Content uploaded by Andrew Bryant
Author content
All content in this area was uploaded by Andrew Bryant on Jun 25, 2015
Content may be subject to copyright.
Metapopulation ecology of Vancouver Island marmots
(Marmota vancouverensis)
by
Andrew Albert Bryant
B.E.S., University of Waterloo, 1984
M.E.Des., University of Calgary, 1990
A Dissertation Submitted in Partial Fulfillment of the
Requirements for the Degree of
DOCTOR OF PHILOSOPHY
in the Department of Biology
© Andrew Albert Bryant, 1998
University of Victoria
All rights reserved. This dissertation may not be reproduced in whole or in part, by photocopying or other
means, without the permission of the author.
Metapopulation ecology of Vancouver Island marmots (Marmota vancouverensis)
Supervisor: Dr. D.S. Eastman
ABSTRACT
Vancouver Island marmots (M. vancouverensis) rank among the world’s most critically
endangered mammals. There were probably fewer than 100 marmots in 1998, with 90%
distributed south of Alberni Inlet, and the remainder on or near Mount Washington. This
represents a 60-70% decline in numbers during the past 10 years, and a considerably reduced
geographic range during the past several decades.
I used data from marked animals, radio-telemetry and population counts to test whether
population dynamics were consistent with predictions made under five hypotheses: habitat tracking,
sink-connectivity, weather, predators and disease. Estimates of demographic rates from intensive
mark-recapture work and population counts were generally consistent, although estimation of adult
survival from counts was problematic because of the difficulty of distinguishing surviving marmots
from immigrants. There was no apparent influence of mark-recapture on survival or reproduction,
and intensively studied colonies showed similar dynamics to colonies that were visited infrequently.
There was little evidence for habitat tracking in natural habitats. Few colonies showed
chronically low reproduction or survival, which would be the predicted result of a gradually
deteriorating environment. Declines were more often abrupt and catastrophic. Marmots did not
colonize clearcuts in proportion to their temporal or spatial availability, and ultimately colonized
only a minuscule fraction of the potential habitat. However, marmots already inhabiting clearcuts
represent a special case of habitat tracking; survival rates were significantly lower at clearcuts of
more advanced seral age (i.e., >11 years after harvest).
Evidence for source-sink and landscape connectivity processes was relatively strong. Marmots
inhabiting clearcuts had chronically lower survival rates (by 5-10%). Per female reproductive
contribution in clearcuts was half that of females inhabiting natural environments. However not all
clearcuts acted as sinks, or acted as sinks in all years. Colonizations of clearcuts were spatially
concentrated and none occurred at distances greater that 5 km from an existing natural colony.
Apparent adult survival was significantly associated with isolation but juvenile survival was not,
which is consistent with the prediction that isolated colonies should receive fewer immigrants.
However the spatial pattern of extinctions was unexpected. Isolated and closely-clustered colonies
had similar probabilities of extinction.
ii
Weather significantly influenced marmot survival and reproduction but explained only small
amounts of variation. Survival was significantly associated with rainfall, temperature and
snowpack depth. Reproduction was negatively associated with snowpack and temperature. Slope
aspect was significantly associated with survival, perhaps suggesting the importance of snowmelt
patterns. Natural and clearcut colonies responded differently to weather.
Indices of wolf and cougar abundance were inconsistent and probably do not reflect true
population sizes. Deer abundance was weakly associated with marmot survival in natural habitats,
which could suggest switching of predator hunting effort. Marmot survival was spatially
correlated, which is consistent with the idea that a few individual predators may focus hunting
efforts at adjacent colonies. Field observations and radio-telemetry corroborated the importance of
predators. In natural habitats, disappearances were uniformly distributed throughout summer, as
predicted. In clearcuts, disappearances were more heavily skewed towards late summer,
suggesting that winter mortality was more important.
Spatial correlation of survival is also consistent with the disease hypothesis. Survival was
lower in colonies with high relative density of adults, which is a predicted result given the
prediction of increased risk of disease transmission. The incidence of high mortality events
increased during the 1990s, and the degree of spatial correlation also increased despite a more
fragmented population structure. These trends are consistent with a hypothesis of a new disease
organism or increased risk of infection.
Forestry appears to be the primary cause of recent population dynamics in the Nanaimo Lakes
region. Logging reduced overall marmot survival, inhibited their ability to re-colonize sites, and
concentrated the population, making colonies more susceptible to predators and disease. The
prognosis for continued survival remains hopeful provided that current plans for captive-breeding
and reintroduction are pursued aggressively.
iii
TABLE OF CONTENTS
Abstract ................................................................................................................................ i.
Table of Contents ................................................................................................................iii.
Appendices........................................................................................................................... v.
List of Figures..................................................................................................................... vi.
List of Tables..................................................................................................................... vii.
Acknowledgments ............................................................................................................. viii.
Frontispiece.......................................................................................................................... x.
1. Introduction
Marmota vancouverensis ...............................................................................................1.
Changes in distribution and abundance ........................................................................... 2.
Environmental tracking hypothesis .................................................................................4.
Weather hypothesis........................................................................................................ 6.
Sink-connectivity hypothesis...........................................................................................6.
Predator hypothesis........................................................................................................ 7.
Disease hypothesis......................................................................................................... 7.
Non-exclusive hypotheses, testable predictions and practical significance........................ 9.
2. Study areas ................................................................................................................. 11.
3. Methods...................................................................................................................... 14.
Population counts.........................................................................................................14.
Population estimation................................................................................................... 14.
Capture, handling and age-assignment .......................................................................... 15.
Surgical implantation and radio-telemetry..................................................................... 17.
Colony-specific demographic rates ............................................................................... 18.
Landscape conditions ...................................................................................................20.
Weather ...................................................................................................................... .21.
Predator-prey trends..................................................................................................... 22.
Statistical analyses....................................................................................................... 23.
4. Results........................................................................................................................ 27.
Part 1: The environment
Weather ....................................................................................................................... 27.
Landscape change ........................................................................................................ 30.
Predator-prey abundance.............................................................................................. 30.
Part 2: Sampling effort
Population count efforts ............................................................................................... 34.
Repeatability of marmot counts .................................................................................... 34.
iv
Mark-recapture effort................................................................................................... 37.
Part 3: Population trends
Probable marmot abundance.........................................................................................38.
Colonizations and extinctions .......................................................................................38.
Population trends among colonies................................................................................. 41.
Part 4: Population ecology
Survival ....................................................................................................................... 41.
Reproduction ............................................................................................................... 46.
Immigration-emigration................................................................................................ 47.
Mortality factors.......................................................................................................... 49.
Congruence among estimates from intensively studied colonies and counts .................... 52.
Part 5: Tests of predictions
Effect of habitat type on demographic rates .................................................................. 54.
Temporal effects on survival and reproduction.............................................................. 59.
Effect of clearcut age on birth and survival rates........................................................... 59.
Density dependence...................................................................................................... 62.
Colonization events in relation to habitat availability.....................................................63.
Extinction and demographic performance in relation to isolation .................................. 65.
Sources and sinks......................................................................................................... 65.
Effects of weather on survival ...................................................................................... 68.
Effects of weather on reproduction ...............................................................................70.
Predator-prey effects .................................................................................................... 70.
Spatial correlation of survival rates ............................................................................. 72.
Incidence of high mortality events................................................................................. 73.
5. Discussion................................................................................................................... 77.
Environmental tracking ................................................................................................ 78.
Weather ....................................................................................................................... 79.
Sink-connectivity .........................................................................................................81.
Predator....................................................................................................................... 84.
Disease ........................................................................................................................ 86.
Converging lines of evidence ........................................................................................ 88.
Lessons from marmots ................................................................................................. 90.
6. Literature cited........................................................................................................... 94.
v
Appendices:
1. Active season temperature and rainfall trends ............................................................. 104.
2. Snowpack trends........................................................................................................ 105.
3. Measures of landscape change.................................................................................... 106.
4. Deer population trends in Management Unit 1-5 ......................................................... 107.
5. Terrestrial predator trends in Management Unit 1-5.................................................... 108.
6. Survey effort and probable count success ................................................................... 109.
7. Age-sex structure at 5 intensively studied colonies. ..................................................... 110.
8. Tagging success and ear-tag loss rates........................................................................ 113.
9. Raw and adjusted marmot population estimates .......................................................... 114.
10. Colony-specific habitat conditions .............................................................................. 115.
11. Ear-tagged and apparent juvenile survival at 5 colonies............................................... 116.
12. Ear-tagged and apparent adult survival at 5 colonies................................................... 117.
13. Birth rates and fecundity at 5 colonies ........................................................................ 118.
14. Variation in female reproductive performance............................................................. 119.
15. Estimates of minimum dispersal distances...................................................................120.
16. Fate of radio-telemetered marmots.............................................................................. 121.
17. Habitat-specific life-tables for M. vancouverensis.......................................................122.
18. Colony-specific juvenile abundance and apparent survival .......................................... 123.
19. Colony-specific adult abundance and apparent survival............................................... 124.
20. Colony-specific birth rates.......................................................................................... 125.
vi
List of Figures
1. Historical and current distribution of Vancouver Island marmots. .................................. .3.
2. The Nanaimo Lakes metapopulation............................................................................. 12.
3. Four measures of summer weather conditions ............................................................... 28.
4. Two measures of snowpack conditions ......................................................................... 29.
5. Extent of forest harvesting over time ............................................................................31.
6. Two measures of landscape change.............................................................................. 32.
7. Predator-prey trends .................................................................................................... 33.
8. Population count extent and intensity............................................................................ 35.
9. Probable accuracy of marmot counts............................................................................ 36.
10. Seasonal effects upon marmot count success ................................................................ 36.
11. Marmot population trends over time............................................................................. 39.
12. Probable abundance in natural and clearcut habitats.. ................................................... 40.
13. Marmot population trends within and among colonies................................................... 42.
14. Sensitivity of finite population growth rate to changes in demographic rates.................. 46.
15. Effect of habitat type on age-specific reproductive performance.................................... 56.
16. Cormack-Jolly-Seber estimates of adult survival........................................................... 57.
17. Timing of last observation of tagged adults in natural and clearcut habitats ................... 58.
18. Temporal changes in adult survival and probability of breeding..................................... 60.
19. Effect of increasing clearcut age on marmot survival..................................................... 61.
20. Effect of relative density on marmot demographics ....................................................... 63.
21. Colonizations and potential colonizations in clearcuts. .................................................. 64.
22. Apparent extinction events at natural colonies from 1985 to 1997. ................................ 66.
23. Colony-specific source-sink analysis............................................................................. 67.
24. Spatial autocorrelation of marmot survival rates........................................................... 73.
25. Colony-specific annual variation in apparent survival ................................................... 74.
26. Incidence of high mortality events................................................................................. 75.
vii
List of Tables
1. Repeatability of count data for adults........................................................................... 37.
2. Jolly-Cormack-Seber estimates of adult survival and recapture probability ................... 44.
3. Cumulative life-table for Vancouver Island marmots .................................................... 45.
4. Demographic rates from intensively studied colonies and non-intensive counts. ............. 53.
5. Effect of habitat, time period and relative density on marmot demographics. ................. 55.
6. Logistic regression of clearcut age against marmot survival. ......................................... 59.
7. Logistic regression of clearcut age against probability of reproducing ........................... 59.
8. Logistic regression of density against survival and probability of reproducing ............... 62.
9. Nearest colony-neighbor distances for marmot colonizations and random sites .............. 65.
10. Effect of increasing isolation on apparent marmot survival............................................ 65.
11. Effect of weather on marmot survival in natural and clearcut habitats ........................... 69.
12. Effect of weather on probability of reproducing ............................................................71.
13. Predator-prey effects on survival.................................................................................. 71.
14. Relative frequency of episodes of low survival.............................................................. 71.
viii
ACKNOWLEDGMENTS:
I thank great good fortune to have met Marmota vancouverensis. Seldom does an ecologist
encounter a problem that involves such beautiful animals, harsh practical, political and theoretical
challenges, and wonderful people. Doug Janz, Ken Armitage, Don Eastman, Bill Munro, Steve
Herrero, and the late Betty McKinnon were mentors in the truest sense of that word. “Thank you”
is vastly inadequate compensation for years of unconditional support.
Important scientific ideas were contributed by many; in particular I thank Walter Arnold, Ken
Armitage, Dan Blumstein, Ilkka Hanski, Colin Laroque , Dave Nagorsen, Sergei Pole, Dirk Van
Vuren, Rik Page, Raymond Ramousse, Ivan Rymalov, Phil Taylor and Viktor Tokarskii.
After eleven field-seasons and a thousand-plus personal field-days, it is humbling to report that
I made only a fraction of the observations discussed in this document. Many people participated in
colony-counts or mark-recapture work, including K. Atkinson (deceased), D. Blumstein, C.
Bryant, L. Campbell, J. Daniels, M. deLaronde, R. Davies (deceased), D. Doyle, L. Dyck, K. Fry,
V. Hiensalu, D. Janz, F. Lockwood, J. Lewis, M. Loedel, B. Mason, G. MacDermott, K.
McDonald, D. Milne, J. Morgan, D. Pemberton, J. Pendergaast, C. Ramsay, G.W. Smith, K.
Sturmanis, W. Swain, J. Voller, L. Wilson and M. Wong. Many others helped on occasion. T.
Chatwin, B. McKinnon, B. Morris, D. Nagorsen, D. Routledge, W. Whitehead and many other
sportsmen, loggers and naturalists provided important field observations. People armed with
binoculars and notebook still play a critical role in conservation biology.
Endangered species work sounds glamourous but is mostly just hard. My technicians (Ludwig
Dyck, Jason Lewis, Donna Milne, and Joan Voller) deserve credit for their “courage of the early
morning”. L. Dyck deserves special praise for field-skills, notes, and five years of effort. K.
Langelier (DVM) taught me blood-sampling and immobilization skills, and pioneered surgical
techniques for M. vancouverensis. Without him I could not have pursued this project. Other
veterinarians (M. McAdie, M. Smith, S. Saksida and H. Schwantje) performed marmot surgeries
and taught me much about the ethics and techniques of handling wild animals. G.W. Smith and M.
deLaronde worked their crews hard to count marmots and find colonies. Their diligence and
devotion to “getting the data right” was exemplary.
Marmots pay scant attention to institutional boundaries, so it is fitting that employees of many
institutions cooperated in exceptional ways to help. B. Kurtz and B. Brough (MacMillan Bloedel
Limited) provided unrestricted field access. D. Lindsay provided similar arrangements on behalf of
TimberWest Forests. E. Meyer (Ministry of Forests) provided weather data, and B. Bevan
ix
(Ministry of Environment, Lands and Parks) and D. Spittlehouse (Ministry of Forests) taught me
to interpret it. Ministry of Environment, Lands and Parks personnel (K. Atkinson, K. Brunt, D.
Doyle, and G.W. Smith) provided marmot and predator-prey data. G. Dunworth (MacMillan
Bloedel), G. Miehn (Pacific Forest Products) and L. Giguere (TimberWest Forests) provided data
that allowed measurement of landscape change. D. Ravenstein (Pacific Spatial Systems)
introduced me to the complex world of digital mapping, M. Maddison spent many long hours
digitizing forest cover maps, and A. Hawryski arranged for necessary GIS facilities at Malaspina
University-College.
Funding came from many sources. In descending importance, these included: Forest Renewal
B.C., Habitat Conservation Trust Fund, Ministry of Environment, Lands and Parks, World
Wildlife Fund (Canada), Forest Alliance of B.C., TimberWest Limited, MacMillan Bloedel
Limited, Cowichan Valley Field Naturalists Society, Vancouver Island Marmot Liaison
Committee, Canadian Wildlife Service, Nanaimo Field Naturalists and a large number of private
donors. Academic funds were relatively small but in some years made the difference between
quitting and continuing (Province of Alberta Graduate Scholarship, University of Calgary Thesis
Research Grant, Canadian Wildlife Service Research Grant, King-Platt Memorial Award, Franc
Joubin Graduate Bursary in Environmental Science, and University of Victoria graduate award).
I’m indebted to the Nature Trust of British Columbia, which until recently administered the
Vancouver Island Marmot Recovery Fund. Special thanks to R. Erickson and H. Torrance, who
kept me from starving on many occasions.
Finally. To my parents, my PhD committee, the Vancouver Island Marmot Recovery Team,
and friends and colleagues on four continents. Thank you. For your patience, guidance, restraint,
insight, and experience. And for teaching me that hypotheses are expendable but that dreams are
not.
x
FRONTISPIECE
The Marmot
On an early spring morning a marmot is born.
It eats grass but not any corn.
It lives in a burrow and not in a tree.
Its life is interesting and carefree.
Now you know a bit about the marmot.
If you read my story you’ll learn a lot.
Alex Dezan (at age 7)
Stanstead Journal, Quebec, Jan. 7 1996
(reprinted by kind permission of the author’s parents)
“...it might be worth while getting to know a little about geology or the movements of the moon
or of a dog’s tail, or of the psychology of starlings, or any of those apparently specialized or
remote subjects which are always turning out to be the basis of ecological problems encountered in
the field.”
Charles Elton (at age 26)
Animal Ecology (1927)
“How often have I said to you that when you have eliminated the impossible, whatever
remains, however improbable, must be the truth?”
Sir Arthur Conan Doyle (at age 31)
The Sign of the Four (1890)
1
INTRODUCTION
Marmota vancouverensis
The Vancouver Island marmot (Marmota vancouverensis: Swarth 1911) is endemic to
Vancouver Island, British Columbia (Nagorsen 1987). Like all 14 currently recognized species in
the genus, M. vancouverensis is fossorial, herbivorous and hibernates during winter (Barash
1989). The species was described from specimens collected in 1910 (Swarth 1912). Marmota
vancouverensis is distinguishable from other marmots by karyotype (Rausch and Rausch 1971),
skull characteristics (Hoffmann et al. 1979), pelage (Nagorsen 1987), behavior and vocalizations
(Heard 1977, D. Blumstein, University of Kansas, pers. comm.). In most respects it is a typical
alpine-dwelling marmot, showing slow maturation, a relatively long life span, and a complex
degree of social organization (Bryant 1996). The species is notable for its highly restricted range
and pronounced metapopulation structure (Bryant and Janz 1996).
Virtually nothing was known about the ecology or distribution of M. vancouverensis prior to
the 1970s (Heard 1977). Since then it has been the subject of systematic population counts
(reviewed by Bryant and Janz 1996), behavioral studies (Heard 1977), habitat and diet
investigations (Milko 1984, Martell and Milko 1986), palaeontological research (Nagorsen et al.
1996), genetic work (Bryant 1990) and demographic analyses (Bryant 1996). These studies
greatly improved our knowledge of the species and its precarious conservation status.
Marmota vancouverensis is listed as endangered under the B.C. Wildlife Act and regulations
(Munro et al. 1985). It is similarly listed by the Committee on the Status of Endangered Wildlife
in Canada (Munro 1979), the U.S. Endangered Species Act (Federal Register, Jan. 23 1984), and
the International Union for the Conservation of Nature (Groombridge and Mace 1994). A
Recovery Team was struck in 1988 and a recovery plan was prepared in 1990 (Bryant 1990),
published in 1994 (Janz et al. 1994) and recently revised (Janz et al. in prep). Marmota
vancouverensis has the dubious distinction of being the only endemic mammal species in Canada
that is listed as endangered (Bryant 1997), and is arguably one of the rarest animals in the world.
The origin and evolutionary history of marmots on Vancouver Island is inextricably linked to
climatic and glacial processes and associated changes in sea levels and habitat conditions. It
remains unclear when marmots first colonized Vancouver Island. Heard (1977) speculated that
marmots crossed to Vancouver Island via land connections that existed during the Illinoian glacial
period, approximately 100,000 years ago, and survived subsequent glacial maxima on nunataks
2
and narrow coastal refugia or both. Nagorsen (1987) suggested the possibility of a more recent
colonization, after the retreat of the Cordilleran Wisconsin glaciation some 10,000 to 13,000 years
ago (see Pielou 1991). Existing evidence does not permit exclusion of either hypothesis. However,
DNA analyses currently underway may clarify the phylogenetic relationships between M.
vancouverensis, Olympic marmots (M. olympus) and hoary marmots (M. caligata), together with
the timing of their divergence (R. Hoffmann, Smithsonian Institution, pers. comm.).
Retreat of the Cordilleran glacier during the past 10,000 years ensured that marmots became
increasingly restricted as forest succession occurred (Nagorsen et al. 1996). Since then M.
vancouverensis has apparently been confined to sites at which snow movement or fire maintained
open meadows in sub-alpine habitats (Milko 1984). Habitat restriction is the fundamental reason
why M. vancouverensis is rare and may also explain some aspects of its morphology and behavior.
For example, Hoffmann et al. (1979) suggested that the rich dark fur of M. vancouverensis
represents a melanistic phase that became genetically fixed in a small founder population.
Similarly, the highly social nature of M. vancouverensis compared to other marmots (Heard 1977)
has been interpreted as reflecting the evolutionary importance of being tolerant towards unrelated
strangers (Bryant and Janz 1996). Under this interpretation, social acceptance of immigrants
would encourage “rescue effects” (Brown and Kodric-Brown 1977) or colonization of unoccupied
habitats.
Changes in distribution and abundance
Location records indicate that Vancouver Island marmots inhabited a considerably broader
geographic range in the recent prehistoric (Nagorsen et al. 1996) and historical past (Bryant and
Janz 1996). They apparently disappeared from substantial portions of Vancouver Island north of
Alberni Inlet within the last several decades (Figure 1). Unfortunately, population data do not exist
with which to evaluate either their recent (post-1900) abundance or the timing of declines on
central Vancouver Island. Currently the species is restricted to fewer than 25 sites on 13
mountains. Apart from small colonies on Mount Washington, all active sites are located within 5
adjacent watersheds on southern Vancouver Island (Nanaimo, Cowichan, Chemainus, Nitinat and
Cameron River drainages).
Population trends on southern Vancouver Island within the last 25 years are intriguing. Many
colonies expanded during the early 1980s and this was accompanied by colonization of some new
habitats created by clearcut logging of forests above 700 m elevation. However the expansion into
clearcuts was limited in geographic and temporal terms (Bryant and Janz 1996).
3
Figure 1: Historical and current distribution of Vancouver Island marmots. Extinction dates
are approximate and based on sighting reports, burrow conditions and specimen
data. Most of the population is found in a small area (150 km²) on private lands
owned by MacMillan Bloedel Limited and TimberWest Forests. A few marmots
live on lands owned by Mount Washington Ski-hill Corporation. Land tenure in
this region has an interesting and convoluted history resulting from the Esquimalt
and Nanaimo Railway Land Grant Act of 1883.
4
Despite evidence of reproduction and survival in new habitats created by forest harvesting,
marmot numbers subsequently declined from a peak of over 300 animals during the mid-1980s to
the present total of fewer than 100 animals. At least five hypotheses have been proposed to explain
recent population dynamics in M. vancouverensis.
1) Habitat tracking hypothesis
2) Weather hypothesis
3) Sink-connectivity hypothesis
4) Predator hypothesis
5) Disease hypothesis
The hypotheses are not mutually exclusive and there is no a priori reason to imagine that a
single factor is responsible for observed population trends. However, such hypotheses are useful in
structuring thought and generating testable predictions. In that sense they are critical to pursuing
what Caughley and Gunn (1996) described as the “diagnostic” phase of endangered species
management. Without understanding there can be no hope of identifying the causes of decline or
reversing them.
Habitat tracking hypothesis
Thomas (1994) suggested that many rare species “track” habitat conditions, becoming locally
extinct when conditions are no longer suitable and colonizing sites when conditions improve.
However, issues of temporal and spatial scale are important to understanding the processes and
potential significance of habitat change. Vancouver Island marmots may be tracking habitat
conditions at a variety of different scales.
Over periods spanning centuries or millennia, habitat tracking could be caused by global
climate change and consequent reduction in the geographic area over which suitable conditions
occur. Discovery of M. vancouverensis remains from sites well outside its historical range
provides support for this idea (Nagorsen et al. 1996), as does similar distribution of alpine marmot
(M. marmota) remains in parts of western Europe (Preleuthner et al. 1995). While undoubtedly
correct, interpretation of habitat tracking at this temporal or spatial scale does not provide useful
insight into recent M. vancouverensis dynamics.
However, tracking could also occur over a temporal scale measured in marmot generations and
a spatial scale measured in hectares. Vegetation changes could result in altered survival or
reproductive rates. In natural sub-alpine meadows, possible mechanisms of habitat change include
5
invasion by trees or bracken ferns (Pteridium spp.), fire or changing food-plant availability (Milko
1984, Martell and Milko 1986, Laroque 1998). Forestry is the principal agent of change for other
habitats relevant to M. vancouverensis. Clearcutting and subsequent forest regeneration are
exceptional because they can act over a temporal scale measured within the lifetime of individual
marmots. Specifically, the extent of clearcuts and timing of their availability would be expected to
influence colonization events because marmots do not inhabit mature forests (Bryant and Janz
1996). Rapid forest regeneration in clearcuts could influence demographic performance or make
habitats unsuitable within a few years.
The tracking hypothesis predicts that marmots will respond to habitat change in deterministic
fashion. However, the speed of the response would necessarily be related to the rate of habitat
change. For colonies in natural sub-alpine meadows, gradual processes such as tree invasion lead
to the expectation of slow decline in survival or birth rates as individual habitats become
increasingly unsuitable. There is no reason to expect that change would occur simultaneously at
all sites. Given the short duration of this study compared to rates of change in sub-alpine meadows
(Kuramoto and Bliss 1970, Schreiner and Burger 1994), one would therefore expect to observe
chronic low birth or survival rates at a subset of natural colonies.
For marmots inhabiting clearcuts the expectations are somewhat different because habitat
change occurs more rapidly. The successional state of regenerating clearcuts could influence birth
or survival in linear fashion (i.e., a gradual reduction as a function of increasing forest age).
Alternatively it could be manifested by a threshold effect, in which conditions become unsuitable
for birth or survival over a period of a few years. Finally, a basic premise of the tracking
hypothesis is that marmots should increase in proportion to habitat availability and decrease when
habitats become unsuitable. The colonization process would necessarily be limited by the number
of potential colonists in the area. However, there exist numerous cases in which marmots
expanded from zero to more than 20 individuals within a short period (Bryant and Janz 1996).
One therefore predicts that marmot populations would increase numerically and spatially in
relation to clearcut availability and population size.
Weather hypothesis
Annual weather patterns could result in altered survival or reproduction, particularly because
marmots are presently restricted to such a small geographic area. One author attributed
Vancouver Island marmot expansion during the early 1980s to a period of “mild winters” although
he did not explain precisely what “mild” meant or how it would relate to hibernating marmots
6
(Smith 1982). However, snow depths, timing of snowpack melt and summer rainfall have been
associated with demographic success in other marmot species (Barash 1973, Van Vuren and
Armitage 1991, Armitage 1994,) and it is possible that weather could exert important and
measurable effects upon M. vancouverensis.
The basic prediction of the weather hypothesis is that one would expect to observe high annual
variance in survival or reproductive rates corresponding to years of “good” or “bad” weather.
Although some effects of weather could be influenced by site-specific conditions (e.g., different
snowmelt patterns at high and low elevation colonies), the expectation is that all colonies within the
Nanaimo Lakes metapopulation would experience similar weather. There should therefore be no
correlation of demographic trends as a function of between-colony distance (i.e., uniform spatial
correlation). In addition, weather patterns would presumably vary randomly over time. The
prediction is therefore that mortality or reproductive rates would show “episodes” of low
performance due to unsuitable conditions, and that these would also occur randomly over time.
Sink-connectivity hypothesis
Pulliam (1988) suggested that populations could be regulated by differential habitat quality.
He demonstrated mathematically that organisms can be most abundant in particular habitat
“patches” but be less successful there (sink habitats) provided that continued influx of individuals
occurs from nearby areas in which organisms do relatively well enough to provide a surplus
(source habitats). Complementary ideas have focused on the ability of organisms to disperse
successfully through a complex landscape (Dunning et al. 1992), form new subpopulations
(Hanski and Gilpin 1991) or “rescue” subpopulations that have experienced poor survival or
reproduction.
For Vancouver Island marmots the sink-connectivity hypothesis followed from suggestion of
reduced marmot survival in clearcuts (Bryant 1990, 1996) together with spatial concentration of
colonization events (Bryant and Janz 1996). The essential idea is that clearcut habitats may
intercept dispersing marmots by offering nearby habitats in which to settle (Bryant 1990). If these
habitats act as “sinks” the result may be to reduce long distance dispersal by intercepting
dispersers and providing them with attractive but sub-optimal habitats in which to settle. If this
hypothesis is correct then M. vancouverensis should not respond in proportion to clearcut habitat
availability. Instead, the prediction is that clearcut colonies should show chronic low demographic
performance. Metapopulation theory also predicts that more isolated colonies should receive fewer
immigrants and show higher extinction rates. Finally, the sink-connectivity hypothesis predicts
7
that colonizations of clearcuts should be spatially concentrated and that colonizations would rarely
occur at more isolated locations.
Predator hypothesis
Predators can play a significant role in regulating prey populations, particularly when prey
populations are low (reviewed by Flowerdew 1987). For example, mustelids apparently spend
much effort hunting voles even when vole density is low (Fitzgerald 1977). Such situations may
result in what has come to be known as the “predator-pit” phenomenon, in which predators exert
pressure on low-density prey populations sufficient to prevent their recovery (Haber 1977).
Predators could act as a limiting factor for Vancouver Island marmots, particularly given the small
size of colonies and their limited geographic distribution.
The predation hypothesis follows from evidence of mortality caused by predators such as
cougars (Felis concolor), wolves (Canis lupus) and golden eagles (Aquila chrysaetos). Marmota
vancouverensis apparently evolved in the presence of these predators, so the problem is not simply
that of exposure to a “novel” predator (see Vitousek 1988 for a description of this problem for
island endemics). However there are several possible mechanisms that may have increased
predation pressure. These include increased predator populations, depressed alternative prey
abundance and consequent “switching” of hunting effort (Bergerud 1983), increased predator
mobility or increased hunting success by individual predators (Bryant 1997).
One prediction of the predation hypothesis is that marmot survival would be associated with
predator abundance (or abundance of alternative prey such as deer if the “switching” idea is valid).
However, spatial or temporal patterns of mortality could also be relevant. It is unlikely that
mortality due to predation would show episodic pattern and be concentrated within particular years
or at individual colonies. Most potential marmot predators are long-lived compared to marmots.
In addition the ability to become successful at hunting marmots presumably represents learned
behavior that would not be exercised sporadically. While most predators are highly mobile
compared to the 150 km² area of the Nanaimo Lakes metapopulation, it also seems logical to
predict that they would focus hunting efforts in areas where success is maximized. For these
reasons a basic prediction of the predation hypothesis is that marmot survival should be spatially
correlated as a function of decreasing between-colony distance. Finally, because predation does
not occur during winter, mortality should be evenly distributed throughout the summer active
season.
8
Disease hypothesis
Recent M. vancouverensis dynamics may be caused by disease or parasites. The disease
hypothesis originated after four marmots died during hibernation after being transplanted (Bryant
et al. in press). In this case the presumptive cause of death was a bacterial infection (Yersinia
fredericksenii). Although Yersinia was detected in other marmot species (Bibikov 1992), to my
knowledge this represents one of the few cases in which marmot mortality has been directly
associated with disease. It remains unclear what factors precipitated the disease event, or whether
the organism is native to Vancouver Island. It is possible that the bacterium was present on the
release site or caused extinction of marmots at sites elsewhere in the past. It is also possible that
the organism represents a new disease, that introduced species such as eastern cottontail rabbits
(Sylvilagus floridanus) are acting as novel disease vectors, or that “normal” low-level marmot
health problems are exacerbated by environmental conditions.
Disease-induced mortality could show episodic pattern if the organism is particularly virulent.
In this case mortality would be expected to be concentrated at particular colonies or years.
Alternatively, mortality due to disease could show chronic pattern if the effect of the organism is to
slightly depress survival rates. This leads to the contrary prediction that mortality would be
temporally correlated within colonies and among years. However, epidemiological theory (e.g.,
May and Anderson 1972) suggests that in either case one would expect mortality to be spatially
correlated. Disease events would be more likely to occur in areas of high marmot density and,
depending on the mode of transmission, would be expected to occur more frequently at nearby
colonies.
Finally, much of the potential impact of disease depends on the nature of the organism and its
history of interaction with M. vancouverensis. For example, pathogens that are native to the
environment would be expected to cause abrupt pulses of mortality followed by a return to normal
conditions after virulence decreases (either because surviving marmots are more resistant or
because other conditions change; e.g., Blake et al. 1991). Alternatively, a non-native organism (or
a non-native means of exposure) might be expected to produce a growing incidence of high
mortality events with no subsequent recovery.
Non-exclusive hypotheses, testable predictions and practical significance
The question of why M. vancouverensis is declining is inherently complex. The five general
hypotheses are not mutually exclusive and result in many testable predictions. Some predictions
are related to only one of the hypotheses, but others relate to two or more. For this reason my
9
approach was to construct sets of predictions that were both amenable to analysis and would allow
inference to be made based on the cumulative “weight of evidence” (Platt 1964).
The predictions and expected nature of the relationships are as follows:
Tracking
Weather
Sink Predator
Disease
1.
Reproduction and survival will be chronically
low at some natural colonies.
yes no - - -
2.
Reproduction and survival will be associated
with habitat type or site characteristics such
as elevation or aspect.
no - yes - -
3.
Colonization of clearcuts will occur in
proportion to their availability.
yes - no - -
4.
Reproduction and survival will be associated
with age of regenerating clearcuts.
yes - - - -
5.
Reproduction and survival will be associated
with weather measurements.
- yes - - -
6.
Reproduction and survival will be chronically
low in clearcuts.
- yes - -
7.
Isolated colonies will show higher extinction
rates.
- - yes - -
8.
Isolated colonies will show lower apparent
survival due to reduced immigration.
- - yes - -
9.
Colonizations of clearcuts will be spatially
concentrated.
no - yes - -
10.
Survival will be spatially correlated. no no no yes yes
11.
Survival will be density-dependent. no no no - yes
12.
Survival will be associated with abundance
of predators or alternative prey such as deer.
- - - yes -
13.
Episodes of high mortality will occur
randomly over time.
- yes - no maybe
10
Tracking
Weather
Sink Predator
Disease
14.
Incidence of high mortality episodes will
increase over time.
- no - maybe
maybe
15.
Most mortality will occur during summer. - - - yes -
The practical significance of the hypotheses is that they suggest different interpretations about
the feasibility of recovering M. vancouverensis populations and about the direction of management
activities.
Specifically, if marmot declines are primarily caused by long-term changes in climate (habitat
tracking) then efforts to re-establish colonies on central Vancouver Island will likely fail and there
may be little that can be done to recover marmot populations. On the other hand, if habitat
tracking is manifested by marmot response to tree invasion, then removal of trees could be a simple
and inexpensive habitat enhancement technique. Retention of the weather hypothesis would yield
few management possibilities but might offer hope that recent dynamics represent a temporary
aberration and that conditions will improve on their own. If forestry has created “sink” habitats
and influenced dispersal (sink-connectivity hypothesis), then marmots on southern Vancouver
Island are in serious jeopardy and the only possible strategy is the one currently proposed: captive-
breeding combined with reintroductions. Retention of the predator or disease hypotheses would
reinforce this interpretation and raise additional management issues such as predator control and
removal or quarantine of animals from the wild.
11
STUDY AREAS
The Nanaimo Lakes marmot metapopulation is located within the Coastal Western Hemlock
and Mountain Hemlock biogeoclimatic zones of the Georgia Depression Ecoprovince (Demarchi
1988). This region is characterized by an effective rain shadow in the lee of the Vancouver Island
Mountains, and consequently is much dryer than sites on the west coast of Vancouver Island
(Campbell et al. 1990a). Mountains are typically lower in elevation and somewhat less rugged
than are the mountains of central and northern Vancouver Island.
Population data were obtained from the entire Nanaimo Lakes metapopulation (Figure 2).
Data from Mount Washington colonies on central Vancouver Island were excluded because of
small sample sizes, because colonies were infrequently sampled, and because it is unlikely that
dispersal occurs between that mountain and the southern metapopulation (Bryant and Janz 1996).
Five intensively studied colonies illustrate the variety of habitats occupied by Vancouver Island
marmots.
The Haley Lake and Green Mountain sites are steeply sloped (30° to 45°) south or southwest-
facing meadows kept free of trees by snow-creep and avalanches. Elevations are 1040 and 1420 m
respectively. Common plant species included Phlox diffusa, Castilleja spp., Erythronium
grandiflorum, Saxifraga ferruginea, S. occidentalis, Anaphalis margaritacea, Aster foliaceus and
Lupinus latifolius (Milko 1984, Milko and Bell 1985). Both sites had numerous boulders and rock
outcrops that marmots use as “loafing spots”. The mountain summits above the marmot meadows
are parklands of mountain hemlock with small ponds and a heavy shrub layer of Vaccinium spp.,
Phyllodoce empetriformis, and Rhododendron albiflorum. Soils on the meadows themselves
consist of colluvial veneers (<1m) overlying bedrock. Bedrock outcrops occur on the upper slopes,
with deeper colluvium on the lower slopes. The Haley Lake and Green Mountain colonies were 8
km apart, but connected by a ridge system that runs north-south. Both sites have a long history of
marmot occupancy, with records dating from 1915 (Haley Lake) and 1954 (Green Mountain).
The Vaughan Road clearcut colony is located 1 km west of the Haley Lake natural colony, in
an area that was logged between 1974 and 1978 (elevation is 940 m). Marmots were first
observed there in 1983. Aspect is west-southwest and the site is surrounded by steep hills to the
east and west. The Pat Lake site is a steep north-facing bowl surrounding a shallow lake 16 km
southeast of Haley Lake and 2 km east of Mount Whymper, where marmots also occur.
12
Figure 2: The Nanaimo Lakes metapopulation. This map illustrates cumulative conditions.
Not all colonies were occupied simultaneously. Locations of intensively studied
colonies are underlined. Snowpack sampling stations () and the Copper Canyon
weather station () are also indicated.
13
Elevation at Pat Lake was 900 meters. The site was logged between 1978 and 1979, and
marmots were first discovered there in 1985. The Sherk Lake site was a south-facing slope at 980
m elevation on the southern flank of Mount Landalt. The area was logged in 1977, and marmots
were first reported there in 1992. The Sherk Lake colony was within 2 km of another Mount
Landalt location where marmots were reported in natural meadows during the 1980s and not
subsequently. Vegetation at clearcut colonies differs from that at natural sub-alpine meadows,
although systematic vegetation work has not been performed on them. Trees were generally
dominated by alder (Alnus sitchensis) and regenerating conifers. Many wildflower species found
at natural meadows were not present in the clearcut sites, although L. latifolius, A. margaritacea
and Epilobium angustifolium were common.
M. vancouverensis inhabits other vegetation types as well. For example, the habitat on Mount
Buttle is dominated by scattered alpine fir (Abies lasiocarpa) and mountain hemlock (Tsuga
mertensiana) interspersed with dwarf juniper (Juniperus communis) and blueberries (Vaccinium
spp.). Marmots on the northwest ridge of "P" Mountain live on steep cliffs and talus slides, while
those on Mount Heather and Westerholm Basin live amidst willow (Salix) thickets interspersed
with rock slides.
14
METHODS
Population counts
Marmot counts were made by many persons (see Acknowledgments). Methods were basic:
scanning meadows and cliffs with binoculars or spotting scopes, listening for marmot whistles and
searching for burrows, scat and mud-stains on rocks, stumps or logs. Marmots were classified as
adults, yearlings or pups (young-of-the-year) based on size and pelage. The latter are readily
identifiable by their small size and dark, almost black, pelage (Nagorsen 1987). Yearlings can be
distinguished by their uniform pelage color and small relative size, although it becomes more
difficult in late summer. Most counts were conducted before 1100 hours to coincide with known
marmot activity rhythms (Heard 1977). Counts of pups were made after early July, when they first
emerge from their natal burrows (Bryant 1996).
Count data provided minimum numbers of adults, yearlings and pups present for each site-year
combination. Daily count tallies were considered as repeated measures (Krebs 1989), and I took
the highest annual count for each age-class to represent minimum population sizes for each colony
(hereafter the “observed” number). For each site I also defined the long-term average number of
adults, yearlings and juveniles across years as the “expected” number. The extent of annual count
coverage was estimated by summing the expected numbers from the colonies that were actually
counted, and expressing this as a proportion of the expected total had every colony been counted.
Count intensity was expressed as the total number of counts made per site-year combination.
Population estimation
To estimate population size I first calculated sums of observed and expected numbers using
those sites that received at least one count. The observed/expected ratio is therefore an index of the
extent to which observed numbers differ from long-term average. I also summed the expected
numbers from all colonies presumed to be occupied (occupied). To do this I assumed that all natural
colonies were occupied even if they did not receive counts (i.e., I included their contribution).
However for clearcuts I assumed that they were not occupied prior to the year of discovery, and
that they would become unsuitable 20 years after logging. If one assumes that trends in the overall
population are reflected by colonies that received counts, then a crude estimate of population size
can be obtained by multiplying observed/expected by occupied.
The assumption that trends at colonies receiving counts are representative of trends elsewhere
is probably reasonable for years in which count effort was extensive (1980-1986 and 1992-1997).
15
It is more tenuous for years in which few colonies were counted (1972-1979 and 1987-1991). To
minimize error I did not calculate observed/expected * occupied for years in which fewer than five
reproductive colonies in natural habitats (25% of the total colonies and 35% of occupied) were
counted.
Finally I applied correction factors to account for probable count underestimation. The
correction factor for adult marmots was based on the average number of counts made per site-year
at non-intensively studied colonies, using a regression formula obtained from the probability of
resighting tagged marmots at intensively studied colonies (Bryant and Janz 1996). In practice, the
correction factor varied from 1.19 to 1.66 (average = 1.40, a value similar to that obtained for
alpine marmots: 1.25: Cortot et al. 1996). Because juvenile marmots typically emerge in July there
is little time in which to conduct repeated counts and the same statistical approach could not be
used to correct the results. Instead a constant multiplier (1.2) was used. This multiplier was
obtained from average litter size at intensively studied colonies divided by average litter size at
other colonies (Bryant and Janz 1996). Correction factors were applied to the total observed
numbers of adults and juveniles within natural and clearcut habitat classes and not to individual
colonies. Correction of individual site-year estimates was unjustifiable because some colonies were
probably counted accurately despite receiving few repeated visits.
Capture, handling and age-assignment
Marmots were ear-tagged and monitored at five colonies from 1987 through 1998 (Green
Mountain, Haley Lake, Pat Lake, Sherk Lake and Vaughan Road). At these sites most animals
were ear-tagged and some animals were radio-telemetered. Numerous repeated visits provided
accurate population estimates for most years (Bryant 1990, 1996). Capture rates were high and
reproductive performance, persistence, and immigration rates could therefore be estimated with
precision.
Marmots were captured using raccoon-sized single-door Havahart traps (model 1079,
Woodstream Corporation, Littitz, PA) baited with peanut butter. Once trapped, marmots were
transferred to a cone-shaped canvas handling bag. The large opening was placed around the
Havahart trap and the door was opened, whereupon the marmot would run into the bag and be
physically restrained as the bag narrowed.
A mixture of Ketamine hydrochloride (Rogarsetic®, Rogets Pharmaceuticals, Vancouver, BC)
and Midazolam (Versed®, Hoffman-La Roche Ltd., Missisauga, ON) was used to facilitate animal
16
handling. Dosage was normally 40 mg/kg of Ketamine and 5 mg/kg of Midazolam, following
guidelines established through experience and veterinary collaboration (Woodbury 1997).
Injections were made intramuscularly, in the lumbar muscles, through the handling bag. Note that
with this dosage and protocol, animals were sedated but not completely immobilized. The drug
normally took effect within five minutes of injection and the animal could then be removed from the
handling sock. A Bacitracin-Neomycin-Polymyxin ophthalmic ointment was used to protect the
animal's eyes during handling (Vatropelycin®; Altana Inc., New York, NY).
Morphological data were recorded at time of capture: sex, weight, total length including tail,
body length excluding tail, tail length, neck circumference, chest circumference, length of hindfoot
from toe to edge of pad, and length of foreleg from toe to elbow. Weights were measured to the
nearest 100 grams using a spring scale; all external measurements were made with a flexible
plastic metric tape to the nearest mm. Sex determination was made by everting the genitalia,
palpating for testes and/or by measuring the distance from anus to genital opening (Heard 1977).
Pelage characteristics, abundance of parasites, fat condition and any external characteristics, such
as scars, which could aid in re-identification were noted. Marmots were placed in one of four age-
classes at time of capture using the following criteria:
Juveniles (young-of-the-year): small body size (body length = 30-47 cm, forearm length =
10.1-13.0 cm, weight = 1-3.75 kg), uniformly dark pelage (Nagorsen 1987) with no faded fur, first
observation in late June or early July (Bryant and Janz 1996, Heard 1977), and observed
emergence from natal burrows.
Yearlings (1 year-olds): Any small, dark marmot captured prior to mid-June was
unquestionably a yearling (Bryant and Janz 1996, Heard 1977). In practice, juveniles and
yearlings were distinguishable well past this date, as yearlings were larger (body length =35-54 cm,
forearm = 12.0-15.5 cm, weight = 2.0-4.75 kg). By late August, most yearlings are either in faded
overall pelage, or are in partial molt (unpublished photographs of known-age yearlings).
Sub-adults (2 year-olds): Most “first-time” captures were assigned to this category by default.
Marmots in this age-class were full-sized (body length = 44.2-55.5 cm, forearm = 12.7-17.1 cm,
weight = 3.5-5.5 kg) but were non-reproductive. In May and June, 2 year-olds have usually
completed their first molt and exhibit a uniformly dark pelage, but often show a patch of faded
(rufous) fur on the dorsal surface at the base of the tail (unpublished photographs of known-age
animals, this study).
17
Adults (3 years and older): Large-bodied males (>60 cm, forearm >16 cm, weight >5.5 kg)
and all reproductive females were initially classified as 3 year-olds. Molt patterns are
unpredictable beyond age 2 (unpublished photographs of known-age animals, this study) but
typically show a mottled appearance of old (faded) and new fur.
Data from animals originally captured as juveniles or yearlings were coded as “known-age”
data, and data from other animals were coded as “presumed-age” data. My aging protocol was
intentionally conservative, and undoubtedly underestimated the true age of older animals. The
reverse is not true. It is unlikely that I overestimated marmot ages using the above criteria.
Marmots were equipped with ear-tags in both ears (monel self-piercing tags, style #1005-3,
National Band and Tag Company, Newport, KY).
Surgical implantation and radio-telemetry
Radio transmitters were surgically implanted in order to determine causes of mortality and
movement patterns. I used two types of radio transmitters (from Custom Telemetry, Watkinsville,
GA, and Telonics, Mesa, AZ). The former were identical to those used by Van Vuren (1989), but
performance was characterized by weak signal strength and relatively short battery life (overall
dimensions = 110 x 20 mm, weight = 35 grams). In 1994 I switched to Telonics units (model IMP
300), which contained a larger battery (overall dimensions = 89 x 23 mm, weight = 40 grams).
Both transmitters featured temperature-dependent pulse rates (50-60 beats per minute at 35 C°).
Transmitters were encased in beeswax and sterilized by soaking in povidone-iodine solution for 24
hours prior to implantation.
Surgical implants were performed in the field by veterinarians (see Acknowledgments). After
preliminary sedation with injectable agents to facilitate handling, marmots were anaesthetized using
2.0-3.0% isofluorine gas (Aerrane®, Anaquest, Missisauga, ON) administered with bottled oxygen
and an Isotec® vaporizer (Ohmeda, Madison WI) mated to a small animal mask. Oxygen flow
rates were 2 to 3 liters per minute. After induction, anesthesia was maintained at reduced
isofluorine concentration (1.5-2.0%). This practice shortened recovery time to 15-30 minutes and
allowed precise control of the depth of anesthesia. Transmitters were implanted in the
intraperitoneal cavity while animals were restrained on a portable steel operating table. Other
deviations from Van Vuren’s procedure included incision through the linea alba to minimize
muscle trauma and blood loss, and the use of methyl-methacrylate glue (Vetbond®, 3M, St. Paul,
MN) to reinforce stitches. Antibiotics were not dispensed routinely, although enrofloxacin
(Baytril®) was used on several occasions after animals received superficial abrasions from traps.
18
Sterile saline solution was used in place of a povidone-iodine wash to irrigate incisions and clean
transmitters prior to implantation.
Instruments, masks, capes and drapes were autoclaved prior to use, and sterile conditions were
maintained as far as was possible. The surgical “drug kit” and “surgery kit” were carried in
backpacks and a waterproof plastic case, weighed <80 kg in total, and could be quickly positioned
by 3-4 persons even in steep terrain. Surgeries in low elevation habitats with good road access
were often performed by two persons. All animals were released within 1.5 hours of initial
capture.
Transmitters were monitored using a Telonics receiver (model TR-2) and either a two-element
“H” (model RA-2A) or folding three-element “yagi” antenna (model RA-3). The former gave
superior directionality. Signal-bounce from steep terrain often made radio-telemetry difficult.
Unless the animal was plainly visible, the normal procedure was to “walk down” the signal rather
than attempt to triangulate from compass bearings (e.g., White and Garrott 1990). Searches for
missing animals were conducted on foot, by road and occasionally by helicopter. Close proximity
to marmots was evaluated by removing the antenna to determine whether the signal was still
audible; in practice this typically occurred when transmitters were within 3-5 meters of the
receiver.
Colony-specific demographic rates
The finite rate of population increase () is the essential measure of colony success. By
definition, a population will increase if >1.0, be stable if =1.0 and decline if <1.0. There are
several methods to calculate but I used Pulliam’s (1988) basic formula because it corresponds
well to the types of data that can be obtained from marmot counts. The formula is:
= PA + (PJ * )
in which PA is the annual probability of adult survival, PJ is the annual probability of juvenile
survival, and is the annual per capita birth rate. Colony-specific annual rates were compiled
from intensively studied colonies and non-intensive marmot counts as follows:
Adult survival: At intensively studied colonies, adult survival rates were estimated from
resightings of previously-tagged adults and yearlings. Given ear-tag loss, dispersal and individuals
that could have been missed due to low sampling effort, resightings yield a minimum adult survival
rate. At other colonies, “apparent” adult survival rates were based on consecutive annual counts of
yearlings plus adults and apparent survivors (i.e., minimum numbers of adults excluding yearlings
19
in the following year). Presence of immigrants ensures that this will yield an inflated survival rate
(using Pulliam’s terminology, this actually represents “i-d-e” or net immigration-death-emigration).
For cases in which numbers of adults increased in the consecutive year, apparent survival was
assumed to be 1.0, with the remainder assumed to represent immigrants.
Juvenile survival: Survival of pups (young-of-the-year) was estimated by comparing counts of
pups with counts of yearlings in the following year. Pups and yearlings apparently do not disperse
(Bryant 1996) so survival estimates should be robust if sufficient sampling effort was made.
Resightings of tagged pups at intensively studied colonies provided an independent estimate of
survival.
Per capita births: For all colonies I defined per capita birth rate as the total number of pups
divided by the total number of non-pups. I also calculated the probability of breeding (number of
litters divided by the number of non-pups) and average litter sizes (pups per litter) to test whether
per capita births accurately reflected these variables.
Relative density: Observed/expected ratios provided a measure of “saturation” or relative
density. This was calculated as the observed number of animals divided by the long-term expected
number for that colony. Relative density was estimated separately for adults and pups.
Immigration-emigration: Data were insufficient to estimate dispersal (emigration) rates.
However, some inference could be obtained from four independent sources of data. First,
resightings of tagged marmots at new colonies provided empirical information about the magnitude
and direction of dispersal movements, and sometimes allowed inference about the timing of
dispersal. Second, measurement of untagged immigrants at intensively studied colonies permitted
inference about the age-sex composition of immigrants. Third, location records for solitary
marmots in low elevation, non-typical habitats were compared with the location of the nearest
colony known to be active in that year. Resulting between-location distances permitted estimates
of minimum dispersal distances (these will be underestimates unless animals originated in the
nearest colony, which is unlikely). Fourth, those cases for which apparent adult survival >1.0 (see
above) permitted assessment of when and where large influxes of immigrants may have occurred.
Mortality: Mortality patterns were impossible to describe with precision. Radio-telemetry
provided useful data about causes of death but sample sizes were small. Disappearances of tagged
marmots yield a maximum mortality rate (because some tagged marmots probably dispersed and
survived but were not seen again). To gain additional insight about possible factors I evaluated the
20
timing of last observation for each marmot that disappeared. My reasoning was that
disappearances that were concentrated at particular times could suggest dispersal (spring
disappearances) or mortality during hibernation (late-season disappearances). Conversely,
disappearances that were distributed throughout the active season could represent the effect of
constant mortality factors such as predation.
Lifetime reproductive performance: Tagged females that disappeared for at least one active
season (and were presumed to have died) were used to calculate lifetime reproductive performance
(i.e., the total number of juveniles produced by that female). Females confirmed to be alive in
1997 were excluded as they could reproduce again.
Colonizations and extinctions: Count records were used to compile discovery dates (earliest
record of occupancy), colonization date (for clearcuts only; many natural colonies probably existed
long before they were first visited by observers and extinction date (previously occupied sites that
had been vacant for at least 2 years prior to the 1997 season). Records of non-reproductive
“potential” colonies (Bryant and Janz 1996) were excluded. It is possible that these records
represented marmots “in transit” that did not remain at the location. For clearcuts, I calculated
longevity (extinction date minus colonization date). This calculation could not be made for natural
colonies because of the uncertainty over dates of colonization. To test whether clearcut location
was important to colonizing marmots, I randomly sampled 30 clearcuts of appropriate age and
elevation, within the apparent dispersal capability of marmots, in order to compare the spatial
location of these sites with those clearcuts that were actually colonized.
Landscape conditions
To measure conditions at marmot colonies, landscape change and potential clearcut marmot
habitats over time, I used a Geographic Information System (ARC/INFO; Environmental Systems
Research Institute, 1994) to create digital landscape maps. These maps contained topographic
features, forest cover data, roads and marmot locations. Software developed by the same
manufacturer (ARCVIEW 3.0) was used to query the resulting maps and measure landscape
conditions.
Colony-specific habitat conditions: Habitat variables included type (natural versus clearcut),
elevation (m above sea level), aspect (degrees of compass bearing) and patch size (in hectares;
clearcut habitats were excluded as it was normally impossible to accurately define the spatial
extent of marmot use). For clearcut colonies, the age of the regenerating forest was measured
21
(current year minus date of logging). Spatial locations were tabulated in UTM units (Universal
Transverse Mercator projection using the 1983 North American Datum). Two measures of
isolation were calculated: isolation (median distance of that colony to all other active colonies,
expressed in km), and nearest neighbor proximity (distance to the nearest active colony, also
expressed in km). To facilitate exploratory analyses, resulting data were then dichotomously coded
based on the median value obtained (i.e., high versus low elevation, exposed versus sheltered
aspect, large versus small habitats, young versus old clearcuts, and isolated versus clustered
colonies).
Landscape conditions: The size of the GIS study coverage was 1006 km² and included all
extant marmot colonies south of Alberni Inlet. Landscape measurements included the annual area
of forests classified by forest companies as mature (old-growth) and the annual area logged above
or below 700 m elevation. This demarcation was selected based on the apparent inability of
marmots to colonize low elevation habitats (Bryant and Janz 1996). Potential clearcut marmot
habitat was defined as the area of logged clearcuts above 700 meters in elevation and between the
ages of 0 and 15 years after logging. This definition probably overestimates the area of habitat
that could actually be used by marmots because it included sites of all slopes and aspect. Most
marmot clearcut colonization events occurred on north-west to south-east-facing slopes and on
relatively steep slopes.
Dates of logging road construction were unavailable from the raw digital data, although the
cumulative (1996) extent of the road network was available. I therefore assumed that roads were
constructed in relative proportion to the extent of logging activities, and queried the digital map for
roads that intersected current and previous clearcuts. This calculation yielded a minimum estimate
of road density. Logging roads deteriorate rapidly in the Vancouver Island climate and typically
become unusable after a few years if not maintained. However, because my purpose was to
explore the possibility of enhanced marmot or predator mobility, I reasoned that animals would
continue to use them long after they became impassable to vehicles, and therefore made no
allowance for forest regeneration along roads. Road densities were expressed as linear km of
roads/km².
Weather
Summer precipitation and temperature data and winter snowpack data were available from
several sources. Summer data included daily midday temperatures and total daily rainfall from an
automated weather station located in a clearcut at 840 m elevation in Copper Canyon (unpublished
22
data, B.C. Ministry of Forests). Average daily temperature and precipitation data were also
available from Nanaimo Airport (unpublished data, Environment Canada). Snowpack data were
available from Green Mountain (1400 m), Heather Mountain (1170 m) and Mount Cokely (1190
m: unpublished data, B.C. Ministry of Environment, Lands and Parks). From these raw data I
calculated several variables of possible relevance to marmots. These data cannot be assumed to
represent conditions at specific colonies but should reflect annual weather trends for the study area
as a whole.
Summer rainfall and temperature: I constructed variables that may influence adults (spring
conditions) or adults and juveniles (late summer conditions). Variables included average midday
temperature during May and June (in °C), average midday temperature during July and August,
total precipitation during May and June (in mm) and total precipitation in July and August.
Drought and nutritional effects upon vegetation could be also caused by differences in the timing of
rainfall. For this reason I also calculated the number of days with significant (>5 mm) rainfall
events in May and June, and number of days with significant rainfall events in July and August.
Finally, because early-spring snowmelt patterns could be driven by both rainfall and temperature, I
constructed additional variables that were “offset” by one year (to evaluate the possibility that next
year’s spring weather might influence survival of this year’s marmot cohort).
Winter snowpack: Monthly snowpack measurements were averaged among sites and
expressed in cm. Two variables were constructed to represent “early” snow conditions that may
influence hibernation physiology (average January-February conditions) and “late” conditions that
could influence hibernation duration or food availability (June). The latter was also offset by one
year to reflect the possibility that next year’s snowmelt affects survival of this year’s marmot
cohort. As with the case of the summer weather station, snowpack sampling locations did not
correspond precisely with marmot colonies and cannot therefore be interpreted to reflect local
conditions at specific marmot colonies.
Predator-prey trends
Predators: Currently the only long-term measure of terrestrial predator abundance on
Vancouver Island comes from sightings made by deer hunters (unpublished data, Ministry of
Environment, Lands and Parks). Numbers of cougars (Felis concolor) and wolves (Canis lupus)
seen by deer hunters were expressed per 100 hunter-days. These “hunter-sighting indices” have
not been tested for reliability against a known population, and data were obtained from an area
considerably larger than the area occupied by marmots (>1500 km²). Additional data were
23
available concerning numbers of animals “removed” due to trapping, hunting, animal control
programs and road-kills. It remains unknown how well these estimators reflect actual abundance.
I used both estimators for both species and constructed two additional variables by pooling relevant
data to estimate “terrestrial predator abundance” and “terrestrial predator removal”. No data were
available with which to assess abundance of hawks or eagles.
Prey: Predator impacts upon marmots could also be influenced by “switching” of hunting
effort. For this reason I included an abundance measure for black-tailed deer (Odocoileus
hemionus), an important prey species for wolves and cougars. I reasoned that deer abundance may
be inversely related to marmot survival. Abundance was estimated from systematic night counts
(Harestad and Jones 1981) and expressed as annual numbers of deer seen per km. The area of deer
counts (~150 km²) was centered on the Nanaimo Lakes metapopulation (unpublished data,
Ministry of Environment, Lands and Parks).
Statistical analyses
Preliminary analyses were conducted to assess the repeatability and coherence of the data.
How consistent were population counts and resulting population estimates? How well did
estimates of demographic rates from counts conform with those obtained from marked animals?
Were there obvious mechanisms (e.g., ear-tag loss, differential capture or dispersal rates) that
rendered the data unusable? Was it possible that mark-recapture efforts themselves caused
population declines? Exploratory analyses were pursued to uncover fundamental patterns in the
data. Were there temporal, spatial or habitat-specific trends in marmot survival or reproductive
rates? Did such trends facilitate testing of more detailed hypotheses? Results from exploratory
analyses permitted finalization of the data sets, removal of outliers and development of detailed
predictions. Final analyses were designed to test these predictions.
I used mean successive difference tests to evaluate consistency of population counts among
successive survey years (Zar 1974). Adult counts, juvenile counts and per capita birth rates were
tested separately. I also used intraclass correlation coefficients to assess the repeatability of adult
counts. Repeatability (R) varies from 0 to 1.0 depending upon the similarity of repeated
measurements (Krebs 1989). For this analysis I used raw count data in which the repeated
measures were multiple counts made for a given site-year. To determine whether count
performance differed with season, I also plotted the “success” of counts against number of days
after 30 April and fitted a locally weighted regression curve (LOWESS; Cleveland 1979). Success
was defined as the ratio of a given count to the highest count obtained for that site-year.
24
I also used LOWESS to plot population trajectories for individual clearcut and natural
colonies. For this I used the maximum adult count. Pup data were excluded because small
numbers of reproductive-age females at most colonies combined with the two-year breeding cycle
of most females (Bryant 1996) would induce high variance and would not assist in evaluation of
trend. Adult counts were square-root transformed prior to plotting to make the data more easily
interpreted. Pearson correlations were used to test for association with year.
To test whether marmots colonized clearcuts in proportion to their availability, I used
Spearman rank correlation of the number of colonizations with availability of clearcut habitat
above 700 meters elevation in any given year. I used Student t tests to determine whether the
nearest-colony-neighbor distance of actual colonizations was different from that of randomly
selected clearcuts within the study area.
Chi-square tests were used to evaluate independence of demographic rates calculated from
intensively studied colonies and non-intensive counts. The same method was used to test for
independence among subsets of colonies (high versus low elevation, exposed versus sheltered,
natural meadows versus clearcuts, and isolated versus clustered). Mean litter sizes, birth rates,
female lifetime reproductive success, and mean age at first reproduction were compared across
habitats using Mann-Whitney U tests or Student t tests for discrete and continuous variables,
respectively. Sex ratios of animals first captured as juveniles were tested against an expected 1:1
ratio using 2 goodness-of-fit methods (Zar 1974).
Demographic trends were evaluated using life-table analysis and mark-recapture models. Life-
tables were constructed using resightings of tagged marmots and raw frequencies of sex and age-
specific disappearances (Caughley 1977, Method 2). All rates were calculated from frequencies of
sex and age-specific disappearances (dx). To verify that the tagged marmot cohort reasonably
reflected actual trends, I re-calculated life-tables using a) tagged pup data alone, b) all pup data
including observations of surviving but untagged yearlings, c) fecundity estimated using the
observed ratio of male:female pups, and d) fecundity estimated assuming a 1:1 sex ratio at birth.
Fecundity was calculated using observed ratios of male:female pups. Because age-specific samples
were so small I did not smooth the Lx values (Krebs 1989) or test for differences in survivorship
curve shape (Pyke and Thompson 1986).
I used Cormack-Jolly-Seber models (Pollock et al. 1990) to estimate parametric survival rates
and 95% confidence limits. This analysis was performed using SURGE (Cooch et al. 1996) and
25
followed the Lebreton et al. (1992) approach in seeking the most parsimonious model that
explained the data. I also performed a basic elasticity analysis (Caswell 1989) to evaluate which
life-history characteristics were most important driving population growth. The analysis involved
modifying each of adult survival, pup survival and fecundity estimates in turn by proportional
amounts and by calculating the relative effect of each variable upon (finite growth rate in years).
I used RAMAS/age to calculate (Akcakaya and Ferson 1990).
I used stepwise logistic regression (Cox 1975) to test whether environmental conditions were
associated with survival or probability of reproduction. Logistic regression is similar to normal
multiple regression techniques except that it is designed to explain variation in a binary response
variable (i.e., a variable with a value of 0 or 1). Independent variables may be either continuous
and categorical. For this analysis I used coded marmot survival data (alive = 1) and reproductive
data (produced a litter = 1) together with sets of continuous or categorical environmental variables
associated with the predictions. Logistic regression uses McFadden’s Rho² statistic to evaluate the
amount of variation explained by independent variables. Like the analogous r² used in linear
regression, larger Rho² values indicate stronger relationships. However, Rho² is generally much
smaller, with values of 0.2 to 0.4 considered to indicate extremely powerful relationship (Hensher
and Johnson 1981). Logistic regression also yields an “odds ratio” that represents the “odds of
making a correct prediction.” Negative influences upon a binary response variable produce odds
ratios <1.0 while positive influences produce odds ratios >1.0. Overall regression significance was
evaluated by likelihood-ratio 2 tests (Hosmer and Lemeshow 1989). Tests were performed using
Systat 7.0 (SPSS Inc. 1997).
I used Moran’s I coefficient (Sokal and Oden 1978, Sawada 1998) to evaluate spatial
autocorrelation of survival rates. This statistic is analogous to Pearson’s R except that it is
designed to evaluate whether events are similar or dissimilar in a spatial context. Coefficients
close to 1.0 indicate that similar values tend to cluster together, and values approaching -1.0
indicate that dissimilar values tend to cluster together. The formula is:
IW
ij ij
ji
in which ij is a measure of the proximity of the variate (in this case apparent survival) between the
ith and jth spatial positions and Wij is a spatial weighting function that is a measure of connectivity
or “contiguity” of the locations. On a regular grid (quadrat data), contiguity (Wij) is typically set to
equal 1 for nearest neighbors and 0 if otherwise (Haining 1990). For spatially complex data (such
26
as locations of marmot colonies), contiguity can be assigned by applying an “effects radius”
relevant to the research question (Smith and Gilpin 1997). In this case only those neighbors within
a pre-defined radius are considered to be contiguous (Wij = 1).
I reasoned that mortality due to weather would occur over the entirety of the study area but
that mortality from predators or disease would occur over progressively smaller areas. I therefore
tested for spatial autocorrelation using cumulative effects radius established at 1 km intervals up to
a maximum of 15 km (at which the entire metapopulation was included in the effects radius).
Reasons of sampling effort (not all colonies were counted in any year), measurement error (annual
variation in survival was high) and predicted effects (predators are long-lived) suggested that I
evaluate change in spatial autocorrelation over two time periods: “early” (1979-1988) and “late”
(1989-1998). Annual survival estimates were treated as repeated measures within the two groups,
with care being taken to ensure that nearest-neighbor distances were measured only within years.
Significance of Moran’s I was evaluated using Z test scores (Sokal and Oden 1978). The null
hypothesis is that the observed distribution of events is no different from a distribution in which
values are randomly assigned to the same set of spatial locations.
Some analyses resulted in tests with low statistical power (Zar 1974) despite being based on a
large fraction of the extant marmot population. I caution that results may resemble parameters
more than they resemble sample statistics. I employed a conventional (=0.05) decision rule to
accept or reject null hypotheses, but leave it to readers to judge whether observed differences might
be biologically significant despite lack of statistical significance (Krebs 1989).
27
RESULTS
Part 1: The environment
Weather
Summer weather patterns on Vancouver Island varied considerably (Figure 3). Late summer
(July-August) measures of cumulative precipitation obtained from Nanaimo airport and Copper
Canyon stations were highly correlated (Pearson r = 0.89, n = 22 years, P < 0.01) but early
summer (May-June) measures were not (r = 0.31, n = 22 years, P > 0.05). Similarly, average
daily temperatures at Nanaimo airport in late summer (July-August) were correlated with average
midday temperatures from Copper Canyon (r = 0.53, n = 22 years, P < 0.05) but early spring
(May-June) temperatures were not (r = 0.37, n = 21 years, P < 0.05). Both temperature and
rainfall data varied more at Copper Canyon. This result is expected. Nanaimo airport is located in
a low elevation (100 m) coastal environment on the leeward side of the Vancouver Island
mountains. The weather station at Copper Canyon is located at higher elevation (840 m) and is
therefore more prone to “mountain weather”. Given the probability of site-specific effects of
weather upon marmots, there was no a priori reason not to use data from Copper Canyon, which is
closer to the mountains inhabited by marmots (Appendix 1).
Overall, some years were extremely hot and dry (1979, 1982, 1985 and 1994) and some years
were relatively cold and wet (1975, 1976, 1991 and 1993). Annual variation was high. In some
years there was snowfall in June (1988 and 1991) and in some years there was virtually no rainfall
during summer (1994 and 1996).
Winter conditions are less easily evaluated. Annual and monthly snowpack conditions from
Mount Cokely, Green Mountain and Heather Mountain were strongly correlated (r values from
0.68 to 0.83, n = 18 years, P < 0.05) but data were available from only a single location (Green
Mountain) after 1995. The general pattern is that snow begins to accumulate in December and
increases through April (Figure 4). There were substantial differences in snowpack accumulation
and melt patterns among study sites and among years (my unpublished photographs). Snowpack
data must therefore be interpreted as reflecting “average” annual conditions and not site-specific
snow depths (Appendix 2).
28
A) Average midday temperature
10
12
14
16
18
20
22
24
26
72 74 76 78 80 82 84 86 88 90 92 94 96 98
TEMPERATURE (C°)
May-June
July-August
B) Average daily rainfall
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
72 74 76 78 80 82 84 86 88 90 92 94 96 98
YEAR
RAINFALL (mm)
May-June
July-August
Figure 3: Four measures of summer weather conditions. Data are (A) daily noon
temperatures averaged across May-June, and July-September, and (B) average
daily rainfall averaged across the same periods at Copper Canyon.
29
A) Snowpack accumulation
0
50
100
150
200
250
300
350
Dec Jan Feb Mar Apr May Jun
SNOWPACK (cm
)
B) Annual snowpack trends
0
100
200
300
400
500
72 74 76 78 80 82 84 86 88 90 92 94 96 98
YEAR
Snowpack (cm)
Jan-Feb
June
Figure 4: Two measures of snowpack conditions. Data are (A) average and SD
monthly snowpack depths, and (B) early and late winter snowpack depths
across years.
30
Landscape change
GIS measurements illustrate the extent of commercial forestry operations on privately-owned
lands in the Nanaimo Lakes region. There was little forest harvesting prior to 1956 and much of
what occurred was concentrated along valley bottoms. This pattern continued through the 1960s.
Harvest rates increased during the 1970s, particularly at higher elevations. By 1976 over 75% of
the annual harvest occurred above 700 meters in elevation. At least 60% of all forests classified as
mature by forest companies within the Nanaimo Lakes region were harvested in a 25 year period
(Figure 5). Road development took place at a similar pace and increased fivefold in density during
the same period (Figure 6). Potential clearcut marmot habitat was first created during the late
1960s and large amounts (>10,000 hectares) became available during the 1970s (Appendix 3).
Predator-prey abundance
Black-tailed deer abundance declined dramatically from the mid-1970s through the mid-1990s.
Current populations are about 40% of the long-term average (Figure 7). Deer abundance estimates
were not highly correlated among the 4 sampled areas in the Nanaimo Lakes region, but all showed
substantial declines (Appendix 4). Within each sub-region estimates were serially correlated
among years (mean square successive difference tests, range of C values = 0.49 - 0.87, P < 0.05).
This is an expected result given deer longevity, and suggests that systematic roadside night counts
provided realistic estimates of deer abundance.
Wolf and cougar abundance indices varied greatly among years (Figure 7). Indices were not
correlated among years (mean square successive difference tests, C = 0.34 for wolves and 0.14 for
cougars, for both species n =15 years, P > 0.05). These results are unexpected given probable
predator longevity and territoriality. Localized predator control activities and incidental kills
occurred in the Nanaimo Lakes region throughout the study but were higher in some years. In
some years substantial numbers of predators were removed from the area (e.g., 24 wolves and 11
cougars in 1985; Appendix 5). However, there was no congruence between “removal” data and
hunter-sighting index in the following year (for wolves, Pearson r = 0.06, and for cougars, r =
0.04, n = 15 years, P > 0.05). For both species the hunter sighting index was just as likely to
increase as it was to decrease following years of high numbers of removals. I conclude that hunter-
sighting indices and removals probably do not reflect actual abundance of cougars and wolves.
31
Figure 5: Extent of forest harvesting over time. Data are unharvested habitats
including non-forested habitats (black areas) and harvested forests (white
areas) sampled at 10 year intervals from 1966 through 1996. Size of the
GIS study coverage was 1006 km², which corresponds almost exactly
with the extent of the Nanaimo Lakes metapopulation shown in Figure 2.
32
1972 1977 1982 1987 1992 1997
0
10,000
20,000
30,000
40,000
50,000
60,000
AREA (ha)
Old-growth
Clearcuts >700m
A) Forest conditions
1972 1977 1982 1987 1992 1997
YEAR
1.0
1.5
2.0
2.5
3.0
ROADS (km/km²)
B) Road density
Figure 6: Two measures of landscape change. Data are A) hectares of old-growth
forest and potential clearcut marmot habitats (clearcuts above 700 m in
elevation and 0-15 years old), and B) road densities (expressed as linear
km of road per km²).
33
1972 1977 1982 1987 1992 1997
0
5
10
15
DEER (per km)
A) Deer abundance
1972 1977 1982 1987 1992 1997
YEAR
0
1
2
3
4
SIGHTINGS (per 100 hunter-days)
Cougars
Wolves
B) Predator abundance
Figure 7: Predator-prey trends. Deer abundance (A) was estimated from nocturnal
counts and expressed as numbers seen per kilometer. Wolf and cougar
abundance (B) was a “hunter sighting index” and expressed as numbers seen
per 100 hunter-days.
34
Part 2: Sampling effort
Population count efforts
Observers counted marmots on 1711 occasions at colonies or potential colonies within the
Nanaimo Lakes metapopulation between 21 April and 7 October during the years 1972-1997.
Count coverage was relatively low prior to the 1980s and during the late 1980s. Not every site
was visited in each year. Coverage was particularly low in years in which I worked alone or with a
single technician (1987-1991). Coverage was higher in years for which dedicated crews were
available to count marmots at known colonies (1979-1986 and 1992-1997). In years of low count
coverage, efforts were focused on the relatively large and well-known colonies within the Gemini-
Green-Haley-Butler core area of distribution. For this reason single-count coverage was obtained
for 30-90% of the expected number of marmots in the Nanaimo Lakes metapopulation in every
year after 1980 (Figure 8).
Count intensity also varied with time. The five intensively studied mark-recapture colonies
typically received more than 10 counts annually with greatly increased effort after 1992. Other
colonies were counted 2-4 times per year, with higher frequencies during the peak of count efforts
in the early 1980s and after 1992. Coverage for pups (counts after 1 July) was generally
consistent with adult coverage (all counts) except for 1985 and 1987, when few late summer
counts were made.
Repeatability of marmot counts
Of the 1711 marmot counts, 206 (12%) constituted the only count for a particular site-year.
An additional 173 counts (10%) resulted in no marmots or fresh burrows observed, and 227 counts
(17%) recorded fresh burrows but no marmots. The remaining 1332 counts constitute repeated
measures of colonies known to be active (grand mean = 3.8 counts per site-year combination). Of
these, 352 counts (26%) produced maximum counts of adults and 799 (60%) recorded fewer than
maximum numbers.
Results from resampled counts of tagged animals suggest that ~ 9 repeated visits are necessary
to obtain accurate population sizes but that 2 to 3 counts provide a reasonable index of
abundance. On average, single counts resulted in detection of 44% of tagged adults known to be
present (Figure 9). Accumulated success for counts repeated 2, 3, and 4 times was 63%, 73% and
78%, respectively. The success of counts is greatly improved if they are made before August
(Figure 10).
35
1970 1980 1990 2000
0
0.2
0.4
0.6
0.8
1.0
PROPORTION OF EXPECTED
Adults
Juveniles
A) Count coverage
1970 1980 1990 2000
0
10
20
30
COUNTS PER SITE
Non-intensive
Intensive
B) Count intensity
YEAR
Figure 8: Population count extent and intensity. Count extent (A) was estimated as the
annual proportion of expected number of marmots contained in habitats that
received at least one visit. Count intensity (B) was expressed as
x
(SE)
number of counts per site-year at intensively studied and other colonies (see
Appendix 6).
36
0 2 4 6 8 10 12
REPEATED COUNTS
0.0
0.2
0.4
0.6
0.8
1.0
PROPORTION OF HIGH COUNT
Figure 9: Probable accuracy of marmot counts. Counts at colonies with known
numbers of adults (n = 437) were resampled in random order to create 100
trials of 10 counts each. Data are mean accumulated success and 95%
confidence limits.
0 16 32 48 64 80 96 112 128 144 160
DAYS AFTER EMERGENCE
0.0
0.2
0.4
0.6
0.8
1.0
PROPORTION OF HIGH COUNT
Figure 10: Seasonal effects upon marmot count success. A locally weighted regression
(LOWESS) line is shown, and slight “jitter” has been introduced to make the
data points more legible. May 1 was defined as the date of emergence.
Count repeatability was moderate (66%). Data from intensively studied colonies showed lower
repeatability despite increased numbers of measurements (Table 1). This result appears counter-
intuitive but in fact is explainable. The intensively studied colonies in which animals were marked
commonly experienced some turnover of individuals within a given season (due to mortality,
37
dispersal and immigration). This had the effect of inflating the annual maximum count because all
marked individuals were considered in the total, despite the fact that not all individuals were
present at any time. For other colonies it was not possible to distinguish such turnover. The
annual maximum count more closely approximated the number present at any given time, with the
result of establishing an artificially low target for repeated measures.
Table 1: Repeatability of count data for adults. Cumulatively the data showed
moderate (66%) repeatability. Single counts are not likely to produce
reasonable estimates of marmot abundance.
SOURCE OF DATA
Cumulative Intensively studied Non-intensive counts
N of counts 1332 450 882
N of site-years 254 30 224
Effective n of counts (n0) * 5.23 14.87 3.93
Num./denom. df 253/1078 29/420 223/658
Repeatability (R) 0.656 0.465 0.700
upper 95% confidence limit 0.695 0.509 0.737
lower 95% confidence limit 0.622 0.430 0.669
* Nomenclature follows Krebs (1989), in which n0 signifies the average effective number of
repeated counts for each site-year combination.
Mark-recapture effort
A total of 144 individual marmots were tagged and monitored at five intensively studied
colonies from 1987 through 1998 (Appendix 7). This involved 635 visits, 817 person-days, 1204
trap-days and 306 marmot captures, with approximately equal effort spent at the 2 natural and 3
clearcut colonies. Most (62%) animals were eventually tagged at these colonies. Ear-tag loss rate
was low (5%) and loss of both tags was very rare (Appendix 8). Given probable abundance it is
likely that ear-tagged individuals represented approximately 10-15% of the entire M.
vancouverensis population over the 11 year mark-recapture period.
38
Part 3: Population trends
Probable marmot abundance
Marmot abundance within the Nanaimo Lakes metapopulation changed dramatically during the
past two decades. Comparison of observed and expected numbers from 1972 through 1998
showed systematic trends (Figure 11). For adults, observed numbers were consistently above
average (134-159%) from 1981 to 1984, and near or below average (53-99%) from 1990 to 1998.
The magnitude of annual change was generally small (
x
absolute change = 19.7%, SD = 15.9).
Application of a correction factor based on count intensity suggests that adult numbers decreased
from a peak of 200-250 during the mid-1980s to fewer than 100 in 1998 (Appendix 9). Probable
adult numbers were highly correlated with values in the preceding year (mean square successive
difference test, C = 0.63, n = 22 years, P < 0.01), which is a predicted result given known marmot
longevity and suspicion that marmots do not disperse after becoming sexually mature and
establishing themselves within a colony. I conclude that population counts provide relatively
consistent estimates of adult abundance.
For pups the ratio of observed to expected numbers fluctuated more dramatically during the
same period (range of annual proportion of expected numbers = 26% to 210%). Magnitude of
annual change was greater (
x
absolute change = 79.9%, SD = 50.3%). Probable abundance of
pups was not correlated with values observed in the preceding year (C = 0.01, n = 18 years, P >
0.05) but this is not surprising. There is no reason to expect consistent annual reproduction. In
fact the reverse is true. Small colony sizes combined with infrequent breeding in adult females
ensure that most colonies would not be expected to reproduce in consecutive years. Limited
population count effort in some years and consequent extrapolation of trends from a small number
of colonies would be expected to exaggerate the variance of observed/expected ratios. I
consequently remain less confident of pup abundance estimates.
Colonizations and extinctions
Recent population dynamics were accompanied by a profound change in the spatial structure
and habitat associations of the metapopulation. Prior to 1981, marmots were confined to steep
natural sub-alpine meadows at elevations above 900 m. Beginning in the early 1980s increased
proportions of the Nanaimo Lakes metapopulation inhabited recently harvested (0-15 year old)
clearcuts. In the last five years 58% of the probable marmot population inhabited clearcuts
(Figure 12).
39
A) Observed/expected marmots
0
50
100
150
200
250
72 74 76 78 80 82 84 86 88 90 92 94 96 98
YEAR
% OF EXPECTED
Adults
Juveniles
B) Probable marmot abundance
0
50
100
150
200
250
300
350
400
72 74 76 78 80 82 84 86 88 90 92 94 96 98
YEAR
PROBABLE NUMBERS
Juveniles
Adults
Figure 11: Marmot population trends over time. Percent of expected values (A) were
calculated using only those sites counted in any year. Probable marmot
numbers (B) were estimated by applying a correction factor based on count
effort, by excluding clearcut habitats for years prior to colonization, and by
excluding years in which fewer than four sites were counted.
40
Marmots apparently first colonized a clearcut in 1981 (Appendix 10). At least seven
additional sites were colonized between 1982 and 1985 and in several cases population increases
were dramatic. Marmots also apparently colonized some natural meadows during the early 1980s,
although it is impossible to confirm which of these represented true colonization events and not
belated discovery dates.
Only two new colonies were discovered during the 1990s despite greatly increased search
effort and public awareness. Both (Mount Franklin and Sherk Lake) were in clearcut habitats and
certainly represent actual colonization events because the habitat was unsuitable for marmots
before forest harvesting occurred. In addition, several new habitat patches were discovered on
several mountains (e.g. Mount Moriarty and Big Ugly). However, in these cases reproduction has
not been confirmed and they may represent sites used only occasionally by marmots. In addition,
for reasons identified above it is impossible to distinguish colonization events from pre-existing
colonies that were only recently discovered.
1971
1975
1979
1983
1987
1991
1995
1999
YEAR
0100200300 100 200 300
Adults
Juveniles
PROBABLE NUMBERS
Clearcuts
Natural habitats
no surveys
Figure 12: Probable marmot numbers in natural and clearcut habitats. Neither
juvenile nor adult abundance was correlated among habitat types
(Pearson r = -0.19 for adults and -0.02 for juveniles). The current
population probably contains fewer than 100 animals, of which ~50%
are found in clearcut habitats.
41
Population trends among colonies
Sequential population estimates could not be made for many colonies because of gaps in
sampling coverage, but LOWESS regressions were useful in exploring trends over time.
Apparently no colony remained stable over the study period (Figure 13). Marmots numbers in
natural habitats declined systematically with year. The data suggest a definite upward bulge that
occurred in 1983, which probably reflects high reproduction in 1982 or earlier. Numbers of
marmots in clearcuts increased greatly during the late 1980s (the Butler “west roads” colony was
the largest) but since about 1992 most colonies declined rapidly.
Part 4: Population ecology
Survival
Population dynamics are the inevitable result of differences in birth and death rates. Robust
estimates of both processes are essential if causal factors are to be understood.
Juvenile survival: At least 25 of 56 (45%) tagged pups at the intensively studied colonies were
confirmed as surviving their first winter. Survival of untagged pups at the same colonies was
slightly higher (88 pups and 46 yearlings; persistence = 52%) but not significantly so (2 = 0.80
with 1 df, P=0.37). From these data there seems little reason to suspect that capture influenced
pup survival. I therefore used pooled data from tagged and untagged pups to estimate annual
survival rates all subsequent analyses. The single exception was that of possible sex bias in
survival, which could only be evaluated from tagged pups. Survival was independent of sex. At
least 12 of 25 (48%) tagged males survived their first winter, as did 13 of 31 (42%) tagged females
(2 = 0.21 with 1 df, P=0.65).
Survival varied dramatically across years and colonies (Appendix 11). Relatively low survival
occurred over the winter of 1989-90 (13 survivors of 31 juveniles; 42%), 1990-91, (0 of 8; 0%)
and 1994-95 (8 of 27; 30%). Relatively high juvenile survival was observed in 1991-92 (10 of 16;
63%) and 1993-94 (17 of 23; 74%).
The short late summer trapping window precluded any meaningful analysis of dates of last
observation to assess the timing of mortality. However, sampling effort was sufficient to confirm
that loss of complete litters occurred on at least 12 occasions. This resulted in loss of 43 of the 73
juveniles (59%) that disappeared. Partial mortality of nine litters was also confirmed. This
resulted in loss of another 13 of the 73 disappearing juveniles (18%). For the remainder of
42
A) Natural habitats
1978 1983 1988 1993 1998
YEAR
0
1
2
3
4
5
6
ADULTS (square-root)
r = - 0.385, n=233, P<0.001
B) Clearcuts
1978 1983 1988 1993 1998
YEAR
0
1
2
3
4
5
6
ADULTS (square-root)
r = - 0.103, n=87, NS
Figure 13: Marmot population trends within and among colonies. The data are
LOWESS regressions of annual adult numbers at natural meadows (A) and
clearcuts (B). Numbers of adults were square-root transformed to facilitate
comparison of trends among colonies. Marmots in natural colonies declined
significantly over time (Pearson correlations) but increased during the early
1980s. Marmots in clearcuts increased later but most colonies subsequently
declined.
43
disappearing juveniles (23%), incomplete trapping or the presence of multiple litters prevented me
from distinguishing complete losses of litters from partial losses. The high mortality involving loss
of complete litters is intriguing because it suggests that the entire group was exposed to a single
mortality factor, as would be the case if animals died during communal winter hibernation. On
four occasions monitoring was sufficient to verify that this was the case (i.e., all animals were seen
in mid September but not in April-May). In all 12 cases involving the loss of complete litters,
disappearance of associated adult female parents increased my suspicion that entire family groups
succumbed during winter hibernation.
Adult survival: In theory, survival estimated from consecutive annual counts (N t+1/N t) of
marmots will provide accurate estimates of survival only if emigration equals immigration. If
successful emigration occurs more frequently than animals immigrate, consecutive counts will
underestimate survival. The reverse is also true. If emigration is lower than immigration, survival
estimates based on N t+1/N t will overestimate true survival.
Following this reasoning, calculating survival from marked animals should underestimate
survival because some individuals emigrate successfully but will never be recorded again. But the
reverse is not true. For a marked population there is no question of confusing immigrants from
surviving animals because immigrants will be untagged (or they will be known immigrants from
other study areas). However, even for a marked population of marmots there remains the problem
of re-observability (i.e., some animals remain undetected in a given sampling period). Low re-
observability would result in underestimates of survival.
Initial results suggested that annual adult survival was independent of whether animals were
tagged (persistence:disappearance ratio = 159:96; persistence rate = 62%) or untagged
(persistence:disappearance = 45:26; persistence rate = 63%; 2 = 0.03 with 1 df, P=0.874).
Similarly, the question of observability of tagged marmots appeared to be essentially irrelevant
provided that sufficient monitoring efforts were made. Only in a few cases (n=7 in a sample of
255 adult-years) did marmots older than juveniles apparently disappear (for at least one year) and
then reappear at the same colony. In no case did a non-juvenile disappear for more than one active
season and then reappear at the same site. Formal testing of recapture (resighting) probability
using Cormack-Jolly-Seber mark-recapture models indicated that the most parsimonious model
was that of assigning a value of 1.0 (Table 2), and I did this for all subsequent analyses.
44
Table 2: Cormack-Jolly-Seber estimates of adult survival and recapture probability.
This analysis was based on a model of constant survival and recapture rates
(i.e., no time or age-dependence). The model including a fixed recapture
probability produced a lower Akaike Information Criterion (AIC) and is
therefore more parsimonious. Nomenclature and methods follow Lebreton et
al. (1992).
Parameter
Estimate
lower 95%
limit
upper 95%
limit
AIC
Sex-biased survival and recapture (,p) 433.3
Male survival probability 0.545 0.451 0. 476
Female survival probability 0.684 0.610 0.750
Male recapture (resighting) probability 0.935 0.776 0.983
Female recapture (resighting) probability 0.940 0.834 0.964
Sex-biased, fixed recapture (,p=1.0) 428.6
Male survival probability 0.545 0.451 0. 656
Female survival probability 0.684 0.609 0.751
Pooled sex survival and recapture (,p) 429.7
Survival probability 0.629 0.570 0.683
Recapture (resighting) probability 0.933 0.867 0.968
Pooled sex, fixed recapture (,p=1.0) 398.7
Survival probability 0.623 0.565 0.678
The sex-biased model was less parsimonious (higher AIC) although survival estimates were
substantially lower for males than for females. Life table analysis also suggests that adult survival
was sex-biased, implying differential mortality or dispersal or both (Table 3). Relatively few
males survived beyond the age of four years compared to females. For females the general pattern
is that of a Type III survivorship curve (e.g., Begon and Mortimer 1986), in which mortality is
concentrated in the younger age classes. For males, lower adult survival suggests a pattern that
more closely resembles a Type II curve, with relatively constant survival among age-classes. As
was the case for juveniles, there was high variation across years and study colonies (Appendix 12).
Sensitivity analysis suggested that the population is more sensitive to changes in adult survival
than to changes in pup survival, probability of breeding or litter size (Figure 14).
45
Table 3: Cumulative life-table for Vancouver Island marmots. Net reproductive rate
(Ro) is congruent with other estimates of a severely declining population.
Nomenclature and methods follow Caughley (1977; Method II)*.
MALES FEMALES
x
dx
qx
px
Lx
x
dx
qx
px
Lx
bx‡
lx*bx
Juveniles † - - 0.51
1.00
1000
144
73
0.51
1.00
1000
0.00
0.00
Yearlings 26
9
0.35
0.49
493
31
10
0.32
0.49
493
0.00
0.00
2 year-olds 30
14
0.47
0.65
322
34
13
0.38
0.68
334
0.07
0.02
3 year-olds 25
12
0.48
0.53
172
35
8
0.23
0.62
206
0.85
0.18
4 year-olds 13
7
0.54
0.52
89
26
7
0.27
0.77
159
0.66
0.11
5 year-olds 6
3
0.50
0.46
41
16
5
0.31
0.73
116
0.57
0.07
6 year-olds 3
2
0.67
0.50
21
10
4
0.40
0.69
80
0.75
0.06
7 year-olds 1
0
0.00
0.33
7
7
2
0.29
0.60
48
0.41
0.02
8 year-olds 1
0
0.00
1.00
7
3
1
0.33
0.71
34
0.96
0.03
9 year-olds 1
1
1.00
1.00
7
3
1
0.33
0.67
23
0.77
0.02
10 year-olds 0
0
0.00
0.00
0
1
1
1.00
1.00
15
0.00
0.00
Ro =
0.50
=
0.88
Notes:
* Data are x (frequency of marked animals per age-class), dx (frequency of disappearances),
qx (disappearance rate), px (probability of persistence), Lx (standardized survivorship), bx (per
female fecundity), lx*bx (reproductive value), and Ro (net reproductive rate).
The data reflect 255 tagged adult- years (n =96 individuals), 56 tagged juvenile-years and 88
untagged-juvenile years at the five intensively studied colonies. An additional 17 adult- years were
included from 6 tagged individuals at non-intensively studied colonies.
† Juvenile survival based on counts of juveniles and yearlings in the following year. I assumed no sex
bias in juvenile survival and used the same rate for males and females.
‡ Fecundity was calculated using the observed juvenile sex ratio (40:28 in favor of females).
Assumption of a 1:1 sex ratio at birth reduced Ro to 0.44.
46
0.93
0.92 0.92
0.92
0.92
0.940.94
0.99
0.94
0.93
1.01
0.96
0.80
0.85
0.90
0.95
1.00
1.05
1.10
Juvenile
surv ival
Adult
surv ival
Prob. of
breeding
Litter size
Lambda
Unadjus ted data
10% improvement
+2 SE improvement
Figure 14: Sensitivity of finite population growth rate () to changes in demographic
rates. Each of the four variables was adjusted by adding either a 10%
proportional increase, or by using the upper 95% confidence limit as the
estimate (the “best-case” interpretation of observed data). Adult survival
exerted a disproportionate effect. This analysis assumed constant survival and
fecundity among adults older than 2 years and therefore exhibits higher
values than observed.
Reproduction
Per capita birth rates () reflect the probability of breeding (Pb) multiplied by the magnitude of
reproductive events (litter size). The probability of breeding at a given colony is influenced by
intrinsic life-history traits (the period over which a marmot can breed and frequency with which it
can breed) and by the age-sex structure and reproductive history of the animals within a colony. It
is therefore not surprising that reproductive rates varied greatly among colonies and years
(Appendix 13).
Most females that eventually reproduced first did so at age three or four (
x
age of first
reproduction = 3.87, SD = 0.92, range = 2 to 5 years). A single female apparently reproduced at
age two. Females were capable of breeding in the oldest age classes (maximum age = nine years)
and age-specific fecundity was relatively stable after the age of two (range of values from 0.41 to
0.98 females per adult female year). The probability of producing a litter was 0.29 (SD = 0.45,
47
n=137 animal years) for tagged females older than 1 year and 0.37 (SD = 0.46, n=103 animal
years) for females older than two years. There was usually a non-reproductive interval of at least
one year between litters (
x
interval = 2.0 years, SD = 0.67, n=10). Litter production in
consecutive years was confirmed twice. Litters normally contained three or four pups, with litters
of two or five observed infrequently (
x
litter size = 3.28, SD = 0.85, n=43). Variation in lifetime
reproductive performance of individuals was high. A few females accounted for the majority of
reproductive events. This was particularly evident at the Haley Lake colony, where three of 12
reproductive-age females (i.e., >2 years old) produced 30 of the 58 pups (52%) born from 1987
through 1997 at that site (Appendix 14).
Emergence of pups was generally synchronous among colonies and years. The earliest date
that pups were seen was 22 June, but most (29 of 34 litters for which emergence data exist) were
first observed aboveground between 28 June and 7 July. Sex ratio of 68 animals initially captured
as pups was skewed towards females (40:28) but this ratio did not significantly differ from 1:1 (2
= 2.12 with 1 df, P=0.14). Addition of 21 animals initially captured as yearlings produced a
similar result (cumulative sex ratio =52:37 in favor of females, 2 = 2.53 with 1 df, P=0.11). Sex
determination is more difficult for pups. On two occasions recapture revealed that I had initially
mis-identified the sex. Despite this problem, results make it difficult to exclude the possibility of a
skewed sex ratio in favor of females.
Immigration-emigration
Dispersal is the “glue” that allows metapopulations to survive (Gilpin 1987). In the case of M.
vancouverensis, the importance of dispersal may be exaggerated because small colony sizes
presumably lead to increased vulnerability from stochastic processes (Hanski and Gilpin 1997).
Timely appearance of immigrants (rescue effects) could therefore be very important. On two
occasions I observed reproduction that became possible only after untagged animals immigrated
into a colony (Haley Lake in 1988 and Sherk Lake in 1993). Conversely, on several occasions I
observed reproductive-age animals that could not reproduce because of the absence of a possible
mate (Green Mountain in 1990 and 1995, Bell Creek in 1996, Pat Lake in 1997).
Five tagged animals dispersed from intensively studied colonies and were observed alive at new
locations. A two-year-old male and a female, probably age two, moved 7.4 km from the Pat Lake
clearcut colony to the Mount Franklin clearcut colony in 1992 or earlier. The female produced a
litter in 1993 and was observed in 1998. The male was still resident on Mount Franklin in 1997
and remains the oldest male recorded during the study (age 10). Another two-year-old male moved
48
5.9 km from the Vaughan Road clearcut colony to the Green Mountain colony in 1994. This
animal was photographed at a hibernaculum entrance on 4 May 1995 but disappeared shortly
thereafter. Two additional tagged animals were seen outside their original colony briefly (at Bell
Creek and Mount Holmes), but tag numbers could not be recorded. It appears that none of the
latter animals became resident at the new location. Their eventual fate remains unknown. Heard
(1977) reported another dispersal movement in which an adult male moved 0.9 km from Haley
Lake to Bell Creek in 1974.
All 39 immigrants observed at intensively studied colonies were judged to be young adults, and
19 captured immigrants were definitely so (8 females, 11 males). These animals were invariably
larger than pups or yearlings but typically smaller than known-age adults of reproductive age.
Most immigrants had relatively uniform pelage color, which is typical of two-year-olds in late
summer or three-year-olds in early spring. No immigrant had the “mottled” molt pattern that is
typical of older adults, and no female had the prominent nipple development that is characteristic of
animals that reproduced in the current or previous year. My data do not preclude the possibility
that some yearlings dispersed, but it appears unlikely. Radio-telemetry provided little additional
information about dispersal. Transmitter limitations, battery failure and small sample sizes
combined to ensure that few dispersal-aged animals were successfully monitored. However, two
radio-equipped marmots at Pat Lake made extensive movements and disappeared for over a month
in 1994, when the brother and sister were two years old. The brother disappeared; the sister
survived and is now in captivity at Toronto Zoo.
Pups and adults that have reproduced apparently do not disperse. Sub-adult (two or three-year
old) marmots made sizable dispersal movements (~10 km) through the landscape. Records of
solitary marmots in low elevation habitats provide another means of estimating the length of
dispersal movements. For example, at least two animals were observed near Nanaimo, including
one animal that hibernated successfully near the Cassidy airport in 1991 (assuming that this animal
originated at the nearest known active colony, it dispersed at least 20 km). The seasonal timing of
dispersal is more difficult to evaluate although records of solitary marmots suggest that dispersal
could occur as early as mid-May (Appendix 15).
Dispersal is an infrequent event and the majority of marmots present at intensively studied
colonies were apparently born there. Of 105 animals in the two-year-old age-class, at least 78
(74%) were born at that site. Using pooled data from two and three-year-old age classes, at least
130 of 166 animals (78%) were not immigrants.
49
Mortality factors
It was rarely possible to determine causes of mortality. Golden eagles, cougars and wolves
take an undeniable toll on marmots. Disease and unsuccessful hibernation are probably equally or
more important.
Avian predators: Observers recorded three cases of predation by golden eagles (Aquila
chrysaetos), all apparently involving pups or yearlings. Many unsuccessful attacks were also seen
and most colonies were frequently hunted by that species. Bald eagles (Haliaeetus leucocephalus)
were rarely seen at high elevations, but a few unsuccessful attacks were observed. Sharp-shinned
hawks (Accipiter striatus) and Cooper’s hawks (Accipiter cooperii) were commonly observed
“dive-bombing” marmots but in these cases the relative sizes of the animals suggested “play” or
“training” behavior on the part of the hawks rather than a serious predation attempt. Other raptors
were occasionally seen throughout summer (especially northern goshawks Accipiter gentilis).
Raptors such as red-tailed hawks (Buteo jamaicensis) and northern harriers (Circus cyaneus) were
sometimes seen, particularly during fall migration. There are no recorded attacks on marmots by
the latter three species.
There is essentially no overlap in diurnal activity patterns of owls and marmots. Pygmy owls
(Glaucidium gnoma) were common in the study area but do not represent a threat to marmots
because of their small size. Great horned owls (Bubo virginianus) were recorded twice and one
barred owl (Strix varia) was seen in a low elevation clearcut colony.
Terrestrial predators: Cougars were observed “stalking” marmots twice, at the Haley Lake
and Bell Creek colonies. Heard (1977) also reported a probable case of cougar predation at the
former site. It is possible that terrestrial predators benefit from easier movement along logging
roads and from learned behavior. On four occasions I observed cougar tracks in the snow
surrounding hibernacula exits in late April or early May, and on two occasions I followed cougar
tracks along roads as they led from the Vaughan Road clearcut colony into the nearby Haley bowl
natural colony. Wolves are another known predator. Wolf scat collected from Gemini Peak in
1984 contained marmot hair (D. Nagorsen, Royal B.C. Provincial Museum, pers. comm.). On two
occasions wolf packs produced pups in areas within 1 km of marmot colonies in clearcut colonies
(at Sherk Lake in 1995 and Green Mountain in 1998).
Radio-telemetry provided useful information about mortality (Appendix 16). Of seven animals
equipped with radio transmitters in 1994, three were confirmed to have been killed by terrestrial
50
predators. Similarly, of six animals equipped with transmitters in 1998, three were killed. In no
case was it possible to definitively identify the predator species. Tooth marks and mangled
transmitters were found on two occasions, perhaps suggesting wolf predation. Similarly, on four
occasions transmitters were found in relatively pristine condition, which may reflect the more
“surgical” nature of cougar feeding habits (D. Doyle, B.C. Wildlife Branch, pers. comm.). The
relative importance of cougar and wolf predation remains unknown. Given the wariness of
Vancouver Island marmots and the fact that they rarely stray far from the safety of a burrow, I
suspect that a “lurk and pounce” technique would be the most effective hunting strategy, and this
might suggest cougars more than wolves. Cougars stalk and pounce on their prey much as
domestic cats do (Banfield 1977), whereas wolves typically run down their prey (Carbyn 1987).
The possibility of predation by mustelids cannot be discounted. Pine marten (Martes
americana) are reasonably common in the study region, although they were rarely reported in high
elevation habitats. In 1990 I photographed an ermine (Mustela erminea) at the Green Mountain
summit colony but that constituted my single observation of this species in a habitat occupied my
marmots. Predation by black bears (Ursus americanus) is probably rare if it occurs at all.
Observers often recorded black bears grazing in close proximity to marmots, particularly in the
early spring at natural subalpine colonies. Marmots were invariably wary but normally did not
respond by whistling or retreating into burrows. In this respect marmots behaved much as they do
while in the presence of humans, which is quite different from their response to predators such as
cougars or golden eagles. Throughout the project there was no evidence of attempts by any
predator to excavate marmots from their burrows.
Some terrestrial predators that are important for other marmot species are irrelevant for M.
vancouverensis. There are no coyotes (Canis latrans), grizzly bears (Ursus arctos), fishers
(Martes pennanti) or long-tailed weasels (Mustela frenata) on Vancouver Island. Predation by
wolverines (Gulo gulo) is unlikely because of the extreme rarity of that species on Vancouver
Island (there are fewer than 20 confirmed sightings in the past fifty years; D. Janz, B.C. Wildlife
Branch, pers. comm.). From the accumulated evidence I conclude that cougars, wolves and golden
eagles are the principal predators upon Vancouver Island marmots, probably in that order of
importance.
Unsuccessful hibernation: Unsuccessful hibernation is a commonly cited mortality factor in
marmot ecology (e.g. Armitage 1994) but is extremely difficult to confirm from field observations.
Incontrovertible evidence of overwinter mortality was obtained only once. Four animals that were
51
transplanted to Mount McQuillan in June of 1996 as part of an experimental reintroduction died
during hibernation (Bryant et al. in press). Three of the four animals were recovered in the
following June. The adult female could not be retrieved but radio-telemetry confirmed her presence
in the same burrow system. Relatively large body mass of recovered corpses suggested that death
did not occur because of depletion of fat reserves. Initial necropsy results suggested that mortality
was due to a bacterial infection (Yersinia frederiksenii) but more recent cultures also identified Y.
enterocolitica and Carnobacterium divergens from the same tissue samples. At present it is
impossible to identify any particular pathogen as the presumptive cause of death, or to exclude the
possibility that bacterial infection occurred post-mortem.
No other animals were confirmed as dying during winter although radio-telemetry provided
some suggestive results. Excluding transplanted animals, the sample of animals that entered
hibernation with functioning transmitters was small (n = 27). In 19 cases both the marmot and the
transmitter survived. In five cases the transmitter failed during winter but the marmot was
confirmed to be alive in the subsequent spring by recapture or resighting. In three cases (11%) the
transmitter failed and the animal was not seen in the following spring. One disappearance of a
telemetered animal involved a yearling that hibernated with a sibling at the Pat Lake clearcut
colony; neither animal was seen again. On another occasion a telemetered pup vanished along with
two siblings and the mother at the Sherk Lake clearcut colony. Another adult male disappeared at
the Sherk Lake colony in the following year, but unfortunately in that instance the hibernaculum
and hibernation group remained unidentified.
On 12 occasions I observed the disappearance of entire family groups that included juveniles.
In all cases animals were observed using particular burrows in late August or early September. In
the following spring these burrows showed no evidence of use by marmots or were occupied by
new marmots. For example, my last day at Haley Lake in 1996 was on 17 September, when four
of the seven juveniles born at that site were confirmed alive. No yearlings were observed in the
following spring despite visits beginning on 22 April and three additional visits made prior to 15
May. Similarly, two of three juveniles born at Vaughan road clearcut colony were recorded as late
as 22 September. Neither the adult female nor surviving yearlings were observed despite
observations that began on 21 April in the following spring. I conclude that evidence for
unsuccessful hibernation is circumstantial but compelling.
Effect of research on mortality: There was little evidence to suggest that research efforts
caused significant mortality. A single fatality occurred in 1992, when an adult male overheated
52
while being moved prior to implantation of a radio transmitter. In addition a yearling male was
killed in 1994 after a trap was set and left open by unknown visitors to the site. For 16 of the 144
tagged animals (11%), dates of first capture and last observation were identical. In these cases the
date of initial capture tended to be late in the season (mean capture date = 13 August, SD = 29.2
days) and few later visits were made to the site before onset of hibernation (mean number of later
visits = 1.6, SD = 0.5). However the possibility cannot be discounted that capture stress
contributed to mortality at a later date. To address this possibility I compared demographic rates
from the intensively studied mark-recapture colonies with those that received only limited visitation
by count crews.
Congruence among estimates from intensively studied colonies and counts
Overall demographic rates obtained from intensively studied colonies and other sites were quite
similar with one exception (Table 4). Average litter size estimated from non-intensive population
counts was significantly lower. This trend probably reflects the relatively late date of litter counts
at many of the non-intensively studied colonies (mean date of count = 29 July, SD = 19.9 days)
compared to the intensively studied sites (mean = 14 July, SD = 16.2 days). When counts made
after 30 July were removed from the data the difference disappeared (Pooled mean litter size = 3.01
juveniles, SD = 1.12, Mann-Whitney U = 1142, P = 0.127).
For the above reason I used the probability of producing a litter rather than per capita births or
litter sizes for subsequent analyses. The latter measure is less likely to be biased by later dates of
observation than is litter size (seasonal bias would be expected if some summer mortality of
juveniles occurs, or if vegetation growth makes it harder to obtain accurate counts, or if expanding
foraging movements make it more difficult to assign litter membership). Probability of
reproducing would be less sensitive to such bias, because a count of two juveniles and one litter
would produce the same probability of reproduction as would a count of five juveniles and one
litter.
Adult survival and pup survival were independent of whether colonies were intensively or non-
intensively studied. Pup survival rates were virtually identical, suggesting that capture and
marking at intensively studied colonies did not increase mortality. Annual demographic rates
among intensively and non-intensively studied colonies were uncorrelated. This probably
represents real variation in birth and survival rates. The exception to this trend was adult relative
density (observed/expected), which was strongly correlated between the two data types and which
53
had the largest sample size (11 years of relevant data). This pattern is consistent with overall
population decline.
Table 4: Marmot demographic rates from intensively studied colonies and non-intensive
counts. Tests were ² tests of independence or comparisons of means using
two-tailed t tests or Mann-Whitney U tests as appropriate. Pearson
correlations were used to compare annual rates. Note that adult survival from
intensively studied colonies reflects “apparent” survival based on counts (net
survival after immigration-emigration) and that not all animals were
necessarily ear-tagged.
Variable and type of data
Breeders: nonbreeders breeder
nonbreeder
gross rate
² P Correlation (annual rates)
Intensively studied 39 296 0.11 0.10 0.749 0.16, n=9 years, NS
Other colonies 142 1146
Juvenile survival
persist
disappear
Intensively studied 71 69 0.49 0.24 0.63 -0.19, n=4 years, NS
Other colonies 98 106
Adult survival (n+i-d-e)
persist
disappear
Intensively studied 194 118 0.65 1.21 0.27 0.65, n=5 years, NS
Other colonies 562 294
Mean litter size N mean SD U P
Intensively studied 39 3.39 0.82 3814 0.007 -0.48, n=5years, NS
Other colonies 155 2.90 1.08
Per capita births N mean SD U P
Intensively studied 335 0.39 1.12 217368
0.978 0.16, n=9 years, NS
Other colonies 1297 0.35 1.01
Adult relative density N mean SD t P
Intensively studied 44 1.02 0.51 -0.854 0.394 0.72, n=11 years, P<0.01
Other colonies 158 1.11 0.68
54
Part 5: Tests of predictions
Effect of habitat on demographic rates
Demographic rates were influenced by some habitat variables and not others (Table 5).
Survival of pups was lower in clearcut versus natural and high versus low elevation habitats (note
that these habitat classes often represented the same colonies, as there are few natural habitats
below 1200 m and no clearcuts above that elevation). Breeder:nonbreeder ratios were independent
of habitat class, although there were more breeders in the early (1983s) period when most colonies
had high relative densities. Adult survival was also higher during the same period, although it is
impossible to evaluate the relative importance of immigrants and surviving residents. There were
no apparent differences in demographic rates among exposed (southeast to southwest-facing) and
sheltered (west-southwest to east-southeast facing colonies) colonies. Peripheral colonies had
relatively low apparent adult survival but birth rates and pup survival were independent of
isolation class.
Life-table analysis from resighting data suggested a reduction of 5-10% in survival of pups
and reproductive-age females (age three and older) living in clearcuts. There was no apparent
difference in survival of yearlings or two-year-olds among the two habitat types (Figure 15). Age-
specific reproductive contribution in clearcuts (Ro =0.35) was slightly more than half that observed
at natural sites (Ro =0.65; Appendix 17). Cormack-Jolly-Seber estimates of adult survival also
suggested site or habitat specific differences in survival (Figure 16). However, sample sizes were
insufficient to resolve these differences at the 95% confidence level. The most parsimonious model
was a hypothesis of no site or habitat specific differences, and therefore a single pooled estimate of
annual survival (survival probability = 0.623, with lower and upper 95% boundaries of 0.565 and
0.678 respectively).
For reasons of low statistical power I caution against rejecting a hypothesis of habitat or
colony-specific effects. The Sherk Lake clearcut colony had high adult survival compared to other
sites, although this estimate was based on the smallest number of years with relevant data.
Similarly, although age-structured data suggest a distinct sex bias in adult survival, mark-recapture
modeling was unable to reject the hypothesis of no difference. Small changes in survival rates
could produce biologically significant results but current data are inadequate to resolve such small
differences. This is problematic because sample sizes are not likely to be materially improved in
many cases. For example, the Vaughan and Pat Lake clearcut colonies are extinct and the Haley
Lake natural colony currently contains only two adults.
55
Table 5: Effect of habitat, time period and relative density on marmot demographics.
Data are pooled from intensively studied and other colonies. Tests are 2
tests of independence. Adult survival is apparent survival based on counts
(net survival after immigration/emigration).
Adult survival Juvenile survival Breeder: non-breeder ratio
Habitat class survive
disap. 2 P survive disap. 2 P breeder
non. 2 P
Natural meadows 484 254 0.57 0.45 103 84 5.81 0.02 70 497 1.11 0.29
Clearcuts 270 156 66 91 111 935
High (>1200 m) 346 193 0.15 0.70 61 46 3.86 0.05 77 665 1.59 0.21
Low (<1200 m) 217 408 108 129 102 720
Sheltered (245-150°) 164 106 2.51 0.11 42 49 0.44 0.51 134 1044 0.02 0.88
Exposed (151-244°) 590 304 127 126 45 341
Core area 598 305 3.70 0.06 133 133 0.36 0.55 141 1063 0.37 0.55
Peripheral 156 105 36 42 38 322
High density (>1.0) 426 261 3.83 0.05 144 151 0.08 0.78 147 960 17.6 0.00
Low density (1.0) 269 127 25 24 31 470
Early (1980s) 377 188 5.43 0.02 59 63 0.00 0.99 87 814 5.28 0.02
Late (1990) 261 177 83 89 69 437
There was a discrepancy in the proportions of immigrants to surviving tagged animals at the
two natural colonies (13 immigrants versus 116 surviving residents) and three clearcut colonies (26
immigrants versus 140 surviving residents; ² = 9.36, P = 0.002). It appears that turnover of
individuals was higher at clearcut colonies. In addition, loss of complete family groups including
juveniles occurred more frequently in clearcuts (8 cases of complete loss and 11 cases of partial or
complete survival; loss rate = 43%) compared to natural habitats (4 cases of loss and 20 cases of
partial or complete survival; loss rate = 17%) and this difference may be significant (2 = 3.41
with 1 df, P = 0.07).
The seasonal timing of disappearances differed among habitat types and could provide insight
about causes of mortality (Figure 16). Dates of last observation of tagged adults were not
independent of month (single-sample goodness-of-fit 2 = 17.1 with 4 df, P=0.002). Most tagged
adults were last recorded alive in August or early September. Dates of last observation do not
necessarily equate with dates of mortality. The September data are probably biased
56
Figure 15: Effect of habitat type on age-specific reproductive performance. Life-
table analysis suggests that female survival rates (A) were 5-10% lower in
clearcut habitats. The consequence was reduced lifetime reproductive
performance (B). Data are from tagged adult females monitored from
1987 through 1998 (n=34 in natural habitats and n=51 in clearcuts).
A) Female survivorship
10
100
1000
0 1 2 3 4 5 6 7 8 9 10
AGE-CLASS
SURVIVORSHIP (Lx )
Natural
Clearcut
B) Age-specific contribution
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0.18
0.20
0 1 2 3 4 5 6 7 8 9 10
AGE-CLASS
lx * bx
Natural
Clearcut
R = 0.65
R = 0.35
o
o
57
Figure 16: Cormack-Jolly-Seber estimates of adult survival. Data are mean annual
survival rates and 95% confidence limits based on 88 tagged adults and 242
marmot-years of observation. Confidence limits overlap and the most
parsimonious model is that of no habitat or colony-specific differences in adult
survival.
clearcut
natural
0.0
0.2
0.4
0.6
0.8
1.0
A) Habitat class
n = 109
133
SURVIVAL
F19 Green Haley Pat
Vaughan
0.0
0.2
0.4
0.6
0.8
1.0
24
14
71
104
29
SURVIVAL
B) Colony
58
downward because of relatively low monitoring effort in that month and because many individuals
enter hibernation in mid-September. August data are probably skewed upwards because of these
trends. Timing of last observation was independent of sex (2 = 5.45 with 4 df, P=0.25). I
conclude that males and females have similar seasonal patterns of mortality and that pooling of the
data was justified for subsequent analyses.
Timing of last observation differed among natural and clearcut habitats (2 = 11.1 with 4 df, P
= 0.03). Last observations in natural habitats were evenly distributed among months (2 = 0.96
with 4 df, P = 0.20). This result could suggest constant mortality pressure throughout the active
season, as might occur from predation. In contrast, most dates of last observation in clearcuts
were concentrated in August and September (2 = 29.4 with 4 df, P < 0.001), which may suggest
a higher incidence of mortality due to unsuccessful hibernation.
Sept.
Aug.
July
June
May
0102030 0 10 20 30
Females
Males
Clearcuts Natural meadows
FREQUENCY
Figure 17: Timing of last observation of tagged adults in natural and clearcut habitats.
Data are based on 48 adults that disappeared from natural habitats and 58
adults that disappeared from clearcuts. Dates of last observation of animals in
clearcuts are skewed towards late summer. Most animals enter hibernation by
mid-September.
59
Temporal effects on survival and reproduction
Adult survival estimated from mark-resighting data varied considerably among years. Low
survival years included 1991 and high survival years included 1992. However 95% confidence
levels overlapped and the time-independent model was more parsimonious (Figure 18). Probability
of producing a litter also varied substantially across years, including some years with low or no
reproduction (1989, 1990 and 1995). However, logistic regression showed no significant effect of
year on the probability of producing a litter. Survival data from tagged juveniles at intensively
studied colonies were too sparse to test.
Effect of clearcut age on birth and survival rates
Seral stage of regenerating clearcuts may represent a special case of environmental tracking.
Logistic regression indicated that apparent adult survival was negatively associated with increased
clearcut age but juvenile survival was not (Table 6). However, clearcuts coded as “old” (older
than 11 years) were negatively associated with survival of all animals (n = 583, coefficient =
-0.343, P = 0.041).
Pup survival results may suggest a nonlinear relationship with seral age (Figure 19). Such a
situation might occur if it took initial immigrants a few years to construct successful hibernacula,
or if predators benefit from forest regeneration, resulting cover and improved experience.
Probability of reproduction was not significantly associated with age of clearcuts (Table 7).
Table 6: Logistic regression of clearcut age against marmot survival.
Type of data N odds ratio coefficient
Rho² P value*
All juveniles (tagged and untagged) 157 0.978 -0.022 0.002 0.532
Non-intensive adult survival (counts)
284 0.910 -0.094 0.035 <0.001
Tagged adults (intensively studied) 121 0.911 -0.093 0.025 0.002
Table 7: Logistic regression of clearcut age against probability of reproducing.
Type of data N odds ratio coefficient
Rho² P value
Adults (non-intensive counts) 412 1.045 0.044 0.007 0.124
Tagged adults (intensively studied) 121 0.984 -0.018 0.001 0.836
60
1986 1988 1990 1992 1994 1996 1998
0.0
0.2
0.4
0.6
0.8
1.0
A) Adult survival (both sexes)
n = 13
28
31 23
20
20
28
35 28
16
SURVIVAL
1986
1988
1990
1992
1994
1996
1998
YEAR
0.0
0.2
0.3
0.5
0.7
0.8
1.0
B) Reproduction
sample sizes are as shown in graph “A”
PROBABILITY OF REPRODUCING
Figure 18: Temporal changes in adult survival and probability of breeding. Data are
mean annual rates and 95% confidence limits based on 88 tagged adults
and 242 marmot-years of observation. Sample sizes are shown. For
survival, confidence levels overlap and the most parsimonious model was
that of no time-dependence. For probability of reproduction, logistic
regression showed no relationship between year and the probability of
producing a litter.
61
A) Juvenile survival
0
5
10
15
20
0.0
0.2
0.4
0.6
0.8
1.0
SURVIVAL
3
4
9
15
13
18
4
8
6
5
9
9
2
36
3
10
13
B) Adult survival
0 5 10 15 20
Number of years after harvest
0.0
0.2
0.4
0.6
0.8
1.0
SURVIVAL
10
6
9
31
19
23
22
33
18
16
17
23
39
46
47
34
23
2
12
Figure 19: Effect of increasing clearcut age on marmot survival. Data are mean
survival estimates with 95% confidence levels for juveniles (A) and adults
(B). Sample sizes are shown. Logistic regression determined that
survival of all animals was reduced in clearcuts coded as “old” (>11
years, see arrows). Clearcuts older than 15-20 years old generally
become unsuitable for marmots.
62
Density dependence
Demographic variables were associated with changes in relative density (Figure 20). Adult
survival was negatively associated with relative density (observed/expected number of adults) for
both tagged and untagged samples, although density explained only a small amount of the variation
in survival (Table 8). Lower survival in high density conditions could reflect increased emigration
of two-year-old marmots or increased mortality from disease or predators.
Pup survival was not associated with adult relative density in linear fashion. Survival peaked
during periods of moderate abundance (75 - 150% of expected) and declined when marmots were
more or less abundant. The result raises questions about hibernacula availability or disease
transmission under high density conditions, and about the role of communal hibernation under low
density conditions.
The probability of producing a litter was positively associated with relative density of adults,
although regression Rho² values were very small. Litter production was lower during periods of
below-average adult density (i.e., <1.0) but quite stable at higher densities. This relationship was
unexpected. I expected that most animals in a high density colony to represent non-reproductive
yearlings and 2 year-olds, which would result in an inverse relationship between probability of
reproducing and relative density.
Table 8: Logistic regression of relative adult density against survival and probability of
reproducing. Samples were marmot-years (an individual alive in one year).
Variable N odds ratio coefficient
Rho² P value
Juvenile survival (all animals) 344 0.814 -0.306 0.002 0.254
Tagged adult survival 254 0.556 -0.547 0.011 0.059
Non-intensive adult survival (counts)
1083 0.684 -0.379 0.013 <0.001
Probability of producing a litter 1608 1.393 0.330 0.017 <0.001
Colonization events in relation to habitat availability
Since 1981 marmots colonized a small fraction (<100 ha of ~10,000 ha) of clearcuts above
700 m that were available to them assuming dispersal capability of 5-10 km. Frequency of
colonization was not correlated with habitat availability (Spearman r = 0.16, n=11, P>0.05). This
result is not surprising giving the small number of colonizations (n=10) and the compressed
temporal period over which they occurred (8 colonizations between 1981 and 1986).
63
The spatial pattern of colonization events was not random (Figure 21). All recorded
colonizations occurred within 5 km of existing natural colonies and 8 of 10 events occurred within
2 km. Colonizations were significantly closer to existing natural colonies than were 30 randomly
selected sites (Table 9). The data do not suggest a “stepping-stone” pattern of colonization.
Instead, results suggest a “wave” of colonizations that began during the early 1980s and stopped
abruptly, perhaps because of a shortage of colonists. Only two colonizations occurred after 1985.
Table 9: Nearest colony-neighbor distances for marmot colonizations and random sites.
Clearcuts of appropriate age (0-15 years) and elevation (>700 m) and timing
(1985) were randomly selected to represent colonization events if all locations
were equally accessible by marmots. Student t test is one-tailed.
Variable N
x
SD t df P value
Nearest colony-distance
Actual colonizations 10 1.47 1.07 -3.61 38 <0.001
Random sample 30 5.47 3.37
0 1 2 3
OBSERVED/EXPECTED
0.0
0.2
0.4
0.6
0.8
1.0
PROBABILITY
Per capita litters
Pup survival
Adult survival
n = 73
n = 272
n = 204
Figure 20: Effect of relative density on marmot demographics. Data are LOWESS
regressions of relative density of adults against annual adult survival, pup
survival and probability of reproducing. Numbers of site-years are shown.
64
Figure 21: Colonizations and potential colonizations of clearcuts. Data are colonizations
() that occurred from 1981 to 1992 and natural colonies () that existed during
that period. The map also shows 30 randomly selected clearcuts above 700 m
and between the ages of 0 and 15 years (shaded polygons) to represent
“potential” colonizations given marmot dispersal capability of at least 5-10 km.
These 30 sites represent only 160 hectares of the more than 10,000 hectares of
potential habitat that were available to marmots in 1986. Most colonizations
occurred in the central portion of the range. Sizes of dots are proportional to
average numbers of adults during the 1980s.
65
Extinction and demographic performance in relation to isolation
Marmots apparently disappeared from 10 historically reproductive natural colonies and 8
potential colonies during the late 1980s and early 1990s (Figure 22). Timing of extinction was
impossible to accurately confirm in the field. Vancouver Island marmots are apparently adapted to
a lifestyle involving extremely small colony sizes and showed extreme tenacity in their persistence
at some sites (e.g., Mount Washington). However it seems unlikely that inventory crews often
missed reproductive colonies for periods of consecutive years. Uncertainty about extinction date
precluded direct testing of extinction probabilities in relation to distance to extant colonies. An
indirect test of isolation effects using apparent survival data as a measure of immigration
supported the idea that isolated colonies receive fewer immigrants (Table 10), although the degree
of variation explained was low (Rho² = 0.070).
Table 10: Effect of increasing isolation on apparent marmot survival. Logistic
regression indicated that “apparent” adult survival was negatively associated
with increased isolation, but juvenile survival was not.
Variable N odds ratio coefficient
Rho² P value
Adult survival + net immigration 1164 0.960 -0.041 0.070 0.002
Juvenile survival 334 1.015 0.015 0.001 0.563
Sources and sinks
Although population trajectories were similar among colonies, some were apparently more
successful than others. Comparison of average annual births and disappearances (net immigration-
death-emigration) suggests that most colonies lost more marmots than they produced between 1979
and 1997. This discrepancy between births and losses is the fundamental cause of observed
population declines, and for this reason most colonies therefore fell below a theoretical “source-
sink” line above which reproduction more than balances survival (Figure 23).
Extinction at three natural colonies (Gemini Peak, Westerholm basin and Mount Whymper)
showed that performance is not guaranteed in natural habitats by virtue of colonies being either
large or isolated. Data from clearcut colonies are also noteworthy. Clearcut colonies tended to be
much larger in size but in only 2 of 10 cases did analysis suggest that these sites were net
producers of marmots. These few positive data are probably also biased. Results from ear-tagged
animals at 3 clearcut colonies (Pat Lake, Vaughan Road and Sherk Lake) suggest that
66
Figure 22: Extinction events at natural colonies from 1985 to 1997. The data reflect sites
with no reproduction or marmots observed despite three consecutive annual counts
(). Persisting colonies are represented by (). Data show no apparent spatial
pattern or relationship between colony size and probability of extinction. Some
large colonies in the center of the current range (e.g., Gemini Peak) suffered
extinction as did some small colonies on the periphery (e.g., “P” Mountain NW
ridge). Dot sizes are proportional to numbers of adults during the mid 1980s.
This map does not show clearcut colonies, of which 5 of 10 suffered extinction
during the same period.
67
apparently high i-d-e rates were disproportionately due to adult immigration compared to that
experienced by natural colonies.
The above method of source-sink analysis based on cumulative data could mask important
temporal or spatial patterns. For example, it is unclear whether the observed patterns reflect
chronic low survival (for example, as would be the case in which habitats are slowly becoming
increasingly unsuitable for marmots) or episodic low survival (as might be the case if a disease
outbreak occurred or if there were “bad weather” years).
-7
-6
-5
-4
-3
-2
-1
0
0 1 2 3 4 5 6 7
Mean annual births
Mean annual i-d-e
Natural
Clearcut
Haley
Butler roads
Sherk Lake
Green summit
Green NW*
Pat Lake*
D13E
Vaughan*
K44A
Franklin
Gemini*
Westerholm*
D15*
Bell*
Whymper*
G2c*
Butler west face
D Rd.*
Butler east alder
"P" NW
Figure 23: Colony-specific source-sink analysis. Data are mean annual births per colony
(b) and mean annual net immigration-death-emigration (i-d-e) rates based on
counts. Labels denote the most numerically important colonies. Sites with
fewer than 5 years of observation were excluded. The dashed line indicates a
predicted b = i-d-e relationship if colonies were stable and there were no
source-sink dynamics. Departure from this line illustrates the degree to which
various colonies acted as “sources” or “sinks” during the 1972-1997 period.
For colonies that are extinct (*) this is a measure of cumulative “lifetime”
colony performance.
68
Effects of weather on survival
Weather significantly influenced marmots but generally explained only small amounts of the
variation in survival (Table 11). Adults and pups responded differently, as did marmots living in
natural and clearcut habitats.
In natural habitats, survival of pups and adults were positively associated with high spring
temperatures, number of days with significant (>5 mm) precipitation in spring, and average June
snowpack depth. The effect of these relationships may be to influence vegetation conditions more
than the marmots themselves, since onset of hibernation does not occur until September, and
juveniles are not born until June. Alternatively, for adults, spring weather conditions could reduce
foraging opportunities or increase metabolic demands beyond levels that their remaining fat
reserves can accommodate. Late-summer temperatures were unimportant but summer days with
significant rainfall were positively associated with pup survival. These results suggest that pups
may be particularly vulnerable to drought conditions.
Snow conditions in the following winter were significantly associated with survival of pups and
adults in natural habitats. Snowpack depth in early winter (December-February following the
active season) was negatively associated with survival. Late-winter (May-June following the
active season) snowpack depth was negatively associated with pup survival but weakly positively
associated with adult survival. Exposed aspect (southeast to southwest) was weakly associated
with adult survival, suggesting that local snow melt patterns could be important. Elevation was
unimportant.
For clearcut habitats, weather again explained only small amounts of the variation in survival.
As was true for natural meadows, adult survival was positively associated with spring
temperatures and days with significant rainfall, but negatively associated with total spring
precipitation. This result is intriguing and could suggest the importance of severe weather events
that deposit a large amount of precipitation over a few days (e.g., the June 1 snowstorm in 1988).
Late summer days with rainfall and late-summer precipitation produced contrary relationships with
survival of adults and pups. Adult survival was positively associated with days with rainfall but
negatively associated with precipitation. For pups the relationships were exactly opposite (i.e.,
negative for days with rainfall and positive for precipitation). These results were unexpected but
could suggest that pups and adults have different physiological vulnerabilities to soil moisture
conditions or by having to curtail foraging during periods of inclement weather.
69
Table 11: Effect of weather on marmot survival in natural and clearcut habitats. Only
significant factors are shown. All logistic regressions were highly significant
but explained only small amounts of the overall variation in survival
(McFadden’s Rho² statistic <0.2). *
Variable
Key to variables
Odds
ratio
Coefficient
P value
Rho²
overall
P value
Natural colonies
Juveniles (n = 162 records) 0.118 <0.001
MAYTEMP May-June average temperature 2.280 0.824 0.037
MAYDAYS May-June days with > 5 mm rainfall 1.443 0.367 0.000
SUMDAYS July-August days with rainfall 1.683 0.521 0.021
SNOEARLY December-February snowpack 0.956 -0.045 0.003
SNOWJUN1 June snowpack (next spring) 0.992 -0.008 0.011
SNOWJUN June snowpack (current year) 1.026 0.026 0.057
Adults (n = 638 records) 0.038 <0.001
ASPECT Aspect (exposed) 1.666 0.510 0.075
MAYTEMP May-June average temperature 1.374 0.318 0.000
MAYRAIN May-June cumulative rainfall 0.462 -0.772 0.002
MAYDAYS May-June days with >5 mm rainfall 1.337 0.290 0.000
SNOEARLY December-February snowpack 0.993 -0.007 0.003
SNOWJUN1 May-June snowpack (next spring) 1.002 0.002 0.069
SNOWJUN May-June average snowpack 1.006 0.006 0.004
Clearcut colonies
Juveniles (n = 142 records) 0.056 0.029
MAYDAYS May-June days with >5 mm of rainfall 1.301 0.263 0.037
SUMRAIN July-August rainfall 9.268 2.227 0.032
SUMDAYS July-August days with >5 mm of rainfall 0.516 -0.662 0.059
SNOEARLY December-February snowpack 0.987 -0.013 0.080
Adults (n = 388 records) 0.055 <0.001
MAYTEMP May-June average temperature 1.211 0.191 0.034
MAYRAIN May-June cumulative rainfall 0.260 -1.020 0.002
MAYDAYS May-June days with rainfall 1.232 0.209 0.006
SUMRAIN July-August cumulative rainfall 0.325 -1.123 0.001
SUMDAYS July-August days with >5 mm of rainfall 1.422 0.352 0.005
SNOEARLY December-February snowpack 1.004 0.004 0.086
* The regression model statement was Survival = Elevation + Aspect + Mayrain + Maydays + Maytemp + Sumrain +
Sumtemp + Sumdays + Snowjun + Snowjun1 + Snoearly
70
In contrast to natural habitats, survival was not associated with late-winter snowpack or aspect
for either adults or pups. I suspect that timing of snow melt may be irrelevant for the relatively
low-elevation clearcut colonies. Snowpack depths in early winter provided contrary results for
adult and pup survival (positive association for adults and negative for pups). These results were
unexpected and are difficult to reconcile; presumably the communally-hibernating marmots
experience very similar microclimate conditions.
Overall, I conclude that weather significantly influences survival but is unlikely to be the
principal cause of recent population trends.
Effects of weather on reproduction
Weather variables were not useful predictors of the probability of reproducing for marmots
living in natural habitats (Table 12). Reproduction was positively associated with low spring
temperatures but the amount of variation explained was small (Rho² = 0.015).
Weather exhibited stronger effects on reproduction in clearcuts (Rho² = 0.111). As with
animals in natural habitats, probability of reproducing was positively associated with low spring
temperatures. I can offer no reasonable explanation for this result except to speculate that estrous
cycles or hormonal processes are somehow influenced by temperature. Adults inhabiting clearcuts
were also significantly influenced by rainfall in spring of that year. Probability of reproducing was
positively associated with cumulative spring precipitation and negatively associated with days
having significant rainfall. Possibly spring rainfall patterns influence snowmelt patterns, a
hypothesis that is supported by the negative influence of late winter (May-June) snowpack.
The probability of breeding was also associated with late-summer rainfall in the previous
active season, a trend that could reflect nutritional composition of food resources. It is likely that
the physiological condition of marmots entering hibernation would influence the likelihood of
producing a litter in the following spring.
Despite some significant relationships, I conclude that weather patterns do not explain recent
reproductive trends.
Predator-prey effects
Indices of cougar and wolf abundance were poor predictors of survival in both habitat types
and for adults and pups (Rho² values < 0.03; Table 13). The hunter-sighting index for wolves was
not associated with survival, whereas for cougars it was weakly and negatively associated with
adult survival in clearcuts. These results are not surprising given the high annual variation
71
Table 12: Effect of weather on probability of reproducing.
Variable
Key to variables
Odds
ratio
Coeff.
P value
Rho²
statistic
overall
P value
Probability of producing a litter (adults only, both sexes)
Natural colonies (n = 799 records) 0.015 0.004
MAYTEMP May-June average temperature 0.865 -0.179 0.050
Clearcut colonies (n = 470 records)
0.111 <0.001
MAYTEMP May-June average temperature 0.613 -0.489 0.002
MAYRAIN May-June days cumulative rainfall 6.155 1.187 0.091
MAYDAYS May-June days with rainfall 0.634 -0.456 0.005
SUMRAIN-1 July-August rainfall (previous summer) 0.292 -1.230 <0.001
SUMDAY-1 July-August rainfall days (previous summer)
1.697 0.529 0.003
SNOWJUN May-June average snowpack 0.985 -0.015 0.001
SNOEARLY December-February average snowpack 1.014 0.014 0.076
* Model statement was Birthcode = Elevation + Aspect + Mayrain + Maydays + Maytemp + Sumrain (previous year)
+ Sumtemp (previous)+ Sumdays (previous) + Snowjun + Snoearly (previous).
Table 13: Effect of predator-prey indices on marmot survival.
Variable
Key to variables
Odds
ratio
Coefficient
P value
Rho²
statistic
overall
P value
Natural meadows
Juvenile survival (n = 124 records) 0.024 0.045
DEER Deer/km 6.537 2.524 0.048
Adult survival (all adults, n = 554 records) 0.020 0.001
COUGREM Cougars removed 0.929 -0.074 0.003
WOLFREM Wolves removed 1.044 0.043 0.005
Clearcuts
Juvenile survival (n=142 records) >0.10
no significant terms
Adult survival (all adults, n = 396 records) 0.026 <0.001
COUGAR Cougars/100 deerhunter-days 0.649 -0.432 0.000
* Model statement was Survival = Cougar + Wolf + Cougrem + Wolfrem + Deer.
72
in hunter-sighting indices; I reiterate my suspicion that such indices may not accurately reflect wolf
or cougar abundance.
Similarly, numbers of cougars and wolves removed from the population produced inconsistent
effects despite their numbers being substantial in some years (e.g., 20 cougars and 9 wolves
removed in 1989, at approximately the time that marmot colonies began to decline). In natural
habitats, adult survival was positively related to numbers of wolves removed and, curiously,
negatively associated with numbers of cougars removed. A suggestion that marmot survival is
improved by increased numbers of cougars would be counter-intuitive to say the least.
Deer abundance was a significant predictor of juvenile survival in natural habitats only. The
amount of variation explained by regressions was very small in all cases, but these results may be
partially due to inconsistencies in the predator abundance data. I cannot reject the hypothesis that
terrestrial predators have exerted important effects upon marmots.
Spatial correlation of survival
Apparent marmot survival (Nt+1/Nt) was spatially autocorrelated during both periods that
received relatively thorough sampling effort (1981-1987 and 1991-1997). Moran’s I coefficient
varied from 0.09 to 0.52, indicating weak to moderate positive correlation at all effects radii (lag
distances). However the strength of correlation varied with distance (Figure 24).
Survival was more strongly correlated at shorter distances. Significant positive correlation
was obtained at the largest radii that encompassed the entire study area (15 km), but strength of the
associations were low (I = 0.09 and 0.25). At shorter distances (< 5 km) stronger correlations
were obtained (I = 0.15 to 0.36). Spatial autocorrelation differed between the early and late
periods although the shape of the curves was similar. Survival rates were more highly correlated
during the late sampling period. This is an interesting result because recent extinctions mean
ensure that the metapopulation was more dispersed in the late period than in the early period.
Results supported the prediction that adjacent colonies would show similar survival rates
within years. Significant positive correlation at the largest effects radii lends support for the “bad
weather” hypothesis although the relationship is not strong. Stronger correlations at shorter
distances are consistent with a hypothesis of localized predator effort or disease events. Most
importantly, change in the magnitude of autocorrelation among sampling periods could suggest a
new mortality factor (e.g., disease outbreak or changing predation patterns).
73
Incidence of high mortality events
For some colonies survival did not vary greatly across years and there was little evidence for
episodes of high mortality (Figure 25). Data from the Pat Lake and Mount Franklin clearcut
colonies suggest relatively constant survival (i.e., low coefficients of variation), although neither
colony showed particularly high survival for years in which they contained more than three
individuals (median survival = 67% and 61%, respectively). At the other extreme, some colonies
with relatively high survival also showed low coefficients of variation (e.g., Sherk Lake, with
median survival of 86%). These results are consistent with a hypothesis of chronic low or high
survival corresponding to “sink” and “source” habitats (Appendix 18-20).
Many other colonies suffered distinct episodes of high mortality. For example, the Gemini
Peak natural colony showed relatively high survival (median = 0.82%) and low annual variability
(CV = 45%) but apparently lost most of its marmots in a single year (1986-1987). Similarly, the
Bell Creek natural colony exhibited high survival and low annual variation (median survival =
0.69%, CV = 36%) but suffered two high mortality events (1979 and 1995).
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
<1
<3
<5
<7
<9
<11
<13
<15
EFFECTS RADIUS (km)
MORAN'S I
Early period (1981-1987)
Late period (1991-1997)
n
= 95 site-years, 903 marmot-years
n = 71 site-years, 704 marmot-years
Figure 24: Spatial autocorrelation of marmot survival rates. Data are Moran’s I
coefficient calculated over increasing lag distances (effects radii) and using
annual survival rates weighted by the numbers of adults and pups present. All
correlations were significant and positive (Z > 1.65, P < 0.05).
74
Episodes of high mortality are a predictable fact of life for marmots because colonies are small
and because marmots hibernate communally. The well-known Haley bowl and Green summit
colonies have long monitoring histories and are therefore illustrative in this regard. Haley Lake
had typical survival rates (median = 0.69, n = 26 years) but on six occasions lost more than half of
its marmots. The Green summit colony showed similar survival trends (median = 0.65, n = 20
years), but on three occasions lost most of its marmots. Occasionally, high mortality episodes have
been catastrophic: for example the apparently well-established Hooper north natural colony
(4 adults, 3 juveniles and 5 active burrows) that was first discovered in 1982 has apparently not
contained marmots since, despite repeated surveys in subsequent years.
High mortality events were apparently not uniformly distributed across time (Figure 26).
Annual survival data coded as high mortality “events” or low mortality “non-events” using a
variety of criteria suggest that the frequency of events increased over time. Using the median
0
10
20
30
40
50
60
70
80
90
100
Sherk Lake
Gemini
Bell Creek
Road D13E
Haley bowl
Green NW
Butler roads
Pat Lake
Green
Westerholm
Mt. Franklin
Road K44A
Vaughan Rd.
COLONY
CV (%)
0
10
20
30
40
50
60
70
80
90
100
SURVIVAL (%)
CV Clearcut
CV Natural
Median survival (%)
6
6
7
15 19
17
9
7
10 11
5
9
6
Figure 25: Colony-specific variation in apparent survival. Bars are coefficients of
variation (CV) based on apparent survival of adults and pups at colonies with
a relatively complete sampling history since 1979. Median annual survival
rate (+) is also shown. Sites with fewer than five years of data were excluded.
Numbers of years in the sample are shown.
75
survival rate (65%), results suggest a gradual increase in the event:non-event ratio. Other event
criteria produce different results. Using the high survival event criterion (<80%), results suggest
that marmots have generally not experienced high survival since the mid 1980s. Of particular
interest are the results based on a more stringent (<50%) event criterion.
Episodes of high mortality occurred often (39 events and 124 non-events; rate = 24%). In 10
years the event:non-event ratio was higher than this, and in 9 years it was lower than this. The
lowest ratio occurred during the 1987-1991 period, but this is probably biased downward because
sampling efforts were made at a few relatively stable colonies. The data suggest that most periods
were characterized by an event:non-event ratio of between 20 and 25%. The ratio has been higher
than 35% since 1994, and it was also higher than this during the 1984-1986 period. The early
sampling period is interesting because it does not suggest high spatial correlation. Inspection of the
data suggests that events occurred at widely separated colonies (e.g., Mount Whymper, “P”
Mountain, Mount Buttle, Green Mountain) during the mid-1980s. It is also interesting that these
7
10
12
15
19
18
14
11
6
7
6
5
5
9
12
10
12
11
12
1978 1983 1988 1993 1998
YEAR
0.0
0.2
0.4
0.6
0.8
1.0
EVENT:NON-EVENT RATIO
<80% survival
<65% survival
<50% survival
Event Criteria
1978 1983 1988 1993 1998
YEAR
0.0
0.2
0.4
0.6
0.8
1.0
Figure 26: Incidence of high mortality events. Data points are event:non-event ratios
using median annual survival (65%) as the event criterion. Lines are
LOWESS regressions using three event criteria. Site-years with fewer than
three marmots were excluded. Numbers of colonies in the sample are shown.
76
events went largely unnoticed by count crews, presumably because attention was focused on
expanding clearcut colonies at that time.
The apparent increase in frequency of high mortality events is disturbing. Data from Haley
Lake are again illustrative. At this colony, three of the four survival years since 1994 represented
high mortality events and the fourth was borderline (survival = 50%). Given the natural history of
M. vancouverensis, no colony could be expected to withstand the demise of 18 of 21 juveniles
produced during a four-year period, and this was the fundamental cause of near-extinction at this
site by 1998. Intriguingly, if the last four years of episodic mortality are excluded, median survival
was increased (to 71%) to the extent that life-table analysis would suggest a relatively stable
population (Ro = 0.94, = 0.985).
From these data I conclude that the frequency, magnitude and spatial correlation of episodes of
high mortality have increased in recent years. These “crashes” are consistent with a hypothesis of
disease outbreak or increased hunting effort by predators within a small geographic area.
77
DISCUSSION
Caughley and Gunn (1996) offered a straightforward model for managing endangered species.
First, determine whether populations are declining or whether other evidence suggests that a problem
exists. Second, learn about the ecology of the organism and use the accumulated knowledge to
construct hypotheses about possible causal factors. Third, subject the resulting hypotheses to
rigorous scientific testing, preferably under controlled experimental conditions. Finally, use the
results to reverse the factors that are causing the problem.
Vancouver Island marmots illustrate many of the difficulties involved in trying to apply
Caughley and Gunn’s model to a “real-world” endangered species issue. Some difficulties are
practical in nature. For example, despite over a decade of count efforts, at the beginning of this
study there was no quantitative information about abundance or population trends because results
had not been mapped, analyzed or tested for consistency. Demography was not well understood, and
as recently as the early 1990s there was considerable uncertainty about whether marmot populations
were increasing or declining (Janz et al. 1994). For these reasons much of my study was necessarily
descriptive and designed to answer basic questions about population ecology. Other issues pose
difficulties of a more philosophical nature. For example, neither marmot populations nor the
landscape remained static during the study. The landscape became increasingly modified by forestry
activities and the marmot population changed in abundance and structure. By definition there could
therefore be no “control” or “treatment” groups with which to test hypotheses using a classical
experimental approach (Popper 1968). For similar reasons we can never know why M.
vancouverensis disappeared from central Vancouver Island. Empty burrows and unoccupied
habitats provide few opportunities to identify causal mechanisms.
My approach was to test whether observed patterns were consistent with predictions made using
a variety of hypotheses. Note that this approach can yield only “strong inference” and not “proof” of
causal factors (Platt 1964). However, I suggest that given ~100 animals left in the world, it is the
only possible approach. Recovering M. vancouverensis from the brink of extinction is
fundamentally a management issue, and managers need to know not only what the problems are, but
what the problems are not. The evidence suggests that Vancouver Island marmots are declining not
because of one factor but because of several. Some factors are more important than others.
78
Habitat tracking
The environmental tracking hypothesis depends on issues of temporal scale. Over the long term,
Vancouver Island marmots are presumably tracking climatic changes and associated vegetation
patterns. Nagorsen et al. (1996) suggested that this could be inferred from finds of prehistoric bones
well outside the core area of current distribution, and this interpretation is probably correct. Extra-
limital finds of prehistoric marmot bones tell a similar story in other parts of the world (e.g.,
Preleuthner et al. 1995, Grayson 1987). Replacement of tundra parkland by forest has greatly
reduced the quantity of marmot habitat available in the Pleistocene-Holocene prehistoric past. Tree-
lines on mountains of western North America have changed substantially over the past 10,000 years
in response to changing climate (Rochefort et al. 1994).
There is little evidence for habitat tracking in historical times. Within the past 100 years a
warmer and drier climate has resulted in tree invasion of sub-alpine meadows in the Olympic (Fonda
and Bliss 1969, Schreiner and Burger 1994) and Cascade mountains (Franklin et al. 1971).
However, dendrochronological work at historic M. vancouverensis colonies north of Alberni Inlet
does not support the hypothesis that similar processes caused recent marmot extinctions there. In
western Strathcona Provincial Park, where marmots apparently disappeared some 10-30 years ago,
most trees are more than 300 years of age and there is little evidence of recent forest succession (C.
Laroque, University of Victoria, unpublished data).
Similarly, fire apparently plays a minor role in maintaining marmot habitat. Milko and Bell
(1985) reported a distinct charcoal layer at Gemini Peak and suggested that an extensive fire created
open meadow habitat at that site. Recent tree-ring analyses do not support this interpretation.
Laroque (1998) determined a large range of tree ages at Gemini Peak and estimated a minimum
disturbance interval of at least 250 years. This is congruent with results from Lertzman et al.
(1998), who suggested that intervals between major fires were relatively short (<300 years) on
southeastern Vancouver Island but substantially longer (700 - 3000 years) in western and central
regions.
The evidence for environmental tracking in recent decades is weak, at least for marmots in
natural habitats. Few of the predictions were strongly supported by the data.
Reproduction and survival will be chronically low at some natural colonies
A few natural colonies (e.g., Westerholm Basin, Meadow #1) had relatively low per capita birth
rates and apparent survival compared to other sites. Low coefficients of variation suggest that these
79
patterns were “chronic”, supporting the idea that habitats are gradually becoming unsuitable because
of tree invasion or more subtle changes in vegetation. However, most natural colonies did not
display similar trends. Colonies such as Hooper North, “P” Mountain, South Green or Gemini Peak
exhibited relatively high survival and reproduction until catastrophic losses occurred. Extinctions
happened abruptly and with no suggestion of a gradual decline in habitat suitability. I conclude that
environmental tracking may be occurring at a small number of natural habitats but this process does
not explain recent population trends.
Reproduction and survival will be associated with age of regenerating clearcuts.
Adult survival was weakly and negatively associated with clearcut age. The apparent reduction
in survival among clearcuts coded as “young” or “old” (>10 years old) was also notable. It appears
that marmots already living in regenerating clearcuts represent a special case of environmental
tracking. However, the data are more suggestive of a threshold effect than of a gradual demographic
response to forest succession. Clearcuts generally become unsuitable for marmots after the age of
approximately 15 years, although site-specific conditions undoubtedly influence this timing.
Colonization of clearcuts will occur in proportion to their availability.
Colonization of clearcuts did not occur in spatial or temporal congruence to habitat availability.
Marmots apparently did not colonize clearcuts until 1981, despite availability of nearby high
elevation clearcuts during the mid 1960s. It appears that marmots are apparently not simply
colonizing places when they become suitable, although they are clearly disappearing when these
habitats become unsuitable.
Overall, the evidence for environmental tracking over the short term is weak. Data from natural
habitats do not support a hypothesis of gradual vegetation change. Seral stage of regenerating
clearcuts apparently represents a special case of environmental tracking. Marmots did not colonize
clearcuts in temporal or spatial relation to their availability. Even allowing for substantial lag
effects that might occur given the difficulty of finding and successfully colonizing clearcuts, the data
are at odds with a fundamental prediction of the tracking hypothesis: marmots did not increase in
numbers despite greatly increased habitat availability.
Weather hypothesis
Weather has been shown to have significant effects on other marmot species (e.g., Barash 1973,
1989, Van Vuren and Armitage 1991, 1994). My results support this interpretation but suggest that
80
weather effects are relatively modest. No prediction made under the weather hypothesis was
strongly supported by the data.
Reproduction and survival will be associated with weather measurements.
Weather variables explained only small amounts of the variation in survival and birth rates.
There was a consistent positive relationship between survival and the number of spring days with
significant rainfall events, yet survival was negatively associated with cumulative spring rainfall.
This result could reflect a threshold effect, in which too much rainfall results in poor survival
conditions, but consistent moderate rainfall results in improved survival conditions. That marmots
in all habitats responded consistently to these variables is particularly interesting, and could reflect
nutritional factors or growth of parasites or disease.
Snowpack conditions were also important, but caution is needed in interpreting this result
because local conditions probably differed considerably from those at snowpack measurement
stations. However, the evidence from studies of other marmot species is compelling. Survival of M.
olympus was positively associated with increasing snow depth, possibly because of increased
insulation value of the snowpack (Barash 1973). Arnold’s (1990a, 1990b, 1992, 1993) work on M.
marmota suggested a possible physiological explanation for this. Indeed he suggested that a driving
force behind the evolution of marmot sociality can be found in the physiological necessity for a
thermally-stable environment in which to hibernate. Recent work on M. marmota lends additional
credence to the snowpack-survival hypothesis (Farand et al. in prep.), but there has been no
additional work concerning whether the mechanism is actually one of snowpack insulation.
Specifically, the depth at which soils experience zero annual temperature amplitude (see Brown
1970) has not been related to the depth at which marmots hibernate, or to soil types. Van Vuren and
Armitage (1991) suggested that the timing of snow melt, and not the depth of snowpack, may be the
critical factor in determining marmot survival.
Reproduction and survival will be associated with habitat type or site characteristics such as
elevation or aspect.
There was a weak association of survival with colony aspect, as would be predicted if snowpack
effects were influenced by site exposure. Marmots inhabiting clearcut and natural habitats
responded differently to weather conditions, which is not surprising given recent work on the
nutritional and hibernation requirements of other marmot species. Physiological studies provide
important clues regarding how short-term climatic variation could influence marmot survival during
winter hibernation (Arnold et al. 1991). The laboratory study by Thorp et al. (1994) may also be
81
very relevant; in this study yellow-bellied marmots were fed diets containing high or low amounts of
polyunsaturated fatty acids. Marmots supplied with a diet deficient in essential fatty acids showed
identical weight gains during summer, but exhibited higher spontaneous arousal rates, shortened
bouts of deep hibernation and higher overall metabolic expenditures. That this could lead to
increased winter mortality is clear (Arnold 1993). It is likely that between-site soil characteristics
and annual weather variation could influence both the availability and chemical composition of
plants eaten by marmots (Sinclair et al. 1982, Walker et al. 1993).
Marmots inhabiting clearcuts do not have access to the same variety of food plants that they do
in natural sub-alpine meadows. Reliance upon a small number of species (especially Anaphalis
margariticea) could result in altered biochemistry or lack of essential nutrients in marmot diets,
particularly if weather conditions such as drought produce important effects on these few plant
species. Armitage (1994) found that a year of particularly low survival for M. flaviventris was
associated with a short snow-free growing season, and with low rainfall during summer. Blumstein
and Foggin (1997) determined that vegetation availability was strongly related to the probability that
red marmots (M. caudata) successfully weaned juveniles in the following spring.
Weather had a marginal effect on the probability of reproducing in natural habitats. However,
animals living in clearcuts were significantly influenced by rainfall and snowpack patterns and
weather explained moderate amounts of the variation in reproductive status. Again there was a
different response to rainfall magnitude and rainfall consistency, which could suggest nutrition or
parasite/disease conditions that are influenced by moisture regimes.
Episodes of high mortality will occur randomly over time.
High mortality events did not occur randomly over time. Regardless of the event criteria used,
the incidence of high mortality events systematically increased with time. This trend is inconsistent
with a hypothesis of population response to random fluctuations in environmental conditions.
Ultimately, weather exerts significant effects upon marmots but does not appear to explain
population trends. I conclude that recent population declines cannot be attributed to a succession of
“bad” weather years.
Sink-connectivity
There is growing empirical evidence from a variety of taxa that some organisms are maintained
in qualitatively heterogeneous “source-sink” metapopulations (e.g., Breininger et al. 1995, Donovan
et al. 1995, Wauters and Dhondt 1990). There is also growing appreciation for the critical role that
82
dispersal or “connectivity” plays in maintaining metapopulations, and consequently for the effects
that fragmentation may have (Fahrig and Merriam 1985, Hanski 1991, Dunning et al. 1992, Taylor
et al. 1993, Fahrig 1997). Some organisms simply change their movement behavior in response to
altered spatial distribution of habitats, seemingly with no effect on reproductive performance or
survival (e.g., Taylor and Merriam 1995, Matthysen 1995). This appears not to be the case for
Vancouver Island marmots. Most of the predictions of the sink-connectivity hypothesis were
supported.
Reproduction and survival will be chronically low in clearcuts.
Mark-recapture results suggested that animals living in clearcuts experience a small (5-10%) but
chronic reduction in annual survival. Population count results corroborated this finding for pups but
were inadequate to evaluate adult survival because of the inability to distinguish survivors from
immigrants. Despite a sample of tagged animals that represented 10-15% of the population,
parametric estimates of survival were statistically insufficient to resolve the small observed habitat-
specific or temporal differences.
Life-table analysis illustrates the biological importance that small changes in survival could
have. Tagged females living in clearcuts produced only half as many offspring as their counterparts
in natural meadows did, and these offspring themselves had a substantially reduced probability of
reaching dispersal age. There are theoretical problems in using uncorrected estimates of net
reproductive rate (Ro) as a measure of the degree to which populations may be increasing or
decreasing (Gregory 1997). Specifically, this rate is valid only if the population in question exhibits
constant birth and death rates (thereby yielding a stable age distribution), an assumption which is
unlikely to be true for most wild populations and especially for Vancouver Island marmots. Despite
such difficulties the data are consistent with a hypothesis of reduced survival in clearcuts and with
observed population trends.
However it would be simplistic to conclude that natural habitats always acted as “sources” or
that clearcuts always acted as “sinks”. Although mortality appears to be concentrated in the
clearcuts, some (e.g., Mount Franklin and Sherk Lake) showed high survival in some years, and
some natural habitats also showed chronically poor performance. The fundamental difference
between the two habitat types is of course temporal. Unlike natural meadows, clearcuts represent
temporary habitats that have a brief “window” of perhaps 10 or 15 years during which they are
capable of contributing dispersers to the overall metapopulation. This is a short period of time in
terms of marmot generations, and represents a temporal habitat restriction that is novel for the
83
species. Because natural meadows do not change so quickly, M. vancouverensis has not had the
opportunity to evolve appropriate behavioral responses to this process. Specifically, adults
apparently do not move as clearcut habitats change with forest succession. In short, behavior that
was adaptive during the past 10,000 years may be maladaptive in habitats created during the last 20
years.
Isolated colonies will show higher extinction rates.
Uncertainty over the timing of extinctions made it impossible to test for the influence of spatial
isolation directly. However, a map of extinction events did not support the prediction that isolated
colonies have a higher extinction probability. Some small and isolated natural colonies became
extinct, but several large colonies near the center of the geographic range also became extinct.
Isolated colonies will show lower apparent survival due to reduced immigration.
Because of small colony sizes, Vancouver Island marmots are extremely vulnerable to random
events. Most mountains apparently contain only 1 or 2 family groups (Bryant and Janz 1996), and
the fate of single individuals causes important population effects. Local extinctions, lack of
reproduction, and immigration “rescue effects” (Brown and Kodric-Brown 1977) were observed on
several occasions that were apparently caused by the appearance or disappearance of single adults.
Isolated colonies exhibited lower apparent adult survival, which is congruent with a hypothesis of
reduced immigration and consequent reduced probability of “rescue effect” or recolonization of
vacant habitat patches.
Colonizations of clearcuts will be spatially concentrated.
The evidence for reduced landscape connectivity is indirect but suggestive. Clearcuts were not
colonized in proportion to their spatial availability. Colonizations appeared to be limited by the
presence of nearby natural colonies at which numerous potential emigrants were produced.
Colonization events were spatially clustered compared to clearcut habitats actually available. Few
tagged marmots made long-distance dispersal movements and only one dispersing female is known to
have eventually reproduced at a new site. All of these results are congruent with a hypothesis that
colonization hinged on successful dispersal from a small number of “source” colonies, that
dispersing marmots were “short-circuited” by the presence of nearby alternative habitats in which to
settle, and that clearcuts consumed more marmots than they produced.
Dispersal is notoriously difficult to study in the wild. Simulation modeling indicates that
metapopulations are extremely sensitive to changes in dispersal, and in particular to changes in rates
84
of long distance movements (e.g., Lindenmayer and Lacy 1995, Lamberson et al. 1992). Yet we still
know relatively little about why real organisms disperse or how they respond to changes in the
physical environment (e.g., Lidicker and Koenig 1997). Like other alpine-dwelling marmots, M.
vancouverensis apparently exhibits presaturation dispersal (animals disperse at a wide range of
population densities) and this may reflect the adaptive tendency of individuals to disperse when they
approach reproductive age (e.g., Arnold 1990a, Barash 1989). However, in the case of M.
vancouverensis, perhaps the question of what makes a marmot begin to move is less important than
what makes it stop.
Apart from the obvious habitat cues furnished by newly available clearcuts, there may be
behavioral issues to consider. Weddell (1991) determined that Columbian ground squirrels
(Spermophilus columbianus) readily moved between habitat patches but invariably settled in
habitats that contained other animals. For this reason recolonization of vacant habitat patches did
not occur. It is tempting to interpret this idea to the highly social M. vancouverensis (Heard 1977),
particularly since dispersing animals would commonly encounter other marmots in nearby clearcuts,
and high mortality in these habitats could lead to unused burrows and therefore perhaps to
“vacancies” in existing social groups.
The evidence for the sink-connectivity hypothesis is relatively compelling. Regenerating
clearcuts in montane sites above 700 m in elevation resemble natural marmot sub-alpine meadow
habitats for at least 10 to 20 years after harvest. This resemblance allowed marmots to colonize
some clearcuts but population expansion was temporary and limited in geographic scope. Only a
small proportion of available clearcuts were eventually colonized. By providing new alternative
habitats in which to settle, forestry apparently changed natural dispersal patterns and probably
reduced the rate at which animals were able to colonize distant habitats. The most important
forestry effect was probably to concentrate the population and therefore exacerbate the “eggs in a
small basket” problem, increasing the risk of mortality from processes that occur within a restricted
geographic area.
Predators and prey
Marmota vancouverensis has evolved in the presence of wolves, cougars and eagles and exhibit
a variety of typical anti-predator strategies that reduce predation risk. They appear to be no less
vigilant than other marmots (D. Blumstein, University of Kansas, unpublished manuscript). To
propose that predators have suddenly (in evolutionary terms) become of critical importance is not
logical unless either predator abundance or hunting effort upon marmots has increased. The
85
evidence suggests that increased predator abundance is unlikely, but that increased hunting effort by
predators is congruent with observed trends.
Most mortality will occur during summer.
In clearcut habitats, the timing of last observation of tagged marmots suggested that most
mortality occurs during hibernation. In contrast, disappearances in natural habitats were more
evenly distributed throughout the active season, which is consistent with the predator hypothesis.
Cougars, wolves and eagles can and do exert significant impacts on marmots. Radio-telemetry and
direct observation provided conclusive evidence of predator-kills, and several colonies declined after
the loss of breeding-age animals. In addition, the frequency of losses to predators was substantial
(16%) compared to the numbers of marmots equipped with radio-transmitters.
Survival will be associated with abundance of predators or alternative prey such as deer.
Merilees (1980) suggested that increasing golden eagle abundance may have had an impact upon
marmots. Unfortunately, data with which to estimate golden eagle population trends in the Nanaimo
Lakes region are non-existent. While it is interesting that the first confirmed nest record for that
species on Vancouver Island was close to a historic marmot colony (Upper Campbell Lake in 1954;
Campbell et al. 1990b), other sightings had been made well before that (Carl 1943) and it is perhaps
significant that Swarth (1912) collected M. vancouverensis from a location known as “Golden Eagle
Basin”. In any case, the small number of observed predation incidents involving eagles makes me
skeptical that marmot population trends could be attributed to changing eagle abundance.
Cougar and wolf abundance indices were poor predictors of marmot survival. However, I
reiterate my concern that predator sightings or removals may not reflect true abundance. Annual
indices fluctuated more dramatically than expected given the life-history characteristics of cougars
and wolves. Certainly the available data cannot be interpreted to suggest significant changes in
predator abundance during the study. It is therefore not surprising that predator indices showed no
consistent relationship with marmot survival.
Predation pressure does not depend on predator abundance alone. Drastic declines in mule deer
abundance raise the possibility that predators might be increasingly “switching” their hunting effort
(Bergerud 1983) to M. vancouverensis. The significant but weak association between deer
abundance and survival of marmot pups in natural habitats supports this idea, but the absence of
similar relationships for adults or pups in clearcuts provides conflicting evidence.
86
Survival rates will be spatially correlated.
Changes in the hunting behavior of predators could also be important. Marmot survival rates
were spatially correlated within an effects radii of several km, which is a predicted result if predators
concentrate their hunting efforts at nearby colonies. Observations of predator tracks leading from
colony to colony are consistent with that possibility, and with the idea that a small number of
individual predators have become particularly adept at hunting M. vancouverensis.
Predator hunting behavior cannot be divorced from the landscape changes produced by forestry
activities. It is possible that terrestrial predators may hunt more successfully in clearcuts because
tree growth limits visibility, and predators may therefore be better able to stalk marmots while
remaining undetected. The difference between survival in “young” and “old” clearcuts is congruent
with that possibility. Logging roads could also function as corridors that provide easier movement to
both predators and marmots, thereby increasing the frequency of interactions (e.g., Simberloff et al.
1992). Regardless of whether such ideas are correct, colonization of clearcuts had the effect of
dramatically increasing marmot abundance in a small (40 km²) area, and this presumably increased
the potential benefits of hunting within that area. The expected consequence would be to
simultaneously increase predation pressure at nearby natural colonies, and this is consistent with the
relatively rapid decline of marmots at natural colonies in the central portion of the Nanaimo Lakes
region.
Disease
The potential importance of disease and parasites has long been recognized (Anderson and May
1979, May and Anderson 1979). There are several cases in which disease devastated populations on
a continental scale (e.g., Dobson and May 1986, Geraci et al. 1982), and the risks for small or
restricted populations are even more extreme. Canine distemper and black-footed ferrets (Mustela
nigripes) must represent the classic case in this regard (Clark 1997), but there are undoubtedly
others (Laurance et al. 1996).
The role of disease or parasites in regulating other species of marmots is not well understood.
There is a large Eurasian literature concerning bubonic plague (Yersinia pestis) in marmots (the
European “black death” of 1347-1348 may have originated from trade in marmot pelts; McEvedy
1988). Many other possible pathogens are carried by marmots, including Tularemia (Zykov and
Dudkin 1996), Leptospirosis, Toxoplasmosis, Ricketsiosis, Listeriosis,
Pseudotuberculosis=Yersiniosis, Salmonellosis and Powassan Encephalitus (Bibikov 1992).
Woodchucks (M. monax) carry a virus similar to human hepatitis B (Summers et al. 1978).
87
External and internal parasite-loads have been studied in several species (e.g., Ageev and Pole 1996,
Callait et al. 1996, Gortazar et al. 1996, Preleuthner et al. 1996). The consistent result of these
studies is that while infections and infestations occur, there is no evidence to suggest that they have
caused significant population declines (M. Callait, Université de Lyon, and S. Pole, Kazak
Antiplague Research Institute, pers. comm.).
However, this does not mean that the risks disease or parasites are insignificant. What may be
unimportant for a widespread or abundant species may be extremely important for a restricted
population of Vancouver Island marmots. The evidence is generally consistent with a hypothesis of
increased incidence of disease or parasitic infestation.
Survival will be density-dependent.
Disease and parasite problems are often considered to be density-dependent because the risks of
infection are higher in high density populations (May and Anderson 1979). The data corroborated
this prediction. Adult survival was negatively associated with observed/expected ratios. Survival of
pups was not associated with relative density in linear fashion, although survival was visibly lower
in colonies of high density (>150% of expected).
Survival rates will be spatially correlated.
Disease outbreaks would be expected to occur more frequently at nearby colonies. Marmot
survival was spatially correlated, and in this case results are especially intriguing because the
magnitude of spatial correlation increased substantially between the early (1981-1987) and late
(1991-1997) sampling periods. The increase in spatial correlation occurred despite the marmot
population becoming much less concentrated during the latter period.
The incidence of high mortality events will increase over time.
One possibility is that Vancouver Island marmots have been recently exposed to a novel
pathogen to which they have no immunity. This presents a frightening scenario in which the
infection would be expected to spread throughout the metapopulation, and in which the incidence of
high mortality events would increase as a result. The evidence is consistent with this prediction.
Episodes of high mortality apparently became more numerous in the 1991-97 period, and perhaps
more disturbingly they became increasingly spatially correlated.
Most large colonies in the center of the geographic range suffered extremely high losses over
short periods (e.g., Gemini Peak in 1986-87, Vaughan Road in 1988-89, Bell Creek in 1993-94, and
Butler roads, Road K44A and Haley Lake in 1994-95). In these cases most disappearances
88
apparently occurred during hibernation, a period when marmots may be particularly vulnerable to
infection. The negative evidence may also be relevant. During the late sampling period several
isolated colonies (e.g., Big Ugly, Mount Franklin) apparently did not suffer similar episodes.
Whether an epizootic is responsible for the disappearance of M. vancouverensis from areas
north of Alberni Inlet remains unknown and untestable. However it is apparent that catastrophic and
abrupt losses from large, apparently healthy colonies contributed significantly to the overall decline
of the Nanaimo Lakes metapopulation. I conclude that the threat of extinction from pathogens is
real, and stress that epidemiological processes should not be viewed in isolation. For example,
increased virulence in the bacterium Y. enterocolitica has been associated with unfavorable climatic
conditions and poor nutrition (Zwart 1993, Blake et al. 1991). In addition, as with the predator
hypothesis, increased density of marmots due to colonization of nearby clearcuts greatly increased
the vulnerability of marmots by increasing the risk of infection.
Converging lines of evidence
Vancouver Island marmots live in a complex world. There is no a priori reason to imagine that
a single factor is responsible for recent population dynamics, or that the same suite of factors acted
uniformly across time or space. M. vancouverensis is and will continue to be influenced by a
multitude of mortality factors including unsuccessful hibernation, predators, and disease. In many
respects it therefore represents an organism caught in the “vortices of extinction” (Gilpin and Soulé
1986), and its story is not dissimilar to that of the heath hen (Tymphanuchus cupido), which became
extinct after disease, predators and wildfires had taken their toll on an already depleted population
(Caughley and Gunn 1996).
South of Alberni Inlet, forestry appears to be the principal factor behind recent marmot
population trends. Logging reduced the ability of marmots to re-colonize or “rescue” isolated
natural colonies by the simple mechanism of creating large amounts of nearby alternative habitat.
Animals living in clearcuts produced fewer than half the number of potential dispersers than did their
counterparts in natural meadows. The fast pace of forest regeneration after clearcut logging posed
new challenges for animals that are not adapted to take advantage of habitats that change over a
period of several years. Both the colonization pattern and the lack of long-term expansion are
consistent with a hypothesis of source-sink regulation, although it appears that clearcuts are not
necessarily sinks, or sinks in all years.
89
The most important impact of forestry was to increase population concentration, thereby making
individual colonies more vulnerable to predators and disease. Sweitzer et al. (1997) documented
near extinction of porcupines (Erithizon dorsatum) from a 15 km² area in Nevada due to cougar
predation and suggested that reduced deer populations were the proximate cause. Beier (1995)
observed that cougars will readily use dirt roads and trails. Observations of marmots and predators
travelling along roads makes it logical to suggest that road networks increased the frequency of
marmot-predator interactions. The growing rate of high mortality events is consistent with a disease
outbreak, and again the implications are made more severe because of population concentration.
Finally, recent population trends in the Nanaimo Lakes metapopulation are disturbing because most
animals occupy habitats that will become unsuitable in a few short years. This structural problem is
not easily resolved except that over time forest regeneration will occur and sites will become
unattractive to marmots. Road networks may well act as a more permanent structural change, at
least as they pertain to facilitating movements of predators and prey.
It remains unknown why marmots disappeared from areas north of Alberni Inlet. Predators,
weather, disease and demographic stochasticity probably all played a role. Landscape connectivity
may have changed for several reasons. Tree invasion may have occurred at important “stepping
stone” colonies such as those that apparently existed on Mount Arrowsmith and in the Beaufort
range prior to the 1980s. Another possible factor may relate to construction of Strathcona Dam at
the north end of Buttle Lake. This dam, completed in 1957, raised the water level by 9 metres, and
increased the area of the lake substantially (it inundated nearly 90% of the lake’s tributary stream
spawning areas; Hyntka 1990). The reservoir could have acted as a new barrier and influenced
dispersal and metapopulation dynamics. Evidence from the Nanaimo Lakes metapopulation
indicates how important dispersal processes are in maintaining colonies, and the timing of dam
construction would seem to be congruent with probable marmot disappearances after the 1950s.
Finally, Nagorsen et al. (1996) documented prehistoric marmot hunting by humans, possibly
significant hunting events also occurred in modern times.
It seems unlikely that we will ever definitively answer the question of why marmots disappeared
from central Vancouver Island. Historical marmot population data simply do not exist with which to
test the relative importance of various environmental effects. However, perhaps focusing on this
question would be less useful than focusing on an essential dichotomy. Specifically, while several
plausible mechanisms can be invoked to explain disappearances, few can be raised to support the
idea that marmots might re-colonize such habitats on their own. There are several sites in western
90
and eastern Strathcona Provincial Park where abandoned burrows and vegetation conditions suggest
that potential natural habitats are still available. Unfortunately, geography and natural and
landscape changes make it unlikely that marmots could successfully re-colonize these sites if left to
their own devices.
Lessons from marmots
Most alpine-dwelling marmots exhibit metapopulation structure (they are colonial, inhabit
discrete habitat patches that suffer local extinctions and recolonizations, and are connected by
dispersing individuals within an ecological time scale). However, with a few notable exceptions
(e.g., Armitage and Downhower 1974, Schwartz et al. 1998) marmots have been largely ignored by
students of metapopulation and landscape ecology. This trend is unfortunate. The same attributes
that make marmots such useful organisms for students of evolution, sociobiology, and demography
also make them good candidates for metapopulation studies (specifically: they are diurnal, large
enough to be easily seen, respond well to capture and marking, and exhibit generally predictable
habitat-affinities and activity-rhythms: see Armitage 1992). Most importantly, there are 14 species
that display a wide range of colony sizes, social assemblages, densities and distributions with which
to test ideas about landscape connectivity and population persistence. Comparative studies among
threatened and non-threatened marmots could indeed build a useful bridge between what Caughley
(1994) termed the “small population paradigm” (what happens to small populations) and the
“declining population paradigm” (what causes populations to become small?”)
The source-sink concept has important ramifications for conservation biology, principally
because it suggests that organisms may not necessarily be most numerous in habitats where they are
most successful in demographic terms (Pulliam 1988, Wauters and Dhondt 1990). Metapopulation
processes that involve changes in landscape connectivity or fragmentation complicate the issue.
Many habitats may be unoccupied not because they are unsuitable but because organisms can’t get
there (Dunning et al. 1992, Fahrig 1997). For many species it is therefore extremely dangerous to
equate high abundance with habitat quality, or to equate low abundance with habitat unsuitability.
Unfortunately, measures of relative abundance still form the basis of many management and land-
use decisions involving threatened species, perhaps because of the mistaken belief that relative
abundance is all that can be measured given practical field difficulties or intrinsic rarity of the
organism in question. The M. vancouverensis experience suggests that such beliefs may be
unwarranted.
This study found close congruence among demographic estimates from marked and unmarked
segments of the study population. Admittedly, marmots represent an unusual case (because
91
juveniles, yearlings and adults are easily distinguished it is possible to estimate birth and survival
rates). However the point remains: carefully designed and tested count methods may be a valuable
adjunct to more intensive (and expensive) monitoring techniques if the basic biology of the organism
is understood. Conversely, unadjusted population counts or estimates of abundance based on
“detection rates” have the potential to provide highly misleading impressions about demographic
performance or population trends if the basic biology is poorly understood.
Recent advances in technology (such as implantable radio-transmitters) and theory (such as
Cormack-Jolly-Seber modeling of survival) are welcome but I caution that there will be inevitable
trade-offs between intensively sampling a small population using such techniques and sampling a
larger population using less intensive methods. Results from M. vancouverensis illustrate the
problem well. Even robust scientific methods may be inadequate to resolve small but potentially
significant changes in demographic rates if only a hundred or so animals exist. For this reason I
believe that much more work is needed to determine valid measures of demographic success for rare
species and to integrate the results of “intensive” and “non-intensive” research methods.
Questions of spatial and temporal scale are critical. I wish to add my name to the long list of
authors who stressed the value of long-term research (e.g., Armitage 1991, Schwartz et al. 1998)
conducted on a geographic scale large enough to address the problem (e.g., Pimm 1991, Caughley
and Gunn 1996). In the case of M. vancouverensis, a study of typical duration for graduate students
(2-5 years) and involving only a few colonies would have produced very different impressions about
what was happening to marmots.
Rarity has been described as the inevitable result of a limited niche pattern (Gaston 1994).
Under this interpretation, niche pattern consists of niche breadth (the range of environmental
conditions in which organisms can survive and reproduce) and niche position (the spatial and
temporal availability of these resources and population abundance). There is no evidence to suggest
that Vancouver Island marmots have a particularly narrow niche breadth. They live and reproduce
in a variety of habitats and exploit a wide variety of food resources. The small amount of
comparative work done with M. vancouverensis and other species suggests that they are not
dissimilar to other alpine-dwelling marmots in antipredator behavior (D. Blumstein, University of
Kansas, pers. comm.) or sociality (Heard 1977).
There is strong evidence for limited niche position. Vancouver Island marmots inhabit a
spatially patchy environment in which the size of available meadows is small compared to the
92
surrounding matrix. However, limited niche position is not a new phenomena. Low rates of natural
disturbance (Lertzman et al. 1996) suggest that niche position has remained essentially unchanged
for at least the last several thousand years. Some habitat losses undoubtedly occurred from forest
succession and some gains occurred after fire or avalanches, but in general there is little support for
the idea that habitat quality or quantity have systematically changed over a temporal scale measured
in centuries. Vancouver Island marmot abundance and distribution in the mid 1900s were probably
quite similar to what they were several hundred years ago.
Hence the paradox of M. vancouverensis. Marmot persistence at particular sites has been quite
remarkable and yet they disappeared from about two thirds of their apparent recent geographic range
within the past 50 years. They successfully colonized some clearcuts and greatly increased numbers
within a few years yet did not expand in proportion to habitat availability. It is difficult to reconcile
recent population trends in the severely human-modified environment on private lands south of
Alberni Inlet with the loss of marmots from seemingly pristine sites on central Vancouver Island. Or
at least it is difficult if one searches for a single causal factor. Perhaps the recognition that there are
many is the most important result of this study.
Given rejection of the tracking and weather hypotheses there seems little reason to expect that
population trends might suddenly reverse themselves. An aggressive program of captive-breeding
combined with reintroduction appears to offer the only realistic hope of preventing extinction of this
species. Fortunately, marmot captive-breeding programs have achieved success both in Russia (with
M. bobac, M. baibacina and M. menzbieri) and the United States (with M. monax, M. broweri and
M. flaviventris). There would appear to be no particular technical difficulties associated with
maintaining and breeding marmots in captivity (e.g., Tokarsky 1996, Rymalov 1996, Concannon et
al. 1989). Similarly, reintroductions of M. marmota and M. bobac have been successful in western
Europe (Ramousse et al. 1992, Dmitriev et al. 1994). In particular, alpine marmots (M. marmota)
disappeared from portions of the Alps (Preleuthner et al. 1995) and from the Pyrenees during
historical times. As is the case for M. vancouverensis, causes of disappearance remained unknown
but reintroduction was spectacularly successful despite this lack of knowledge. Reintroduction of
several groups of wild-captured M. marmota in the 1960s has resulted in an apparently stable
population of several hundred animals (Herrero et al. 1994). In the Alps, substantial portions of the
extant population is thought to have been derived from transplants that occurred since the 1940s
(Preleuthner et al. 1995, Ramousse et al. 1992). There appears to be no reason why reintroduction
93
could not be equally successful on Vancouver Island, provided that sufficient reintroduction habitat
still exists.
Additional research is required in several broad areas of study. Much more work is required in
the field of epidemiology, with the objective of identifying pathogens of relevance to M.
vancouverensis. Emphasis must be placed on developing diagnostic tools and effective treatments.
Further work involving spatial patterns of survival could provide important clues about possible
modes of transmission (e.g., Mantel 1967, Rossi et al. 1992). The idea that hibernation success
could depend on the nutritional composition of food plants also needs to be tested, particularly as it
relates to dietary differences among habitat types and possible weather effects (e.g., Thorp et al.
1994). Finally, more work is needed to identify and prioritize reintroduction habitats on Vancouver
Island, and to test for the possibility of systematic environmental change either through tree invasion
or more subtle vegetation change in natural sub-alpine meadows (e.g., Walker et al. 1993).
Successful management of endangered species requires an understanding of the biology of the
organism and the processes that shape its environment (Caughley and Gunn 1996). For many years
little progress was made towards understanding M. vancouverensis, largely because of a lack of
carefully framed and tested hypotheses. Much speculation was therefore published about the
supposed effects of ski-hill development, all-terrain vehicles, people or dogs (e.g., Dearden and Hall
1983). Conjectures of natural population regulation from weather or natural processes of vegetation
change were accepted with little or no supporting evidence (Milko 1984). Severe population crashes
were sometimes discounted because they were presumed to be something that “just happened”.
Conversely, the relationship between forestry and marmots remained unexplored because population
counts suggested that marmots were increasing in clearcut habitats. The result of these impressions
was to convince many people that either no problem existed, that the problem was somehow
unsolvable, or that the problem could be solved by creation of small protected areas (e.g., the 127 ha
Haley Lake Ecological Reserve).
In their assessment of wildlife reintroductions, Griffiths et al. (1989) suggested that the
possibility of success is highest when animals are taken from donor populations that are expanding.
However they ruefully noted that such conditions “are the ones that tend to make endangered species
biologists relax”. This is an apt description of much of the recent human history involving
Vancouver Island marmots. The tragedy is that populations declined precipitously while the causes
of decline remained untested and mis-understood. While I remain convinced that full recovery of this
species is ecologically feasible, recovery will be a more lengthy and expensive process as a result.
94
LITERATURE CITED
Ageev, V.S., and S.B. Pole. 1996. Fleas of the marmots in the plague enzootic areas of Tien Shan
and Pamiro-Alai mountains. Pages 89-94 in M. LeBerre, R. Ramousse and L. Le Guelte
(Editors): Biodiversity in Marmots. International Marmot Network (Moscow-Lyon). 276
pp.
Akcakaya, H.R. and S. Ferson. 1990. RAMAS/age. Applied Biomathematics (Setauket, NY).
Anderson, R.M., and R.M. May. 1979. Population biology of infectious diseases. Part I.
Nature. 280: 361-367.
Armitage, K.B. and J.F. Downhower. 1974. Demography of yellow-bellied marmot populations.
Ecology. 55: 1233-1245.
Armitage, K.B. 1991. Social and population dynamics of yellow-bellied marmots: results from
long-term research. Annual Review of Ecology and Systematics. 22: 379-407.
Armitage, K.B. 1992. Preface. Pages III-IV in B. Bassano, P. Durio, U. Gallo Orsi and E.
Macchi (Editors): Proceedings, 1st International Congress on Alpine Marmot and Genus
Marmota. Dipartimento di Produzioni Animali, Epidemiolgia ed Ecologia (Torino, Italy).
268 pp.
Armitage, K.B. 1994. Unusual mortality in a yellow-bellied marmot population. Pages 5-13 in
V.Yu. Rumiantsev (Editor): Actual problems of marmots investigation. Russian Academy of
Science (Moscow, Russia). 256 pp.
Arnold, W. 1990a. The evolution of marmot sociality: I. Why disperse late? Behavioural Ecology
and Sociobiology. 27: 229-237.
Arnold, W. 1990b. The evolution of marmot sociality: II. Costs and benefits of joint hibernation.
Behavioural Ecology and Sociobiology. 27: 239-246.
Arnold, W. 1992. Social evolution and obligatory group hibernation in marmots. in B. Bassano,
P. Durio, U. Gallo Orsi and E. Macchi (Editors): Proceedings, 1st International Symposium
on Alpine Marmot and Genus Marmota. Dipartimento di Produzioni Animali, Epidemiologia
et Ecologia (Torino, Italy). 268 pp.
Arnold, W. 1993. Energetics of social hibernation. Pages 65-80 in C. Carey, G. L. Florant, B.
A. Wunder and B. Horwitz (Editors): Life in the cold: ecological, physiological, and
molecular mechanisms. Westview Press (Boulder, CO). 214 pp.
Arnold, W., G. Heldmaier, S. Ortmann, H. Pohl, T. Ruf and S. Steinlechner. 1991. Ambient
temperatures in hibernacula and their energetic consequences for alpine marmots. Journal of
Thermal Biology. 16: 223-226.
Banfield, A.W.F. 1977. The Mammals of Canada (Revised edition). University of Toronto Press
(Toronto, ON). 438 pp.
Barash, D.P. 1973. The social biology of the Olympic marmot. Animal Behaviour Monographs.
6: 171-245.
95
Barash, D.P. 1989. Marmots: social behavior and ecology. Stanford University Press (Stanford,
CA). 360 pp.
Begon, M., and M. Mortimer. 1986. Population Ecology: a unified study of animals and plants.
Second edition. Blackwell Scientific Publications (Oxford, England). 220 pp.
Beier, P. 1995. Dispersal of juvenile cougars in fragmented habitat. Journal of Wildlife
Management. 59: 228-237.
Bergerud, A.T. 1983. Prey switching in a simple ecosystem. Scientific American. 249: 130-141.
Bibikov, D.I. 1992. Marmots are zoonosis provoking carriers. Pages 25-29 in B. Bassano,
P. Durio, U. Gallo Orsi and E. Macchi (Editors): Proceedings, 1st International Symposium
on Alpine Marmot and Genus Marmota. Dipartimento di Produzioni Animali, Epidemiologia
et Ecologia (Torino, Italy). 268 pp.
Blake, J.E., B.D. McLean, and A. Gunn. 1991. Yersiniosis in free-ranging muskoxen on Banks
Island, Northwest Territories, Canada. Journal of Wildlife Diseases. 27: 527-533.
Blumstein, D.T., and J.M. Foggin. 1997. Effects of vegetative variation on weaning success,
overwinter survival and social group density in golden marmots (Marmota caudata aurea).
Journal of Zoology (London). 243: 57-69.
Breininger, D.R., V.L. Larson, B.W. Duncan, R.B. Smith, D.M. Oddy and M.F. Goodchild.
1995. Landscape patterns of Florida scrub jay habitat use and demographic success.
Conservation Biology. 9: 1442-1453.
Brown, J.H., and A. Kodric-Brown. 1977. Turnover rates in insular biogeography: effect of
immigration on extinction. Ecology. 58: 445-449.
Brown, R.J.E. 1970. Permafrost in Canada. University of Toronto Press (Toronto, ON).
234 pp.
Bryant, A.A. 1990. Genetic variability and minimum viable populations in the Vancouver Island
marmot (Marmota vancouverensis). M.E.Des. Thesis, University of Calgary (Calgary, AB).
101 pp.
Bryant, A.A. 1996. Reproduction and persistence of Vancouver Island marmots (Marmota
vancouverensis) in natural and logged habitats. Canadian Journal of Zoology. 74: 678-687.
Bryant, A.A. 1997. Updated Status of the Vancouver Island marmot (Marmota vancouverensis)
in Canada. Committee on the Status of Endangered Wildlife in Canada (Ottawa, ON). 21
pp.
Bryant, A.A., and D.W. Janz. 1996. Distribution and abundance of Vancouver Island Marmots
(Marmota vancouverensis). Canadian Journal of Zoology. 74: 667-677.
Bryant, A.A., H.M. Schwantje and N.I. deWith. In press. Disease and unsuccessful
reintroduction of Vancouver Island marmots (Marmota vancouverensis). Proceedings, 3rd
International Congress on Genus Marmota (Cheboksary, Russia, 25-30 August 1997).
96
Callait, M.P., D. Gauthier, C. Prud’homme and B. Sabatier. 1996. Impact of the parasitic fauna
of digestive tract of alpine marmots (Marmota marmota) on their population dynamic. Pages
97-104 in M. LeBerre, R. Ramousse and L. Le Guelte (Editors): Biodiversity in Marmots.
International Marmot Network (Moscow-Lyon). 276 pp.
Campbell, R.W., N.K. Dawe, I. McTaggart-Cowan, J.M. Cooper, G.W. Kaiser and M.C.E.
McNall. 1990a. The Birds of British Columbia. Volume 1. Royal B.C. Museum (Victoria,
BC). 514 pp.
Campbell, R.W., N.K. Dawe, I. McTaggart-Cowan, J.M. Cooper, G.W. Kaiser and M.C.E.
McNall. 1990b. The Birds of British Columbia. Volume 2. Royal B.C. Museum (Victoria,
BC). 636 pp.
Carbyn, L.N. 1987. Gray wolf and red wolf. Pages 359-376 in Novak, M., J.A. Baker, M.E.
Obbard and B. Malloch (Editors): Wild Furbearer Management and Conservation in North
America. Ontario Ministry of Natural Resources (Toronto, ON). 1150 pp.
Carl, G.C. 1943. The natural history of the Forbidden Plateau area, Vancouver Island, British
Columbia. Pages 18-40 in Report of the B.C. Provincial Museum for 1943 (Victoria, BC).
78 pp.
Caswell, H. 1989. Matrix population models. Sinauer Associates (Sunderland, MA). 328 pp.
Caughley, G. 1977. Analysis of vertebrate populations. 1977. John Wiley and Sons, (New
York, NY). 234 pp.
Caughley, G. 1994. Directions in conservation biology. Journal of Animal Ecology. 63: 215-
244.
Caughley, G., and A. Gunn. 1996. Conservation Biology in Theory and Practice. Blackwell
(Cambridge, MA). 459 pp.
Clark, T.W. 1997. Prairie Night. Smithsonian Institution Press (Washington, DC). 348 pp.
Cleveland, W.S. 1979. Robust locally weight regression and smoothing of scatterplots. Journal
of the American Statistical Association. 74: 829-836.
Concannon, P.W., L.A. Fullam, B.H. Baldwin and B.C. Tennant. 1989. Effects of induction
versus prevention in captive male and female woodchucks (Marmota monax). Biology of
Reproduction. 41: 255-261.
Cooch, E.G, R. Pradel and N. Nur. 1996. Practical Guide to Mark-Recapture Analysis using
SURGE. Centre d’Ecologie Fonctionnelle et Evolutive (Montpellier, France).
Cortot, H., M. Fancou, D. Juan, L. Tron, M. Le Berre and R. Ramousse. 1996. Development of
a census method for alpine marmots in the Ecrins National Park. Pages 23-28 in M. LeBerre,
R. Ramousse and L. Le Guelte (Editors): Biodiversity in Marmots. International Marmot
Network (Moscow-Lyon). 276 pp.
Cox, D.R. 1975. Partial likelihood. Biometrika. 62: 269-276.
97
Dearden, P., and C. Hall 1983. Non-consumptive recreation pressures and the case of the
Vancouver Island marmot (Marmota vancouverensis). Environmental Conservation. 10:
63-66.
Demarchi, D.A. 1988. Ecoregions of British Columbia. B.C. Ministry of Environment and Parks
(Victoria, BC). Map.
Dimitriev, A.V., M.N. Leontieva, I.B. Abrakhina, M.A. Barmin, A.Yu. Isaev, V.A. Kuvshinov,
V.G. Marphin, Z.N. Plechova and R.I. Siyan. 1994. Modern state and prospects of
reacclimatization of the bobac (Marmota bobac) in the Volga region. Pages 27-44 in V.Yu.
Rumiantsev (Editor): Actual problems of marmots investigation. Russian Academy of
Science (Moscow, Russia). 256 pp.
Dobson, A.P., and R.M. May. 1986. Disease and conservation. Pages 345-365 in M.E. Soulé
(Editor): Conservation Biology: the Science of Scarcity and Diversity. Sinauer Associates
(Sunderland, MA). 584 pp.
Donovan, T.M., F.R. Thompson, J. Faaborg and J.R. Probst. 1995. Reproductive success in
habitat sources and sinks. Conservation Biology. 9: 1380-1395.
Doyle, A.C. 1890. The Sign of the Four. Reprinted in: The Complete Sherlock Holmes (1973).
Doubleday (Garden City, NY). 186 pp.
Dunning, J.B., B.J. Danielson and H.R. Pulliam. 1992. Ecological processes that affect
populations in complex landscapes. Oikos. 65: 169-175.
Fahrig, L. 1997. Relative effects of habitat loss and fragmentation on population extinction.
Journal of Wildlife Management. 61: 603-610.
Fahrig, L., and Merriam, G. 1985. Habitat patch connectivity and population survival. Ecology.
66: 1762-1768.
Farand, E., D. Allaine and J. Coulon. In prep. Survival of the Alpine marmot (Marmota
marmota): a capture-recapture study of the population from La Saussiere, France.
Contributed paper, “Holarctic Marmots as a Factor of Biodiversity”, 3rd International
Congress on Genus Marmota, Cheboksary, Russia, 25-30 August 1997.
Fitzgerald, B.M. 1977. Weasel predation on a cyclic population of the montane vole (Microtus
montanus) in California. Journal of Animal Ecology. 46: 367-397.
Flowerdew, J.R. 1987. Mammals: their reproductive biology and population ecology.
Cambridge University Press (Cambridge, England). 241 pp.
Fonda, R.W., and L.C. Bliss. 1969. Forest and vegetation of the montane and subalpine zones,
Olympic Mountains. Ecological Monographs. 39: 371-301.
Franklin, J.F., W.H. Moir, G.W. Douglas and C. Wiberg. 1971. Invasion of subalpine meadows
by trees in the Cascade Range, Washington and Oregon. Arctic and Alpine Research. 3:
215-224.
Gaston, K.J. 1994. Rarity. Chapman and Hall (London, England). 205 pp.
98
Geraci, J.R., D.J. St. Aubin, I.K. Barker, R.G. Webster, V.S. Hinshaw, W.J. Bean, H.L. Ruhnke,
J.H. Prescott, G. Early, A.S. Baker, S. Madoff and R.T. Schooley. 1982. Mass mortality of
harbour seals: pneumonia associated with influenza B virus. Science. 215: 1129-1131.
Gilpin, M.E., and M.E. Soulé. 1986. Minimum viable populations: processes of species
extinction. Pages 19-34 in M.E. Soulé (Editor): Conservation Biology: the Science of
Scarcity and Diversity. Sinauer Associates (Sunderland, MA). 584 pp.
Gilpin, M.E. 1987. Spatial structure and population vulnerability. Pages 125-139 in M.E. Soulé
(Editor): Viable Populations for Conservation. Cambridge University Press (Cambridge,
England). 189 pp.
Gortazar, C.J., J. Herrero, A.V. Bezverkhni, E.S. Kopbaev and V.A. Mosko. 1996. Preliminary
data on the parasitic fauna of the digestive system of Marmota marmota in the western
Pyrenees. Pages 105-108 in M. LeBerre, R. Ramousse and L. Le Guelte (Editors):
Biodiversity in Marmots. International Marmot Network (Moscow-Lyon). 276 pp.
Grayson, D.K. 1987. The biogeographic history of small mammals in the Great Basin;
observations on the last 20,000 years. Journal of Mammalogy. 68: 359-375.
Gregory, P.T. 1997. What good is Ro in demographic analysis (and what is a generation
anyway?) Wildlife Society Bulletin. 25: 709-713.
Griffith, B., J. M. Scott, J. W. Carpenter and C. Reed. 1989. Translocation as a species
conservation tool: status and strategy. Science. 245: 477-480.
Groombridge, B., and G. Mace. 1994. The 1994 list of threatened animals. International Union
for the Conservation of Nature (Gland, Switzerland). 286 pp.
Haber, G.C. 1977. Socio-ecological dynamics of wolves and prey in a subarctic ecosystem. PhD
Dissertation, University of British Columbia (Vancouver, BC). 786 pp.
Haining, R. P. 1990. Spatial Data Analysis in the Social and Environmental Sciences.
Cambridge University Press (Cambridge, England). 312 pp.
Hanski, I. 1991. Single-species metapopulation dynamics: concepts, models and observations.
Pages 17-38 in M. Gilpin and I. Hanski (Editors): Metapopulation dynamics: empirical and
theoretical investigations. Academic Press (London, England). 336 pp.
Hanski, I., and M.E. Gilpin. 1991. Metapopulation dynamics: brief history and conceptual
domain. Pages 3-16 in M. Gilpin and I. Hanski (Editors): Metapopulation dynamics:
empirical and theoretical investigations. Academic Press (London, England). 336 pp.
Harestad, A.S., and G.W. Jones. 1981. Use of night counts for censusing black-tailed deer on
Vancouver Island. Pages 83-96 in F.L Miller, A. Gunn and S.R. Hieb (Editors): Symposium
on census and inventory methods for populations and habitats. University of Idaho
(Moscow, ID). 220 pp.
Heard, D.C. 1977. The behavior of Vancouver Island marmots (Marmota vancouverensis).
M.Sc. Thesis, University of British Columbia (Vancouver, BC) 129 pp.
99
Hensher, D., and L.W. Johnson. 1981. Applied Discrete Choice Modelling. Croom Helm
(London, England). 227 pp.
Herrero, J., R. Garc’a-Gonzalez, and A. Garc’a-Serrano. 1994. Altitudinal distribution of alpine
marmot (Marmota marmota) in the Pyrenees, Spain/France. Arctic and Alpine Research.
26: 328-331.
Hoffmann, R.S., J.W. Koeppl and C.F. Nadler. 1979. The relationships of the amphiberingian
marmots (Mammalia: sciuridae). Occasional Papers, Museum of Natural History.
University of Kansas (Lawrence, KA). 83:1-56.
Hosmer, D.W., and S. Lemeshow. 1989. Applied Logistic Regression. Wiley and Sons (New
York, NY). 307 pp.
Hyntka, J. 1990. Strathcona Park Master Plan Background Report. Unpublished report to B.C.
Parks (Victoria, BC). 141 pp.
Janz, D., C. Blumensaat, N.K. Dawe, B. Harper, S. Leigh-Spencer, W.T. Munro and D.
Nagorsen. 1994. National Recovery Plan for the Vancouver Island Marmot. Report No. 10.
Recovery of Nationally Endangered Wildlife Committee (Ottawa, ON). 32 pp.
Janz, D.W., A.A. Bryant, N.K. Dawe, H. Schwantje, B. Harper, D. Nagorsen, D. Doyle, M.
deLaronde, D. Fraser, D. Lindsay, S. Leigh-Spencer, R. McLaughlin and R. Simmons. In
prep. Revised Recovery Plan for the Vancouver Island Marmot (1998). RENEW Recovery
of Nationally Endangered Wildlife Committee (Ottawa, ON).
Krebs, C.J. 1989. Ecological methodology. HarperCollins (New York, NY). 654 pp.
Kuramoto, R.T., and L.C. Bliss. 1970. Ecology of subalpine meadows in the Olympic
Mountains, Washington. Ecological Monographs. 40: 317-347.
Lacy, R.C., and D.B. Lindenmayer. 1995. A simulation study of the impacts of population
subdivion on the mountain brushtail possum in south-eastern Australia. II. Loss of genetic
variation within and between subpopulations. Biological Conservation. 73: 131-142.
Lamberson, R.H., R. McKelvey, B. Noon and C. Voss. 1992. A dynamic analysis of northern
spotted owl viability in a fragmented forest landscape. Conservation Biology. 4: 505-512.
Laroque, C.P. 1998. Tree invasion in subalpine Vancouver Island marmot meadows.
Unpublished report to the B.C. Environmental research Scholarship Committee (Victoria,
BC) 36 pp.
Laurance, W.F., K.R. McDonald and R. Speare. 1996. Epidemic disease and the catastophic
decline of Australian rain forest frogs. Conservation Biology. 10: 406-413.
Lebreton, J-D., K.P. Burnham , J. Clobert and D.R. Anderson. 1992. Modeling survival and
testing biological hypotheses using marked animals: a unified approach with case studies.
Ecological Monographs. 62: 67-118.
100
Lertzman, K, D. Gavin, D. Hallett, L. Brubaker, D. Lepfosky, and R. Mathews. 1998. Long-term
fire histories and the dynamics of wet coastal forests. Pages 93-94 in J. A. Trofymow and A.
MacKinnon (Editors): Structure, processess and diversity in successional forests of coastal
British Columbia. Canadian Forest Service (Victoria, BC). 151 pp.
Lidicker, W.Z., and W.D. Koenig. 1996. Responses of terrestrial vertebrates to habitat edges and
corridors. Pages 85-109 in D.R. McCullough (Editor): Metapopulations and Wildlife
Conservation. Island Press (Washington, DC). 429 pp.
Mace, A.E. 1964. Sample Size Determination. Rheinhold (New York, NY).
Mantel, N. 1967. The detection of disease clustering and a generalized regression approach.
Cancer Research. 27: 209-220.
Martell, A.M. and R.J. Milko. 1986. Seasonal diets of Vancouver Island marmots. Canadian
Field-Naturalist. 100:241-245.
Matthysen, E., F. Adriaensen and A.A. Dhondt. 1995. Dispersal distances of nuthatches, Sitta
europaea, in a highly fragmented forest habitat. Oikos. 72: 375-381.
May, R.M., and R.M. Anderson. 1979. Population biology of infectious diseases. Part II.
Nature. 280: 455-461.
McEvedy, C. 1988. La peste bubonique. Pour la Science. 126: 72-77.
Merilees, W.J. 1980. Vancouver Island fauna best left alone. B.C. Naturalist. 18: 19-20.
Milko, R.J. 1984. Vegetation and foraging ecology of the Vancouver Island marmot (Marmota
vancouverensis). M.Sc. Thesis, University of Victoria (Victoria, BC). 127 pp.
Milko, R.J., and A.M. Bell. 1985. Subalpine meadow vegetation of south central Vancouver
Island. Canadian Journal of Botany. 64: 815-821.
Munro, W.T. 1979. Status of the Vancouver Island marmot in Canada. Committee on the Status
of Endangered Wildlife in Canada (Ottawa, ON). 8 pp.
Munro, W.T., D.W. Janz, V. Heinsalu and G.W. Smith. 1985. The Vancouver Island Marmot:
status and management plan. B.C. Ministry of Environment Wildlife Bulletin B-39.
(Victoria, BC). 24 pp.
Nagorsen, D.W. 1987. Marmota vancouverensis. Mammalian Species. 270 :1-5.
Nagorsen, D.W, G. Keddie and T. Luszcz. 1996. Vancouver Island marmot bones from subalpine
caves: archaeological and biological significance. Occasional Paper #4. B.C. Ministry of
Environment, Lands and Parks (Victoria, BC). 58 pp.
Pielou, E.C. 1991. After the Ice Age: the return of life to glaciated North America. Chicago
University Press (Chicago, IL). 366 pp.
Pimm, S. 1991. The Balance of Nature: ecological issues in the conservation of species and
communities. Chicago University Press (Chicago, IL). 434 pp.
Platt, J.R. 1964. Strong inference. Science. 347-353.
101
Pollock, K.H, J.D. Nichols, C. Brownie and J.E. Hines. 1990. Statistical inference for capture-
recapture experiments. Wildlife Monographs #107. 97 pp.
Popper, K.R. 1968. The Logic of Scientific Discovery. Second edition. Hutchinson (London).
127 pp.
Preleuthner, M., W. Pinsker, L. Kruckenhauser, W.J. Miller and H. Prosl. 1995. Alpine marmots
in Austria: the present population structure as a result of the postglacial distribution history.
Acta Theriologica (Supplement). 3: 87-100.
Preleuthner, M., H. Prosl, A. Bergmann and W. Pinsker. 1996. Infestation by endoparasites in
different Austrian populations of M. marmota. Pages 111-112 in M. LeBerre, R. Ramousse
and L. Le Guelte (Editors): Biodiversity in Marmots. International Marmot Network
(Moscow-Lyon). 276 pp.
Pulliam, H.R. 1988. Sources, sinks and population regulation. American Naturalist. 132: 652-
661.
Pyke, D.A., and J. N. Thompson. 1986. Statistical analysis of survival and removal rate
experiments. Ecology. 67: 240-245.
Ramousse, R., J.P. Martinet and M. Le Berre. 1992. Twenty years of re-introduction policy of
alpine marmots from the national park of La Vanoise (French Alps). Pages 171-177 in B.
Bassano, P. Durio, U. Gallo Orsi and E. Macchi (Editors): Proceedings, 1st International
Symposium on Alpine Marmot and Genus Marmota. Dipartimento di Produzioni Animali,
Epidemiologia et Ecologia (Torino, Italy). 268 pp.
Rausch, R.L., and V.R. Rausch. 1971. The somatic chromosomes of some North American
marmots. Extrait de Mammalia. 35: 85-101.
Rochefort, R.M., R.L. Little, A. Woodward, and D.L. Peterson. 1994. Changes in the tree-line
distribution in western North America: a review of climatic and other factors. The Holocene.
4: 89-100.
Rossi, R. E., D. J. Mulla, A. G. Journel and E. H. Franz. 1992. Geostatistical tools for modelling
and interpreting ecological spatial dependence. Ecological Monographs. 62: 277-314.
Rymalov, I.V. 1996. Hibernation and breeding of Marmota bobac. Pages 203-208 in M.
LeBerre, R. Ramousse and L. Le Guelte (Editors): Biodiversity in Marmots. International
Marmot Network (Moscow-Lyon). 276 pp.
Sawada, M. 1998. Rook's Case Version 0.9. Laboratory of Paleoclimatology and Climatology,
University of Ottawa (Ottawa, ON).
Schreiner, E.G., and J.E. Burger. 1994. Photographic comparisons: a qualitative appraisal of the
influence of climate and disturbances on vegetation, 1915-1990. Pages 139-172 in D.B.
Houston, E.G. Schreiner and B.B. Moorhead (Editors): Mountain Goats in Olympic National
Park: biology and management of an introduced species. United States Department of the
Interior Scientific Monograph NPS/NROLYM/NRSM-94/25.
102
Schwartz, O.A., K.B. Armitage and D. Van Vuren. 1998. A 32-year demography of yellow-
bellied marmots (Marmota flaviventris). Journal of Zoology (London). 246: 337-346.
Simberloff, D., J.A. Farr, J. Cox and D.W. Mehlman. 1992. Movement corridors: conservation
bargains or poor investments? Conservation Biology. 6: 493-504.
Sinclair, A.R.E., C.J. Krebs and J.N.M. Smith. 1982. Diet quality and food limitation in
herbivores: the case of the snowshoe hare. Canadian Journal of Zoology. 60: 889-897.
Smith, G.W. 1982. Vancouver Island marmots: present status - 1981. Unpublished report to the
B.C. Fish and Wildlife Branch (Nanaimo, BC). 19 pp.
Smith, A. T., and M. E. Gilpin. 1997. Spatially correlated dynamics in a pika population. Pages
407-428 in I.A. Hanski and M.E. Gilpin (Editors): Metapopulation Biology: ecology,
genetics and evolution. Academic Press (New York, NY).
Sokal, R. R., and N. L. Oden. 1978. Spatial autocorrelation in biology 1: Methodology.
Biological Journal of the Linnean Society. 10: 199-228.
SPSS Ltd. 1997. Systat 7.0. SPSS Ltd., (Chicago, IL).
Summers, J., J.M. Smolec and R. Snyder. 1978. A virus similar to human hepatitus B virus
associated with hapatitus and hepatoma in woodchucks. Proceedings of the National
Academy of Sciences. 75: 4533-4537.
Swarth, H.S. 1911. Two new species of marmots from Northwestern America. University of
California Publications in Zoology. 7 :201-204.
Swarth, H.S. 1912. Report on a collection of birds and mammals from Vancouver Island.
University of California Publications in Zoology. 10: 1-124.
Sweitzer, R.A., S.H. Jenkins and J. Berger. 1997. Near-extinction of porcupines by mountain
lions and consequences of ecosystem change in the Great Basin desert. Conservation
Biology. 11: 1407-1417.
Taylor, P.D., L. Fahrig, K. Henein and G. Merriam. 1993. Connectivity is a vital elemement of
landscape structure. Oikos. 68: 571-573.
Taylor, P.D., and G. Merriam. 1995. Wing morphology of a forest damselfly is related to
landscape structure. Oikos. 73: 43-48.
Thomas, C.D. 1994. Extinction, colonization and metapopulations: environmental tracking by
rare species. Conservation Biology. 8: 373-378.
Thorp, C.R., P.K. Ram and G.L. Florant. 1994. Diet alters metabolic rate in the Yellow-bellied
marmot (Marmota flaviventris) during hibernation. Physiological Zoology. 67: 1213-1229.
Tokarsky, V.A. 1996. Biology, behaviour and breeding black-capped marmots in captivity.
Pages 257-260 in M. LeBerre, R. Ramousse and L. Le Guelte (Editors): Biodiversity in
Marmots. International Marmot Network (Moscow-Lyon). 276 pp.
103
Van Vuren, D. 1989. Effects of intraperitoneal transmitter implants on yellow-bellied marmots.
Journal of Wildlife Management. 53: 320-323.
Van Vuren, D., and K. B. Armitage. 1991. Duration of snow cover and its influence on life-
history variation in yellow-bellied marmots. Canadian Journal of Zoology. 69: 1755-1758.
Vitousek, P.M. 1988. Diversity and biological invasions of oceanic islands. Pages 181-189 in
E.O. Wilson (Editor): Biodiversity. National Academy Press (Washington, DC). 521 pp.
Walker, D.A., J.C. Halfpenny, M.D. Walker and C.A. Wessman. 1993. Long-term studies of
snow-vegetation interactions. Bioscience. 43: 287-301.
Wauters, L., and A.A. Dhondt. 1990. Red squirrel population dynamics in different habitats.
Zeitschrift Saugetierkunde. 55: 161-175.
Weddell, B.J. 1991. Distribution and movements of Columbian ground squirrels (Spermophilus
columbianus): are habitat patches like islands? Journal of Biogeography. 18: 385-394.
White, G.C., and R.A. Garrott. 1990. Analysis of Wildlife Radio-tracking Data. Academic
Press (San Diego, CA). 620 pp.
Woodbury, M.R. (Editor). 1997. The Chemical Immobilization of Wildlife. Canadian
Association of Zoo and Wildlife Veterinarians (Winnipeg, MB).
Zar, J.H. 1974. Biostatistical analysis. Prentice-Hall (Englewood Cliffs, NJ). 620 pp.
Zwart, P. 1993. Yersiniosis in nondomestic birds and mammals. Pages 52-53 in Fowler, M.E
(Editor): Zoo and Wild Animal Medicine. W.B. Saunders (Toronto, ON). 286 pp.
Zykov, A.E., and O.V. Dudkin. 1996. Steppe marmot (Marmota bobac) as a tuleremia carrier in
Ukraine. English abstract. Page 107 in Marmots of Northern Eurasia: the Biodiversity
Saving. Moscow State University (Moscow, Russia). 111 pp.
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
