................. 74
Acknowledgements ................................................................................................................... 78
Literature Cited ......................................................................................................................... 79
Chapter 3 Figures ...................................................................................................................... 82
CHAPTER 4 ................................................................................................................................. 87
viii
LIST OF FIGURES
INTRODUCTION
Figure 0.1 Southwest Kodiak Island study site……………………………..................................9
CHAPTER 1
Figure 1.1. A) Map of study area. B) Photo of camera system. C) Photo of contrast
panels…………………………………………………………………………………………….36
Figure 1.2. Regressions between time lapse and video counts………………………………..…37
Figure 1.3. Estimates of in-stream salmon abundance…………………………………………..38
Figure 1.4. Estimated hourly and cumulative sockeye passage and in-stream abundance………39
Figure 1.5. In-stream salmon abundance with varying stream-life estimates……………………40
CHAPTER 2
Figure 2.1. Map of study area……………………………………………………………………58
Figure 2.2. Date of salmon availability by habitat type………………………………………….59
Figure 2.3. Seasonal use of salmon sites by GPS collared bears………………………………...60
Figure 2.4. A) Number of salmon sites used by bears. B) Number of days a bear consumed
salmon as a function of the number of salmon sites used………………………………………..61
CHAPTER 3
Figure 3.1. A) Photo of spawning sockeye salmon. B) Photo of ripe red elderberry.
C) Example of bear activity cam photo and elderberry phenology photo……………………….82
Figure 3.2. Map of study area showing elderberry land cover…………………………………..83
Figure 3.3. Predicted historical red elderberry phenology……………………………………….84
Figure 3.4. Comparison of bear activity, salmon abundance, and elderberry phenology across
three years………………………………………………………………………………………..85
Figure 3.5. Salmon and elderberry use of GPS collared bears…………………………………..86
ix
LIST OF TABLES
Table 1.1. Comparison of salmon enumeration methods………………………………………..34
Table 1.2. Salmon enumeration model details and model selection metrics…………………….35
1
INTRODUCTION
BACKGROUND
The Problem
Coastal brown bears (Ursus arctos middendorffi), such as those found on the Kodiak
Archipelago, Alaska, are perhaps the most iconic salmon (Oncorhynchus spp.) consumers.
Although the Kodiak bear population seems to be stable and productive overall (there are an
estimated 3500 bears across the archipelago), recent data suggests a population decline in the
Karluk basin in southwest Kodiak Island, which has historically supported a density of 0.48
bears/km2 (William Leacock/FWS, unpublished data).
One of the only sources of bear abundance data in Kodiak is the Intensive Aerial Survey
(IAS), a sightability corrected aerial count of bears conducted jointly by the U.S. Fish and
Wildlife Service (FWS) and the Alaska Department of Fish and Game (ADF&G). Generally,
one IAS is completed each spring before leaf-out (to ensure consistent sightability), although in
some years, logistical problems or very early spring leaf-out have prevented surveys. Thus,
surveys in a given drainage occur only every 4-7 years. The IAS results in 2010 indicated a
severe decline in the bear density in the Karluk basin. The estimated number of bears/1000 km2
dropped from 483 ± 61 (90% confidence interval) in 2003 to 252 ± 61 in 2010, a 48% decline
(William Leacock/USFWS, unpublished data). There was some speculation that the 2010
estimate was biased low because of late den emergence cause by the harsh winter. Rather than
wait another 7 years for the next Karluk basin survey, it was repeated in 2013 to check the 2010
results. The 2013 density estimate result was 248 ± 20 bears, which corroborated the 2010 data.
2
There was much speculation about the cause of such a rapid decrease in bear densities;
hunting, disease, human development activities, and research activities have all been considered.
Although each of these could have played a role, changes in resource availability and foraging
behavior likely played the largest role in the decline. The IAS bear data and salmon spawner
data collected by ADF&G supports this contention: over the period of the bear decline, sockeye
salmon escapement declined steadily from over 1,000,000 in 2003 to under 350,000 in 2010.
Although brown bear population density, body size, and fecundity all strongly correlate with
salmon consumption (Hilderbrand et al. 1999b), ecologists know surprisingly little about the
factors that mediate the bear-salmon relationship. While we might expect overall salmon
abundance to control the relationship between salmon abundance and consumption by bears,
several other factors add complexity to this predator-prey relationship. First, salmon are not
passive prey and are only vulnerable to bears in certain habitat types. Second, bears are
omnivores that can switch to other foods when salmon abundance is low, competition with other
bears is fierce, or a better resource is available. Finally, spawning salmon are very patchily
distributed across space and through time, and bears must navigate this resource mosaic to
maximize their consumption of salmon. The overarching goal of my dissertation was to fill these
gaps in our knowledge of bear foraging in the face of variation in resource abundance. This
project required a systems ecology perspective because bears and salmon are valued
economically, culturally and ecologically, which creates a conflict between those who want to
harvest salmon, and those who want many large bears for hunting or wildlife viewing. In this
dissertation, I tried to both contribute to ecological theory, and use a systems ecology approach
to consider how lessons about brown bear foraging behavior can be used by managers to
maximize bear abundance while minimizing impacts on salmon harvest.
3
Salmon and Bears
Annually, hundreds of millions of salmon swim up the rivers and streams of the North
Pacific Rim. The runs are critical to local economies and culture, and provide an important
influx of nutrients to often nutrient limited systems. Alaska Department of Fish and Game
(ADF&G) recorded an average annual harvest of almost 157 million fish from 2000-2004,
valued in excess of $230 million (ADF&G website, 2012). Because of this high monetary value,
the majority of early research into the interaction between salmon and their predators focused on
the presumed detrimental effects on salmon populations (Gard 1971). More recently, research
has attended to the vital role that spawning salmon play as a link between marine, freshwater,
and terrestrial ecosystems (Schmidt et al. 1998, Schindler et al. 2005, Claeson and Li 2006,
Piccolo et al. 2009). This linkage takes two forms: salmon are important nutrient vectors,
injecting a relatively large subsidy of nutrients derived from the ocean into often nutrient limited
systems (Ben-David et al. 1998), and they serve as a source of food for a variety of mammal,
avian, and aquatic consumers (Willson and Halupka 1995). The overall effect is substantial:
where salmon are abundant, they drive freshwater primary production (Schindler et al. 2005) and
have a strong impact on nearby riparian and terrestrial areas (Willson and Halupka 1995,
Chaloner et al. 2002, Naiman et al. 2002, Helfield and Naiman 2006, Morris and Stanford 2011).
The most conspicuous salmon predator in the Kodiak Archipelago, Alaska, is the Kodiak
brown bear (Ursus arctos middendorffi), which spends considerable time and energy locating,
catching and consuming salmon through the summer and fall (Barnes 1990). Salmon have such
a large influence on coastal brown bears, including Kodiak brown bears, they are considered
distinct from the otherwise similar grizzly bear (Ursus arctos horribilis), that do not have access
to salmon (Pasitschniak-Arts 1993, Hilderbrand et al. 1999a, 1999b).
4
Because of the strong link between salmon consumption and bear fitness, researchers and
managers often focus on salmon abundance as a tool for bear conservation (Hilderbrand et al.
2004, Levi et al. 2012). One approach has been changing salmon management to explicitly
account for the dietary needs of consumers like bears. There are concerns that escapement goals
targeting maximum sustainable yield (MSY) for fisheries may not be sufficient to sustain
historical populations of animal consumers (Levi et al. 2012). Ecosystems based fisheries
management (EBFM) has emerged as an alternative to traditional MSY fisheries management.
Minimizing the impact of fisheries on non-target species is a key goal of EBFM.
Many salmon managers are mandated to consider the needs of consumers and the ecosystem.
For example, the State of Alaska’s Policy for the Management of Sustainable Fisheries states
“The role of salmon in ecosystem functioning should be evaluated and considered in harvest
management decisions and setting of salmon escapement goals (5 AAC 39.222, section
(c)2(G)).” Although this policy is in place, managers must have quantitative tools to estimate
how salmon abundance affects bear population productivity in order to implement it (Levi et al.
2012). One of the biggest challenges of quantifying the effect of varying salmon escapement on
bears is bears operate on a different scale than fisheries and salmon management. Salmon
management occurs over large spatial scales; commercial fishers intercept salmon as they
approach river mouths and the remaining salmon are counted at weirs located near the mouths of
main stem rivers. To achieve escapement goals, fisheries managers use data from the entire river
to decide when to open and close fisheries. In contrast, bears and other consumers primarily prey
on sockeye once they are segregated into smaller spawning sub-populations that are distributed
across heterogeneous landscapes. In a given year, millions of sockeye salmon may spawn in
southwest Kodiak Island lakes, rivers, and streams, but their availability to bears is patchy in
5
both space and time. While human fishers capture salmon from stock aggregations, bears must
move across the landscape to access salmon from multiple subpopulations. How bears respond
to this dynamic resource mosaic determines the impact of anthropogenic and natural fluctuations
of salmon on bears.
In order to consume salmon, bears must identify streams where they are currently spawning,
a potentially difficult task given the high variation in run timing and widely distributed spawning
sites. Spawn timing is variable because it is influenced by abiotic conditions such as water
temperature and groundwater flux (Olsen 1968). In the small tributaries, those most important to
foraging bears, salmon presence is relatively fleeting, sometimes as short as two weeks. A bear
with access to only one stream would have a very short window for consuming salmon.
However, by integrating across multiple streams with asynchronous run timing, bears may
consume salmon for multiple months (Ruff et al. 2011, Schindler et al. 2013). Although there is
some evidence of bears moving among salmon populations (Barnes 1990, Schindler et al. 2013),
there has been no direct evidence from GPS collared individuals that has allowed us to quantify
this behavior.
Foraging Ecology
One of the fundamental areas of ecological inquiry is the study of foraging behavior.
Optimal foraging theory states that animals forage in a manner that maximizes their energy
intake. Within this framework, ecologists hypothesize that movements between patches
(Heinrich and Oecologia 2014), time spent at patches (Charnov 1976), and prey choice are
optimized to maximize energy intake (Pyke et al. 1977). Although they were critical to
furthering ecological understanding, much of the early optimal foraging papers treated predator-
prey systems as static, ignoring the complexities common in real-world examples (Holling 1961,
6
Charnov 1976, Pyke et al. 1977). Additional complexity can arise in natural systems due to:
prey vulnerability that changes across space and time, the spatial pattern of resource patches, and
nutritional constraints on diet selection.
Recently, researchers have made progress towards understanding how foraging animals
select resource patches when the resource changes across space and through time. For example,
some ungulates migrate in response to landscape-scale gradients in vegetation phenology,
moving along the “green wave” of freshly sprouting vegetation (Sawyer and Kauffman 2011,
Bischof et al. 2012). Although this shows animals can track resources across space and time,
vegetation phenology varies along predictable gradients (e.g. elevation, latitude). It is less clear
how consumers respond to resource mosaics that are more patchily distributed, vary less
predictably, and whose availability is more ephemeral. Here, I demonstrate that spawning
salmon function as a resource wave because of spatial variation in spawning phenology. In
contrast to existing examples, however, salmon spawning phenology is driven by water
temperature regimes, which do not vary along a continuous gradient. As a consequence,
consumers may have a harder time tracking this resource wave. My results show that bears
indeed track the resource wave, using it to increase their access to salmon, a resource strongly
linked to fitness (Hilderbrand et al. 1999b).
One limitation of existing foraging theory is the assumption that animals forage to
maximize their energy intake. In reality, foods exert different costs on organisms; for example,
diets overly high (Soucy and Leblanc 1998) or low (Robbins et al. 2007) in protein can increase
digestive costs (dietary induced thermogenesis) which increases an organism’s maintenance cost
(energy consumption needed to offset basal metabolism and digestion). The increased
maintenance costs reduce net energy gain compared to diets with lower digestive costs. The
7
macronutrient optimization hypothesis (MOH) attempts to improve foraging theory by
recognizing dietary costs. According to the MOH, instead of only foraging to maximize energy
consumption, animals also regulate their intake of macronutrients (protein, fat, carbohydrates)
towards specific multidimensional intake targets (Simpson et al. 2004). The increased
maintenance costs reduce net energy gain compared to diets at the optimal macronutrient target.
Research on captive bears has shown bears forage according to both macronutrient optimization
and energy maximization principles; they select foods which maximize their net energy gain, by
both selecting high energy foods and foods which result in diets near their macronutrient targets
(Robbins et al. 2007, Erlenbach et al. 2014). This work found bears mixed food items to
consume a diet with an intermediate amount of protein, specifically, 17 ± 4% of the
metabolizable energy, which produces some counterintuitive predictions about the foraging
behavior of wild bears. Specifically, the MOH predicts Kodiak brown bears faced with a choice
between abundant red elderberries and abundant salmon, will gain more weight by selecting
elderberries. Here we tested the MOH by analyzing bear distributional data across three years
with varying red elderberry phenology. We present evidence suggesting that Kodiak bears, some
of the largest in the world, foraged according to the MOH, consuming red elderberries despite
the presence of hundreds of thousands of highly accessible sockeye salmon.
OVERVIEW OF CHAPTERS
This dissertation consists of four chapters presented in manuscript format. In chapter one, I
present a novel method for monitoring salmon abundance in small streams. Although there are
many existing salmon enumeration methods, they were too expensive, time consuming, or
disruptive to bears for our purposes. Our method increases the breadth of sites where salmon can
be efficiently monitored. This is important given the increasing recognition of the importance of
8
small scale salmon dynamics to large scale population stability (Schindler et al. 2010). I used the
data collected using this method in chapters two and three.
In chapter two, I used movement data from GPS collared female brown bears to quantify
their use of different salmon spawning sites. The timing of salmon spawning varied across the
study area, but generally grouped by habitat: salmon spawned first in streams, then rivers, and
finally in lake beaches. Collared bears used an average of three different sites, and visited them
in the order of salmon spawning. More importantly, the bears that used more sites were able to
access salmon for longer than those that used few sites.
In chapter three, I used natural variation in red elderberry phenology to test whether wild
bears forage according to the macronutrient optimization hypothesis (MOH). Research on
captive bears showed bears forage to balance macronutrient (protein, fat, and carbohydrates)
intake rather than simply maximize energy consumption. The results from the captive bear
studies predicted bears would prefer red elderberries over salmon because of differences in
protein content, however, these bears did not face the same foraging constraints as wild bears. I
monitored bear activity in relation to salmon spawning streams and habitat use of collared bears
across three years where red elderberry (Sambucus racemosa) phenology varied. The results
strongly suggest bears prefer red elderberry over salmon, showing wild bears facing real world
foraging constraints such as competition and movement costs still forage according to the MOH.
Chapter four is a synthesis of the work as a whole. I summarize the results from the other
three chapters and explain how they provide complementary information about the foraging
behavior of Kodiak bears. I finish by highlighting the primary management implications of this
work and by outlining future avenues of research.
9
STUDY AREA
This project focused on the Frazer, Ayakulik, and Karluk drainages on the southwestern
end of Kodiak Island (Figure 0.1). These drainages contain many tributaries that serve as
spawning habitat for sockeye salmon, and habitat that supports some of the highest densities of
brown bears in the world (Miller et al. 1997). Karluk Lake is 19 km long by .8 km wide and has
11 tributaries, most of which are short and steep with only very short reaches accessible to
spawning salmon (Berns et al. 1980). The exceptions are O’Malley and Thumb creeks, which
have relatively high discharge and drain large valleys. Karluk Lake drains into the Karluk River,
which is 39 km long and terminates at the ocean on the north side of the island. The Frazer
drainage contains Frazer Lake which is 14 km long by 1.3 km wide, and has four tributaries.
Figure 0.1- Map of study area. Focal streams are in light blue. Rivers draining each lake are in dark blue.
10
From Frazer Lake, the Dog Salmon River drains southward into Olga Bay after a 14 km run.
The Frazer drainage is unique among the three focal drainages because it supports an introduced
sockeye salmon stock. Historically, there were no anadromous salmonids in Frazer Lake
because a waterfall downstream from the lake prevented upstream migration. In 1951, salmon
were introduced to Frazer Lake, and in 1962, the Frazer Fish Pass was constructed, which
provided access to spawning and juvenile rearing habitat it Frazer Lake. Currently, the Alaska
Department of Fish and Game (ADF&G) operates the fish pass near Frazer Lake outlet and a
weir just upstream of the river’s mouth in Olga Bay. The third drainage consists of the Ayakulik
River and the smallest of the three lakes, Red Lake. Red Lake is 6 km by 1.3 km and has two
significant tributaries, Connecticut Creek and Southeast Creek. From Red Lake the Ayakulik
River runs 25 km to its mouth on the west side of the island where ADF&G operates a weir.
Escapement data collected by ADF&G indicates that the highest salmon abundance in the
Kodiak Archipelago is associated with the three large lake-river systems of southwestern Kodiak
Island. These drainages contain all five species of pacific salmon found in Alaska, but are
dominated by pink (Oncorhynchus gorbuscha) and sockeye salmon (Oncorhynchus nerka).
Collectively SW Kodiak systems account for an average of 51% of the total Kodiak Archipelago
salmon escapement (Van Daele et al. 2013). Historically the Karluk sockeye run has been the
most productive of the three drainages, yielding over 3 million fish at the turn of the last century,
one of the highest returns per unit area on earth (Schmidt et al. 1998). Analysis of nitrogen
isotopes using sediment cores from Karluk Lake indicated large fluctuations in salmon
abundance during the last 500 years (Finney 1998). Harvest and weir data also document
significant variation in abundance, as well as an unprecedented decrease within the past 100
years (ADF&G records).
11
BROADER IMPACTS
Overall, this work contributes to both foraging theory and bear management. I used the
movements of GPS collared bears to provide the first direct evidence of bears tracking the
salmon resource wave. I also used opportunistic variation in bear resources to test whether diet
optimization theory developed using captive bears could predict the diet preferences of wild
bears faced with real-world foraging constraints. Finally, I developed a new method for
monitoring salmon in small streams, which will allow previously impractical studies of bear
salmon interactions at small scales. For bear management, this work highlights the importance
of resource diversity to bears. My chapter 2 results suggest that managers should protect salmon
population diversity, because bear use it to extend their access to salmon. Managers should be
particularly careful to protect bear access to salmon runs that provide salmon at unique times.
Specifically, Kodiak managers should avoid impacts from bear viewing on both Red Lake River
and the Lower Falls of the Dog Salmon River because they are the only sites where bears can
access salmon early in the summer. In chapter 3, I demonstrated the importance of resources
other that salmon to bears. The importance of red elderberry to bears gives managers another
tool for encouraging high bear population productivity. For example, if introduced deer are
found to negatively impact elderberry abundance, increased deer harvests could enhance the
resources available to bears.
LITERATURE CITED
Barnes, V. G. 1990. The Influence of Salmon Availability on Movements and Range of Brown
Bears on Southwest Kodiak Island. Bears : Their Biology and Management 8.
Ben-David, M., T. a. Hanley, and D. M. Schell. 1998. Fertilization of Terrestrial Vegetation by
Spawning Pacific Salmon: The Role of Flooding and Predator Activity. Oikos 83:47.
Berns, V. D. VD, G. C. G. Atwell, and D. D. L. Boone. 1980. Brown bear movements and
habitat use at Karluk Lake, Kodiak Island. Bears: Their Biology and Management 4:293–
296.
12
Bischof, R., L. E. Loe, E. L. Meisingset, B. Zimmermann, B. Van Moorter, and A. Mysterud.
2012. A migratory northern ungulate in the pursuit of spring: jumping or surfing the green
wave? The American naturalist 180:407–24.
Chaloner, D., M. Wipfli, and J. Caouette. 2002. Mass loss and macroinvertebrate colonisation of
Pacific salmon carcasses in south-eastern Alaskan streams. Freshwater Biology:263–273.
Charnov, E. 1976. Optimal foraging, the marginal value theorem. Theoretical population biology
9:129–136.
Claeson, S., and J. Li. 2006. Response of nutrients, biofilm, and benthic insects to salmon
carcass addition. Canadian Journal of … 1241:1230–1241.
Van Daele, M., C. T. Robbins, B. X. Semmens, E. J. Ward, L. J. Van Daele, and W. B. Leacock.
2013. Salmon consumption by Kodiak brown bears (Ursus arctos middendorffi) with
ecosystem management implications. … Journal of Zoology 174:164–174.
Erlenbach, J. a., K. D. Rode, D. Raubenheimer, and C. T. Robbins. 2014. Macronutrient
optimization and energy maximization determine diets of brown bears. Journal of
Mammalogy 95:160–168.
Finney, B. 1998. Long-term variability of Alaskan sockeye salmon abundance determined by
analysis of sediment cores. North Pacific Anadromous Fish Commission Bulletin.
Gard, R. 1971. Brown bear predation on sockeye salmon at Karluk Lake, Alaska. The Journal of
Wildlife Management 35:193–204.
Heinrich, B., and S. Oecologia. 2014. Resource Heterogeneity and Patterns of Movement in
Foraging Bumblebees. Oecologia 40:235–245.
Helfield, J. M., and R. J. Naiman. 2006. Keystone Interactions: Salmon and Bear in Riparian
Forests of Alaska. Ecosystems 9:167–180.
Hilderbrand, G., S. Farley, C. Schwartz, and C. Robbins. 2004. Importance of salmon to wildlife:
implications for integrated management. Ursus:1–9.
Hilderbrand, G. V, S. G. Jenkins, C. C. Schwartz, T. A. Hanley, and C. T. Robbins. 1999a.
Effect of seasonal differences in dietary meat intake on changes in body mass and
composition in wild and captive brown bears. Canadian Journal of Zoology 1630:1623–
1630.
Hilderbrand, G. V, C. C. Schwartz, C. T. Robbins, M. E. Jacoby, T. a Hanley, S. M. Arthur, and
C. Servheen. 1999b. The importance of meat, particularly salmon, to body size, population
productivity, and conservation of North American brown bears. Canadian Journal of
Zoology 77:132–138.
Holling, C. 1961. Principles of insect predation. Annual review of entomology:163–182.
Levi, T., C. T. Darimont, M. MacDuffee, M. Mangel, P. Paquet, and C. C. Wilmers. 2012. Using
Grizzly Bears to Assess Harvest-Ecosystem Tradeoffs in Salmon Fisheries. PLoS Biology
10:e1001303.
Miller, S., G. White, and R. Sellers. 1997. Brown and black bear density estimation in Alaska
13
using radiotelemetry and replicated mark-resight techniques. Wildlife ….
Morris, M., and J. Stanford. 2011. Floodplain succession and soil nitrogen accumulation on a
salmon river in southwestern Kamchatka. Ecological Monographs 81:43–61.
Naiman, R. J., R. E. Bilby, D. E. Schindler, and J. M. Helfield. 2002. Pacific Salmon, Nutrients,
and the Dynamics of Freshwater and Riparian Ecosystems. Ecosystems 5:0399–0417.
Pasitschniak-Arts, M. 1993. Mammalian Species: Ursus arctos. The American Society of
Mammologists 439:1–10.
Piccolo, J., M. Adkison, and F. Rue. 2009. Linking Alaskan salmon fisheries management with
ecosystem-based escapement goals: a review and prospectus. Fisheries:37–41.
Pyke, G., H. Pulliam, and E. Charnov. 1977. Optimal foraging: a selective review of theory and
tests. Quarterly Review of Biology 52:137–154.
Robbins, C. T., J. K. Fortin, K. D. Rode, S. D. Farley, L. A. Shipley, and L. A. Felicetti. 2007.
Optimizing protein intake as a foraging strategy to maximize mass gain in an omnivore.
Oikos 116:1675–1682.
Ruff, C. P., D. E. Schindler, J. B. Armstrong, K. T. Bentley, G. T. Brooks, G. W. Holtgrieve, M.
T. McGlauflin, C. E. Torgersen, and J. E. Seeb. 2011. Temperature-associated population
diversity in salmon confers benefits to mobile consumers. Ecology 92:2073–84.
Sawyer, H., and M. J. Kauffman. 2011. Stopover ecology of a migratory ungulate. The Journal
of animal ecology 80:1078–87.
Schindler, D., J. Armstrong, K. Bentley, K. Jankowski, P. Lisi, and L. Payne. 2013. Riding the
crimson tide: mobile terrestrial consumers track phenological variation in spawning of an
anadromous fish. Biology Letters 9:2–6.
Schindler, D. E., R. Hilborn, B. Chasco, C. P. Boatright, T. P. Quinn, L. a Rogers, and M. S.
Webster. 2010. Population diversity and the portfolio effect in an exploited species. Nature
465:609–12.
Schindler, D., P. Leavitt, and C. Brock. 2005. Marine-derived nutrients, commercial fisheries,
and production of salmon and lake algae in Alaska. Ecology 86:3225–3231.
Schmidt, D. C. D., S. S. R. Carlson, G. B. Kyle, and B. P. Finney. 1998. Influence of carcass-
derived nutrients on sockeye salmon productivity of Karluk Lake, Alaska: importance in the
assessment of an escapement goal. North American Journal … 18:743–763.
Simpson, S. J., R. M. Sibly, K. P. Lee, S. T. Behmer, and D. Raubenheimer. 2004. Optimal
foraging when regulating intake of multiple nutrients. Animal Behaviour 68:1299–1311.
Soucy, J., and J. Leblanc. 1998. Protein meals and postprandial thermogenesis. Physiology and
Behavior 65:705–709.
Willson, M. F., and K. C. Halupka. 1995. Anadromous Fish as Keystone Species in Vertebrate
Communities. Conservation Biology 9:489–497.
14
CHAPTER 1
A TIME-LAPSE PHOTOGRAPHY METHOD FOR MONITORING SALMON
(ONCORHYNCHUS SPP.) PASSAGE AND ABUNDANCE IN STREAMS
WILLIAM DEACY1, WILLIAM LEACOCK
2, LISA EBY3, JACK A. STANFORD
1
1 Flathead Lake Biological Station, University of Montana, Polson, MT, USA
2 Kodiak National Wildlife Refuge, Kodiak, AK, USA
3 Wildlife Biology Program, University of Montana, Missoula, MT, USA
ABSTRACT
Accurately estimating population sizes is often a critical component of fisheries research and
management. Although there is a growing appreciation of the importance of small-scale salmon
population dynamics to the stability of salmon stock-complexes, our understanding of these
populations is constrained by a lack of efficient and cost-effective monitoring tools for streams.
Weirs are expensive, labor intensive, and can disrupt natural fish movements. While
conventional video systems avoid some of these shortcomings, they are expensive and require
excessive amounts of labor to review footage for data collection. Here, we present a novel
method for quantifying salmon in small streams (<15m wide, <1m deep) that uses both time-
lapse photography and video in a model-based double sampling scheme. This method produces
an escapement estimate nearly as accurate as a video-only approach, but with substantially less
labor, money, and effort. It requires servicing only every 14 days, detects salmon 24 hrs. /day,
costs less than $3000 per system, and produces escapement estimates with confidence intervals.
In addition to escapement estimation, we present a method for estimating in stream salmon
abundance across time, data needed by researchers interested in predator-prey interactions or
nutrient subsidies. We combined daily salmon passage estimates with stream specific estimates
15
of daily mortality developed using previously published data. To demonstrate proof of concept
for these methods, we present results from two streams in southwest Kodiak Island, Alaska in
which high densities of sockeye salmon spawn.
INTRODUCTION
Accurately estimating population sizes is often a critical component of fisheries research and
management. Managers use salmon (Oncorhynchus spp.) escapement estimates (salmon
remaining after harvest that enter freshwater to spawn) to develop stock-recruit curves and to
decide when to open and close fisheries. Researchers often need escapement data for studies
involving productivity, nutrient subsidies, and predator-prey dynamics. Although we have good
escapement data for many main-stem rivers used by migrating salmon, we have little escapement
data at smaller scales, including small streams where many salmon ultimately spawn. This is
regrettable given that large salmon stock-complexes are composed of dozens or hundreds of
distinct salmon populations, many of which spawn in first and second order streams. A
collection of small salmon populations spawning at different times and in different locations
tends to have more stable interannual abundance than a single homogenous population due to
“portfolio effects,” which results in more reliable returns and fewer closures for commercial
fisheries (Schindler et al. 2010). This stability arises from population diversity occurring at
small spatial scales (i.e. first and second order streams), so it is important that we have the tools
to investigate and understand these populations in order to effectively manage salmon for human
and wildlife consumers.
Watershed-scale escapement estimates do not effectively characterize the resources
available to wildlife consumers, because they do not tell us how long salmon are available to
consumers. In many watersheds, consumers cannot catch salmon while they migrate up the
16
relatively deep water of main-stem rivers; they must wait until salmon enter shallow spawning
streams where they are more easily caught. As a result, consumers interact with individual
salmon populations rather than entire stock complexes, and thus, watershed scale escapement can
be a poor estimate of the salmon available to consumers of conservation concern such as eagles,
bears and trout (Bentley et al. 2012, Schindler et al. 2013, Levi et al. 2015; Deacy et al. in press).
Also, consumers are easily satiated by even modest densities of spawning salmon, so the
duration of spawning activity is likely just as important to consumers as the abundance of salmon
(Jeschke 2007). Despite the importance of small tributary salmon escapement to salmon
management, ecosystem function, and salmon conservation, existing methods of monitoring
salmon abundance do not perform well at these sites, because they are expensive, time
consuming, and alter salmon behavior.
Traditionally, anadromous salmonids (Oncorhynchus spp.) moving into large rivers or
streams have been counted by observers stationed at fish weirs, fences, and observation towers,
or by use of sonar stations (Table 1; Cousens et al. 1982). These methods can produce reliable
estimates; however, high labor and equipment costs make them too expensive for simultaneously
monitoring many streams. To fill this gap, researchers have experimented with systems that
record video of passing salmon using either under or above water cameras (Hatch et al. 1994,
Davies et al. 2007, Van Alen 2008). These video weir methods have three key advantages: 1)
footage can be counted long after the data are collected, allowing a small crew to monitor several
runs simultaneously; 2) periods with high salmon abundance can be counted more accurately by
reducing playback speed; and 3) fewer site visits reduce impacts on wildlife caused by human
presence. Although these benefits have made video enumeration an increasingly popular method
for counting salmonids, reviewing large amounts of video is required. The resulting personnel
17
costs make video weir methods impractical for many applications. A method is needed for
collecting escapement data that produces reliable estimates without thousands of hours of video
review or frequent site visits. Furthermore, some enumeration methods (i.e. weirs) can obstruct
natural movements of salmon and other fishes. This may not be a problem on main-stem rivers
if salmon tend to move consistently upstream, however, it is problematic in small streams where
diel movements into and out of streams is common (Bentley et al. 2014).
In addition to total escapement, studies focused on consumer responses to availability of
salmon need to know the number of living salmon in streams (hereafter in stream abundance)
across time. In stream abundance across time represents foraging opportunities better than gross
escapement when consumers are swamped by a pulsed resource, which is often the case for
consumers of spawning salmon (Armstrong and Schindler 2011). Typically, in stream
abundance data are collected using ground (Quinn et al. 2001) or aerial surveys (Neilson and
Geen 1981) which are repeated several times during a salmon run. Ground surveys work well on
streams that are easy to access, small enough to survey in a reasonable amount of time, and
where disturbing wildlife is not a concern. Aerial surveys may work well for less accessible sites
if visibility from the plane is not impeded by riparian vegetation or complex channel
geomorphology. Moreover, because salmon abundance in streams tends to change rapidly, these
methods only work well when the survey frequency is high. Furthermore, to collect reliable data
using aerial surveys, researchers need to correct for differences among observers (Bue et al.
1998). Here, we present an alternative method for estimating the number of living salmon in a
stream through the full duration of the run. The approach combines daily estimates of salmon
passage, collected using our time-lapse camera system, with a model of spawning salmon
mortality.
18
Our system requires service only every 14 days, detects salmon 24 hrs. /day, costs less
than $3000 per system, and produces escapement estimates with confidence intervals. This
system works on rivers and streams up to ~15m wide and ~1m deep. In addition, we present a
method for estimating in stream salmon abundance, data which are important for studies focused
on the response of wildlife consumers to salmon runs and nutrient subsidies. To demonstrate
proof of concept, we present results from two small streams with very high densities of salmon.
METHODS
Approach
To harness the advantages of remote camera systems without time-consuming video
enumeration, we utilized a “double sampling” scheme, which is often used when a variable of
interest is costly to measure, but an auxiliary variable is more easily measured and has a
predictable relationship to the variable of interest (Cochran 1977). The cheaper variable can be
measured for all of the sample units while the expensive variable is measured on a subsample of
units in order to model the relationship between the variables. Here, our variable of interest is
the number of salmon that pass into and out of a stream each hour, which we can accurately
quantify with an above-water video camera. The related auxiliary variable is the number of
salmon detected in time-lapse images each hour. The total time required to review footage is
low relative to video-only approaches because we only have to enumerate salmon in a subset of
the hour long sample units. We can determine the salmon passage for the remaining hours by
modelling the relationship between the subsample of hourly video counts and photo counts and
then using the model to predict salmon passage across the entire salmon run.
Study Streams
19
We developed this method on two streams used by spawning sockeye salmon: Meadow
and Southeast Creeks (Fig. 1.1a) in southwest Kodiak Island, Alaska. Meadow Creek is a
second order tributary to Karluk Lake. It has a mean width of 4.50 m and depth of 13 cm in the
lower 0.8 km used by spawning salmon. Southeast Creek is a first order tributary to Red Lake
that flows out of a small spring pond. It has a mean width of 3.90 m and depth of 9.1 cm in the
lower 2.7 km used by spawning salmon. Tens of thousands of salmon enter these streams
annually to spawn and bidirectional movement and pre-spawn mortality is common owing to a
large number of brown bears that prey on the salmon.
Time lapse camera system
To record time lapse images of passing salmon, we used a Reconyx® Hyperfire PC800
camera, programmed to take 3 photos in rapid succession (<1 sec. between frames) each minute,
24 hrs./day. Each three frame burst allowed us to detect the number and direction of travel (up
or downstream) of salmon passing the camera. We suspended the time lapse camera above the
stream using steel electrical conduit attached to a steel Big Game® Pursuit tripod tree stand
positioned adjacent to the stream (Fig. 1.1b). We attached the camera to the conduit with a
Camlockbox® ball mount which allowed us to easily aim the camera. To light the streambed at
night we secured an LTS® IR50 850nm infrared (IR) light to the tripod platform. Although
visible light would have worked well, we used IR light to avoid changing the behavior of salmon
and/or their predators with visible light. The Reconyx camera and infrared light were powered
by an 80 amp-hour deep-cycle battery charged by a 100W solar panel secured to the south side
of the tower.
To record video, we secured a video camera to the top of the tower. The video footage
was stored by a Digital Video Recorder (DVR) set to record D1 resolution, 30 frames per second
20
video from 12pm-8pm, the periods with the best quality video (good light) and the majority of
salmon movement activity. The video camera and DVR were powered by its own battery/solar
power system, identical to the one powering the Reconyx camera and IR light. To make passing
salmon easier to see, we secured 5.08 cm X 76.2 cm white High Density Polyethylene (HDPE)
contrast panels to the bottom of the stream below the cameras by attaching them to a heavy chain
(Alaska Department of Fish and Game Permit # FH-14-II-0076). The HDPE panels are buoyant
in water and the chain prevents the panels from floating off of the streambed. Using stainless
steel carabineers, we attached the chain to T-posts which we pounded into the margins of the
streambed. To prevent salmon from swimming under the panels, we pinned the chain to the
stream bed using several steel stakes.
We visited each camera system every two weeks from early June through early
September to switch out data cards and remove algae and debris from the contrast panels. Back
at our field station, we separately counted the number of salmon moving up and downstream past
the contrast panels during each three-photo burst. We only counted a salmon as passing if it
moved at least ½ the length of the panels; we did not count stationary fish. Finally, we summed
upstream and downstream counts separately for each hour of the monitoring season. To ensure
consistent counting technique, each stream was counted by the same person for the entire season.
Modelling salmon escapement (abundance)
We used a model-based double sampling approach to estimate salmon escapement. We
modelled the relationship between video salmon counts and photo salmon counts for a non-
random subsample of hours, and then used this model to predict salmon passage for the entire
season. This is different from the “sampling-design approach” more commonly used to double
sample (Cochran 1977). If we had used the sampling-design approach, we would have counted
21
the salmon passing in a simple random subsample of video hours, and then calculated the total
escapement by multiplying the time lapse salmon count by the ratio of video counts to photo
counts in the subsample. However, the sampling design-based approach has two requirements
which are difficult to satisfy. First, to be random, every hour of the salmon run must be available
for sampling, meaning that video must be recorded throughout the entire run. A single day of
missed video (due to a power outage, insects sitting on the lens, etc.) could significantly bias the
resulting abundance estimates if the outage occurred on a day with relatively few or many
passing salmon. Second, the video must be high enough quality to assume 100% salmon
detection. This requirement can be difficult to meet because of glare and poor night-time video
quality. Rather than attempt to design a system that meets these strict requirements, we used a
model-based approach, where we model the relationship between video counts and time lapse
counts (Stephens et al. 2012). This framework allows us to select our sample of video-
enumerated hours non-randomly; our estimate of abundance is unbiased as long as the model is
correctly specified (Hansen et al. 1983, Gregoire 1998).
We selected 70 hours that spanned the full range of hourly time-lapse salmon counts,
from the hours with many salmon swimming downstream to hours with strong upstream
movement. Also, we selected hours where we were confident of nearly 100% detection,
excluding hours with bad glare or poor lighting. In total, we watched 70 hours of video for each
stream, however, because we considered up and downstream salmon movement independently,
this gave us a sub-sample of 140 values for each stream (70 upstream counts and 70 downstream
counts).
Next, we modelled video counts as a function of time-lapse photo counts for the
subsample. We compared four different models for each stream: first and second order linear
22
regressions and first and second order segmented or “split-point” linear regressions (Table 2).
The segmented regression allows the slope to differ across ranges of the predictor variable. This
makes sense for salmon swimming in a stream; salmon swimming upstream (positive values)
might move slower, and thus have a greater chance of being detected in a time-lapse burst. In
contrast, salmon swimming downstream (negative values) might move faster and have a lower
likelihood of detection. To address this possibility, we including segmented regression models
with the split-point (slope inflection point) constrained to zero. To assess relative model fit, we
compared Akaike’s Information Criterion values (AICc; Akaike 1974). To validate models and
test for over-fitting, we performed leave one out cross validation (LOOCV; Kohavi 1995), and
used the resulting predictions to calculate the precision (mean squared error, MSE) and accuracy
(the percent difference between the predicted and actual escapement of the 70 hours for which
we watched video). Based on these metrics, we selected a top model for each stream.
Using the top model for each stream, we predicted the salmon passage for all of the hours
of the monitoring period. The sum of these predictions is the estimated escapement. Because we
did not use random sampling to select our modelling subsample, it is inappropriate to use the
model variance to calculate confidence intervals for total escapement. Instead, we bootstrapped
our subsample with replacement (140 values to match our original subsample), refit our model
using the top model structure, and re-predicted the total escapement (Efron and Tibshirani 2003).
We repeated this 10,000 times and used the 2.5 and 97.5 percentile values as upper and lower
95% confidence intervals of total escapement.
Modelling number of living salmon in streams
To model in stream abundance across time, we took daily escapement estimates
(upstream moving salmon minus downstream moving salmon), and applied mortality estimates
23
from the literature. Carlson et al. (2007) investigated the relationship between stream
width/depth and stream life (number of days from salmon stream entry to death) on a range of
tributaries to Nerka and Aleknagik Lakes, Alaska which are morphologically similar to our focal
streams. The three main sources of mortality for spawning sockeye salmon were senescent
death, predation (mostly by bears), and stranding. They found that salmon spawning in
wider/deeper streams tended to have longer stream lives. The authors’ explanation was that
salmon in shallow/narrow streams experienced higher predation rates which selects for more
rapid reproductive cycles and consequently earlier deaths. Because of this interaction between
stream morphology and salmon stream life, it is probably inappropriate to use a single estimate
of stream life across streams with varying morphology. We used the results of Carlson et al.
(2007) to create a model of stream life as a function of stream morphology.
Assuming salmon in our streams were equally likely to die by stranding, predation, and
senescence as they were in the Carlson study, we calculated a weighted average of the mean
stream life for each of the Carlson et al. (2007) streams. We then used this weighted average
stream life as the response variable and stream width and depth as predictor variables in a simple
linear regression model. Because stream depth and width were strongly correlated (r =0.90),
including both variables in the model resulted in collinearity. We thus selected between depth-
only and width-only models by comparing AICc scores. We then used the top model to predict
the mean stream life of sockeye salmon in Meadow and Southeast Creek, using field
measurements of stream morphology measured in 2014 as predictors. There was a strong
positive correlation between the mean and pooled standard deviation (Hedges 1981) of stream
life in the Carlson data (r =0.95, p=0.004); therefore, rather than model the standard deviation
24
(SD) of stream life separately from the mean, we assumed stream life SD was proportional to the
mean (SD=.499 * mean stream life).
To calculate in stream abundance each day, we summed the number of salmon that
entered on that day with the predicted number of surviving salmon from the previous days:
𝐿𝑖𝑣𝑖𝑛𝑔 𝑆𝑎𝑙𝑚𝑜𝑛 𝑂𝑛 𝐷𝑎𝑦 𝑥 = ∑ 𝑃𝑥
𝑁
𝑡=1
+ 𝑃𝑥−𝑡𝑆𝑡
where Px is the number of salmon that passed into the stream on day x, Px-t is the number of
salmon that passed into the stream t days before day x, St is the proportion of those salmon
surviving to day x, and t is an index of days. The values of St are from the cumulative
distribution function of survival which we modelled above. N is the number of days it takes for
survival (St) to reach zero, which varies based on the survival model (it will be larger on deeper
streams where stream life is greater).
To understand the sensitivity of in stream abundance models to changes in stream life
estimates, we calculated in stream abundance for each stream across a range of stream life
values. We then used percent change in maximum abundance to assess the impact of changing
stream life. Because the amount of time consumers have access to salmon is at least as important
as peak abundance, we also calculated the duration of the salmon run, defined as the number of
days where abundance was at least ten percent of the maximum in stream estimate from the un-
altered model. This (admittedly arbitrary) ten percent threshold was an attempt to set a lower
limit on the salmon density below which benefits to consumers decline.
RESULTS
Salmon Escapement
25
Of the suite of models relating video counts to time lapse counts for Meadow Creek, the
top model was the segmented first order model (Table 1.2, Fig. 1.2). It had the lowest AICc
(1534.3), best precision (MSE=3556), and best accuracy (+3.0%). The segmented models likely
explained more variation than the unsegmented models because salmon had different detection
rates while swimming upstream versus downstream (salmon swim slower against the current), in
the relatively steep gradient of Meadow Creek. Using the top model, the predicted escapement
for Meadow Creek was 30,509 ± 9,494 (95% confidence intervals).
The top model for Southeast Creek was the first order regression which had the lowest
AICc (1732.2), best precision (MSE=14167), and best accuracy (+3.1%). In contrast to Meadow
Creek, the segmented model only explained slightly more variation than the first order model,
but required an additional parameter. This suggests salmon in Southeast Creek have a similar
detection rate whether they are swimming up or downstream, which is likely because Southeast
Creek has a relatively flat gradient and low velocity. The total escapement for Southeast Creek
was 65,355 ± 4,305 (95% confidence intervals). For Southeast Creek, the escapement estimates
were not very sensitive to the model selected (maximum difference of only 4.4%) (Fig. 1.3). This
contrasts with Meadow, where the difference between the highest and lowest estimate was 38%.
Modelling number of living salmon in streams
The model with depth as a predictor (AICc = 27.5) explained more variation than the
width model (AICc = 31.9), so we used this model to predict mean stream life for our two
streams. Meadow Creek had a predicted mean stream life of 7.1 days (SE=3.5) while Southeast
Creek (which is shallower), had a predicted stream life of 5.9 days (SE=3.0). Using these values,
we found the predicted salmon abundance over time in each stream were quite different;
abundance peaked at just over 11,000 sockeye on July 11th in Meadow Creek and the run was
26
finished around August 16th (Fig. 1.4). In contrast, Southeast Creek had two distinct peaks in
abundance: the first on July 21st with just over 15,000 sockeye and the second peaking at 4,645
on August 29th. Thus, although the total escapement in Southeast Creek was more than double
that of Meadow Creek, the peak salmon abundance was only 29% higher in Southeast.
In general, the in stream abundance models were quite sensitive to changes in stream life
estimates. Increasing mean stream life in Meadow Creek by 2 days, from 7.1 to 9.1 days,
increased the estimated maximum abundance by 14% (Fig. 1.5). The effect was even greater on
Southeast Creek, with a 22% increase in abundance from a 2 day increase in mean stream life.
Increasing the standard deviation had the opposite effect: a 1 day increase in SD of stream life
decreased the maximum abundance by 5% and 3% on Meadow and Southeast Creeks,
respectively. The sensitivity of salmon run duration (defined as the number of days with at least
10% of the maximum salmon abundance), to changes in mean and SD of stream life was less
clear. On Meadow Creek, increasing mean stream life by 2 days increased the salmon run
duration by 2 days (from 40 to 42 days) and increasing stream life SD by 1 day resulted in no
measurable increase in salmon run duration. In contrast, the same changes on Southeast Creek
resulted in an 5 day and 2 day increase in salmon run duration for changes to the mean and SD of
stream life, respectively. This difference is likely because Southeast Creek has two distinct
peaks in salmon abundance, and a 2 day increase in stream life is a larger proportional change
compared to Meadow creek.
DISCUSSION
Researchers and managers increasingly acknowledge the important role of small salmon
populations in generating stable returns for commercial fisheries and for supporting wildlife of
high economic and commercial value (Schindler et al. 2010, Beacham et al. 2014). Many
27
existing salmon monitoring tools were designed primarily for large streams and rivers and are
ineffective or too expensive for monitoring the salmon populations that use small streams for
spawning. The time-lapse salmon counting system presented here proved to be a low-cost, time-
efficient, and accurate method for counting salmon in streams less than 15m wide. This method
only required bi-weekly site visits, which is ideal for remotely monitored sites and studies
involving the response of wildlife to spawning salmon. These benefits will allow managers and
researchers to quantify salmon in streams where it was previously too difficult or expensive. In
addition, we presented a method for estimating the number of living salmon in a stream across
the run, data which are particularly important for consumer-resource studies.
To estimate in stream salmon abundance, we developed a model of salmon stream life
(number of days a salmon survives following spawning stream entry) based upon data collected
in the Wood River system, Alaska (Carlson et al. 2007). These data are specific to the sites and
years where they were collected; differences in water level, intensity of predation, and salmon
abundance are all likely to change these values. For these reasons, future users of the method we
demonstrated here should estimate stream life in their own systems, rather than relying on the
model developed using the Carlson et al. (2007) data. This is particularly important because a
sensitivity analysis showed our in stream salmon estimates were quite sensitive to changes in
estimated stream life (Fig. 1.5); a two day increase in stream life increased the estimated
maximum abundance by 14% on Meadow Creek and 22% on Southeast Creek.
Similarly, a good escapement estimate is only possible if users accurately model the
relationship between time lapse and video counts (Hansen et al. 1983). This is critical given the
large differences in abundance estimates resulting from small differences in model structure or
fit (Table 2, Fig. 1.4). It is important to consider multiple model shapes; different stream
28
morphologies or salmon species may produce different salmon run patterns. For example, steep
streams are likely to produce models with different slopes for salmon swimming upstream and
downstream. The segmented model structure can account for this pattern, and thus should
always be included in the candidate model set. Also, a polynomial model might be appropriate
for streams that experience high densities of spawners. In general, a polynomial model is needed
if time-lapse detection of passing salmon changes with salmon run intensity. For example, as
salmon reach high densities, they may not be able to move upstream very quickly because of
crowding. This could result in relatively higher detection at high run intensities. In this case, a
polynomial model would likely model the relationship better than a first order model.
Regardless of the model shape, it is important that users use standard model diagnostics and
good sense to fit the best model possible.
From four years of testing this method on different streams and different sites within
streams we have learned several important lessons. First, this counting system is most accurate
and requires the least effort when located where flow is rapid but the water surface is smooth.
The rapid flow prevents salmon from loitering above the contrast panels (which can introduce
noise into the time-lapse counts), while the smooth water surface makes it easy to see passing
salmon. Second, this system works best in shallow streams. Deep streams (>1 m) were
problematic because salmon were more likely to swim at different depths, which caused their
outlines to overlap and made counting more difficult. It was also more difficult to light deep
streams at night. We found that our infrared lights did not light passing salmon adequately if
streams were more than one meter deep. Using conventional flood lights (visible light) solves
this problem; however, it negates the advantages of using IR lights, which is invisible to humans,
fishes, and most wildlife. Third, it is important to orient the camera away from the sun
29
(northward in the northern hemisphere), because otherwise the surface of the water reflects glare
towards the camera.
Although this new method increases the breadth of sites that can be monitored, it has
some limitations. As with other methods, the turbidity associated with high flow events can
make seeing passing salmon difficult or impossible. Fortunately, these events tend to be brief in
the small streams for which we designed this system. Also, it can be difficult to distinguish
among species if a site has multiple species migrating at the same time. Finally, this system can
only monitor streams up to 15m wide. Beyond this width, counting accuracy is likely to
decrease as the salmon in the images become more distant. One potential solution is to use two
camera towers on opposite banks, each viewing one half of the stream.
Using this system, it can be difficult to accurately model the relationship between time-
lapse counts and salmon passage if escapement is less than two or three thousand salmon. This
is because at low escapement, hourly time-lapse counts tend to vary little, regardless of the
relative intensity of the run. This makes it difficult to effectively model the relationship between
time-lapse photo counts and video counts. One solution to this problem is to increase observer
effort by either increasing the length of the sampling unit (e.g. from one to two hours) or by
increasing the sampling frequency (e.g. 3-photo burst every 30 seconds). This would increase
the contrast between weak and strong runs, but also increase the time required to review photos
and/or video. Another solution is to use a model from a stream with similar features (width,
depth, velocity, etc.), although we know from the data presented here that models can differ
greatly among streams (Table 2). For example, if we had used the Southeast Creek model to
estimate Meadow Creek escapement, we would have overestimated by 89% compared to the
Meadow Creek top model.
30
In general, salmon researchers should strive to minimize their impact on natural salmon
behavior. In small streams such as those monitored here, spawning salmon tend to move up and
downstream frequently (Fig. 1.4, top), a behavior that may be a strategy for avoiding predators
(Bentley et al. 2014). Salmon monitoring methods such as weirs have the potential to limit these
movements. This could allow predators such as bears to catch salmon more easily, which could
decrease salmon spawning success rates and alter trophic interactions with salmon consumers. A
key strength of the method presented here is that it allows salmon to move freely and allows
natural interactions with salmon consumers.
As with many resources used by wildlife, salmon availability is very patchy in space and
time (Armstrong and Schindler 2011). This presents a challenge for researchers and managers
interested in using sampling to estimate their abundance; the more patchy or pulsed the salmon
run, the less accurate a random sampling method will be without large amounts of effort. Here,
we overcame this challenge by using a model-based design instead of a random sampling-based
design. This allowed us to relax the demands on our camera system; rather than requiring
complete video coverage, we merely needed hours of video that represented the full range of
salmon run intensities. Given the ubiquity of patchy (in space) or pulsed (in time) resource
availability, we suspect that this approach to double sampling could be usefully employed in a
variety of natural resources applications.
The salmon counting method that we present here expands the range of salmon spawning
habitats that can be realistically monitored. Compared to existing methods, our solution is less
expensive, less time consuming, and less detrimental to salmon and the wildlife that use them.
The data produced can help improve our understanding of how population dynamics at small
scales creates stability at the watershed scale. Lastly, due to their low cost and relative
31
portability, these systems would be ideal for monitoring salmon populations of conservation
concern. For example, they could produce baseline and ongoing data on the abundance of
salmon spawning downstream of mines or other resource development projects.
ACKNOWLEDGEMENTS
This work evolved from the early video enumeration efforts of Mat Sorum. His experimentation
with remote video systems greatly accelerated the development of the methods presented here.
We appreciate the staff at the Kodiak National Wildlife Refuge and Flathead Lake Biological
Station for their committed support and assistance for this project. We thank Caroline Cheung,
and volunteer field technicians Barbara Svoboda, Alex May, Bill Dunker, Tim Melham, Tyler
Tran, Marie Jamison, Louisa Pless, Francesca Cannizzo, Isaac Kelsey, Mark Melham, Jane
Murawski, Prescott Weldon, Shelby Flemming, Andy Orlando, and Kristina Hsu for their hard
work in the field and lab. We thank pilots Kurt Rees and Kevin VanHatten for their skilled
flying.
LITERATURE CITED
Akaike, H. 1974. A new look at the statistical model identification. IEEE Transactions on
Automatic Control 19:716–723.
Van Alen, B. W. 2008. Kook Lake sockeye salmon stock assessment, 2005-2007. Anchorage,
Alaska.
Anderson, T., and B. McDonald. 1978. A portable weir for counting migrating fishes in rivers.
Armstrong, J. B., and D. E. Schindler. 2011. Excess digestive capacity in predators reflects a life
of feast and famine. Nature 476:84–7.
Beacham, T. D., S. Cox-Rogers, C. MacConnachie, B. McIntosh, and C. G. Wallace. 2014.
Population Structure and Run Timing of Sockeye Salmon in the Skeena River, British
Columbia. North American Journal of Fisheries Management 34:335–348.
Bentley, K., D. Schindler, and J. Armstrong. 2012. Foraging and growth responses of stream-
dwelling fishes to inter-annual variation in a pulsed resource subsidy. Ecosphere 3.
Bentley, K. T., D. E. Schindler, T. J. Cline, J. B. Armstrong, D. Macias, L. R. Ciepiela, and R.
Hilborn. 2014. Predator avoidance during reproduction: diel movements by spawning
32
sockeye salmon between stream and lake habitats. The Journal of animal ecology:1478–
1489.
Bue, B. G., S. M. Fried, S. Sharr, D. G. Sharp, J. A. Wilcock, and H. J. Geiger. 1998. Estimating
salmon escapement using area-under-the curve, aerial observer efficiency, and stream-life
estimates: The Prince William Sound pink salmon example. NPAFC Bulletin 1:240–250.
Carlson, S. M., R. Hilborn, A. P. Hendry, and T. P. Quinn. 2007. Predation by bears drives
senescence in natural populations of salmon. PloS one 2:e1286.
Cochran, W. G. 1977. Sampling Techniques.
Cousens, N. B. F., G. A. Thomas, C. G. Swann, and M. C. Healey. 1982. A Review of Salmon
Escapement Estimation Techniques. Canadian Technical Report of Fisheries and Aquatic
Sciences.
Davies, T. D., D. G. Kehler, and K. R. Meade. 2007. Retrospective sampling strategies using
video recordings to estimate fish passage at fishways. North American Journal of Fisheries
Management 27:992–1003.
Efron, B., and R. J. Tibshirani. 2003. Introduction to the Bootstrap. Chapman and Hall.
Gregoire, T. G. 1998. Design-based and model-based inference in survey sampling: appreciating
the difference. Canadian Journal of Forest Research 28:1429–1447.
Grifantini, M. C., R. Teubert, R. Aschbacher, and M. Mitchell. 2011. Video Weir Technology
Pilot Project Final Project Report. Anderson, CA.
Hansen, M. H., W. G. Madow, and B. J. Tepping. 1983. An evaluation of model-dependent and
probability-sampling inferences in sample surveys. Journal of the American Statistical
Association 78:776–793.
Hatch, D. R., M. Schwartzberg, and P. R. Mundy. 1994. Estimation of Pacific Salmon
Escapement with a Time-Lapse Video Recording Technique. North American Journal of
Fisheries Management 14:626–635.
Hedges, L. V. 1981. Distributional theory for Glass’s estimator of effect size and related
estimators. Journal of Educational Statistics 6:107–128.
Holmes, J., G. Cronkite, H. Enzenhofer, and T. Mulligan. 2006. Accuracy and precision of fish-
count data from a “dual-frequency identification sonar” (DIDSON) imaging system. ICES
Journal of Marine Science 63:543–555.
Jeschke, J. M. 2007. When carnivores are “full and lazy”. Oecologia 152:357–64.
Kohavi, R. 1995. A Study of Cross-Validation and Bootstrap for Accuracy Estimation and
Model Selection. International Joint Conference on Artificial Intelligence 14:1137–1143.
Levi, T., R. E. Wheat, J. M. Allen, and C. C. Wilmers. 2015. Differential use of salmon by
vertebrate consumers: implications for conservation. PeerJ 3:e1157.
Maxwell, S. L., and N. E. Gove. 2007. Assessing a dual-frequency identification sonars’ fish-
counting accuracy, precision, and turbid river range capability. The Journal of the
Acoustical Society of America 122:3364–3377.
33
Neilson, J. D., and G. H. Geen. 1981. Enumeration of Spawning Salmon From Spawner
Residence Time and Aerial Counts. Transactions of the American Fisheries Society
110:554–556.
Quinn, T. P., L. Wetzel, S. Bishop, K. Overberg, and D. E. Rogers. 2001. Influence of breeding
habitat on bear predation and age at maturity and sexual dimorphism of sockeye salmon
populations. Canadian Journal of Zoology 79:1782–1793.
Schindler, D., J. Armstrong, K. Bentley, K. Jankowski, P. Lisi, and L. Payne. 2013. Riding the
crimson tide: mobile terrestrial consumers track phenological variation in spawning of an
anadromous fish. Biology Letters 9:2–6.
Schindler, D. E., R. Hilborn, B. Chasco, C. P. Boatright, T. P. Quinn, L. a Rogers, and M. S.
Webster. 2010. Population diversity and the portfolio effect in an exploited species. Nature
465:609–12.
Stephens, P. R., M. O. Kimberley, P. N. Beets, T. S. H. Paul, N. Searles, A. Bell, C. Brack, and
J. Broadley. 2012. Airborne scanning LiDAR in a double sampling forest carbon inventory.
Remote Sensing of Environment 117:348–357
34
CHAPTER 1 TABLES
Table 1.1. Comparison of salmon enumeration methods.
Method Typical sites Advantages Disadvantages Man-ned?
Refs
Rea
l Tim
e C
ou
nts
Weir
Large clear rivers/ streams
Easy sampling of age, sex, length, genetics
Expensive (equipment/personnel); May hinder natural fish movements
Yes Anderson and McDonald 1978
Observat-ion Tower
first to fifth order clear streams/ rivers
Does not hinder fish passage Expensive (personnel); turbulence or bad light can make counts difficult
Yes Cousens et al. 1982
Ret
rosp
ecti
ve C
ou
nts
Sonar Large clear or opaque rivers
Not affected by turbulence; Records of run can be saved and reviewed; Playback can be slowed and counts repeated for QA/QC; Does not obstruct fish passage
Expensive (equipment/personnel); Lengthy footage review; Accuracy suffers at highest densities
Yes Holmes et al. 2006; Maxwell and Gove 2007
Video net weir
medium to small rivers and streams
Records of run can be saved and reviewed; Playback can be slowed and counts repeated for QA/QC; Does not obstruct fish passage
Expensive(equipment/personnel); Lengthy footage review; May hinder natural movements of fish; Nets can catch debris
usually Van Alen 2008, Grifantini et al. 2011
Above water video
Medium to small clear streams
Records of run can be saved and reviewed; Playback can be slowed and counts repeated for QA/QC; Does not obstruct fish passage
Expensive; Time consuming footage review
varies Hatch et al. 1994
Time-lapse double sampling
Medium to small clear streams
Inexpensive; Can be left unattended for 14 days; Records of run can be saved and reviewed; Playback can be slowed and counts repeated for QA/QC; Does not obstruct fish passage; lower human presence decreases impacts on wildlife
Limited to smaller streams (<15m)
No Current study
35
Table 1.2. Model descriptions, escapement estimates and model validation metrics. AICc is
Aikaike’s Information Criterion adjusted for small sample size (Akaike 1974). 95% confidence
intervals on escapement were calculated using bootstrap resampling methods. Accuracy is the
percent difference between the leave-one-out cross-validation predicted escapement and the
actual escapement for the 70 hours where escapement was counted using video recording.
Precision is the mean squared error (MSE). The top model for each stream is in bold.
Model Name
Model of Sockeye Passage k AICc Escapement
estimate (±95% CI)
Accuracy Precision
Mea
do
w
segmented first order
pass=(x>0)*15.242+ (x<0)*18.920)
2 1534.303 30,509
(± 9,494) +3.0% 3556
segmented polynomial
pass =(x>0)*16.096+(x>0)2 * -0.0206+(x<0)*18.454+(x<0)2 *-0.0206
3 1535.021 30,064
(± 14,211) +10.1% 3557
first order pass =x*16.1441 1 1551.562 41,539
(± 3,692) +37.5% 3917
polynomial pass =x* 17.3137+
(x2 * -0.04552) 2 1535.943
31,830 (± 11,628)
+24.4% 3591
Sou
thea
st
segmented first order
pass=(x>0)*22.78+ (x<0)*22.18)
2 1733.43 68,253
(± 15,759) +7.2% 14932
segmented polynomial
pass=(x>0)*22.555+(x>0)2*0.0045+ (x<0)*22.653+(x<0)2 *-0.0045
3 1735.26 66,303
(± 23,052) +5.7% 15569
first order pass =x* 22.505 1 1732.17 65,355
(± 4,305) +3.1% 14167
polynomial pass =x*22.589 +
x2 * -0.0040) 2 1733.44
66,610 (± 8,045)
+6.2% 14669
36
CHAPTER 1 FIGURES
Figure 1.1. A) Map of southwest Kodiak Island, Alaska showing the locations of streams where
salmon were counted using time-lapse double sampling. B) Salmon counting system including
foldable steel tower holding a time-lapse camera (box at top), video camera, and solar panels.
The tower is surrounded by an electric fence to prevent equipment disruption by bears. White
high density polyethylene (HDPE) plastic panels were placed on the stream bed to improve
sightability of sockeye salmon (Oncorhynchus nerka). C) An image of two sockeye salmon
passing across the contrast panels, with the video camera in the foreground.
A B
C
37
Figure 1.2. Relationship between hourly time-lapse and video counts of salmon passage for two
streams in southwest Kodiak Island, Alaska. The lines show the top model for each stream,
selected using Akaike’s Information Criterion adjusted for small sample size (AICc)(Akaike
1974): a segmented first order relationship for Meadow Creek, and a simple first order linear
relationship for Southeast Creek. The segmented model (Meadow) has a different slope above
and below the origin, which is indicated by crossed vertical and horizontal dashed lines.
38
Figure 1.3. Comparison of estimates of the number of living sockeye salmon in two streams
(Meadow Creek at top, Southeast Creek at bottom) derived using four different models (model
details in Table 1).
39
Figure 1.4. Estimated hourly sockeye salmon passage (top), estimated cumulative passage
(middle), and estimated in-stream salmon abundance (bottom) in Meadow and Southeast Creeks.
In the salmon passage plots (top row), positive numbers indicate salmon moving into the stream
from the downstream lake, while negative numbers indicate salmon leaving the stream and
entering the lake.
40
Figure 1.5. Effect of altering mean and standard deviation (SD) of sockeye salmon survival (in
days) on the estimated in-stream salmon abundance in Meadow and Southeast Creeks. In the top
row the mean was manipulated, while the SD was altered in the bottom row of plots. In all plots,
the unaltered model is shown in black.
41
CHAPTER 2
KODIAK BROWN BEARS SURF THE SALMON RED WAVE: DIRECT
EVIDENCE FROM GPS COLLARED INDIVIDUALS
WILLIAM DEACY1, WILLIAM LEACOCK
2, JONATHAN B. ARMSTRONG3, JACK A. STANFORD
1
1Flathead Lake Biological Station, University of Montana, Polson, Montana 59860 USA
2U.S. Fish and Wildlife Service, 1390 Buskin River Road, Kodiak, AK 99615 USA
3University of Wyoming, Wyoming Cooperative Fish and Wildlife Research Unit, Laramie, Wyoming 82071 USA
ABSTRACT
A key constraint faced by consumers is achieving a positive energy balance in the face of
temporal variation in foraging opportunities. Recent work has shown that spatial heterogeneity in
resource phenology can buffer mobile consumers from this constraint by allowing them to track
changes in resource availability across space. For example, salmon populations spawn
asynchronously across watersheds, causing high quality foraging opportunities to propagate
across the landscape, prolonging the availability of salmon at the regional scale. However, we
know little about how individual consumers integrate across phenological variation or the
benefits they receive by doing so. Here, we present direct evidence that individual brown bears
track spatial variation in salmon phenology. Data from 40 GPS collared brown bears show that
bears visited multiple spawning sites in synchrony with the order of spawning phenology. The
number of sites used was correlated with the number of days a bear exploited salmon, suggesting
the phenological variation in the study area influenced bear access to salmon, a resource which
strongly influences bear fitness. Fisheries managers attempting to maximize harvest while
maintaining ecosystem function should strive to protect the population diversity that underlies
the phenological variation used by wildlife consumers.
42
Key words: Kodiak; brown bear; salmon; GPS collar; resource wave; phenological
tracking; portfolio effect; resource subsidy; grizzly bear; sockeye; landscape; foraging.
INTRODUCTION
One of the central themes in ecological theory is that biodiversity enhances and stabilizes
ecosystem services (Tilman et al. 1996, Kennedy et al. 2002, Hooper et al. 2005). While most
biodiversity research and conservation efforts have focused on species diversity, finer levels of
biodiversity (i.e., intraspecific diversity) are far more threatened—for example extinction rates for
populations are roughly 1000-times higher than those for species (Hughes et al. 1997). Thus, a
critical challenge in ecology is to understand the functional significance of intraspecific diversity.
There has been recent interest in the potential for intraspecific variation to generate
“portfolio effects”, in which asynchronous dynamics among populations have emergent
properties expressed at higher levels of biological organization (Schindler et al. 2015). For
example, asynchrony in the population dynamics of sockeye salmon (Oncorhynchus nerka)
dampens levels of temporal variation expressed across the aggregate of populations. This can be
seen in sockeye salmon stock-complexes where the boom of one population compensates for the
bust in another, resulting in more stable commercial fisheries harvests (Schindler et al. 2010).
Asynchrony among populations occurs not only in the interannual trends of abundance, but also
in the intraannual timing of life-cycle events (i.e., phenology). For example, populations that
occur in different habitats may exhibit different seasonal patterns of birth, migration, and
reproduction, often due to local adaptation. There is increasing interest in whether phenological
asynchrony among populations (or other scales of biological organization) can generate
ecologically significant emergent properties. For example, an accumulating body of evidence
43
shows that asynchronous phenology among prey resources can have strong positive effects on
wide-ranging consumers by triggering resource waves (Armstrong et al., in press).
Resource waves are important when a prey species is only available (or is of high quality)
during a specific developmental stage and its phenology varies across prey subpopulations
(variation at other levels of biological organization may also cause resource waves). For
example, migrating ungulates and waterfowl take advantage of spatial variation in the timing of
spring vegetation growth (the so called “green wave”) in order to consume high quality forage
for a longer period than is available at a single foraging site (Sawyer and Kauffman 2011, van
Wijk et al. 2012). Similarly, surf scoters (Melanitta perspicillata) and rainbow trout (O. mykiss)
track spatial variation in spawning phenology of herring (Clupea pallasi) and sockeye salmon
(O. nerka), respectively, to extend their access to energy dense eggs (Ruff et al. 2011, Lok et al.
2012). Although there is rapidly increasing interest in this topic, we often lack data to address
how individuals track resources. Commonly, tracking is inferred from consumer distributional
data (Fryxell et al. 2004, Lok et al. 2012, Schindler et al. 2013) or assumed based on the
existence of a resource wave (Coogan et al. 2012). Using these methods, it is difficult to
determine whether changes in consumer distribution and abundance are due to individuals
aggregating around a local resource (only using a single prey subpopulation) or individuals
tracking resources across the landscape (using several prey populations). Individual movement
data is needed to provide conclusive evidence of resource tracking and to directly quantify the
functional significance of resource waves to consumers.
Populations of spawning Pacific salmon (Oncorhynchus spp.) provide an example of how
population diversity can prolong the temporal extent of prey availability across landscapes
(Schindler et al. 2010, 2013, Ruff et al. 2011). Salmon breeding phenology is related to
44
freshwater thermal regimes that vary spatially due to heterogeneity in geomorphology and
hydrology (Lisi et al. 2013). Across an aggregate of salmon populations, spawning activity often
spans several months, however, each individual population may only spawn for as little as two to
three weeks (Gende et al. 2004, Carlson et al. 2007, Schindler et al. 2010). These brief periods of
salmon spawning are spread across space and through time, creating resource waves that
potentially benefit mobile consumers, however, the actual benefit depends on the degree to
which mobile consumers can track the shifting mosaic of salmon resources.
Of the large number of predators and scavengers that feed on seasonally available
spawning salmon (Shardlow and Hyatt 2013), brown bears (Ursus arctos) are perhaps the most
iconic and have a well-documented dependence on salmon; fecundity, body size, and population
density are all strongly correlated with salmon consumption (Hilderbrand et al. 1999). Given the
importance of salmon to bears, their keen sensory abilities, and their mobility, one would expect
them to be highly capable of tracking spatiotemporal variation in salmon abundance across
landscapes. Schindler et al. (2013) revealed strongly suggestive evidence that bears surf salmon
resource waves and the potential for this behavior to prolong foraging opportunities for bears.
However, no direct evidence exists nor do we understand the degree to which individual bears
track salmon, or how much individual variation exists in tracking behavior. In this paper we : 1)
quantify the salmon resource wave; 2) track individual bear movements in relation to the wave;
and 3) quantify the degree to which individual bears extend their foraging opportunities by
surfing the resource wave. We provide the first direct evidence of bears tracking salmon
phenology and show that salmon phenological diversity prolongs the duration of bear foraging
opportunities by an average of 1.7 times.
METHODS
45
Study site
This work was conducted in southwestern Kodiak Island, in the western Gulf of Alaska
(Fig. 2.1). The Kodiak Archipelago has an estimated population of 3500 brown bears, hundreds
of rivers, lake shoals, and streams used by spawning Pacific salmon (Oncorhynchus spp.), and
limited human activity. The area has a rich history of bear–salmon research (Gard 1971, Barnes
1990, Van Daele et al. 2013). Barnes (1990) showed the home ranges of bears in southwest (SW)
Kodiak often overlap multiple drainages and many salmon spawning sites, providing the first
evidence that individual bears may exploit multiple salmon populations. The majority of the SW
portion of the island is within the Kodiak National Wildlife Refuge, which is managed by the US
Fish and Wildlife Service. Human activity in the study area is limited and consists primarily of
sport fishers, bear viewers, and hunters. The bears on the Kodiak archipelago are hunted during
the fall and spring each year. Approximately 190 bears were harvested annually from 2000–2009.
Although five species of salmon spawn in SW Kodiak waters, sockeye and pink salmon
(Oncorhynchus gorbuscha) are the most abundant (Van Daele et al. 2013). From 2000–2009,
over half of the salmon returns for the Kodiak Archipelago occurred in the SW region, with an
average escapement (fish remaining after harvest) of over 3.2 million. Pink salmon spawn
primarily in main stem rivers and estuaries at the mouths of rivers. Sockeye salmon spawn
mainly in headwater streams, on lake beaches with interstitial flow of groundwater and in lake-
outlet rivers. Most of the stream habitats are narrow (<5 m), shallow (<0.5 m) and flow into
lakes, rivers, or directly into the ocean. Sockeye juveniles typically rear in lakes downstream of
tributary spawning streams. Four large salmon-producing, stream-lake systems exist in the study
area: Karluk, Red, Akalura, and Frazer. Preliminary results from ongoing genetic studies have
detected population genetic differences in spawning sockeye salmon at the level of habitat types
46
within a watershed (i.e., river, lake shore, tributary stream), but not within a habitat type (e.g.,
tributary streams within the Karluk watershed) ( Jeff Olson, personal communication).
In addition to salmon, bears routinely consume several species of berries, including red
elderberry (Sambucus racemosa L.), salmonberry (Rubus spectabilis Pursh), crowberry
(Empetrum nigrum L.) and blueberry (Vaccinium spp.) and many species of grasses, sedges, and
forbs (Van Daele et al. 2013).
Variation in timing of salmon availability for bears
We used the Alaska Anadromous Streams Catalog (AASC) and field observations to
identify the water bodies in SW Kodiak where bears have access to salmon. We found 7 rivers
and 68 streams listed in the AASC as salmon spawning habitat. The AASC does not list beach
spawning sites; we identified 19 beach sites where sockeye salmon spawn through weekly aerial
surveys. In addition to spawning sites, we included one site where a salmon-passable cataract
called Dog Salmon Falls makes migrating salmon vulnerable to bear predation. These 95 sites
include all of the sites where bears can access salmon within the study area, however, bear
telemetry data suggests that only a subset of these sites are regularly visited by bears. We
characterized the average spawning phenology at 32 of these sites using nine years of aerial,
boat, and ground observations (William Leacock, unpublished data). The order of salmon
availability we observed among habitat types (the falls, lake-tributary streams, lake-outlet rivers,
lake beaches) matched the patterns documented in similar systems driven by water temperature
variation (Doctor et al. 2010, Schindler et al. 2010).
Bear movements in relation to salmon abundance
A challenge in many behavioral studies is to infer foraging behavior from movement and
habitat use when data on trophic resources are not available across the entire landscape.
47
Although we could not determine the timing of salmon availability at every site across the
landscape, we characterized salmon phenology data across a large number of spawning sites
(n = 32; 34% of all sites). Inferring bear foraging opportunity from movement behavior would be
problematic if bears resided at salmon spawning sites for purposes other than salmon foraging.
To test whether this was the case, we monitored streams with remote cameras and evaluated how
bear presence responded to salmon abundance. In 2013, we deployed 1–3 time lapse trail
cameras (Reconyx PC800) along six streams in the study area. The cameras were programmed to
take a photo every 5 minutes, 24 hours/day from June through September. We counted the
number of bears in each time lapse frame, counting sows with cubs as a single independent bear.
Similar to the results of Schindler et al. (2013) and Quinn et al. (2014), the peak spawning date at
each site was positively correlated with the median date of bear detections (R2 = 0.33, Appendix
A). This indicated that SW Kodiak bears responded to seasonal changes in salmon availability.
Most importantly, bears were virtually absent (0.6 +/- 1.8 bear detectionsday-1 stream-1 (mean
+/- standard deviation (s))) when salmon were not spawning, but became ephemerally super
abundant (36.6 +/- 66.8 detectionsday-1 stream-1) during the salmon run. These data, in
addition to prior studies (Schindler et al. 2013, Shardlow and Hyatt 2013) confirm that it is
reasonable to assume that (1) bears present at spawning salmon sites were foraging on salmon
and (2) the number of days spent at salmon sites accurately reflected the duration of salmon
foraging opportunities for bears. To account for the occasional use of salmon spawning habitats
as movement corridors, we differentiated between bear passage and residence by considering
individuals to be exploiting salmon only when they exhibited GPS locations within 50 m of a
salmon spawning site at least twice a day for at least five days in a year. For each bear, we
48
calculated the number of salmon spawning sites attended and the total number of days spent
foraging on salmon.
Movements of collared bears
Seasonal changes in bear distribution (Appendix A, Fig. A1) may be due to local bears
aggregating at a nearby spawning site (only using a single salmon subpopulation) or individuals
tracking salmon spawning phenology across the landscape (using multiple salmon populations).
Distributional data cannot distinguish between these scenarios nor quantify the functional
significance of salmon resource waves to bears; therefore, we collected movement data from
individual bears using GPS collars.
We captured adult female brown bears in the SW region of Kodiak Island, Alaska by
firing immobilization darts from a helicopter. We fitted each bear with a GPS radio collar
programmed to record a location every hour from early June through mid-November. Collars
contained a UHF (ultrahigh frequency) transmitter and were downloaded using an airplane fitted
with a UHF receiver. From 2008 to 2014, 143 284 GPS locations were recorded from 43
individuals over 67 bear-years (some bears carried collars for more than one year). We screened
GPS locations for accuracy, removing relocations with a positional dilution of precision (PDOP)
greater than 10 (Lewis et al. 2007). We excluded bears from the analysis if their collars failed
before acquiring at least 1500 relocations in a year. Following these quality control measures,
133 085 relocations from 52 bear-years and 40 unique bears remained for analysis.
To determine the order of habitat use for each bear, we first produced empirical
cumulative distribution functions (ECDF) for each habitat and each bear. Next, we used the
median date of each ECDF to determine the first and last habitat used by each of the bear-years
where a bear used more than one habitat (N=41). Finally, we tabulated these values in a
49
contingency table and used a chi-squared test of independence (α=.05) to determine whether the
pattern of habitat visit order was random (H0) or not (Ha).
RESULTS
Salmon were available to bears (i.e., on spawning grounds or migrating past the
waterfall) at different times in different habitats. Median occupancy date for the waterfall,
tributary streams, lake-outlet rivers, and lake beaches was 14 July, 3 August, 23 September, and
23 October, respectively (Fig. 2.2). Most of the sites visited by bears were salmon spawning
grounds, however, salmon availability to bears was further prolonged by point habitat features
that made fish vulnerable to predation. At the Lower Falls of the Dog Salmon River, a small
waterfall where bears intercept salmon as they migrate upriver, salmon were available as early as
3 June. Thus, habitat heterogeneity and phenological diversity of salmon prolonged their duration
of availability to bears from approximately 40 days for a single stock, to roughly 150 days for the
aggregate.
We documented considerable variation in the number of spawning populations exploited
by collared female bears. On average, each female bear exploited 3.1 populations of spawning
salmon in a year (median = 3.0, n = 52, s = 1.5, Fig. 2.4a). The maximum used by a single bear
was 7 sites, while one bear used no salmon sites. In general, the order in which bears visited
spawning sites matched the sequence of salmon run timing (Fig. 2.3a, b); bears tended to visit
habitats with early salmon availability first (falls and streams) and habitats with late availability
last (river and lake beaches, χ2 = 31.7, n = 41, P < 0.0001, Appendix A). Furthermore, the
median date that individual bears used the habitat with the latest availability (lake beaches) was
48 days later than the median date they used the site with earliest salmon availability (the Lower
Dog Salmon Falls).
50
The mean number of days each bear exploited salmon was 67 (n = 52, s = 33.5), whereas
the average spawning population was only available for approximately 40 days. Seventy-three
percent of bears spent more than 40 days fishing for salmon. Regression analysis indicates the
number of spawning populations exploited was positively correlated with the number of days
each bear fished (Fig. 2.4b, R2 = 0.36, P < 0.0001).
DISCUSSION
Although each individual subpopulation spawned for a brief period (~40 days), spawning
activity spanned several months across all of the salmon subpopulations. The timing of salmon
availability varied by habitat: salmon first appeared while migrating past waterfalls, then while
spawning in streams, rivers and, finally, lake beaches. Counts of bears from time lapse images
showed that bears were unlikely to be detected at streams when salmon were not spawning (0.6
detections/day) compared to when they were spawning (31 detections/day). Given this pattern,
we used GPS relocations from collared female bears to indicate bear foraging behavior. These
data showed the number of sites used by bears varied from zero to seven (mean = 3.1, s = 1.5)
and they tended to visit sites in their order of availability, using the falls (available in June/July)
an average of 48 days earlier than lake beaches (available Sept./Oct.). Although spawning
salmon were only available at individual sites for ~40 days, bears foraged for an average of 67
days, 1.7x longer than if there was no variation in run timing. Ruff et al. (2011) documented a
similar effect; rainbow trout in their study had access to salmon 1.5x longer due to phenological
variation among salmon populations. The degree to which our collared bears moved among sites
correlated with their access to salmon: as bears increased the number of sites they attended, they
significantly prolonged their access to salmon (R2 = 0.36, P < 0.0001). Given that a bear’s
consumption of abundant prey such as salmon is limited by duration of access (because of
51
digestive constraints on foraging rates) rather than merely abundance, our results strongly
suggest that bears directly benefit from salmon life history diversity. Because we only studied
the foraging habits of female bears, the results of this study are likely conservative; females have
smaller home ranges than males, particularly when they have cubs (Berns et al. 1980), and their
smaller body sizes make them less dependent on high calorie foods such as salmon (Welch et al.
1997, Rode et al. 2001).
Population diversity in salmon and the corresponding asynchrony in spawn timing
increases the duration of salmon availability for bears. In addition, physical features along
salmon migration routes, such as waterfalls or cataracts, can extend the life history phases in
which salmon are vulnerable to include not only spawning, but also migration to upstream
spawning sites. While point features (e.g., McNeil and Brooks Falls, Alaska) are recognized as
important because they make salmon vulnerable to bears in large rivers where they are otherwise
inaccessible (Quinn et al. 2001, Peirce et al. 2013), their significance in regards to timing are
much less appreciated. In the Karluk system, salmon were available at the Lower Falls of the
Dog Salmon River almost a month before spawners in streams.
Loss of life history diversity has the potential to erode the ecosystem services important
to humans. Schindler et al. (2010) simulated the effects of loss of population diversity on the
reliability of commercial fishing harvests and found that population homogenization would result
in ten times more frequent fisheries closures. Our results indicate that loss of population diversity
would also impact wildlife consumers such as bears: the average bear in our study would have
48% less time to consume salmon if all of the salmon in our study area spawned at the same
time. A challenge for fisheries management is to conserve diversity at the population level while
managing harvest at coarser levels (i.e., watersheds consisting of dozens of populations).
52
Population diversity is not explicitly considered in the maximum sustained yield paradigm of
salmon fisheries management, yet it clearly mediates the long-term reliability of fisheries
(Hilborn et al. 2003, Schindler et al. 2010) and likely the energy flows from fish to consumer
species in freshwater food webs (Ruff et al. 2011, Schindler et al. 2013). Many salmon fisheries
are temporally biased, substantially increasing harvest rates once escapement goals are met
(Quinn et al. 2007). Given evidence for population-level variation in salmon migration
phenology (Boatright et al. 2004, Doctor et al. 2010, McGlauflin et al. 2011), temporally biased
fisheries may diminish population diversity by selecting against stocks with late migration
phenologies (Quinn et al. 2007), which are likely associated with late spawning phenologies and
thus availability to bears (Boatright et al. 2004, Doctor et al. 2010).
Given the well-documented benefits of salmon consumption, it is interesting that several
bears (23% of bear-years) used salmon for fewer than forty days and one bear was never
relocated within 50 m of a salmon site. An earlier study on Kodiak Archipelago,(Van Daele et al.
2013), found that salmon accounted for an average of 48% of assimilated diets of adult female
bears and 16% of females had diets consisting of less than 10% salmon (based on stable isotopes
and mercury analysis). Some bears may eat few salmon because salmon availability varies across
the study area. In some areas, a bear could attend multiple spawning sites with only short
movements, while in others the costs of moving among sites are greater. It may also be a result
of intraspecific competition at salmon sites; due to higher bear densities, there is a heightened
risk of aggressive encounters (Gende and Quinn 2004) and infanticide for sows with cubs (Ben-
David et al. 2004). This may cause some bears to eschew salmon for less energy-dense, but less-
risky, foods such as vegetation or berries.
53
Researchers have noted the amount of salmon consumed by bears varies by sex, age, and
maternal status, with dominant males consuming the most salmon and subdominant bears the
least (Van Daele et al. 2013). This may be due to allometric scaling between body mass and
nutritional requirements (Welch et al. 1997, Rode et al. 2001), but likely also reflects the
tendency for dominant bears to exclude less dominant bears from preferred salmon foraging sites
(Gende and Quinn 2004). Although bears adopt strategies to limit competitive interactions at
spawning sites, for example, by partitioning use across space and through time (Nevin and
Gilbert 2005), competition may be reduced further when several populations of salmon are
spawning at the same time in multiple locations. Thus, while phenological diversity increases the
duration of salmon access for bears, this benefit may only be realized by the most dominant
bears unless salmon are spawning across a sufficiently large area to limit competition. In this
context, it is not surprising that 73% of tracked bears used at least one stream site, while only
10% used the Lower Falls, the site that provides the earliest access to salmon.
The SW Kodiak Island study site has limited human development and recreational
activity. In many other parts of Alaska, landscapes face increasing pressure for resource and
infrastructure development. Recent evidence suggests that such habitat alteration often results in
permeable barriers that may maintain habitat connectivity, yet interfere with the ability of
consumers to track resource waves (Sawyer et al. 2013). Bears in the most productive
populations often rely on salmon for the majority of their annual energy intake (Hilderbrand et
al. 1999). Our results suggest that tracking of phenologically diverse salmon populations plays
an important role in allowing bears to acquire energy from ephemeral salmon resources. Human
actions that reduce salmon population diversity or inhibit bear movements reduce the potential
for bears to eat salmon, which would likely decrease bear population productivity (Hilderbrand
54
et al. 1999). The corollary for salmon restoration efforts is that restoring salmon abundance with
homogenous hatchery stocks, in heavily fragmented landscapes, is unlikely to restore the
functional link between salmon and culturally, commercially, and ecologically important
consumers such as brown bears.
ACKNOWLEDGEMENTS
W. Deacy and J. Stanford were supported by the Jessie M. Bierman Professorship, J.
Armstrong was supported by the David H. Smith Conservation Research Fellowship, and W.
Leacock was supported by the USFWS. Funding for fieldwork was provided by the USFWS
Inventory and Monitoring and Refuge Programs. We appreciate the staff at the Kodiak National
Wildlife Refuge and Flathead Lake Biological Station for their committed support and
assistance. We thank Mat Sorum, Caroline Cheung, and volunteer field technicians Barbara
Svoboda, Alex May, Bill Dunker, Tim Melham, Tyler Tran, Marie Jamison, Louisa Pless,
Francesca Cannizzo, Isaac Kelsey, Mark Melham, Jane Murawski, Prescott Weldon, Shelby
Flemming, Andy Orlando, and Kristina Hsu for assisting with data collection. We thank
helicopter pilot Joe Fieldman and fixed-wing pilots Kurt Rees, Kevin Vanhatten, Kevin Fox, and
Issac Beddingfield for their skilled flying during bear captures and aerial telemetry. Thanks also
to Marie Kohler for her assistance in preparing and submitted this manuscript. Comments by
Mark Hebblewhite, Kate Searle, and one anonymous reviewer improved this manuscript. This
work is dedicated to the memory of Tip Leacock, who we miss dearly.
LITERATURE CITED
Armstrong, J. B., G. T. Takimoto, D. E. Schindler, M. M. Hayes, and M. J. Kauffman. In press.
Resource waves: phenological diversity enhances foraging opportunities for mobile
consumers. Ecology.
Barnes, V. G. 1990. The influence of salmon availability on movements and range of brown
bears on Southwest Kodiak Island. Bears: Their Biology and Management 8.
55
Ben-David, M., K. Titus, and L. R. Beier. 2004. Consumption of salmon by Alaskan brown
bears: a trade-off between nutritional requirements and the risk of infanticide? Oecologia
138:465–74.
Berns, V., G. Atwell, and D. Boone. 1980. Brown bear movements and habitat use at Karluk
Lake, Kodiak Island. Bears: Their Biology and Management 4.
Boatright, C., T. Quinn, and R. Hilborn. 2004. Timing of adult migration and stock structure for
sockeye salmon in Bear Lake, Alaska. Transactions of the American Fisheries Society
133:911–921.
Carlson, S. M., R. Hilborn, A. P. Hendry, and T. P. Quinn. 2007. Predation by bears drives
senescence in natural populations of salmon. PloS one 2:e1286.
Coogan, S. C. P., S. E. Nielsen, and G. B. Stenhouse. 2012. Spatial and temporal heterogeneity
creates a “Brown Tide” in root phenology and nutrition. ISRN Ecology 2012:1–10.
Doctor, K. K., R. Hilborn, M. Rowse, and T. Quinn. 2010. Spatial and temporal patterns of
upriver migration by sockeye salmon populations in the Wood River System, Bristol Bay,
Alaska. Transactions of the American Fisheries Society 139:80–91.
Fryxell, J., J. Wilmshurst, and A. Sinclair. 2004. Predictive models of movement by Serengeti
grazers. Ecology 85:2429–2435.
Gard, R. 1971. Brown bear predation on sockeye salmon at Karluk Lake, Alaska. The Journal of
Wildlife Management 35:193–204.
Gende, S., and T. Quinn. 2004. The relative importance of prey density and social dominance in
determining energy intake by bears feeding on Pacific salmon. Canadian Journal of Zoology
85:75–85.
Gende, S., T. Quinn, and R. Hilborn. 2004. Brown bears selectively kill salmon with higher
energy content but only in habitats that facilitate choice. Oikos 104:518–528.
Hilborn, R., T. P. Quinn, D. E. Schindler, and D. E. Rogers. 2003. Biocomplexity and fisheries
sustainability. Proceedings of the National Academy of Sciences USA 100:6564–6568.
Hilderbrand, G. V, C. C. Schwartz, C. T. Robbins, M. E. Jacoby, T. A. Hanley, S. M. Arthur, and
C. Servheen. 1999. The importance of meat, particularly salmon, to body size, population
productivity, and conservation of North American brown bears. Canadian Journal of
Zoology 77:132–138.
Hooper, D. U., F. S. Chapin, J. J. Ewel, A. Hector, P. Inchausti, S. Lavorel, J. H. Lawton, D. M.
Lodge, M. Loreau, S. Naeem, B. Schmid, H. Setälä, A. J. Symstad, J. Vandermeer, and D.
A. Wardle. 2005. Effects of biodiversity on ecosystem functioning: A consensus of current
knowledge. Ecological Monographs 75:3-35.
56
Hughes, J., G. Daily, and P. Ehrlich. 1997. Population diversity: its extent and extinction.
Science 278:689–692.
Kennedy, T. A, S. Naeem, K. M. Howe, J. M. H. Knops, D. Tilman, and P. Reich. 2002.
Biodiversity as a barrier to ecological invasion. Nature 417:636–8.
Lewis, J. S., J. L. Rachlow, E. O. Garton, and L. a. Vierling. 2007. Effects of habitat on GPS
collar performance: using data screening to reduce location error. Journal of Applied
Ecology 44:663–671.
Lisi, P. J., D. E. Schindler, K. T. Bentley, and G. R. Pess. 2013. Association between
geomorphic attributes of watersheds, water temperature, and salmon spawn timing in
Alaskan streams. Geomorphology 185:78–86.
Lok, E., D. Esler, J. Takekawa, S. De La Cruz, W. Boyd, D. Nysewander, J. Evenson, and D.
Ward. 2012. Spatiotemporal associations between Pacific herring spawn and surf scoter
spring migration: evaluating a “silver wave” hypothesis. Marine Ecology Progress Series
457:139–150.
McGlauflin, M. T., D. E. Schindler, L. W. Seeb, C. T. Smith, C. Habicht, and J. E. Seeb. 2011.
Spawning habitat and geography influence population structure and juvenile migration
timing of sockeye salmon in the Wood River Lakes, Alaska. Transactions of the American
Fisheries Society 140:763–782.
Nevin, O. T., and B. K. Gilbert. 2005. Perceived risk, displacement and refuging in brown bears:
positive impacts of ecotourism? Biological Conservation 121:611–622.
Peirce, J., E. Otis, M. Wipfli, and E. Follmann. 2013. Interactions between brown bears and
chum salmon at McNeil River, Alaska. Ursus 24:42–53.
Quinn, T., S. Hodgson, and L. Flynn. 2007. Directional selection by fisheries and the timing of
sockeye salmon (Oncorhynchus nerka) migrations. Ecological Applications 17:731–739.
Quinn, T. P., L. Wetzel, S. Bishop, K. Overberg, and D. E. Rogers. 2001. Influence of breeding
habitat on bear predation and age at maturity and sexual dimorphism of sockeye salmon
populations. Canadian Journal of Zoology 79:1782–1793.
Quinn, T. P., A. J. Wirsing, B. Smith, C. J. Cunningham, and J. Ching. 2014. Complementary
use of motion-activated cameras and unbaited wire snares for DNA sampling reveals diel
and seasonal activity patterns of brown bears (Ursus arctos) foraging on adult sockeye
salmon (Oncorhynchus nerka). Canadian Journal of Zoology 92:893–903.
Rode, K., C. Robbins, and L. Shipley. 2001. Constraints on herbivory by grizzly bears.
Oecologia 128:62–71.
Ruff, C. P., D. E. Schindler, J. B. Armstrong, K. T. Bentley, G. T. Brooks, G. W. Holtgrieve, M.
57
T. McGlauflin, C. E. Torgersen, and J. E. Seeb. 2011. Temperature-associated population
diversity in salmon confers benefits to mobile consumers. Ecology 92:2073–84.
Sawyer, H., and M. J. Kauffman. 2011. Stopover ecology of a migratory ungulate. The Journal
of Animal Ecology 80:1078–87.
Sawyer, H., M. J. Kauffman, A. D. Middleton, T. A. Morrison, R. M. Nielson, and T. B.
Wyckoff. 2013. A framework for understanding semi-permeable barrier effects on
migratory ungulates. Journal of Applied Ecology 50:68–78.
Schindler, D., J. Armstrong, K. Bentley, K. Jankowski, P. Lisi, and L. Payne. 2013. Riding the
crimson tide: mobile terrestrial consumers track phenological variation in spawning of an
anadromous fish. Biology Letters 9: 20130048.
Schindler, D. E., J. B. Armstrong, and T. E. Reed. 2015. The portfolio concept in ecology and
evolution. Frontiers in Ecology and the Environment 13:257–263.
Schindler, D. E., R. Hilborn, B. Chasco, C. P. Boatright, T. P. Quinn, L. A. Rogers, and M. S.
Webster. 2010. Population diversity and the portfolio effect in an exploited species. Nature
465:609–612.
Shardlow, T. F., and K. D. Hyatt. 2013. Quantifying associations of large vertebrates with
salmon in riparian areas of British Columbia streams by means of camera-traps, bait
stations, and hair samples. Ecological Indicators 27:97–107.
Silverman, B. W. 1986. Density estimation for statistics and data analysis. Monographs on
Statistics and Applied Probability. Chapman and Hall, New York, New York, USA.
Tilman, D., D. Wedin, and J. Knops. 1996. Productivity and sustainability influenced by
biodiversity in grassland ecosystems. Nature 379:718–720.
Welch, C., J. Keay, K. Kendall, and C. Robbins. 1997. Constraints on frugivory by bears.
Ecology 78:1105–1119.
Van Daele, M. B., C. T. Robbins, B. X. Semmens, E. J. Ward, L. J. Van Daele, and W. B.
Leacock. 2013. Salmon consumption by Kodiak brown bears (Ursus arctos middendorffi)
with ecosystem management implications. Canadian Journal of Zoology 91:164–174.
van Wijk, R. E., A. Kölzsch, H. Kruckenberg, B. S. Ebbinge, G. J. D. M. Müskens, and B. A.
Nolet. 2012. Individually tracked geese follow peaks of temperature acceleration during
spring migration. Oikos 121:655–664.
58
CHAPTER 2 FIGURES
FIG. 2.1. Map of study area on southwest Kodiak Island, Alaska. There are 95 water bodies in
this area used by spawning Pacific salmon (Oncorhynchus spp.). These sites are colored by
habitat category, each of which corresponds with a different periods of salmon availability to
bears. Salmon availability in streams, rivers, and lakes occurs during salmon spawning while
availability at falls occurs during salmon migration. All of these sites are assumed to be available
to bears in the study area.
Stream
Lake Beach
Falls
River
59
FIG. 2.2. Dates of salmon availability in four aquatic habitats. Inset shows salmon spawning
phenology of seven streams within the study area. Solid lines indicate periods with salmon in all
years, while dotted lines indicate less frequent salmon observations. Salmon are available at a
single site for an approximately 40 days while overall availability spans at least 150 days.
60
FIG. 2.3. Seasonal use of salmon spawning/migration sites by GPS collared bears as a function
of habitat type. A) Location data pooled across individuals and grouped by habitat type. Data
were smoothed using a kernel density estimator with a bandwidth of 7.97, which was arbitrarily
selected because it highlights the general pattern in bear habitat use (Silverman 1986). B)
Individual timelines of bear use of salmon spawning/migration sites. Each row corresponds to at
bear-year. Colors indicate habitat class attended each day, whereas the absence of any marker
indicates periods where bears were not attending salmon sites.
61
FIG. 2.4. A) Histogram of the number of spawning populations exploited by GPS collared bears.
Median = 3.0 populations. B) Number of days collared female bears ate salmon as a function of
the number of salmon populations exploited. Simple linear regression shown; P < 0.0001,
R2 = 0.36. Red dashed line corresponds with the maximum length of time a bear could eat salmon
if there was no phenological variation among salmon populations.
62
APPENDIX A
Appendix A: Fig. A1. The relationship between salmon spawning phenology and bear activity
across 6 streams in 2012. Peak bear activity is defined as the median of the cumulative
distribution function (CDF) of bear detections, whereas peak salmon is the date of highest
observed abundance. R2 = 0.3262.
Appendix A: Table A1. Contingency table of habitats visited first and last by the 41 bears that
visited at least 2 habitats.
7/19
7/20
7/21
7/22
7/23
7/24
7/25
7/26
7/27
6/30 7/5 7/10 7/15 7/20 7/25 7/30
Pe
ak B
ear
Act
vity
Peak Salmon Spawning
Visited
First
Visited
Last
Falls 5 0
Stream 25 5
River 6 21
Lake 5 15
63
CHAPTER 3
KODIAK BROWN BEARS SELECT BERRIES DESPITE HIGH SALMON
AVAILABILITY: AN OPPORTUNISTIC TEST OF THE MACRONUTRIENT
OPTIMIZATION HYPOTHESIS IN WILD BEARS
ABSTRACT
In Kodiak, Alaska, in both 2014 and 2015, we observed periods of very low brown bear (Ursus
arctos middendorffi) activity along streams filled with spawning sockeye salmon (Oncorhynchus
nerka). The common explanation was bears were consuming seasonally available red elderberry
(Sambucus racemosa). This seems maladaptive from an energy maximization perspective
because large bears face constraints which make it difficult to gain weight eating berries, and
salmon are an energy rich food. The macronutrient optimization hypothesis (MOH) is a more
nuanced view of diet selection that takes dietary costs into account. This is important, because
certain foods require more energy to digest than others, and this energetic cost decreases the net
energy gained by an animal. Captive bears mix foods to achieve a diet where protein provides
17 ± 4% of the metabolizable energy. Since protein provides ~65% and ~16.5% of
metabolizable energy in salmon and red elderberries, respectively, the MOH provides a potential
explanation for the bear distribution patterns we observed in 2014 and 2015. We used bear
distributions from three years with natural variation in elderberry phenology to see whether wild
bear foraging patterns supported MOH predictions. Elderberry phenology was relatively early in
2014 and 2015, overlapping the second half of the salmon runs, whereas the elderberry crop and
salmon were separate in time in 2013. In both 2014 and 2015, bear detection along streams
dropped when elderberries became ripe, while in 2013 bear activity was more synchronous with
salmon abundance. During the lull in bear activity on streams, collared bears were using
elderberry habitat. Together, these data suggest wild bears facing real-world foraging constraints
forage according to the MOH. Although bears preferred berries to salmon, salmon were
available for much longer, and likely contribute more to brown bear annual energy budgets.
INTRODUCTION
Organisms often depend on resources that vary greatly in abundance across space and through
time. Despite this, organisms have many strategies for meeting their energy needs, including
regulating digestive tract size to match food availability (Armstrong and Schindler 2011),
tracking spatial variation in resource phenology (Deacy et al. in press), caching foods when they
are abundant for later consumption (Dearing 1997), and hibernating during food scarcity
64
(Humphries et al. 2003). Omnivory can be viewed as another coping strategy; by increasing
their dietary breadth, omnivores buffer themselves from variation in the availability of foods
(Singer and Bernays 2003).
Bears (Ursus spp.) are often regarded as archetypical omnivores because they consume a
very wide variety of foods. Brown bear (Ursus arctos) populations fall along a spectrum of
dietary breadths, from very wide to relatively narrow, with diets differing dramatically among
populations and individuals (Bojarska and Selva 2012). Recent research found grizzly bears in
the Greater Yellowstone Ecosystem consumed at least 266 food items (Gunther et al. 2014). In
contrast, the diets of some coastal brown bear populations consist largely of salmon
(Oncorhynchus spp.): research shows 76% of adult male Kodiak brown bears (Ursus arctos
middendorffi) derive over 60% of their annual nourishment from salmon (Van Daele et al. 2013).
Brown bear consumption of migrating and spawning salmon is among the most widely
recognized predator-prey interactions, both because bears disperse marine-derived nutrients into
terrestrial and riparian ecosystems (Helfield and Naiman 2006) and because bear fecundity, body
size, and population productivity are all strongly linked to salmon consumption (Hilderbrand et
al. 1999). Because of the strong correlation between bear density and salmon consumption,
researchers and managers often emphasize the need to manage salmon abundance to maintain
high-density bear populations (Hilderbrand et al. 2004, Levi et al. 2012).
The strong evidence for the dependence of bears on salmon and the intuitive notion of
salmon as a particularly energy and nutrient rich food seem inconsistent with observations of
bears occasionally forgoing salmon for other foods. In southwest Kodiak, Alaska in both 2014
and 2015, we observed periods of very low bear activity along streams filled with spawning
sockeye salmon (Fig. 3.1A). The common explanation for this behavior is bears move off
65
streams to consume seasonally abundant berries, notably red elderberries (Sambucus racemosa)
that are widely distributed on Kodiak Island and throughout North America (Clark 1957). From
an energy maximization perspective, this behavior seems incongruous where salmon are
abundant; salmon are high in protein and energy rich (an average unspawned female sockeye
salmon contains over 4,200 Kcal; Hendry and Berg 1999), while berries tend to be low in
protein, small and spatially dispersed (Welch et al. 1997). Furthermore, bear weight gain from
berry consumption is often constrained by limitations on consumption rate, especially for large
bears (Welch et al. 1997). One hypothesis explaining this seemingly maladaptive behavior is
healthy, productive bears require compounds in berries that are not found in salmon. For
example, berries may provide key vitamins or minerals that may limit growth, maturation or,
ultimately, fitness (Robbins 1983). Alternatively, it has been suggested that dispersed berry
patches in a complex landscape may reduce behavioral conflict amongst bears compared to
salmon foraging (Gende and Quinn 2004, Ben-David et al. 2004, Rode et al. 2006). These
explanations could have merit, however, recent work suggests Kodiak bears likely have excellent
energetic reasons for selecting elderberries over salmon.
Animals forage in a way that maximizes their life-history energy balance while avoiding
mortality (Hall et al. 1992). The simplest model for how animals forage is energy maximization:
they maximize their energy intake by selecting foods that result in the highest energy
consumption rate. However, the macronutrient optimization hypothesis (MOH) suggests a more
nuanced foraging approach. According to the MOH, instead of only foraging to maximize
energy consumption, animals also regulate their intake of macronutrients (protein, fat,
carbohydrates) towards specific multidimensional intake targets (Simpson et al. 2004). The idea
is different macronutrient combinations exact different digestive costs. For example, diets overly
66
high (Soucy and Leblanc 1998) or low (Robbins et al. 2007) in protein can increase digestive
costs (dietary induced thermogenesis) which increases an organism’s maintenance cost (energy
consumption needed to offset basal metabolism and digestion). The increased maintenance costs
reduce net energy gain compared to diets at the optimal macronutrient target.
Research on captive bears has shown bears forage according to both macronutrient
optimization and energy maximization principles; they select foods which maximize their net
energy gain, by both selecting high energy foods and foods which result in diets near their
macronutrient targets (Robbins et al. 2007, Erlenbach et al. 2014). For brown bears, the
multidimensional intake target is dominated by protein (Rode and Robbins 2000, Robbins et al.
2007). Robbins et al. (2007) found that captive bears offered high carbohydrate and high protein
foods ad libitum, mixed these foods to achieve diets with intermediate protein levels. At
intermediate levels, bears benefited by having lower maintenance energy costs (because of lower
digestion costs) which increased their net energy intake, and thus, increased rates of mass gain.
Erlenbach et al. (2014) further investigated this phenomenon by adding high fat foods to the
choices presented to bears. They found that even with a high lipid food available in excess,
bears still mixed food items to consume a diet with an intermediate amount of protein,
specifically, 17 ± 4% of the metabolizable energy or 22 ± 6% of dry matter.
In many field studies, researchers have documented bears mixing food items, presumably
to reach optimal macronutrient targets. Bears in Alaska mix salmon with grasses, forbs, and
several berry species, including blueberry (Vaccinium ovalifolium), crowberry (Empetrum
nigrum), and lowbush cranberry (Vaccinium vitis idaea; Robbins et al. 2007), while bears
without access to salmon mix berries with herbaceous vegetation and other foods (Welch et al.
1997, Rode and Robbins 2000). Although food mixing is often needed to achieve a diet where
67
protein is ~22% of digestible dry matter, many high-quality bear foods approach this target:
whitebark pine (Pinus albicaulis) seeds, army cutworm moths (Euzoa auxiliaris), and harp seals
(Pagophilus groenlandicus) contain 21%, 27%, and 21% dry matter protein, respectively.
Similarly, protein supplies 16.5% of the metabolizable energy in red elderberry (Sambucus
racemosa), the food purportedly luring Kodiak bears away from salmon stream (Welch et al.
1997). This value is very close to the optimal protein target found for captive brown bears (17%
± 4% of metabolizable energy) and much more optimal than salmon (where protein provides
62% of metabolizable energy). Thus, the MOH provides a possible explanation for why bears
would eschew salmon for abundant red elderberries. It suggests bears eating elderberries would
experience lower maintenance costs and therefore have higher energy gain efficiencies. Even
though energy gain is more efficient while eating elderberries, bears must still be able to
consume enough berries to exceed the net energy gain from salmon consumption.
Even when they are abundant, consumption rates of fruits are constrained by the time
required to effectively consume them. Welch et al. (1997) used experiments on captive bears to
understand the constraints experienced by bear foraging on berries. Berry density, size, and
clustering were all important factors in determining berry mass intake rates, and these factors
interacted with bear size. As bears become larger, relative efficiency of bite size and rates of
consumption decrease in relation to maintenance requirements. They concluded large bears (e.g.
Kodiak brown bears), need dense aggregations of clumped berries in order to achieve
consumption rates high enough for maintenance and growth. Although each individual
elderberry is small (.09 g; Welch et al. 1997), they occur as clusters of hundreds of berries (Bill
Pyle/USFWS, unpublished data) presented on the surface of the shrub (Fig. 3.1B, 3.1C). These
68
factors may relieve fruit foraging constraints, allowing even large Kodiak bears to gain more
energy by consuming red elderberries instead of salmon.
In this paper we use bear responses to natural variation in interannual red elderberry
phenology to test whether wild bears with access to both salmon and red elderberries select
elderberries as predicted by the macronutrient optimization hypothesis. We present evidence
suggesting that Kodiak bears, some of the largest in the world, foraged according to the MOH,
consuming red elderberry fruits despite the presence of hundreds of thousands of highly
accessible sockeye salmon.
METHODS
Study Area
We conducted this work in southwestern Kodiak Island, in the western Gulf of Alaska (Fig. 3.2).
The Kodiak Archipelago has an estimated population of 3500 brown bears, hundreds of rivers,
lake shoals, and streams used by spawning Pacific salmon (Oncorhynchus spp.), and limited
human activity. The majority of the southwest portion of the island is within the Kodiak National
Wildlife Refuge, which is managed by the US Fish and Wildlife Service (USFWS). Human
activity in the study area is limited and consists primarily of sport fishers, bear viewers, and
hunters. The bears on the Kodiak archipelago are hunted during the fall and spring each year.
Approximately 190 bears were harvested annually from 2000–2009.
Although five species of salmon spawn in SW Kodiak waters, sockeye (O. nerka) and
pink salmon (O. gorbuscha) are the most abundant (Van Daele et al. 2013). From 2000–2009,
over half of the salmon returns for the Kodiak Archipelago occurred in the SW region, with an
average escapement (fish remaining after harvest) of over 3.2 million. Pink salmon spawn
primarily in main stem rivers and estuaries at the mouths of rivers, and their abundance tends to
69
peak every other year on even or odd year cycles. Sockeye salmon spawn mainly in headwater
streams, on lake beaches with interstitial flow of groundwater and in lake-outlet rivers. Most of
the stream habitats are narrow (<5 m), shallow (<0.5 m) and flow into lakes, rivers, or directly
into the ocean.
In addition to salmon, bears routinely consume several species of berries, including red
elderberry (Sambucus racemosa L.), salmonberry (Rubus spectabilis Pursh), crowberry
(Empetrum nigrum L.) and blueberry (Vaccinium spp.) and many species of grasses, sedges, and
forbs (Van Daele et al. 2013). Red elderberry is found primarily at mid-elevation slopes, with
greater concentrations on northern aspects (Fig. 3.2). Elderberry is often mixed with alder,
salmonberry, and forb meadows. Non-native sitka black-tailed deer (Odocoileus hemionus
sitkensis) and snowshoe hare (Lepus americanus) both forage on red elderberry stems and leaves
and were introduced in the 1880s and 1930s, respectively.
Salmon Abundance
For this study, we focused on four tributaries to Karluk Lake, the area most used by collared
bears (Fig. 3.2). To estimate salmon abundance in the four tributaries, we used a time-lapse
double sampling method explained in detail in Deacy et al. (in review; see chapter 1).
Elderberry phenology
We acquired daily minimum and maximum temperatures for our field site (Karluk Lake) and the
Kodiak Benny Benson State Airport in Kodiak, Alaska from the National Climatic Data Center.
Data for Karluk Lake was limited to spring-fall of 1965-1969, while daily records starting in
1949 were available for the airport. There was a strong positive correlation between Karluk
Lake and airport minimum (Pearson’s r =0.82, p<0.0001) and maximum (Pearson’s r =0.81,
p<0.0001) air temperatures. To benefit from the longer airport time series, we fitted simple
70
linear regression models of Karluk max and min temperatures as functions of airport
temperatures and then used these models to predict 1949-2015 Karluk Lake max/min
temperatures.
To estimate elderberry phenology we reviewed images from a time-lapse camera placed
near Canyon Creek in 2010, and 2013-2015. There were several red elderberry shrubs within the
image viewshed (Fig. 3.1C). For each year, we noted the first date where ripe red elderberries
were visible on at least half of the shrubs. We defined this as the “ripening date” for each year.
To understand how phenology varies with climate, we modeled observed ripening date as
a function of growing degree days (GDDs; McMaster and Wilhelm 1997). Growing degrees are
calculated by taking the average of a day’s high and low, and then subtracting a base temperature
which specifies the minimum temperature needed for growth. Growing degree days are the
cumulative sum of growing degrees across the growing season. We calculated a GDD time
series for each of the years from the temperature record. Because we found no published data on
the minimum temperature required for red elderberry growth, we calculated GDDs with a base
temperature ranging from 0-10 °C, and then used the coefficient of variation (CV) of the number
of GDDs that accumulated for each of our four observed ripening dates as a model selection
index (Yang et al. 1995). We assumed the correct base temperature would result in the least
variation in GDDs among our observed years. We selected the base temperature that produced
the smallest CV of GDDs (2 °C) as the best base for our red elderberry phenology model.
Bear habitat use
We monitored bear activity along the four study streams using time lapse photography. Each
year, twelve trail cameras were placed in the same location with the same viewshed to ensure
count effort was consistent across years. Cameras were deployed in late May or early June
71
before salmon runs began and were removed in late August in 2013/2014 and late September in
2015. Cameras were programmed to take a photo every five minutes during daylight hours. We
counted the number of independent bears (all bears minus cubs) in each photo and then summed
counts by day. Although these counts cannot tell us exactly how many individual bears used
these streams because we did not view the entire stream and we did not account for repeated use
by the same bears, the counts index overall bear use and tell us how use changes across time.
To test whether bears were using elderberry when it was available, we analyzed the space
use of individually collared female brown bears. We captured bears in the SW region of Kodiak
Island, Alaska by firing immobilization darts from a helicopter. All captures procedures were
approved by the Fish and Wildlife Service Institutional Animal Care and Use Committee
(IACUC permit # 2012008, 2015-001). We fitted each bear with a GPS radio collar
programmed to record a location every hour from early June through mid-November. Collars
contained a UHF (ultrahigh frequency) transmitter and were downloaded using an airplane fitted
with a UHF receiver. Over 101 000 GPS locations were recorded from 36 individuals over 47
bear-years (some bears carried collars for more than one year). 20, 12, and 15 bears carried
collars during 2010, 2014 and 2015, respectively. We screened GPS locations for accuracy,
removing relocations with a positional dilution of precision (PDOP) greater than 10 (N=6297,
Lewis et al. 2007).
To capture temporal patterns of bear habitat use we divided the location data by week.
For each week, we calculated the proportion of locations associated with salmon and the
proportion associated with red elderberry. Salmon associated locations were within 50m of a
stream, river, or lake beach known to be used by spawning salmon (Fig. 3.2 shows stream and
river sites). To evaluate use of elderberry, we used percent elderberry values from the Kodiak
72
Land Cover Classification (KLCC), a 30 X 30m raster of percent canopy cover of the dominant
plant species in the study area (Fig. 3.2; Fleming and Spencer 2007). These values ranged from
0-50 percent elderberry cover, however, to be more confident elderberry was present in the plots
we restricted the elderberry class to 10 percent or greater cover. Elderberry points were defined
as any bear relocation within a plot with at least 10 percent elderberry cover. We think it is valid
to assume bears were in these locations to eat elderberry as opposed to another food given past
evidence of elderberry as a heavily used bear food. For example, a 1957 fecal content survey
found in 69 of 264 scats (Clark 1957), while a 2011 fecal survey found elderberries in 24% of
scats collected from August-early October (Sorum 2013).
RESULTS
Elderberry Phenology
Red elderberry phenology varied across years, with relatively early phenology in 2014 (July 21st)
and 2015 (July 20th) and late phenology in 2013 (August 8th) (Fig 3.3). We saw no ripe
elderberries in time lapse images in 2010. This matched other observations made from the air
and ground; except for the rare plant with some berries, fruit production appeared to be very low
in 2010. Using these dates in the elderberry phenology model resulted in a mean of 926 GDDs
(SD = 31.5) needed to produce ripe berries with a base temperature of 2 °C. Predictions of berry
ripening date for the temperature record starting in 1949 show that ripening date is highly
variable (range = July 14-Sept. 28; SD= 13 days) and trending towards earlier dates at a rate of
2.2 days per decade (R2= 0.10, p<.01; Fig. 3.3). Comparing all predicted ripening dates, 2013,
2014, and 2015 had the 19th, 5th, and 4th earliest ripening dates, respectively, of the 66 year
temperature record.
73
Salmon Abundance
The peak of total salmon abundance in the four study streams varied across years (Fig. 3.4).
Abundance peaked earliest in 2014 on July 15th with just over 95,000 sockeye salmon.
Abundance peaked on July 17th in 2013 with almost 75,000 sockeye, while the peak in
abundance was quite a bit later in 2015 with 92,000 sockeye on July 28th.
Bear habitat use
Bear detection patterns along the four study streams varied greatly among years (Fig. 3.4). In
2013, daily bear detections roughly tracked salmon abundance. In contrast, bear detections were
much lower in 2014 and 2015 and did not track salmon abundance through the entire spawning
period as occurred in 2013. In particular, bears were essentially absent during the second half of
the salmon run in 2014 and relatively scarce during the second half of the run in 2015. These
periods of low bear detections relative to salmon abundance correspond with periods of ripe red
elderberry availability.
The habitat use of collared female brown bears also varied greatly among years (Fig.
3.5). In 2014 and 2015, bear use of salmon habitat (streams, rivers, and lake shoals where we
assumed they were eating salmon) had two distinct peaks, the first in mid-July, and the second in
mid-September. Although salmon were still abundant at the time, there was a distinct drop in
selection for salmon in early August in these two years. Use of berry habitat (areas consisting of
at least 10% red elderberry) followed a roughly inverse pattern, peaking in early to mid-August.
Early August, when use of berries was peaking and use of salmon was low, corresponded with
the observed period of red elderberry availability. Habitat use was different in 2010 when almost
no ripe red elderberries were observed: use of salmon was higher than in any other year for the
entire summer and berry use remained below 20% of locations.
74
DISCUSSION
Red elderberry phenology was sensitive to interannual variation in spring and early summer
temperatures. In years with average temperatures (e.g. 2013, Fig. 3.3), elderberries ripen
towards the end of the stream salmon runs. In contrast, elderberries ripened relatively early in
warm years (e.g. 2014, 2015), which caused elderberry availability to overlap with the period of
high salmon availability. Bear activity data from four streams showed that despite robust runs
(>70,000 sockeye salmon), bears abandoned salmon once red elderberries were available, which
corroborates anecdotal accounts of bears being absent on streams despite high salmon
abundance. The locations of GPS collared female bears confirmed that bears switched from
using salmon habitat to elderberry habitat during periods of berry availability, rather than moving
to another source of salmon or habitat associated with other vegetation resources. These data
strongly suggest that wild bears faced with real world foraging constraints forage according to
the macronutrient optimization hypothesis.
Studies of bear behavior often reveal substantial variation among individuals. For example, a
recent study of bear use of salmon resources documented several collared bears using salmon for
over 100 days in the summer/fall, and one bear that was never detected using nearby streams
filled with spawning salmon (Deacy et al. in press). In comparison, our 2014 results point
towards unusual consistency in foraging strategies: during the abundant salmon period from
7/25/14 – 8/20/14 bear detection rates along streams were similar to periods when no salmon
were present, suggesting nearly all bears, regardless of demographic, switched from salmon to
elderberries (Fig 3.4). We noted that even the largest bears, which normally have a much higher
proportion of salmon in their diets (Van Daele et al. 2013), seemed to switch to foraging on red
elderberry.
75
Bears in our study preferentially used elderberry over salmon, a resource strongly linked to
bear fitness. So, is red elderberry the perfect bear food? Although it has almost exactly the
optimal protein content (~17% of metabolizable energy), Erlenbach et al. (2014) showed bears
selected diets where fat supplied most of the metabolizable energy, not carbohydrates. Although
elderberries are preferred over salmon, their high carbohydrate and low lipid content means they
are not the perfect food. Foods available to other bear populations, such as whitebark pine seeds,
and army cutworm moths are better candidates for bear superfoods because they approach the
macronutrient protein target and have high lipid contents. However, elderberries have the
advantage of being in excess when they are abundant, which eliminates intraspecific
competition, allowing all demographics to access berries for as long as they are available. In
contrast, young bears are less likely to gain access to a preferred concentrated food source like a
prime fishing location because of exclusion by more dominant bears (Sellers and Aumiller
1994).
Researchers have often observed large, dominant bears foraging for salmon more than
subdominant bears (Gende and Quinn 2004, Van Daele et al. 2013). The most common
explanation is dominant bears exclude others from accessing salmon, which is generally thought
to be the universally preferred resource. But, Welch et al. (1997) and Robbins et al. (2007)
found smaller bears used berries more efficiently for weight gain than large bears. They
proposed this size dependent foraging constraint as an alternative explanation for observations of
demographic differences in foraging behavior (i.e. lower use of salmon by smaller bears). Our
results support the Robbins et al. (2007) hypothesis that dietary optimization drives foraging
patterns and occurs regardless of bear size, age or sex (Fig 3.4). We show that at Kodiak, which
has some of the largest bears in the world, nearly all bears appeared to be eating elderberries
76
when berries were ripe and generally available, even during peak salmon abundance (Fig. 3.4,
2014). This suggests dietary constraints rather than competition shapes demographic differences
in foraging behavior.
Our predictions of timing of ripe berry availability show that ripening date is highly variable
(range = July 14-Sept. 28; SD= 13 days), and trending towards earlier dates. If the trend
continues, mean elderberry phenology will become 2.2 days earlier per decade. This pattern
raises interesting questions about the future of Kodiak bear food resources. First, will salmon
and berry resources become progressively more synchronized with time? It is difficult to predict
whether synchrony between salmon runs and elderberry will increase given the large uncertainty
surrounding the impact of climate change on sockeye spawn timing (Kovach et al. 2015), and red
elderberry abundance. This is important because bears likely benefit most when their preferred
resources are asynchronous because asynchrony increases the overall duration of high value food
resources.
Because of the strong link between bear population productivity and meat intake
(Hilderbrand et al. 1999), bear conservation research often focuses on increasing or maintaining
salmon abundance as a tool for maintaining bear population productivity (Levi et al. 2012).
However, our results strongly suggest elderberries are an important resource that appears to be
preferred by bears when ripe berries are abundant. Nonetheless, salmon availability likely has a
larger influence on bear population productivity, for several reasons: 1) salmon are available for
a longer period (3 months in southwest Kodiak) than berries (~1 month); 2) salmon abundance is
likely more stable than elderberry abundance because of portfolio effects and salmon
management policy (Clark et al. 2006, Schindler et al. 2010); and 3) net energy intake from
foraging on elderberries may drop below net intake from salmon foraging if elderberry
77
abundance is lower than we observed in 2014 and 2015 because of constraints on berry foraging
efficiency (Welch et al. 1997). Thus, while bears seem to preferentially eat red elderberries,
salmon are likely required to maintain dense populations of large productive bears.
Nonetheless, given the strong selection for elderberries we observed, managers should be aware
of the impact of management decisions and climate change on this resource. Although managers
cannot prevent climate change impacts from occurring, they may be able to mitigate these
impacts through management of other species. For example, Sitka black-tailed deer (Odocoileus
hemionus sitkensis) were introduced to Kodiak Island in the 1890s. Although we know Sitka
deer forage on red elderberry, little is known about how deer impact berry productivity, or
whether deer have caused long-term changes in elderberry abundance. If, however, Sitka deer
negatively impact elderberry abundance, managers could increase deer harvest to increase berry
abundance. Clearly, the interactions among deer, climate change and elderberry abundance
deserve more research.
One benefit of omnivory is it buffers organisms from temporal and spatial variation in
abundance of individual resources (Singer and Bernays 2003); as dietary breadth increases, it is
increasingly likely some food item will be available at any given time. Coastal brown bear
populations with access to salmon tend to be more productive (Hilderbrand et al. 1999), but also
tend to have more specialized diets (Van Daele et al. 2013, Gunther et al. 2014). This may make
them more vulnerable to fluctuations in resource abundance compared to lower density bear
populations with less specialized diets. For example, whitebark pine seeds are often considered a
preferred food for grizzly bears in Yellowstone, a population with wide dietary breadth, yet a
recent crash in pine seed abundance seems to have had little to no effect on their population
productivity, likely because Yellowstone grizzlies switched to alternative foods (Interagency
78
Grizzly Bear Study Team 2013). Wide dietary breadth likely buffers these bears from high
interannual variation in their preferred resources (Gunther et al. 2014, Costello et al. 2014). In
contrast, most of the annual energy budgets of bears in southwest Kodiak comes from salmon
(Van Daele et al. 2013), and recent data suggests local bear populations are sensitive to temporal
variation in this resource. The estimated density of bears/1000 km2 in the Karluk Lake drainage
dropped from 483 ± 61 (90% confidence interval) in 2003 to 252 ± 61 in 2010, a 48% decline
(William Leacock/USFWS, unpublished data). Although other factors may have contributed,
(e.g. concurrent poor berry production), a steady 64% decline in annual sockeye salmon returns
over the same period was a likely key contributor (ADF&G weir records). This comparison
suggests bear populations with relatively specialized diets may experience more interannual
variation in overall food abundance than populations with relatively generalized diets.
In striving for simple answers to complicated management challenges, it is tempting to focus
on a single resource required by wildlife populations. For coastal brown bears, the focus is on
salmon, for interior bears the focus is on berries, ungulates, or pine nuts depending on the
population. This paper is a reminder that even bears with relatively specialized diets integrate
foods from across landscapes; they are omnivores with the dietary flexibility and plastic foraging
behavior that allows them to switch to alternative resources when they are available. To better
manage omnivorous wildlife, we must identify and conserve key resources while also
recognizing that many complementary resources can stabilize temporal variation in resource
availability.
ACKNOWLEDGEMENTS
W. Deacy and J. Stanford were supported by the Jessie M. Bierman Professorship, J.
Armstrong was supported by the David H. Smith Conservation Research Fellowship, and W.
79
Leacock was supported by the USFWS. Funding for fieldwork was provided by the USFWS
Inventory and Monitoring and Refuge Programs. We appreciate the staff at the Kodiak National
Wildlife Refuge and Flathead Lake Biological Station for their committed support and
assistance. We thank Mat Sorum, Caroline Cheung, and volunteer field technicians Barbara
Svoboda, Alex May, Bill Dunker, Tim Melham, Tyler Tran, Marie Jamison, Louisa Pless,
Francesca Cannizzo, Isaac Kelsey, Mark Melham, Jane Murawski, Prescott Weldon, Shelby
Flemming, Andy Orlando, and Kristina Hsu for assisting with data collection. We thank
helicopter pilot Joe Fieldman and fixed-wing pilots Kurt Rees, Kevin Vanhatten, Kevin Fox, and
Issac Beddingfield for their skilled flying during bear captures and aerial telemetry. Many
thanks to Chris Servheen and Charlie Robbins for early discussions about our surprising bear
distribution data and detailed discussions about bear diets.
LITERATURE CITED
Armstrong, J. B., and D. E. Schindler. 2011. Excess digestive capacity in predators reflects a life
of feast and famine. Nature 476:84–7.
Ben-David, M., K. Titus, and L. R. Beier. 2004. Consumption of salmon by Alaskan brown
bears: a trade-off between nutritional requirements and the risk of infanticide? Oecologia
138:465–74.
Bojarska, K., and N. Selva. 2012. Spatial patterns in brown bear Ursus arctos diet: The role of
geographical and environmental factors. Mammal Review 42:120–143.
Clark, J. H., R. D. Mecum, A. McGregor, P. Krasnowski, and A. M. C. N.-A. 12. 1. 00.-146
Carroll. 2006. The commercial salmon fishery in Alaska. Alaska Fishery Research Bulletin
12:1–146.
Clark, W. K. 1957. Seasonal food habits of the Kodiak bear. Transactions of the North American
Wildlife and Natural Resources Conference 22:145–151.
Costello, C. M., F. T. van Manen, M. a Haroldson, M. R. Ebinger, S. L. Cain, K. a Gunther, and
D. D. Bjornlie. 2014. Influence of whitebark pine decline on fall habitat use and movements
of grizzly bears in the Greater Yellowstone Ecosystem. Ecology and evolution 4:2004–
2018.
Crossin, G., and S. Hinch. 2004. Energetics and morphology of sockeye salmon: effects of
upriver migratory distance and elevation. Journal of Fish …:788–810.
80
Van Daele, M., C. T. Robbins, B. X. Semmens, E. J. Ward, L. J. Van Daele, and W. B. Leacock.
2013. Salmon consumption by Kodiak brown bears (Ursus arctos middendorffi) with
ecosystem management implications. … Journal of Zoology 174:164–174.
Dearing, D. 1997. The function of haypiles of pikas (Ochotona princeps). Journal of
Mammalogy 78:1156–1163.
Erlenbach, J. a., K. D. Rode, D. Raubenheimer, and C. T. Robbins. 2014. Macronutrient
optimization and energy maximization determine diets of brown bears. Journal of
Mammalogy 95:160–168.
Fleming, M. D., and P. Spencer. 2007. Kodiak Archipelago Land Cover Classification Users
Guide.
Gende, S., and T. Quinn. 2004. The relative importance of prey density and social dominance in
determining energy intake by bears feeding on Pacific salmon. Canadian Journal of Zoology
85:75–85.
Gunther, K. A., R. R. Shoemaker, K. L. Frey, M. A. Haroldson, S. L. Cain, F. T. Van Manen,
and J. K. Fortin. 2014. Dietary breadth of grizzly bears in the Greater Yellowstone
Ecosystem. Source: Ursus Management Ursus 25:60–7260.
Hall, C., J. Stanford, and F. Hauer. 1992. The distribution and abundance of organisms as a
consequence of energy balances along multiple environmental gradients. Oikos.
Helfield, J. M., and R. J. Naiman. 2006. Keystone Interactions: Salmon and Bear in Riparian
Forests of Alaska. Ecosystems 9:167–180.
Hendry, A. P., and O. K. Berg. 1999. Secondary sexual characters, energy use, senescence, and
the cost of reproduction in sockeye salmon. Canadian Journal of Zoology 77:1663–1675.
Hilderbrand, G., S. Farley, C. Schwartz, and C. Robbins. 2004. Importance of salmon to wildlife:
implications for integrated management. Ursus:1–9.
Hilderbrand, G. V, C. C. Schwartz, C. T. Robbins, M. E. Jacoby, T. a Hanley, S. M. Arthur, and
C. Servheen. 1999. The importance of meat, particularly salmon, to body size, population
productivity, and conservation of North American brown bears. Canadian Journal of
Zoology 77:132–138.
Humphries, M., D. Thomas, and D. Kramer. 2003. The role of energy availability in mammalian
hibernation: a cost‐ benefit approach. Physiological and Biochemical … 76:165–179.
Interagency Grizzly Bear Study Team. 2013. Response of Yellowstone grizzly bears to changes
in food resources : a synthesis . Final Report to the Interagency Grizzly Bear Committee.
Kovach, R. P., S. C. Ellison, S. Pyare, and D. A. Tallmon. 2015. Temporal patterns in adult
salmon migration timing across southeast Alaska. Global Change Biology 21:1821–1833.
Levi, T., C. T. Darimont, M. MacDuffee, M. Mangel, P. Paquet, and C. C. Wilmers. 2012. Using
Grizzly Bears to Assess Harvest-Ecosystem Tradeoffs in Salmon Fisheries. PLoS Biology
10:e1001303.
McMaster, G. S., and W. W. Wilhelm. 1997. Growing degree-days: One equation, two
81
interpretations. Agricultural and Forest Meteorology 87:291–300.
Robbins, C. T. 1983. Wildlife Feeding and Nutrition. First edition. New York, NY.
Robbins, C. T., J. K. Fortin, K. D. Rode, S. D. Farley, L. A. Shipley, and L. A. Felicetti. 2007.
Optimizing protein intake as a foraging strategy to maximize mass gain in an omnivore.
Oikos 116:1675–1682.
Rode, K. D., S. D. Farley, and C. T. Robbins. 2006. Sexual dimorphism, reproductive strategy,
and human activities determine resource use by brown bears. Ecology 87:2636–46.
Rode, K. D., and C. T. Robbins. 2000. Why bears consume mixed diets during fruit abundance.
Canadian Journal of Zoology 78:1640–1645.
Schindler, D. E., R. Hilborn, B. Chasco, C. P. Boatright, T. P. Quinn, L. a Rogers, and M. S.
Webster. 2010. Population diversity and the portfolio effect in an exploited species. Nature
465:609–12.
Sellers, R., and L. Aumiller. 1994. Brown bear population characteristics at McNeil River,
Alaska. Bears: Their Biology and Management 9.
Simpson, S. J., R. M. Sibly, K. P. Lee, S. T. Behmer, and D. Raubenheimer. 2004. Optimal
foraging when regulating intake of multiple nutrients. Animal Behaviour 68:1299–1311.
Singer, M. S., and E. A. Bernays. 2003. Understanding omnivory needs a behavioral perspective.
Ecology 84:2532–2537.
Sorum, M. 2013. Behavior-Specific Resource Selection by Kodiak Brown Bears. University of
Idaho.
Soucy, J., and J. Leblanc. 1998. Protein meals and postprandial thermogenesis. Physiology and
Behavior 65:705–709.
Welch, C. A., J. Keay, K. C. Kendall, and C. T. Robbins. 1997. Constraints on frugivory by
bears. Ecology 78:1105–1119.
Yang, S., J. Logan, and D. L. Coffey. 1995. Mathematical formulae for calculating the base
temperature for growing degree days. Agricultural and Forest Meteorology 74:61–74.
82
CHAPTER 3 FIGURES
Fig. 3.1- A) Spawning sockeye salmon (Oncorhynchus nerka) entering a small tributary within
the study area. By the time they enter spawning grounds, salmon contain very little lipids, and
~65% protein (Crossin and Hinch 2004). B) Close up of red elderberry shrub (Sambucus
racemosa) with ripe berries. Although each individual berry is small, clustering likely makes
foraging by Kodiak brown bears (Ursus arctos middendorffi) much more efficient C) Picture
from a time-lapse camera at Canyon Creek, Kodiak Island, Alaska. A Kodiak brown bear is in
the stream and the fruits of red elderberry are in the foreground at right and to the right of the
stream.
B A
C
83
Fig. 3.2- Map of the Karluk Lake, Alaska study area showing elderberry land cover and the
streams/rivers used by spawning salmon. Black dots indicate the four study tributaries where
salmon abundance was quantified and bear activity was monitored. Green indicates 10 percent
elderberry cover, orange indicates 20-39 percent elderberry, and red is ≥40 percent elderberry.
Values were derived from the Kodiak Land Cover Classification (KLCC; Fleming and Spencer
2007).
84
Fig. 3.3- Predicted onset of ripe red elderberry (Sambucus racemosa) presence in southwest
Kodiak Island, Alaska. Height of each bar indicates the amount of uncertainty for each
prediction. Predictions resulted from a plant growth model informed by observations of
elderberry phenology in the study area paired with a time series of daily air temperatures. The
line is the best simple linear regression fit to the mean value for each year, with 95% confidence
intervals shown in grey (R2= 0.10, p<.01). If the trend shown by this regression continues, we
can expect mean date of elderberry ripening to be 2.2 days earlier every 10 years into the future.
85
Fig. 3.4- Kodiak brown bear (Ursus arctos middendorffi) activity and estimated abundance of
sockeye salmon (Oncorhynchus nerka) in four Karluk Lake, Alaska tributaries. Bear activity
(black line) shows the daily sum of detections of independent bears (not cubs) recorded by 12
time-lapse cameras (black line) placed along the tributaries. Cameras took photos every 5
minutes during daylight hours. Salmon abundance was measure using a double sampling camera
system (Deacy et al. in review). The peak period of ripe red elderberry (Sambucus racemosa)
availability is shown by the pink square. The start of this period is the date where at least half of
the Elderberry plants in time lapse images had ripe berries. The end is 30 days after the start, the
approximate length of ripe berry availability.
86
Fig. 3.5- Proportion of hourly female Kodiak brown bear (Ursus arctos middendorffi) locations
recorded by Global Positioning System (GPS) radio collars associated with either salmon (blue)
or red elderberry (red; Sambucus racemosa) habitat patches. We considered a location to be
associated with salmon if it was within 50m of a stream known to contain salmon and elderberry
if it was located within a patch containing at least 10% elderberries, as determined by the Kodiak
Detailed Land Cover Classification (KDLCC; Mike Fleming unpublished data). The grey bars
towards the top of each panel indicate the periods during which ripe red elderberries were
available to bears, observed using time lapse phenology cameras. No berry phenology bar is
shown for 2010 because phenology cameras recorded no images showing ripe elderberries.
Essentially no ripe berries were seen in ground and aerial observations of many other parts of the
study area in 2010.
87
CHAPTER 4
SYNTHESIS
Bears responded flexibly to a highly variable resource mosaic. Rather than being hindered by
significant spatial and temporal variation in their resources, bears used variation to their
advantage. For example, salmon spawned at different times in different habitats, which created a
resource wave. In chapter 2, I showed that bears tracked this resource wave, which allowed
them to consume salmon for 70% longer than if no spawning variation had occurred. In
addition, bears rapidly responded to alternative resources when they became available, switching
to feeding on red elderberries, even when salmon were available (chapter 3). Although salmon
provide a greater share of annual energy budgets, bears can likely gain weight faster using red
elderberry because of lower digestive costs.
This work has important implications for foraging theory. Although researchers have
documented several examples of animals tracking resource waves, in all of these examples the
resource wave propagated along a continuous gradient (i.e. elevation, latitude, or precipitation),
and tracking consisted of starting at the correct time/place and altering movement rates in a
continuous direction. In chapter two, I presented direct evidence of Kodiak bears tracking a
resource wave caused by variation in salmon spawning phenology, which does not vary along a
continuous spatial gradient, and thus creates a more complex mosaic of resource availability
through time. As far as I am aware, this is the first documented example of animals tracking a
more complex resource wave. This work also contributes to the diet selection literature. Studies
on captive bears resulted in models of bear foraging behavior, including the constraints they face
when consuming different foods. These included the observation that bears select diets by
balancing macronutrient mixtures rather than simply maximizing energy intake (Erlenbach et al.
2014), and large bears have difficulty gaining weight through frugivory because harvesting rates
do not scale allometrically with body size (Welch et al. 1997). In chapter 3, I presented evidence
suggesting Kodiak bears of all sizes (including males which are among the largest bears in the
world), chose to eat elderberries instead of abundant salmon. These results agree with the
prediction of a preference for diets with intermediate protein (Erlenbach et al. 2014), but seem to
contradict the prediction of frugivory constraints on large bears (Welch et al. 1997). Actually,
bears eating red elderberries is the exception that proves the rule: elderberries grow in tight
clumps which allow bears to consume hundreds of individual berries with a single bite. This
likely relieves the harvest rate constraints normally faced by large bears. Altogether, chapter 3
confirms that predictions based on analyses of captive bear diets can predict the diets and
distributions of wild bears faced with intraspecific competition, travel costs, and complex
resource mosaics.
Bear densities on Kodiak Island are very high (0.48 bears/km2). The results presented in this
dissertation highlight some possible reasons for this. First, Kodiak Island has many salmon
populations spawning in different habitats, which creates a resource wave that extends bear
88
access to salmon. Bears are able to track this variation because human development is minimal
and thus, habitat connectivity is excellent. Second, bears in Kodiak have access to two sources
of food with which they can rapidly gain weight, salmon and red elderberries. In years with
average temperatures, red elderberries become ripe between the early and late runs of sockeye
salmon, which forms a continuous stream of preferred resources from early July through mid-
October. This allows bears to gain weight across a longer period of time. Finally, the key
resources used by bears, red elderberry and salmon, are both available across very large areas
during most of the summer and fall. This likely alleviates intraspecific competition, allowing all
bear demographics to access these resources.
This work also has important lessons for bear managers and conservationists. Overall, the
results presented here strongly argue for the importance of diversity in resources used by bears.
This diversity takes many forms. Managers must protect aquatic habitat heterogeneity and
salmon population diversity, because they generate the salmon resource wave which increased
bear access to salmon by 70%. Managers should also focus on the diverse foods used by bears.
Surprisingly, red elderberry is the food that maximizes bear growth rates, not salmon. This
additional key resource likely buffers bears from interannual variation in salmon abundance,
making them less susceptible to salmon run failures. Although some impacts to bear resources
are hard to predict and even harder to control (e.g. climate change), managers and bear
conservationists can mediate these unavoidable impacts by protecting resource heterogeneity.
For example, altering salmon management to protect salmon population diversity, and reducing
deer densities to increase red elderberry production would likely increase bear resilience in the
face of future resource changes.
As with most research, this study produced more questions than it answered. First, there is
much we don’t yet understand about the costs and benefits of resource synchrony for wildlife.
On the one hand, multiple resources available at the same time each year can reduce the risk of
famine at the interannual time scale. On the other hand, animals are often digestion limited, so
resources might also be more beneficial if they are asynchronous. These tradeoffs are complex
and likely play out over long time scales, especially for omnivores that consume hundreds of
different foods. Second, we know little about how bears track resources like salmon across
space. Do they use their memory of past salmon run timing to anticipate when/where they
should travel? Do they use proximate cues like temperature, smells, or activity of other animals
(e.g. gulls, eagles)? Or, do they sample the landscape by checking streams periodically? Finally,
bear managers and conservationists need to know how their management actions or natural
changes in resources will impact bear populations. For example, how much would a 30%
reduction in salmon abundance change bear abundance? And how much would bear populations
change if we reduced deer density by 50%? These questions are critical and deserve further
research.