CLIMATIC AND HABITAT DRIVERS OF AMERICAN PIKA (OCHOTONA PRINCEPS) OCCUPANCY AND POPULATION DENSITY DYNAMICS IN THE SOUTHERN ROCKY MOUNTAIN REGION by LIESL PETERSON ERB B.A., Colorado College, 2004
A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirement for the degree of Doctor of Philosophy Department of Ecology and Evolutionary Biology 2013
This thesis entitled:
Climatic and habitat drivers of American pika (Ochotona princeps) occupancy and population density dynamics in the Southern Rocky Mountain Region
written by Liesl Peterson Erb
has been approved for the Department of Ecology and Evolutionary Biology
Dr. Robert Guralnick
Dr. Chris Ray
Dr. Christy McCain
Dr. Daniel Doak
Date
The final copy of this thesis has been examined by the signatories, and we find that both the content and the form meet acceptable presentation standards of scholarly work in the above mentioned discipline.
iii
Erb, Liesl Peterson (Ph.D., Ecology and Evolutionary Biology)
Climatic and habitat drivers of American pika (Ochotona princeps) occupancy and population density dynamics in the Southern Rocky Mountain Region
Thesis directed by Associate Professor Robert P. Guralnick and Research Associate Chris Ray
ABSTRACT Climate change is affecting ecosystems worldwide. Among those ecological communities most affected are those inhabiting alpine habitats. These communities have evolved key adaptations to thrive in cold, wet environments. As temperatures warm and precipitation patterns become more variable due to global climate change, many alpine species are expected to be impacted. This dissertation research focuses on the American pika, a small lagomorph inhabiting broken rock slopes in the mountains and high plateaus of western North America. Population declines in the Great Basin region at the end of the 20th Century caused concern for populations elsewhere in the species range. The goal of this dissertation work was to document pika occupancy and density throughout the Southern Rocky Mountain region. Occupancy and density trends were modeled using potential climate- and habitat-based predictors known to impact pikas elsewhere in the species’ range. Survey sites were selected from among hundreds of locations known to be occupied by pikas prior to 1980. In 2008, modeling of the resurvey results from 69 of these sites indicated that mean annual precipitation plays an important role in maintaining pika populations in this region. Further surveys of 19 of these sites in 2009-2011 showed a shift toward mean summer temperature and forage quality as the top predictors of occupancy, though sites lacking pikas also remained drier than those with pikas throughout this survey period. Pika occupancy in this region was relatively high, with Southern Rockies occupancy rates ranging from 74% to 94%. Among the extant populations, variability in population densities were best explained by patch area and vegetation quality: the highest density populations were reported in regions with small patches of talus, high forb diversity, and low graminoid to forb ratios. These results suggest that intraspecific competition for food resources strongly influences pika density. iv
As climate change continues, vegetation quality is expected to decline in pika habitats. Given this species’ reliance on cool, wet climates with high forb content, continued changes toward drier, hotter, and more graminoid-dominant habitats are likely to lead to declines in both pika densities and occupancy throughout the Southern Rockies and the western United States.
v
DEDICATION
This dissertation is dedicated to my grandmother, Jean “Tommy” Peterson.
.
vi
ACKNOWLEDGEMENTS
This dissertation would not have been possible without the wise and patient guidance of
my co-advisors, Robert Guralnick and Chris Ray. I am also grateful for the guidance and advice
of my dissertation committee, Christy McCain, Andrew Martin, and Dan Doak. I would like to thank Erik Beever, Andrew Smith, and other anonymous reviewers for their constructive comments on the manuscripts that compose this document.
The research presented here would not have been possible without the hard work and
positive attitudes of my amazing teams of undergraduate field assistants, including Justine Smith,
Lizzy Studer, Gavin Dean, Nate Kleist, Aaron Stecker, and Kira Powell, whose tireless efforts
and attention to detail were invaluable. Many thanks also to my collaborators at Colorado Parks and Wildlife, including Amy Seglund and the many dedicated research biologists in offices throughout the state. Thank you also to the staff of the parks, reserves, and wild places in which I conducted research, including Jennie Reithel, Bob Parmenter, Craig Allen, Judy Visty, Jeff
Connor, Ralph Swain, Melanie Woolever, Wyoma Hansen, Missy Dressen, and many other hard working employees at public agencies across the Southern Rockies. I would also like to the financial supporters of this work: the CU Ecology and Evolutionary Biology Department and
Museum of Natural History, Audubon Society of Greater Denver, Colorado Mountain Club,
Mountain Studies Institute, National Geographic Society, and National Science Foundation.
Outside of the research realm, many members of the CU-Boulder community made
positive contributions to my educational experience in graduate school. These include: David
Armstrong, Carol Wessman, Laura Border, Michael Breed, Bob Hermanson, John Basey,
Alexander Cruz, Jeffry Mitton, Barbara Demmig-Adams, William Adams, Mindy Sclaro, Jill
Skarstad, Julie Graf, Tammy Maldonado, and Kristin Swihart. In addition, my graduate student vii
peers and their families were instrumental in my happiness and success in this venture. Some of the many key members of this support group are: Leigh Cooper, Sara Hellmuth Paull, Kelly
Ramirez, Kallin Tea, Sarah Wagner, Brian, Audrey, and Cole Buma, Eve Gasarch, Loren
Sackett, SeJin Song, Monica Madronich, Joanna Hubbard, Clint Francis, Marcus Cohen, Mari
Elise Ewing, Preston Cumming, Natalie Robinson, Sarah Orlofske, Robert Jadin, Stower Beals,
Abbey Paulson, Samantha Weintraub, Ashwin Ravikumar, Mike and Heather Robeson, and
Sierra Love Stowell.
Perhaps even more admirable was the support of friends and family outside of academia, where it can at times be hard to understand what is happening inside the ivory tower. I would like to thank Amy Belcastro, Jeremy Sueltenfuss, Autumn Rivera, Kayly and Matt Newland, and
Jaemey and Bill Bush for their friendship, love, and support. My heartfelt thanks also go to my biggest cheerleaders: my family. A big thanks to the Ambrose Aunties for their love and hugs and the Peterson clan for the support and encouragement. Thank you also to my “extra” families, who I love so dearly: Pam Peterson and the Allenger and Allen crews for letting me join in your fun and loving family; and Bob and Marilyn Erb, leaders of the amazing Erb clan, for teaching me so much about life. While I’ve already named a long cheerleading squad, my head cheerleaders have always been Phil and Polly Peterson, who taught me to love nature, hold myself to high standards, and always fight for what is right.
The most credit for my success in this venture goes to my rock, my best friend, and my
husband: Peter Erb. His love and encouragement have carried me through my darkest days and
lifted me to be the best version of myself. And last but not least, I would be remiss in not
thanking the “person” who got me out of bed each morning and kept my priorities straight:
Bridger Erb. viii
CONTENTS
CHAPTER
1 On the generality of a climate-mediated shift in the distribution of the American pika
(Ochotona princeps)
Abstract……………………………………………………………………………………1
Introduction………………………………………………………………………………..1
Methods……………………………………………………………………………………3
Results……………………………………………………………………………………..8
Discussion………………………………………………………………………………..13
2 Interactive effects of climate and vegetation on multi-year occupancy of the American
pika (Ochotona princeps)
Abstract…………………………………………………………………………………..18
Introduction………………………………………………………………………………19
Methods…………………………………………………………………………………..21
Results……………………………………………………………………………………27
Discussion………………………………………………………………………………..31
3 Determinants of pika population density versus occupancy in the Southern Rocky
Mountains
Abstract…………………………………………………………………………………..36
Introduction………………………………………………………………………………36
Methods…………………………………………………………………………………..38
Results……………………………………………………………………………………48
Discussion………………………………………………………………………………..50 ix
4 Conclusion……………………………………………………………………………….53
BIBLIOGRAPHY………………………………………………………………………………..57
APPENDICES:
I Histograms of precipitation and summer temperature at pika study sites……………….65
II Boxplot of mean pika occupancy predictor coefficients…………………………………66
III Comparison of climatic predictors at 18 pika latrine density study sites………………..67
IV Pika latrine density at 18 sites in the Southern Rocky Mountains……………………….68
V Models predicting pika density with ΔAIC<2, along with the null model………………69
x
TABLES
Chapter 1:
1. 2008 historical resurvey alternative hypotheses and candidate model covariates……7
2. Results of 2008 historical resurvey logistic regression modeling…………………...11
Chapter 2:
1. Relative support for candidate predictors of mean pika occupancy, 2008-2011…….26
2. Top models of mean pika occupancy compared with the null model……………….29
Chapter 3:
1. Predictor descriptions, expected relationships to pika population density, weighted
average coefficients, and Akaike weights……………………………………………45
xi
FIGURES
Chapter 1:
1. Map of 69 sites historically occupied by the American pika (Ochotona princeps) in the
Southern Rocky Mountains………………………………………………………………..4
2. 2008 pika persistence as observed and predicted by logistic regression modeling……...10
3. Mean annual precipitation and change in mean annual precipitation vs. occupancy in
2008………………………………………………………………………………………12
Chapter 2:
1. Map of 19 sites historically occupied by the American pika (Ochotona princeps),
differentiated by occupancy status during 2008-2011…………………………………...22
2. Pika occupancy during 2008-2011 as observed and as predicted by model averaging
based on the five top occupancy models…………………………………………………30
3. Mean summer surface temperature, annual precipitation, and gram:forb at occupied,
transient, and unoccupied pika survey sites during 2008-2011…………………………35
Chapter 3:
1. Map of 18 study sites at which Ochotona princeps population density was evaluated….39
2. Pika latrine density versus the top three density predictor variables, determined by
Akaike weights…………………………………………………………………………...49
Chapter 1
ON THE GENERALITY OF A CLIMATE-MEDIATED SHIFT IN THE DISTRIBUTION
OF THE AMERICAN PIKA (OCHOTONA PRINCEPS)
ABSTRACT
Alpine species are among those most threatened by climatic shifts due to their physiological and geographic constraints. The American pika (Ochotona princeps), a small mammal found in mountainous, rocky habitats throughout much of western North America, has experienced recent population extirpations in the Great Basin linked to climatic drivers. It remains unclear whether these patterns of climate-related loss extend to other portions of the species’ range. We investigated the distribution of the American pika and the climatic processes shaping this distribution within the Southern Rocky Mountain region. Results from a survey of 69 sites historically occupied by pikas indicate that only four populations have been extirpated within this region over the past few decades. Despite relatively few extirpations, low annual precipitation is implicated as a limiting factor for pika persistence in the Southern Rockies.
Extirpations occurred only at sites that were consistently dry over the last century. While there was no climate change signal in our results, these data provide valuable insight into the potential future effects of climate change on O. princeps throughout its range.
INTRODUCTION
Climate change is affecting alpine communities worldwide (Krajick 2004; Hughes 2003).
Empirical evidence from alpine plants in Europe (Lenoir et al. 2008) and alpine mammals in western North America (Moritz et al. 2008) clearly shows range retraction in many species as a
1
response to climate change, generally due to an upslope shift in their lower elevational limits.
However, few studies have demonstrated local extinctions at lower elevations, and fewer still
have investigated specific climatic drivers that may lead to local extinction.
One species for which population extirpations have been documented is the American pika, Ochotona princeps. The American pika is a small, herbivorous lagomorph that resides primarily in talus (rocky debris) found in mountain ranges and high plateaus of western North
America. Pikas worldwide benefit from metabolic and behavioral adaptations allowing them to survive cold winters without hibernating (Li et al. 2001, Sheafor 2003). However, because their resting body temperature is only a few degrees below lethal body temperature (Li et al. 2001), pikas are sensitive to temperature extremes. This sensitivity, coupled with high habitat specificity and low vagility in contemporary climates (Smith and Weston 1990), suggests that climate change could contribute to extirpation of pika populations.
American pikas have become a bellwether species for alpine taxa in peril (Krajick 2004), partly because they are conspicuous, charismatic denizens of alpine communities, and partly because population declines have been attributed to climatic changes (Beever et al. 2003 &
2010; Grayson 2005). Distributional shifts and population extirpations in the Great Basin and
Sierra Nevada have been linked to recent climatic trends (Beever et al. 2003 & 2010; Moritz et al. 2008) as well as climate change over the last glacial-interglacial period (Grayson 2005).
As evidenced by the recent decision not to protect this species under the Endangered
Species Act, it remains unclear exactly how climate change is affecting the American pika across its full geographic range (Crist 2010). Although pika populations have been lost from fragmented, lower elevation habitats in the Great Basin (Beever et al. 2003 & 2010), it is unknown whether local extinctions are occurring in more contiguous habitats where populations
2
are more likely to be rescued through dispersal. The Southern Rocky Mountains support some of the southern-most populations of the American pika, and represent the largest, most continuous region of habitat occupied by the species, including great heterogeneity in elevation and vegetation. This region is also climatically heterogeneous, and the intensity of climate change – experienced thus far and projected in future – varies greatly across the region (Mote et al. 2005;
Knowles et al. 2006).
Here we analyze change in the distribution of the American pika throughout the Southern
Rocky Mountains to assess the generality of a climate-mediated range shift in this species. This study also improves the spatial extent and resolution of data on the current range of this species and its climatic drivers. Though naturalists recorded pika populations in the Rocky Mountains as early as 1872, the current regional distribution of the species is unknown. Here we examined whether the distribution of pikas in the Southern Rockies has changed in the last century by documenting where pikas were found prior to 1980 and surveying a subset of these locations in
2008. We sought to explain the persistence pattern seen in this region by surveying potential covariates of extirpation, including landscape characteristics (e.g., elevation), microhabitat features (e.g., talus properties), and climatic trends. We used these covariates to construct and evaluate alternative models of pika persistence within an information theoretic framework.
METHODS
Study area and design
Our research was conducted at 69 sites historically occupied by pikas in the Southern Rocky
Mountains of southern Wyoming, Colorado, and New Mexico (41º 30’ to 35º 20’ N and -104º
54’ to -108º 17’ W; Figure 1). Historically occupied sites were defined as those with documented
3
Figure 1. Map of 69 sites historically occupied by the American pika (Ochotona princeps), differentiated by recent occupancy status in 2008 (occupied = circles, unoccupied = triangles).
4
pika presence prior to 1980, after which anthropogenic climate change became prominent in
many datasets (IPCC 2007). Nearly 800 historical records of pikas were found in the region,
gleaned from georeferenced museum specimens (from GBIF data portal; http://data.gbif.org),
individual museum records (Denver Museum of Natural History and University of Colorado
Museum of Natural History), and literature sources. Our 69 sites were selected based on
geographic accuracy; selected sites had a mean georeferenced radius location error estimate of
1.3 km and maximum error of 3 km. Sites varied in elevation from 2703 to 4340 m and dates of
historical records range from 1872 to 1979. The most common vegetation communities consisted
of alpine forbs and grasses, but communities dominated by willow (Salix spp.), conifers, or aspen
(Populus tremuloides) were also relatively common.
Climate data
Local climate data for each site were compiled for the years 1908-2007. The climatic data necessary for site-specific climate calculations were obtained from PRISM (2007), which provides these grid-based estimates at a 4 km2 resolution over the time scales in question. As
with any interpolated data, particularly in a mountainous region, PRISM estimates may not be
accurate for the exact coordinates of our historical records. However, because the precision of
our historical records averaged over 1 km in radius and individual pikas commonly disperse over
2 km (Tapper 1973) we feel PRISM data were at an appropriate scale for this analysis. In a post-
hoc analysis to account for fine-scale effects of solar gain within PRISM grid cells, we estimated
insolation at each site as sin(mean slope)*cos(mean aspect), similar to Martinuzzi et al. (2009).
Resurveys
Crews visited 69 sites in the summer of 2008 to determine current pika occupancy. Resurveys
were comprised of searches for fresh pika sign – detection of individuals by sight and sound and
5
fresh pika food stores (“haypiles”). Any one of these signs was considered evidence of current
occupancy. Due to the difficulty of determining scat age, scat was not used as evidence for
current pika occupancy. If visited early in the season (July), sites lacking fresh sign were
revisited in fall (October or November) to verify site status. Where no fresh sign was found,
exhaustive searches were conducted in all talus within the precision estimate for each
georeferenced location, to a maximum distance of 3 km in all directions from the estimated
historical coordinates. A minimum of 0.5 person-hours per hectare and was spent searching talus for pika sign at each of these extirpation sites. In addition to pika occupancy data, we collected data on suspected drivers of pika persistence at a site, including microhabitat variables.
Analysis
Using maximum likelihood estimates and an information-theoretic approach for model assessment, we compared models of pika persistence incorporating elevation, maximum summer temperature, annual precipitation, and site characteristics with potential climate-buffering effects such as rock type, talus depth, porosity of individual rocks, and evidence of persistent soil moisture beneath the talus. Talus depth was estimated visually at the deepest crevice found at a site, and rocks were defined as porous or not based on the presence of natural holes and pits in their surfaces. Visible or audible running water or pools under the talus and riparian vegetation at the base of the talus slope were all considered evidence of persistent sub-talus moisture.
The models explored in this study (Table 1) represent hypotheses derived from previous literature in other portions of the species' range, suggesting temperature and precipitation
(Beever et al. 2003 & 2010; Millar and Westfall 2010; Wilkening et al. 2011) as strong predictors of pika occupancy and persistence. Our hypotheses were also influenced by the results of Millar and Westfall (2010), Hafner (1994), and Smith (1974b), suggesting talus properties
6
Table 1. Alternative hypotheses and candidate model covariates. Each candidate model contained 1-3 covariates.
Hypothesis Covariates used in constructing candidate models
Pika persistence is related to:
• Change in mean annual precipitation (1908-1979 vs. The severity of changes in 1980-2007) temperature and precipitation • Change in maximum annual summer (June-August) since initial pika detection temperature averaged over 1908-1979 vs. 1980-2007
• Mean annual precipitation (1908-2007) Prevailing climatic • Maximum annual summer (June-August) temperature conditions averaged over 1908-2007
• Coefficient of variation in annual precipitation (1908-
Variation in climatic trends 2007) over the last century • Coefficient of variation in maximum annual summer
(June-August) temperature (1908-2007)
Elevation • Average elevation of talus habitat found at a site
• Each climatic variable above
Habitat quality (as a climatic • Talus depth buffer) • Porosity of rock substrate
• Presence of water under talus
7
provide climate-buffering effects for pikas. Elevation was included as a covariate because it
often varies with a suite of environmental variables. We did not explore effects of latitude, as we
felt this variable was redundant with climatic data used to test more specific hypotheses. Models
represented the following hypotheses: Pika persistence is related to (H1) the severity of changes
in temperature and precipitation since initial pika detection (pikas persist in locations where the
least change in climate has occurred); (H2) prevailing climatic conditions (pikas persist where
the dominant climate over the last century was relatively wet and cool); (H3) variation in
climatic trends over the last century (pikas persist where they have been exposed to the least
climatic variability); (H4) habitat quality (pikas persist in locations with the deepest talus, most
porous/insulating rock, and/or where water or ice persist under the talus); (H5) elevation (pikas
persist at higher elevation locations).
Thirty models with < 3 uncorrelated (|r| < 0.5) predictor variables were evaluated using
the program PRESENCE (Mackenzie et al. 2002), allowing us to incorporate detection probability into our models. This study and others have found the probability of detecting this species to be quite high (>0.90; Beever et al. 2010; Rodhouse et al. 2010). Given this high
detectability, we report the results of traditional logistic models, but model ranks were identical
2 2 in PRESENCE. Model fit was evaluated using Nagelkerke's max-rescaled R (RN ), which
provides a measure of the proportion of variance explained through logistic regression
(Nagelkerke 1992).
RESULTS
The results of our pika surveys indicate that local population extirpations have been relatively
few in the Southern Rockies: only four of 69 sites lacked recent sign of O. princeps in 2008
8
(Figure 1). Despite the low number of extirpations recorded, analyses of these data indicate that
2 the pattern of extirpation was not random. The model that best explained persistence (RN =
0.72) included mean precipitation (average annual precipitation from 1908 to 2007) and decline
in precipitation (the difference, in mm, in average annual precipitation for the period 1908-1979
versus 1980-2007). A similar model with additional effects of persistent moisture under the talus
2 (positively associated with pika occupancy) obtained similar support (ΔAICc = 0.94; RN = 0.68).
These two top models have been averaged for depiction in Figure 2. A full list of models with
ΔAICc <4 can be found in Table 2.
In the Southern Rockies, climatic factors, rather than landscape or most microhabitat
variables, appear to be the most influential in driving pika population extirpations. While our
four extirpation sites were all roughly South to Southwest-facing (170° to 258°), a post-hoc analysis of insolation indicated that this factor was not influential in pika extirpation. This result, combined with the fact that nearly half of our sites (n = 31 of 69) were roughly South-facing, suggests that aspect is not predictive of American pika distribution in this region. Models incorporating elevation, talus depth, and rock porosity were not supported by the persistence pattern found. Our four extirpation sites ranged from 2700 to 3400 m ( =3116 m) in elevation and from 0.5 to >1.5 m in maximum talus depth. The talus at all four extirpation sites, along with
57 additional sites, was not comprised of porous rock. Vegetation type varied among sites, but reflected the dry nature of these locations: all extirpations occurred at locations dominated by montane grasses, shrubs, evergreens, or aspens, rather than alpine or riparian vegetation.
Pika populations have been extirpated from among the driest pika habitats in this region
(Figure 3a). Climate at our 69 sites varied dramatically, with maximum summer (June-August) temperatures for the period 1908-2007 averaging 13.4-22.9°C and mean annual precipitation
9
Figure 2. Pika persistence as observed (open circles) and as predicted (line) by a model based on local mean precipitation (MeanPrecip), change in precipitation (ΔPrecip), and presence of sub- talus water (Water). 1 Observed and modObserved 0.5 0
0 10 20 30
Linear predictor: MeanPrecip+Precip+Water
10
Table 2. Results of logistic regression modeling. Only models with ΔAICc < 4 and the null model are shown. Mean Precip = Average
annual precipitation 1908-2007. ΔPrecip = Difference between mean annual precipitation for 1908-1979 and 1980-2007. Sub-Talus
Water = Presence of water under talus. Mean Max Temp = Mean of maximum June through August temperatures for 1908-2007.
2 2 Porous = Porosity of individual rocks. RN is Nagelkerke's max-rescaled R .
2 Model AICc ΔAICc Number of Parameters -2LogLikelihood Akaike Weight RN
Mean Precip, ΔPrecip 17.70 0 3 11.33 0.49 0.68
Mean Precip, ΔPrecip, Sub-Talus Water 18.64 0.94 4 10.01 0.35 0.72 11
Mean Precip 21.04 3.34 2 16.86 0.08 0.50
Mean Max Temp, Porous 21.23 3.53 3 14.86 0.08 0.57
Constant (Null) 30.61 12.91 1 30.55 0.00 -
Figure 3. (a) Mean annual precipitation (1908-2007) and (b) change in mean annual precipitation (comparing 1908-1979 and 1980-2007) vs. occupancy in 2008 for 69 sites historically occupied by pikas.
a)
1
Observed oc Observed 0.5
0
600 800 1000 1200 1400 Mean annual precipitatio 1908-2007 (mm)
b)
1
Observed oc Observed 0.5
0
-50 0 50 100 150 200 Change in mean annual p 1908-1979 vs. 1980-200
12
ranging 461-1407 mm for the same period. The mean annual precipitation across all sites
between 1908 and 2007 was 884 mm (sd = 184 mm; range = 461-1406 mm), while the mean
across extirpation sites was 593 mm (sd = 137 mm; range = 461-717 mm), a significant
difference (Welch’s t-test for populations with unequal variance: t = 4.06; df = 3.66; p = 0.02).
Extirpation sites have also shown relatively little change in mean annual precipitation since 1980
(Figure 3b). In addition, extirpation sites are among the 47 sites apparently lacking a sub-talus
water source. These results support our hypotheses H2 (prevailing climatic conditions) and H4
(habitat quality). While change in climate (H1) was influential for pika persistence, the trend was
opposite our expectation: pikas were extirpated from sites that did not experience climatic
change. Hypotheses represented by models incorporating average maximum summer
temperatures, change in summer maximum temperatures, and variation in both summer
maximum and mean annual precipitation were not supported by the persistence pattern found.
Since 1980, maximum summer temperatures at our study locations have averaged 0.48°C
warmer than they were from 1908 to 1979. Changes in maximum temperature varied widely
among sites (-1.2°C to +2.5°C), as did changes in annual precipitation (-80mm to +202mm, -
8.2% to +24.1%). The overall trend in this region appears to be toward an increase in annual precipitation: across all sites, precipitation has increased an average of 46mm (+5.6%).
DISCUSSION
In this study, we investigated landscape, microhabitat, and climate characteristics as possible drivers of population extirpation for O. princeps. Determination of these factors is particularly important in light of the recent consideration of this species for protection under the Endangered
Species Act (ESA). The species was not listed under the ESA, in part because it remains unclear
13
whether the patterns of climate-related loss observed in the Great Basin extend to other portions
of the species’ range. This study serves to narrow this knowledge gap and improve our
understanding of pika distribution and the factors determining this pattern.
In the Great Basin, sub-talus high mean summer temperatures and low minimum winter
temperatures were implicated as drivers of population extirpation (Beever et al. 2010). Millar
and Westfall (2010) examined Sierra Nevada talus slopes and found that warmer, drier sites were
less likely to support current pika populations than cooler, wetter sites. Here, we consider
whether such patterns are consistent with results from the eastern portion of the species' range.
Extant pika populations in the Southern Rockies are experiencing substantial climatic
change. Maximum summer temperatures are highly heterogeneous, but indicate a notable
warming trend. Despite these changes in summer temperature, we did not find an effect of
temperature on pika persistence in our study area.
While a recent impact of temperature on O. princeps populations does not appear
universal throughout the species’ range, an apparent impact of precipitation is more consistent.
Our results indicate that water, in the form of precipitation and sub-surface moisture, is the
primary driver of pika persistence patterns in our study region. Not only were the four extirpation
sites among the driest of our sites, but they also lacked sub-talus water sources, a trend
corroborated in the Sierra Nevada, where Millar and Westfall (2010) found a strong relationship
between pika sign and high precipitation, as well as sub-talus ice and water reserves. If present,
these water sources could buffer pikas and the plant communities on which they depend from the
effects of low precipitation.
If low precipitation drives extirpation in the Southern Rockies, one might expect populations experiencing a decrease in annual precipitation due to modern climate change to be
14
more prone to extirpation. Thus far, however, this is not the case. Extirpation sites have
experienced relatively little change in precipitation between the periods 1908-1979 and 1980-
2007. The fact that sites experiencing decreasing precipitation continue to support pikas may seem contradictory with the finding that populations at dry sites are more prone to extirpation.
However, the sites that have decreased in precipitation since 1980 were among the wettest locations previously ( =968 mm vs. =849 mm at sites increasing in precipitation; p=0.04, t=2.25, d.f.=18.95). Consequently, these 13 drying sites did not differ from the 56 remaining sites in mean post-1980 precipitation (950mm/yr and 910mm/yr, respectively; p=0.46, t=0.76, df=20.15). It will be important to monitor these locations in the coming years and decades, as continued drying trends could place the pika populations at these sites at risk for future extirpation.
Although our documented extirpation sites represent the driest of the sites surveyed, pikas were detected in these locations in the past century. What, then, has changed? These extirpation locations may be marginal pika habitats. As such, these sites have likely always housed “sink” populations, requiring immigration from adjacent populations in order to maintain populations of their own (Pulliam 1988). We propose that these current extirpation sites likely support populations only when ideal conditions can facilitate recolonization by individuals dispersing from adjacent sites. Local climate histories show that each of the four extirpation sites experienced a year in which annual precipitation exceeded the site’s upper 99% CI for this variable just 1-4 years before the site’s historical record of pika presence. This evidence suggests that dispersal may be facilitated by anomalously high precipitation conditions.
Our data indicate that precipitation is a driver of pika distribution in the Southern
Rockies, but why are dry sites unable to sustain pika populations? The mechanism driving the
15
trends we have found could be explained by several possible scenarios, including: 1) low-
precipitation sites do not provide adequate vegetation moisture content to sustain pika
populations (Morrison and Hik 2007 and 2008); 2) low-precipitation sites do not provide adequate snowpack insulation to buffer pikas from sub-zero temperatures (Beever et al. 2010); 3) a combined effect of plant water content and winter insulation. While Beever et al. (2010) did not investigate precipitation directly, they hypothesized that if sufficient snow cover is present during extreme cold events, pika populations are buffered from these events by the insulating properties of snow cover. Our results may support this hypothesis, given that our extirpation sites, which exhibited persistently low precipitation, may also lack sufficient snow cover. Our extirpation sites also lacked sub-talus water, a likely correlate of both low snow cover and
reduced moisture in local vegetation. Thus, the potential mechanisms proposed are difficult to
tease apart with current data. More detailed analyses of plant water content, sub-talus
temperatures, and pika survival are needed to fully examine this question.
Conclusions
Pikas in the Southern Rocky Mountains have not experienced the severe declines in site
occupancy seen in the Great Basin. While these results suggest cause for optimism concerning
the current status of the pika, the future of the species remains uncertain. Our data, combined with evidence from other regions within the species’ range (Beever et al. 2010; Millar and
Westfall 2010), indicate that the American pika’s distribution is limited by climatic factors:
populations in areas with chronically low precipitation and lacking sub-talus water sources have
been extirpated, supporting previous observations that these dry habitats are marginal for this
species (Hafner 1993 and 1994). Though the Rocky Mountains provide habitats that are higher in
elevation and more contiguous than those in the Great Basin, as the severity of climate change
16
increases in the American West, population extirpations may become more frequent throughout the species’ range. Projected declines in snowpack throughout the western United States (Mote et al. 2005) suggest that apparently stable pika populations in regions such as the Southern Rockies may soon be facing drier conditions. Further monitoring of both pika populations and climatic trends should be conducted throughout the range of O. princeps to facilitate a range-wide analysis of trends and threats to this species. Such an exercise would not only provide a better understanding of threats to the species, but may also aid managers in identifying extant pika populations that are at greatest risk of future extirpation.
17
Chapter 2
INTERACTIVE EFFECTS OF CLIMATE AND VEGETATION ON MULTI-YEAR
OCCUPANCY OF THE AMERICAN PIKA (OCHOTONA PRINCEPS)
ABSTRACT
Documenting species distributional change is challenging for species known to exhibit source- sink dynamics. For such species, a single year’s absence may reflect short term response to environmental stochasticity or the beginnings of long term decline. To distinguish between normal population fluctuations and a serious conservation crisis, species must be monitored for multiple years. Here we examine multi-year occupancy in the American pika (Ochotona princeps), a species known for its source-sink dynamics, and link that occupancy to potential climatic and habitat drivers. Surface climate and microclimatic conditions, vegetation community composition, and talus features were investigated as potential drivers of average pika occupancy from 2008 to 2011. While the majority of our survey sites (74%) maintained pika populations throughout our study period, two of 19 sites lacked pikas from 2008 to 2011 and an additional four sites lacked pikas in at least one of our four survey years. Logistic regression modeling results indicate that high summer average temperatures and low vegetation quality, in the form of high graminoid to forb ratios, have led to pika population extirpations at several locations in the Southern Rocky Mountains. Temperatures are predicted to continue their upward trend in the Southern Rocky Mountains. As temperatures increase, pika populations and the vegetation communities on which they depend are likely to show continued declines in this and other regions of western North America.
18
INTRODUCTION
Multiple meta-analyses show that many species will be committed to extinction during this
century due to global environmental changes (e.g. Thomas 2004). A single-year study of habitat
occupancy can reveal environmental correlates of species distribution, which can then be used
for predicting trends under, for example, climate change scenarios. However, source-sink
dynamics (Pulliam 1988) and environmental stochasticity can obscure these species-habitat
relationships. Species exhibiting source-sink dynamics require monitoring over multiple years to separate typical stochastic dynamics from longer term declines (Jonzen et al. 2005, Rhodes and
Johnson 2011).
The American pika (Ochotona princeps) is one such species well known for source-sink dynamics (Smith 1974a; Smith and Gilpin 1997). Pikas are a mountain- and plateau-dwelling lagomorph residing in rocky debris in most western US states and Canadian provinces (Smith and Weston 1990). As is typical of lagomorphs, pikas do not hibernate, instead depending on metabolic and behavioral adaptations which allow them to survive cold winters (Li et al. 2001,
Sheafor 2003). Among the adaptations that allow pikas to survive cold winters are a high metabolic rate and a subsequent resting body temperature only a few degrees below lethal body temperature (Li et al. 2001). As a consequence, pikas have been shown to be sensitive to temperature extremes (Smith 1974b).
Due to this climatic sensitivity, O. princeps has also become a bellwether for the ecological consequences of climate change in recent years (Krajick 2004; Guralnick et al. 2011).
Distributional shifts as well as population extirpations in the Great Basin have been linked to recent climatic trends (Beever et al. 2003 & 2010; Wilkening et al. 2011). Climatic factors have
19
also been documented as important predictors of pika occupancy and/or persistence patterns in
the Southern Rockies, Sierra Nevada, and in several US national parks (Millar and Westfall
2010; Erb et al. 2011; Jeffress et al. 2013).
Studies in the Great Basin have been able to assess long-term drivers of pika occupancy,
showing that population extinction is not being counterbalanced by new colonization in that
region (Beever et al. 2011). Due to the geography of the Great Basin, characterized by a basin-
and-range topography, movement between patches and mountain ranges is limited for many of
the region’s mammal species (Waltari and Guralnick 2009). While the initial studies of
population persistence and distribution in such regions have been integral to our understanding
of climate’s role in structuring the American pika’s past and current distribution, still needed are
multi-year studies of the species’ distribution in areas where pika habitat and pika populations
are much more extensive, such as the Southern Rocky Mountains.
In contrast to the Great Basin, regions featuring more contiguous high elevation habitats,
such as the Southern Rocky Mountains, are likely to facilitate higher rates of recolonization
following population extirpations. This potential for recolonization makes highly connected
habitats such as the Southern Rockies important testing grounds for ecological change due to
climatic shifts. These highly connected regions, featuring an abundance of high elevation habitat,
are likely to continue to serve as refugia in the face of climatic change.
The goal of the current study was to determine the spatially and temporally general
relationship between pika occupancy and habitat variables in the Southern Rocky Mountain
region. We endeavored to answer two questions: (1) What are the drivers of multi-year occupancy patterns in this region? and (2) Are these drivers consistent with the factors predicting
20
occupancy in other regions? If these long-term drivers are consistent throughout the species’
range, managers may be able to better predict the habitats of highest conservation concern in the
face of continued climatic changes.
MATERIALS AND METHODS
Study area.
Our research was conducted at 19 sites in the Southern Rocky Mountains of New Mexico,
Colorado, and southern Wyoming (35º 20’ N to 41º 30’ and -104º 54’ to -108º 17’ W; Figure 1).
These sites were selected from among 69 historically pika-occupied sites (Erb et al. 2011) via random sampling stratified by latitude, longitude, and elevation. All 2008 absences were also selected for monitoring purposes. This subset of sites represented a similar distribution of summer temperature and mean annual precipitation as our original 69 sites (Appendix I). Sites varied in elevation from 2703 to 3708 m (µ= 3240, σ=298), and consequently vegetation varied from upper montane forest to alpine meadow. Most sites were dominated by alpine forbs and grasses, but some sites were dominated by willows (Salix spp.), conifers, or aspen (Populus tremuloides).
Occupancy Surveys.
Crews of 2 to 4 individuals visited the 19 study sites each summer during 2008-2011 to determine current pika occupancy. Crews recorded fresh pika food stores (“haypiles”) and pikas were detected by sight or sound. Any one of these signs was considered evidence of current occupancy. Due to the difficulty of determining scat age, scat was not used as evidence for current pika occupancy. In most cases, pika sign was found within a matter of minutes of arrival.
Where no fresh sign was found, exhaustive searches were conducted in all rocky debris
21
Figure 1. Map of 19 sites historically occupied by the American pika (Ochotona princeps), differentiated by occupancy status during 2008-2011 (occupied = blue, transient occupancy = yellow, unoccupied = red).
22
(hereafter, “talus”) patches meeting the historical description of the site (see Chapter 1), to a
maximum distance of 3 km in all directions from the estimated historical coordinates (Erb et al.
2011). A minimum of 0.45 person-hours per hectare was spent searching talus for pika sign at
each potential extirpation site.
Climate data.
Sub-surface temperatures were used to characterize the microclimate in the talus at each site.
Four HOBO temperature data loggers (Onset U10-003) were buried in the talus at each site, one under each of four randomly selected haypiles. Each logger was placed at the maximum depth possible, averaging 75 cm below the average surface level. Where fewer than four fresh haypiles were available (n=4 sites), remaining loggers were buried under old haypiles or latrines. Where no pika sign occurred (n=2 sites), loggers were buried in locations with large rocks and deep talus to mimic pika haypile site selection. Loggers were installed between July 2008 and July
2009 and recorded temperature every 30 minutes until removal in September 2011. From these temperature data we determined 1) the proportion of values below -10°C as a measure of winter cold stress, and 2), the proportion of temperatures above 25°C and (alternatively) the average temperature between June and August as measures of summer heat stress. The data from multiple loggers were averaged to obtain a single value for each locality.
Local above-surface climatic data for each site were compiled from the PRISM interpolated climate data set for the years 2008-2011 (4 km grid cell; www.prism.oregonstate.edu). Average values for annual precipitation and summer high
temperature were calculated for each site. As with any interpolated data, particularly in a
mountainous region, PRISM estimates may not be accurate for the exact coordinates of our
23
historical records. To account for fine-scale effects of solar gain within PRISM grid cells, we estimated an insolation predictor for each site as sin(mean slope)*cos(mean aspect), similar to
Martinuzzi et al. (2009). Using this method, steep, south-facing slopes generate large negative values representing the most solar input, while steep north-facing slopes produce the largest positive values, indicating the lowest insolation. The insolation predictor was then used in models that included PRISM-based temperature estimates (see “Statistical Analysis”).
Habitat Data.
Talus and vegetation features are known or suspected to affect pika occurrence (Millar and
Westfall 2010; Rodhouse et al. 2010; Wilkening et al. 2011; Jeffress et al. 2013). Maximum talus depth was estimated visually at the deepest crevice observed within each site. Sub-surface water was categorized as present where it was seen or heard or where riparian vegetation was observed at the base of the site (Erb et al. 2011).
Vegetation communities were characterized using methods modified from Wilkening et al. (2011). Three of the four data logger locations at each site were randomly selected for vegetation sampling. At each logger, vegetation was sampled along three parallel transects, each
50 m long: a central transect, centered on the logger and running perpendicular to the dominant aspect of the site, a parallel transect 15 m upslope, and a parallel transect 15 m downslope of the logger. Vegetation was quantified at one-meter intervals using the line-point-intercept method as in Wilkening et al. (2010), resulting in 150 points per logger location and 450 points per site. All trees, shrubs, and forbs were identified to species level, while grasses, sedges, and rushes were classed as graminoids. We considered several vegetation metrics representative of forage quality:
Percent forb cover, forb species richness, and the ratio of graminoids to forbs (gram:forb).
24
Statistical Analysis.
Extinction-recolonization dynamics may be common in this species, at least at small spatial
scales (Moilanen et al. 1998). We focused on mean occupancy over a four year survey period
(2008-2011) to minimize the effects of inter-annual environmental stochasticity and source-sink
dynamics. Number of years of presence and absence of pikas at each site was tallied for use in a
binomial model (see below). Total number of years surveyed was four for all sites.
Using maximum likelihood estimates and an information-theoretic approach for model
assessment, we compared models of pika persistence incorporating the climatic and habitat
variables described in the previous section. Our goal was to evaluate the relative support for each
of 10 predictor variables (Table 1). We considered all possible models incorporating three or
fewer predictors. By considering each predictor in a variety of contexts, we developed an
unbiased ranking to facilitate the comparison of predictors of population density (Chapter 3) with
predictors of species occupancy. The 121 models explored in this study represent hypotheses
regarding factors that affect habitat suitability for pika occupancy. These hypotheses fall into
four major categories derived from previous literature on American pika occupancy drivers: 1)
Above-surface climatic suitability, tested using models incorporating PRISM-generated data for mean summer maximum temperature and mean annual precipitation (Beever et al. 2003 & 2010;
Millar and Westfall 2010; Wilkening et al. 2011; Erb et al. 2011); 2) Microhabitat suitability,
tested using sub-surface summer and winter temperature metrics (Beever et al. 2011), as well as
talus depth, presence of water under the talus, and insolation (Hafner 1994, Millar and Westfall
2010, Erb et al. 2011); 3) Vegetation quality, assessed using gram:forb, forb species richness,
and percent forb cover (Rodhouse et al. 2010, Wilkening et al. 2011, Jeffress et al. 2013); and 4)
Interactive effects of above-surface climate, microhabitat, and vegetation.
25
Table 1. Relative support for candidate predictors of pika occupancy. A boxplot of all coefficient values is provided in Appendix II.
Predictor Summed Akaike Weight Weighted Average Coefficient
Mean Summer Maxima 0.863 -1.584
Gram:Forb 0.803 -1.632
Sub-Talus Water 0.235 0.249
Mean Annual Precipitation 0.225 0.387
26 Forb Cover 0.187 0.263
Sub-Surface Temperature Values < -10°C 0.167 -0.176
Forb Richness 0.146 0.170
Talus Depth 0.054 0.013
Mean Sub-Surface Summer Temperature 0.016 -0.013
Sub-Surface Temperature Values > 25°C 0.007 -0.004
Microclimate was assessed in several different ways in this study, allowing evaluation of
alternative models. For example, models incorporated either subsurface temperatures or microhabitat features such as talus depth. The insolation predictor was used as a correction for the coarse scale of PRISM, and therefore was included only in models that also included above- surface summer maximum temperature. The three above- and sub-surface summer temperature metrics each represented alternative heat stress hypotheses, and therefore these metrics were not run in models together.
Binomial logistic regression models representing the above hypotheses for species occupancy were developed and compared using AICc (Burnham and Anderson 2002). Models
were fitted in R 2.13.0 using function glm (R Development Core Team 2011). The probability of
detecting this species is quite high (>0.90; Beever et al. 2010; Rodhouse et al. 2010; Erb et al.
2011), allowing for the use of traditional logistic models without the addition of detection
probability. Several variables were log-transformed to reduce skew. Each continuous predictor variable Xi was standardized as , allowing for intuitive interpretation of model coefficients: each βi measures the effect of Xi in units of standard deviation in Xi. Model
rank was determined using AICc and predictor influence was assessed via calculation of Akaike
2 2 weights. Model fit was evaluated using Nagelkerke’s max-rescaled R (R N), which measures
model performance relative to the null model (Nagelkerke 1992).
RESULTS
Overall, pika occupancy is high in the Southern Rocky Mountains, and 79% (15 of 19) of our surveyed sites were occupied in 2008, 89% (17 of 19) in 2009, 84% (16 of 19) in 2010, and 74%
27
(14 of 19) in 2011 (Figure 1). Two of 19 sites were continuously unoccupied in our survey period, while four sites lacked pikas for between one and three of our four years of study. As these variable occupancy records indicate, we documented both recolonization events and population extirpations between 2008 and 2011. Two sites experienced recolonizations in 2009 following pika absences in 2008. While one of these recolonized sites lost its pika(s) again prior to the 2010 field season and remained unoccupied in 2011, the other site, Bighorn Peak in Rocky
Mountain National Park, has retained its pikas and has demonstrated a growing population since
2009 (Erb et al. in press).
Mean PRISM summer maximum temperatures ranged 15.9-22.3°C (µ=18.8, SE=0.38), while average sub-surface summer maximum temperatures ranged 8.57-20.9°C (µ=13.7,
SE=0.77). Average sub-surface summer temperatures ranged 6.4-16.8°C (µ=10.6, SE=0.64), and precipitation ranged 432.9-1476.5 mm (µ=888.4, SE=58.0). Seven sites experienced no sub- surface temperatures above 25°C, and the proportion of sub-surface temperatures above 25°C at the remaining 12 sites ranged 0.003-0.6% (µ=0.15%, SE=0.06%). Only one site experienced no sub-surface temperatures below -10°C, and the proportion of sub-surface temperatures below -
10°C at the remaining 18 sites ranged 0.01% - 8% (µ=1.7%, SE=0.5%). Vegetation composition was highly variable, with forb cover ranging 9-59% (µ=32%, SE=3.9%), gram:forb ranging 0.3-
5.1 (µ=1.3, SE=0.27), and forb species richness ranging 2-23 (µ=12.5, SE=1.2).
The five top models (ΔAIC<2; Table 2) all included above-surface summer maximum temperature and graminoid to forb ratio (gram:forb), and all top models reported high pseudo-R2
2 values (RN ranged 0.91-0.93). These top models were averaged for depiction in Figure 2.
28
Table 2. Top models of mean pika occupancy compared with the null model.
2 Model AICc DeltaAICc R N y~ Mean Summer Maxima + Gram:Forb 29.98 0 0.91 y~ Mean Summer Maxima + SubTalusWater + Gram:Forb 30.04 0.06 0.93 y~ Mean Summer Maxima + Gram:Forb + Values < -10°C 30.13 0.15 0.93 y~ Mean Precip + Mean Summer Maxima + Gram:Forb 31.22 1.25 0.92 y~ Mean Summer Maxima + Forb Richness + Gram:Forb 31.54 1.56 0.92
… y~1 63.76 33.78 -
29
Figure 2. Pika occupancy during 2008-2011 as observed (open circles) and as predicted (line) by model averaging based on the five top models in Table 2.
30
Above-surface summer maxima and gram:forb were the top predictors of pika occupancy in our
survey period, with Akaike weights of 0.86 and 0.80, respectively (Table 1). Both of these
predictors demonstrated negative relationships with occupancy in our study area (Appendix II).
DISCUSSION
Pikas continue to persist in the majority of historically occupied locations surveyed in the
Southern Rocky Mountains. Occupancy rates in this region, ranging at least 74-89% within the
time period examined, remain higher than those in the Great Basin, where only 60% of historical
sites remain occupied (Beever et al. 2010). Despite overall high occupancy rates in the Southern
Rockies, not all populations surveyed between 2008 and 2011 exhibited consistent occupancy.
Fifteen sites demonstrated consistent pika presence or absence throughout our study period,
while the occupancy status of the remaining four sites was transient, changing between 2008 and
2011.
By modeling mean occupancy over a four year period, we were able to determine that the
interactive effects of above-surface climate and vegetation quality are likely driving occupancy
dynamics in the Southern Rockies. Our results suggest that high summer maximum temperatures
and low quality forage resources have caused some locations to become unsuitable for pikas in
recent years. Although our historical data indicate that pikas were present at these sites prior to
1980, we cannot say with certainty that these sites were consistently occupied historically.
However, given the known presence of pikas at these sites at some point prior to 1980, and the well-documented recent increases in temperature in this region (Ray et al. 2008, Rangwala and
31
Miller 2010, Clow 2010), it is possible that climate change is contributing to the current pika occupancy pattern seen in the Southern Rockies.
The importance of summer temperature and forage quality for pika occupancy in this region is consistent with findings from elsewhere in the species’ range. Individual pikas near
Bodie, California have demonstrated high sensitivity to prolonged elevated temperatures (Smith
1974b). Mean summer temperature and days above 28°C have been implicated in pika population extirpations in the Great Basin (Beever et al. 2010). High graminoid cover has negatively predicted occupancy in some regions (Rodhouse et al. 2010, Jeffress et al. 2013), and higher forb cover positively predicted pika occupancy in others (Wilkening et al. 2011, Jeffress et al. 2013).
Higher temperature sites with healthy populations may be buffered from the effects of climate by high quality forage resources and selective caching behavior, as documented further in Smith and Erb (2013) and Erb et al. (in press). Of the six sites lacking pikas at some point during our survey period, all exhibited relatively high summer average temperatures (>18.7°C), and four of the six also had poor quality forage resources (gram:forb > 1). It remains unknown, however, the extent to which selective caching behavior can buffer pikas against a changing climate. The two remaining absence sites featuring high quality forage resources (gram:forb
~0.5) both experienced transient occupancy, supporting populations in all but the 2011 season.
Since 2011 was our final field season, we are currently unable to determine if these extirpations are a sign of the weakening buffering capacity of forage quality, or if other environmental or population stochasticity drove these population declines. Four additional sites with high
(>18.7°C) summer temperatures and high quality forage resources maintained their populations throughout our study period. The future trajectory of these sites, along with existing transient
32
sites, are likely to be most informative regarding the role of vegetation buffering as warming
continues.
While summer high temperatures were the dominant climatic predictor in our analysis, it
is notable that precipitation was not a top predictor of long-term pika occupancy. Precipitation
was an important driver of pika persistence when comparing pre-1980 Southern Rockies
occupancy with that in 2008 (Erb et al. 2011). The shift away from precipitation as a top
predictor is likely due to the strong trend toward increasing temperatures (Pederson et al. 2013,
Ray et al. 2010) and the high variability in the direction and magnitude of changes in
precipitation in this region (Mote et al 2005). Despite the absence of mean annual precipitation in
our top predictor list, it is likely that precipitation continues to influence pika occupancy, if
indirectly. Unoccupied sites and those sites supporting transient populations are not only hotter
than occupied sites, but drier as well (Figure 3). Furthermore, the importance of both local
temperature and precipitation in structuring plant communities should not be overlooked (Box
1996), particularly given the crucial role of vegetation composition in determining pika
occupancy in this region.
As the higher elevations of Southern Rocky Mountains become hotter and drier, as is
projected by many models (e.g. Rangwala et al. 2012), we expect vegetation to shift as well. We
are likely in the early stages of such a shift, resulting in transient pika populations, where sites
with higher forb cover may be able to shield pikas from extreme climatic events. As these
climatic shifts take their toll on vegetation as well, however, these sites are unlikely to continue
to provide their residents with a buffer against stressful, hot summers. Unfortunately, such
summers are predicted to increase in frequency in the coming century (Rangwala et al. 2012), and species’ ability to adapt to the current rapid rate of change is unlikely (Davis et al. 2005).
33
Given these predicted trends and the consistency across studies in the determinants of pika occupancy throughout the western US, we anticipate higher rates of extirpation in the Southern
Rockies in the coming decades.
34
Figure 3. Mean summer maximum surface temperature (left panel), annual precipitation (center panel), and gram:forb (right panel) at
occupied, transient, and unoccupied pika survey sites during 2008-2011. 35
Chapter 3
DETERMINANTS OF PIKA POPULATION DENSITY VERSUS OCCUPANCY IN THE
SOUTHERN ROCKY MOUNTAINS
ABSTRACT
Species distributions are responding rapidly to global change. While correlative studies of local extinction have been vital to understanding the ecological impacts of global change, more mechanistic lines of inquiry are needed for enhanced forecasting. The current study assesses whether the predictors of local extinction also explain population density for a species apparently impacted by climate change. We tested a suite of climatic and habitat metrics as predictors of relative population density of the American pika (Ochotona princeps) in the Southern Rocky
Mountains. Population density was indexed as the density of pika latrine sites. Negative binomial regression and AICc showed that the best predictors of pika latrine density were patch area
followed by two measures of vegetation quality: the diversity and relative cover of forbs. In
contrast with previous studies of habitat occupancy in the Southern Rockies, climatic factors
were not among the top predictors of latrine density. Populations may be buffered from decline
and ultimately from extirpation at sites with high quality vegetation. Conversely, populations at
highest risk for declining density and extirpation are likely those in sites with poor quality
vegetation.
INTRODUCTION
Climate change is a key driver of species distributional shifts worldwide (e.g. Lenoir et al. 2008;
Chen et al. 2011). Occupancy studies documenting local extinctions and range shifts have been
36
vital to our understanding of the ecological effects of climate change (Chen et al. 2011).
However, it is not enough to understand which habitat metrics best predict occupancy; some occupied sites may be “sinks” that receive immigrants from suitable habitat (Pulliam 1988).
Studies of population density may be required to presage populations at risk (Pimm et al. 1988).
In order to better manage species threatened by global change, scientists must move toward a more mechanistic line of inquiry (Guralnick et al. 2011). Explaining patterns in population density represents one step toward inferring the mechanics of range shift, given that dwindling populations are often a precursor to local extinction (Pimm et al. 1988, Gaston et al.
2000). More generally, studies of population density are needed to better illustrate the occupancy-abundance relationship, which has been characterized as mainly positive but sometimes variable across taxa (Gaston et al. 2000, Holt et al. 2002). Here we ask whether the determinants of occupancy align with those of population density using data from a species widely noted for recent declines in occupancy.
The American pika (Ochotona princeps), a small lagomorph, is closely associated with rocky debris such as talus slopes and lava beds and is further limited to relatively cool summer and wet winter climates in western North America (Smith and Weston 1990, Hafner 1994).
Though pikas can escape extreme weather by using the sub-surface microclimates created by large-diameter rocky debris (MacArthur and Wang 1974, Millar and Westfall 2010), they are sensitive to variation in summer temperature and winter snowpack (Smith 1974, Beever et al.
2011). Pika distribution is often predicted by both climatic and vegetation-based factors (Millar and Westfall 2010, Jeffress et al. 2013) and there is accumulating evidence linking decline in pika occupancy to climatic and microclimatic factors (Beever et al. 2010, Erb et al. 2011).
However, frequency of extirpation varies by region. In the Southern Rocky Mountains, relatively
37
few extirpations have been documented (Erb et al. 2011), providing an opportunity to investigate
the drivers of density among intact populations.
Occupancy patterns in the Southern Rockies indicate that climatic and habitat factors
structure pika distribution. In an assessment of the predictive power of multiple habitat and
climatic variables, recent (2008-2011) occupancy patterns were best explained by precipitation,
vegetation quality, and summer temperature metrics (Erb et al. 2011 and in prep). Here we
employ the same suite of potential predictors to model pika latrine density, as a metric of relative
population density, using generalized linear models ranked via AICc (Akaike’s information
criterion corrected for small sample sizes). If similar factors predict both occupancy and relative
density, these predictors could be used to determine which populations are at highest risk of
extirpation in the near future.
METHODS
Study Area. Pika latrine density was assessed at 18 sites in the Southern Rocky Mountains of
New Mexico, Colorado, and southern Wyoming (35°20’ to 41°30’ N and 104°54’ to 108°17’
W). These sites were selected from among 69 historically occupied sites (see Erb et al. 2011) via random sampling stratified by latitude, longitude, and elevation. All 2008 absences were also selected for monitoring purposes. These 18 sites represented a similar distribution of climatic
predictors as the original 69 (Appendix III). Vegetation varied among sites from upper montane
forest to alpine meadow. Selected sites ranged 2703-3708 m in elevation (µ= 3250, σ=302; Fig
1) and each was sampled for pika latrine density in both 2009 and 2010. At sites with multiple
talus patches, the largest talus patch was selected for pika sampling as well as habitat
characterization (Erb et al. 2011).
38
Figure 1. Map of 18 study sites at which Ochotona princeps population density was evaluated.
All study sites were occupied by pikas at some point prior to 1980.
39
Quantifying Pika Latrine Density. The density of fresh scat piles (latrines) was assessed using a transect-based design, with transects positioned relative to a “control point” located at the lowest instance of rocky debris at the site. A base transect was established 10 m upslope of the control point running perpendicular to the dominant site aspect (e.g., toward 120º/300º if the dominant site aspect was 30°). Subsequent transects were established upslope and parallel to the base transect at 60 m intervals (distances are actual, not elevational). Number of transects per site varied between one and six, depending on the extent of rocky debris (hereafter, “talus”). Sixty meter spacing between transects ensured that each pika sign was counted only once in our surveys, given an average pika territory diameter of 25 m (Smith and Weston 1990). Each transect was just as long as the talus was wide. Most transects were <150 m long (µ=62.3 m,
σ=46.1 m), however, where transects were longer than 150 m (n=3), only the middle 100 m and
25 m at each end were sampled.
In our pilot year of 2009, crews recorded all fresh haypiles (pika food stores), fresh latrines, and individual pikas observed within 30 m of transects. Fresh hay and latrines were determined by the presence of green pigment in constituent plant material. Pikas are coprophagous and produce both caecal feces, which are reingested, and fecal pellets, which are not. In our study of latrines, only fecal pellets were examined for evidence of deposition within the current year. The green color within fecal pellets has been documented to disappear within months of deposition (Nichols 2010). Paired observers walked each transect, with one observer focused downslope and the other focused upslope, while both observed sign directly on the transect. Paired observers conferred frequently and recorded each sign only once. Independent observations were obtained by having two independent pairs walk each transect on different days. Averaging across sites in 2009, 82% of fresh latrines were recorded by the observer
40
looking upslope. Because pika scat is generally deposited under sheltering rocks, making latrines
easier to see when looking upslope, we discontinued downslope observations in 2010 and
omitted downslope data from all analyses.
Fresh latrine counts were the most consistently available sign of current pika activity,
likely due to variation in pika behavior at different population densities (C. Ray and L. Erb, pers.
obs.) and in different weather conditions (Hayes and Huntly 2005) as well as climatic and
seasonal effects on pika haying behavior (Smith and Weston 1990, Dearing 1996).
Consequently, we used the density of fresh pika latrines as our metric of relative population
density.
It is notable that scat was deemed an inconsistent metric of pika occupancy in the multi-
crew data collection descriptions of Chapters 1 and 2, but was considered the best metric of
density determination for Chapter 3. This inconsistency was due to the large number of different
field crews used in occupancy estimations in 2008. Inter-observer variability among crews in determining scat freshness made this an inconsistent metric for occupancy analyses. In 2009 and
2010, however, a single, closely-monitored crew was assessing scat freshness. This greatly reduced inter-observer variability and greatly increased our confidence in scat as an index for pika population density.
To limit the influence of observer error, density analyses were based on observations
within 2 m of transects, where most latrines (mean±SE = 84.4%+15.8%) were observed in this
study. Given a standard transect length of 150 m, latrine density (number per standard transect)
was calculated as y = N/L*150, where N was the total number of fresh latrines observed at the
site and L was the total length of transects surveyed at the site. All density estimates (two from
41
2009 and two from 2010) were averaged to generate our response variable, minimizing effects of
environmental stochasticity and observer bias.
Characterization of Climate. For each site, we calculated five climate metrics representing
potential stressors (Table 1). Mean annual precipitation and mean summer temperature maxima
were obtained via the PRISM interpolated climate data set for the years 2008-2011 (4 km grid cell; www.prism.oregonstate.edu). Pika occupancy has been predicted in part by mean annual precipitation in the Southern Rocky Mountains (Erb et al. 2011 and in prep) and high summer temperatures in the Great Basin (Beever et al. 2011, Wilkening et al. 2011).
Sub-surface temperatures were used to characterize the microclimate in the talus at each site. Four HOBO temperature data loggers (Onset U10-003) were buried in the talus at each site,
one under each of four randomly selected haypiles. Each logger was placed at the maximum
depth possible, averaging 75 cm below the average surface level. Where fewer than four fresh
haypiles were available (n=4 sites), remaining loggers were buried under old haypiles or latrines.
Where no pika sign occurred (n=2 sites), loggers were buried in locations with large rocks and
deep talus to mimic pika haypile site selection (C. Ray and L. Erb, pers. obs.). Loggers were
installed between July 2008 and July 2009 and recorded temperature every 30 minutes until
removal in September 2011.
Cold stress was quantified as the proportion of sub-surface temperatures below -10°C recorded at each site, a metric that predicted pika occupancy patterns in the Great Basin (Beever et al. 2010). Sub-surface heat stress was quantified as the proportion of temperatures above 25°C
and (alternatively) as the mean summer (June-August) temperature recorded at each site. Mean
summer temperature has explained pika occupancy patterns in the Great Basin (Beever et al.
42
2010) and the Southern Rockies (Erb et al. in prep), and pikas held above the talus have died at
surface temperatures as low as 25.5ºC (Smith 1974b).
Characterization of Habitat. Maximum talus depth was estimated visually at the deepest crevice observed within each site. Sub-surface water was categorized as present where it was seen or heard or where riparian vegetation was observed at the base of the site (Erb et al. 2011). Deeper
taluses and those with sub-surface water sources should provide a more stable microclimate (Ray
et al. 2012).
Vegetation communities were characterized using methods modified from Wilkening et
al. (2011). Three data logger locations at each site were randomly selected for vegetation
sampling. At each of these three logger locations, vegetation was sampled along three parallel
transects, each 50 m long: a central transect, centered on the logger and running perpendicular to
the dominant aspect of the site, along with transects 15 m upslope and downslope of the logger.
Vegetation was quantified at one-meter intervals using the line-point-intercept method (as in
Wilkening et al. 2011), resulting in 150 points per logger location and 450 points per site. All
trees, shrubs, and forbs were identified to species level, while grasses, sedges, and rushes were
classed as graminoids. We considered several vegetation metrics consistent with recent studies of
pika occupancy (Rodhouse et al. 2010, Wilkening et al. 2011, Jeffress et al. 2013). Total percent
vegetation cover was our measure of available forage for this generalist species. Forage highly
accessible to pikas was represented by the combined percent cover of forbs, graminoids, and
shrubs, omitting only tree cover. Percent forb cover, forb species richness, and the ratio of
graminoids to forbs (gram:forb) were all considered metrics of forage quality.
Cover at ground level was also classed at each transect point as soil, rock, or litter.
43
Although highly associated with talus, pikas occur mainly at the edges of rocky habitats where
forage is easily accessed (Smith and Weston 1990). Thus, we anticipated a negative relationship
between rock cover and latrine density, with lower rock cover sites providing greater foraging
opportunities. Similarly, talus patch size effects were expected to follow Kawamichi (1982) who
showed a trend in O. princeps toward smaller home ranges resulting in higher densities in small
patches. Patch size was measured as the area of large-diameter (>20 cm) rocky debris contained
within each talus patch surveyed for latrine density.
Statistical Analysis. Negative binomial models of latrine density were developed and compared
using AICc (Burnham and Anderson 2002). Our goal was to evaluate the relative support for
each of 15 predictor variables (Table 1). We considered all possible models incorporating three
or fewer predictors. By considering each predictor in a variety of contexts, we developed an
unbiased ranking to facilitate the comparison of predictors of latrine density with predictors of
habitat occupancy. Models were fitted in R (R Development Core Team 2012) using function
glm.nb (Venables and Ripley 2002). We considered a mixed-distribution model to address pika
occupancy and density as separate processes, but a Kolmogorov-Smirnov test indicated no
excess zeros (absences) in our response variable. Several variables were log-transformed to
reduce skew, and each continuous predictor variable Xi was standardized as
allowing for intuitive interpretation of model coefficients (βi): each βi measures the effect of Xi in units of standard deviation in Xi. Model rank was determined using AICc and predictor influence
was assessed via Akaike weights. Model fit was evaluated using Nagelkerke’s max-rescaled R2
2 (R N), which measures model performance relative to the null model (Nagelkerke 1992).
44
Table 1: Predictor descriptions, expected relationships to pika population density, weighted average coefficients, and Akaike weights.
Top predictors are highlighted in gray.
Expected Weighted CV of Akaike Predictor Description Relationship Average Coefficient Weight with Density Coefficient Estimate
Climate
45 Mean Annual Precipitation, PRISM 2008-2011 (mm) + 0.539 0.217 0.209
Sub-Surface Summer Mean Temperature; Mean June, July, August - -0.340 -0.710 0.113 temperatures from data loggers 2009-2011 (°C)
Sub-Surface Summer Maximum Temperature; Proportion of recorded - -0.096 -4.212 0.055 temperatures above 25°C
Surface Summer Maximum Temperature; Mean June, July, August - -0.292 -0.87 0.092 monthly PRISM maxima 2008-2011 (°C)
Sub-Surface Winter Minimum Temperature; Proportion of recorded - -0.442 -0.383 0.204 sub-surface temperatures below -10° C
Habitat Quality: Talus
Talus depth (m) + 0.198 0.783 0.067
Sub-talus water source + -0.078 -0.713 0.057
Habitat Quality: Vegetation
46 Percent vegetation cover (%) + -0.188 2.64 0.078
Gram:Forb (cover ratio) - -0.642 -0.367 0.251
Forb species richness (Number of species) + 0.594 0.335 0.260
Percent cover of forbs, grasses, and shrubs (%) + 0.415 0.410 0.137
Percent of vegetation consisting of forbs (%) + 0.427 2.39 0.173
Habitat Quantity
Rock Cover (%) - -0.504 -0.520 0.131
Talus Area (m2) - -0.698 -0.15 0.661
Overall Habitat Quality
Rock Cover*Vegetation Cover + -0.434 -0.415 0.123
47
RESULTS
Across our 18 study sites, pika latrine densities ranged 0-57 per 200 m2 (µ=10, σ=14.3;
Appendix IV). Two of our sites lacked any fresh pika sign throughout our survey period (Figure
1). Mean PRISM summer maximum temperatures ranged 15.9-22.3°C (µ=18.9, σ=1.6), while
mean sub-surface summer maximum temperatures ranged 8.57-20.6°C (µ=13.3, σ=3.1). Mean
sub-surface summer temperatures ranged 6.4-16.8°C (µ=10.6, σ=2.9) and precipitation ranged
432.9-1476.5 mm (µ=890.9, σ=259.8). One site experienced no sub-surface temperatures below -
10°C. At the remaining 17 sites, the proportion of sub-surface temperatures below -10°C ranged
0.01% - 8% (µ=1.7%, σ=2.2%).
Talus patch area ranged 495-144,000 m2 (µ=17,180, σ=33,831) and rock cover ranged
24-83% (µ=54%, σ=17%) among sites. Vegetation cover ranged 18%-55% (µ=35%, σ=11%)
and accessible forage cover (omitting trees) ranged 7-54% (µ=28%, σ=11%). Vegetation composition was highly variable, with cover values ranging 9-59% (µ=32%, σ=18%) for forbs,
0.5-5.1% (µ=1.4%, σ=1.2%) for gram:forb, and 2-23% (µ=12.3%, σ=5.4%) for forb species
richness.
Results of modeling showed that the best predictor of pika latrine density in the Southern
Rockies was patch area. After area, vegetation quality metrics were the best predictors among
other climate and habitat variables. Twelve of our 15 predictors demonstrated expected
relationships with density (Table 1). Sites with smaller patch area, higher forb species richness,
and lower gram:forb ratios demonstrated the highest pika densities (Table 1; Figure 2). The top
2 model included patch area and sub-surface minimum temperature, with R N=0.47; however,
2 there were 13 additional models that were equally explanatory, with ΔAIC<2 and R N>0.30
48
Figure 2. Pika latrine density versus the top three density predictor variables, determined by Akaike weights. Predictor variable units
appear in Table 1.
49
(Appendix V). Predictor weights supported the importance of patch area and forage quality in determining pika latrine densities (Table 1).
DISCUSSION
Our results indicate that patch area and vegetation metrics determine pika latrine density in the
Southern Rocky Mountains. The importance of patch area, the ratio of graminoids to forbs, and forb diversity indicates that forage accessibility and quality drive relative pika density patterns in this region. This finding is consistent with studies of occupancy in the western US, which indicate that O. princeps occupancy and/or persistence are predicted by graminoid cover
(negative effect; Rodhouse et al. 2010, Jeffress et al. 2013, Erb et al in prep), forb cover (positive effect; Wilkening et al. 2011, Jeffress et al. 2013, Erb et al. in prep) and forb richness (positive effect; Erb in prep). Our results support the likelihood of a positive occupancy-abundance relationship for pikas, as has been found for many other taxa (Gaston et al 2000; Holt et al.
2002).
It is notable that vegetation structure, especially gram:forb, may reflect local climate
(Box 1996) and may be affected by the pika’s selective foraging behavior (Huntly et al. 1986,
Dearing 1997). If gram:forb is a more comprehensive metric of local climate than PRISM or our own local data, pikas may be responding directly to climate rather than vegetation quality. While we cannot rule out this possibility, we can rule out effects of pika density on gram:forb. Pikas are known to prefer forbs over graminoids (Huntly et al. 1986, Dearing 1997). As suggested by
Rodhouse et al. (2010), the selective removal of forbs should generate a positive relationship between gram:forb and pika occurrence, and our study is the first to confirm that relative pika density scales negatively with gram:forb. This result, combined with observations discussed below, suggest that there are direct effects of forage quality mediating the pika’s response to
50
climate and climatic changes.
The negative relationship between patch area and latrine density is likely the result of
edge effects at the rock-meadow interface. Kawamichi (1982) found a similar area-density result
for this species, suggesting that reduced territory size is sustainable where a larger perimeter-to- area ratio increases access to forage. Other herbivores have been documented showing preference for smaller patches of habitat (e.g. Hester et al. 1999) and landscape ecology studies have found a similar affinity for edges among small mammals (e.g. Salek et al. 2010).
Supporting the importance of access to forage, our results also suggest that forage quality is an important predictor of relative pika density. Gram:forb likely represents the relative abundance of low (graminoid) to high (forb) quality forage for pikas (Dearing 1997, Smith and
Erb 2013). The dominant forbs in these habitats contain higher water and nitrogen content than the dominant graminoids (Smith and Erb 2013). Sites with higher forb diversity and lower gram:forb are more likely to provide a variety of highly nutritive forage, reducing the effects of intraspecific competition and allowing relatively higher densities of pikas at such sites.
Due to the high metabolic demands on pikas (Sheafor 2003), nutrient and water availability should limit pika survival and density, especially where high summer temperatures limit foraging time. Smith and Erb (2013) found a positive correlation between mean summer temperature and pika selectivity for plants with higher water content. Thus, selective foraging behavior may mask effects of a warming climate. However, it is an open question whether, and for how long, such behavior can counter effects of climate change. Higher summer temperatures have predicted pika extinctions in the Great Basin (Beever et al. 2010) and Southern Rockies
(Erb et al. in prep.).
51
Our results suggest that vegetation accessibility and quality are important for both occupancy and density patterns in the Southern Rockies. We did not find strong evidence that relative pika density is predicted directly by climate. These results have theoretical and management implications, especially because summer temperature and mean annual precipitation have been important predictors of pika occupancy throughout the western US (Beever et al. 2010, Erb et al.
2011). Combining results from pika occupancy and density studies in this region, our findings suggest that some previously suitable habitats are becoming climatically unsuitable (Erb et al.
2011), but the accessibility and quality of forage mediates population response to climate.
Though selective foraging can mask the effects of climate to some extent, the buffering ability of vegetation may be limited, especially as vegetation communities show stronger responses to continuing climate change. Given the buffering capacity of higher quality sites, pika populations in regions of low forb diversity and cover are likely to be at more immediate risk of extirpation as climate change accelerates.
.
52
Conclusion
In the context of evolutionary time, 30 years is a blink of an eye. Just a century ago, one would
have thought that thirty trips around the sun, when the Earth has made millions, surely couldn’t
produce changes that altered the course of life on Earth. But just a century ago, most humans
couldn’t foresee the consequences of our energy acquisition needs. The consequences of
accelerating human hunger for resources became readily apparent in the 1980s. Since that time,
each trip around the sun has been, on average, warmer than the previous. While Earth is used to
and thrives on change, the rate at which this change is happening is leading to unprecedented
ecological consequences.
Ochotona princeps has been the subject of a media blitz in recent years regarding its
potential role as a bellwether for the ecological effects of climate change. While many saw this
attention as overhyped and premature, several years of population monitoring across its range
indicates that pikas may indeed be aptly characterized as such. While declines have not been
severe in the Southern Rockies, the importance of summer high temperatures in the maintenance
of pika populations and their densities, paired with the projections of continued increases in
temperatures in this region, suggest challenging times lie ahead for many populations of this
species. Regions such as the Southern Rockies also provide some hope, however. The highly
connected, high elevation landscape present in this region provides pikas with potential refugia from warming temperatures and potential for recolonization of marginal habitats from these refugia. While the distribution of this species is rapidly changing throughout western North
America, refugia such as those found in the Southern Rockies may allow O. princeps to resist
53
extinction in the near future. How much optimism we have may be proportional to our own active efforts to curb CO2 emissions.
Following four years of field work and six years of processing that data, we have learned a great deal about the status of the American pika in the Southern Rocky Mountains.
Unsurprisingly, we are also left with many more questions about the species, its limitations, and its resilience in the face of global change. Occupancy patterns indicate that populations of the
American pika have maintained themselves in the Southern Rockies better than other regions such as the Great Basin. Occupancy rates in both our historical resurvey and our surveys from
2008 to 2011 are higher than those documented in the Great Basin.
The extirpation of several populations in the Southern Rockies during our study period allowed us to determine the drivers of these patterns. After modeling these data using multiple potential climatic and habitat predictors, it is clear that temperature and vegetation quality are important drivers of these patterns. We gained some insight into the mechanism behind these relationships when similar predictors were found to be important drivers of pika densities as well. The importance of high vegetation quality and low summer temperatures in maintaining pika populations should assist managers in determining critical habitat and populations that may be threatened by future changes in climate and the subsequent changes to vegetation composition.
Despite these higher occupancy rates and a lack of significant decline in occupancy during our survey period, it will be imperative that populations continue to be monitored over the coming years and decades. The disappearance of two new populations in 2011 is cause for concern. In both cases, pikas were found in moderate to high densities in previous surveys. Due
54
to the fact that 2011 was our final survey year, it remains unknown if those sites have been
recolonized since. Further, long term monitoring of populations will provide still-needed data on
rates of recolonizations and whether the source-sink system upon which pika population dynamics relies has been compromised in some way.
The important question of the factors limiting recolonization of potential pika habitat can also be addressed through landscape genetics methods. I am currently collaborating with Dr.
Loren Sackett in an NSF Dissertation Improvement Grant-funded effort to assess the role of climate and habitat features in limiting pika dispersal in this region. Genetic samples were acquired via scat samples from throughout the Southern Rockies region in 2009 and 2010. These samples were successfully sequenced at multiple microsatellite loci by the USDA Forest
Service’s Wildlife Genetics Lab in Missoula, MT. Current work includes developing GIS models of climate and vegetation for the region, which will be paired with the genetic analyses to determine the limitations to O. princeps gene flow in the Southern Rockies. While this analysis
is not included as part of this dissertation, we intend to submit work for publication on the
subject in early 2014.
Now that this dissertation research has shed some light on the drivers of pika extirpations in the region, the pursuit of answers regarding our pika recolonization questions will be vital to future management of this species. The Rocky Mountains have served as a refuge for many cold-
loving species in past warming periods. The region is ideal for such a role due not only to its
high-elevation habitat, providing a wet and cool climate and diverse forb communities, which
facilitate pika persistence, but also due to the connectivity between patches. While individual
American pikas have been documented to travel long distances during dispersal events, such
events are only possible in regions providing suitable matrix habitat for survival during the
55
dispersal period. The fast pace of current climate warming is not only raising sea levels, but also
ecotone boundaries, creating more discrete habitat islands and isolating populations of species
such as O. princeps.
The fate of O. princeps will depend on pikas and humans alike; pikas’ survival will depend on their own ability to sustain themselves on these islands and disperse between them, but it will also depend on human resource use. Our environmental decisions and indecisions in the coming years and decades will determine the rate at which alpine habitat islands shrink and recolonization between them become infeasible. As species such as pikas decline, our resource reserves also deplete, from fresh water to clean air. In our first 200,000 laps around the sun, humans have shown the power we have to change the course of Earth’s history. What we choose to do with our next 20 laps will determine the fate of many of Earth’s species, Homo sapiens
included.
56
Bibliography
Barnett, T.P., D.W. Pierce, H.G. Hidalgo, C. Bonfils, B.D. Santer, T. Das, G. Bala, A.W. Wood,
T. Nozawa, A.A. Mirin, D.R. Cayan, M.D. Dettinger. 2008. Human-induced changes in the
hydrology of the western United States. Science 5866:1080-1083.
Bartlein, P.J., C. Whitlock, S.L. Shafter. 1997. Future climate in the Yellowstone National Park
region and its potential impact on vegetation. Conservation Biology 11:782-792.
Beever, E.A., P.F. Brussard, and J. Berger. 2003. Patterns of apparent extirpation among isolated
populations of pikas (Ochotona princeps) in the Great Basin. Journal of Mammalogy 84: 37-
54.
Beever, E.A., C. Ray, P.W. Mote, and J.L. Wilkening. 2010. Testing alternative mechanisms of
climate stress in the ecoregional collapse of an alpine mammal. Ecological Applications 20:
164-178.
Beever, E.A., C. Ray, J.L. Wilkening, P.F. Brussard, and P.W. Mote. 2011. Contemporary
climate change alters the pace and drivers of extinction. Global Change Biology 17:2054–
2070.
Bonfils, C., B.D. Santer, D.W. Pierce, H.G. Hidalgo, G. Bala, T. Das, T.P. Barnett, D.R. Cayan,
C. Doutriaux, A.W. Wood, A. Mirin, and T. Nozawa 2008. Detection and attribution of
temperature changes in the mountainous western United States. Journal of Climate 23: 6404-
6424.
Box, E.O. 1996. Plant functional types and climate at the global scale. Journal of Vegetation
Science 7:309-320.
Burnham, K.P. and D.R. Anderson. 2002. A Practical Information-Theoretic Approach. 2nd
edition. Springer-Verlag, New York, NY.
57
Clow, D.W. 2010. Changes in the timing of snowmelt and streamflow in Colorado: a response to
recent warming. Journal of Climate 23: 2293-2306.
Crist, L. 2010. Endangered and threatened wildlife and plants; 12-month finding on a petition to
list the American pika as threatened or endangered. FWS-R6-ES-2009-0021. U. S. Fish and
Wildlife Service, Utah Ecological Services Field Office, West Valley City, UT, USA.
Davis, M.B., R.G. Shaw, and J.R. Etterson. 2005. Evolutionary responses to changing climate.
Ecology 86: 1704-1714.
Dearing, M.D. 1996. Disparate determinants of summer and winter diet selection of a generalist
herbivore, Ochotona princeps. Oecologia 108:467-478.
Dearing, M.D. 1997. The manipulation of plant toxins by a food-hoarding herbivore, Ochotona
princeps. Ecology 78:774-781.
Erb, L.P., R. Guralnick, and C. Ray. In prep. The influence of climate and vegetation on multi-
year occupancy of the American pika. Global Change Biology.
Erb, L.P., C. Ray, and R. Guralnick. 2011. On the generality of a climate-mediated shift in the
distribution of the American pika (Ochotona princeps). Ecology 92:1730-1735.
Erb L.P., C. Ray, R. Guralnick. In press. Determinants of pika population density versus
occupancy in the Southern Rocky Mountains. Ecological Applications.
Gaston, K.J., T.M. Blackburn, J.J.D. Greenwood, R.D. Gregory, R.M. Quinn, and J.H. Lawton.
2000. Abundance-occupancy relationships. Journal of Applied Ecology 37:39-59.
Grayson, D.K. 2005. A brief history of Great Basin pikas. Journal of Biogeography 32: 2103-
2111.
Guralnick, R.P., L.P. Erb, and C. Ray. 2011. Mammalian distributional responses to climatic
changes: A review and research prospectus. In J. Ballant and E. Beever (Eds.), Ecological
58
Consequences of Climate Change: Mechanisms, Conservation, and Management. Taylor and
Francis, Boca Raton, FL.
Hafner, D.J. 1993. North American Pika (Ochotona princeps) as a Late Quaternary
Biogeographic Indicator Species. Quaternary Research 39: 373-380.
Hafner, D.J. 1994. Pikas and permafrost: Post-Wisconsin Historical Zoogeography of Ochotona
in the Southern Rocky Mountains, U.S.A. Arctic and Alpine Research 26: 375-382.
Hayes A. and N. Huntly. 2005. Effects of wind on the behavior and call transmission of pikas
(Ochotona princeps). Journal of Mammalogy 86:974-981
Hester, A.J., I.J. Gordon, G.J. Baillie, and E. Tappin. 1999. Foraging behaviour of sheep and red
deer within natural heather grass mosaics. Journal of Applied Ecology 36: 133-146.
Hickling, R., D.B. Roy, J.K. Hill, R. Fox, C.D. Thomas. 2006. The distributions of a wide range
of taxonomic groups are expanding polewards. Global Change Biology 12:450-455.
Holt, A.R., K.J. Gaston, and F. He. 2002. Occupancy-abundance relationships and spatial
distribution. Basic and Applied Ecology 3:1-13.
Hughes, L. 2003. Climate change and Australia: Trends, projections and impacts. Austral
Ecology 28: 423-443.
Huntly, N. J., A. T. Smith, and B. L. Ivins. 1986. Foraging behavior of the pika (Ochotona
princeps), with comparisons of grazing versus haying. Journal of Mammalogy 67:139-148.
Intergovernmental Panel on Climate Change (IPCC). 2007. Climate change 2007: The Physical
Science Basis. Cambridge University Press, New York, NY, USA.
Jeffress, M.R., T.J. Rodhouse, C. Ray, S. Wolff, C.W. Epps. 2013. The idiosyncrasies of place:
geographic variation in the climate-distribution relationships of the American pika.
Ecological Applications 23: 864-878.
59
Jonzen, N, J.R. Rhodes, H. P. Possingham. 2005. Trend detection in source-sink systems: when
should sink habitats be monitored? Ecological Applications 15: 326-334.
Kawamichi, T. 1982. Factors affecting sizes of home range and territory in pikas. Japanese
Journal of Ecology 32:21-27.
Knowles, N., M.D. Dettinger, and D.R. Cayan. 2006. Trends in snowfall versus rainfall in the
western United States. Journal of Climate 19: 4545-4559.
Krajick, K. 2004. All downhill from here. Science 303: 1600-1602.
Lenoir, J., J.C. Gegout, P.A. Marquet, P. de Ruffray, and H. Brisse. 2008. A significant upward
shift in plant species optimum elevation during the 20th century. Science 320:1768–1771.
Li, Q., R. Sun, C. Huang, Z. Wang, X. Liu, J. Hon, J. Liu, L. Cai, N. Li, S. Zhang, and Y. Wang.
2001. Cold adaptive thermogenesis in small mammals from different geographical zones of
China. Comparative Biochemistry & Physiology Part A, Molecular & Integrative Physiology
129: 949–961.
MacArthur, R.A. and L.C.H. Wang. 1974. Behavioral thermoregulation in the pika Ochotona
princeps: a field study using radio telemetry. Canadian Journal of Zoology 52:353-358.
MacKenzie, D., J.D. Nichols, G.B. Lachman, S. Droege, J.A. Royle, and C.A. Langtimm. 2002.
Estimating site occupancy when detection probabilities are less than one. Ecology 83, 2248-
2255.
Martinuzzi, S., L.A. Vierling, W.A. Gould, M.J. Falkowski, J.S. Evans, A.T. Hudak, K.T.
Vierling. 2009. Mapping snags and understory shrubs for a LiDAR-based assessment of
wildlife habitat suitability. Remote Sensing Environment 113: 2533-2546.
Millar, C.I. and R.D. Westfall. 2010. Distribution and climatic relationships of American pika
(Ochotona princeps) in the Sierra Nevada and Western Great Basin, USA; Periglacial
60
landforms as refugia in warming climates. Arctic, Antarctic, and Alpine Research 42:76-88.
Moilanen A., A.T. Smith, I. Hanski. 1998. Long-term dynamics in a metapopulation of the
American pika. The American Naturalist 152: 530-542.
Moritz, C., J.L. Patton, C.J. Conroy, J.L. Parra, G.C. White, and S.R. Beissinger. 2008. Impact of
a century of climate change on small-mammal communities in Yosemite National Park,
USA. Science 322: 261-264.
Morrison, S.F. and D.S. Hik. 2007. Demographic analysis of a declining pika Ochotona collaris
population: linking survival to broad-scale climate patterns via spring snowmelt patterns.
Journal of Animal Ecology 76: 899-907.
Morrison, S.F. and D.S. Hik. 2008. Discrimination of intra- and inter-specific forage quality by
collared pikas (Ochotona collaris). Canadian Journal of Zoology 86: 456-461.
Mote, P.W., A.F. Hamlet, M.P. Clark, and D.P. Lettenmaier. 2005. Declining mountain
snowpack in western North America. Bulletin of the American Meterological Society 86: 39-
50.
Moyer-Horner, L., M.M. Smith, and J. Belt. 2012. Citizen science and observer variability
during American pika surveys. The Journal of Wildlife Management 76:1472-1479.
Nagelkerke, N.J.D. 1992. Maximum likelihood estimation of functional relationships. Springer-
Verlag, Berlin, Germany and New York, USA.
Parmesan, C. 2006. Ecological and evolutionary responses to recent climate change. Annual
Review of Ecology Evolution and Systematics 37:637-669.
Pederson, G. T., J. L. Betancourt, G. J. McCabe. 2013. Regional patterns and proximal causes of
the recent snowpack decline in the Rocky Mountains, US. Geophysical Research Letters 40:
1811-1816.
61
Pimm, S. L., H. L. Jones, and J. Diamond. 1988. On the risk of extinction. The American
Naturalist 132: 757-785.
PRISM Group, Oregon State University, http://www.prismclimate.org, created 7 Dec 2007.
Pulliam, H.R. 1988. Sources, sinks, and population regulation. American Naturalist 132:652-661.
R Core Team. 2012. R: A language and environment for statistical computing. R Foundation for
Statistical Computing. Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-
project.org/.
Rangwala I., J.R. Miller. 2011. Long-term temperature trends in the San Juan mountains. In:
Blair R, Bracksieck G (eds) Eastern San Juan Mountains: Their geology, ecology, and human
history. University Press of Colorado, Boulder.
Rangwala, I, J. Barsugli, K. Cozzetto, J. Neff, J. Prairie. 2012. Mid-21st century projections in
temperature extremes in the southern Colorado Rocky Mountains from regional climate
models. Climate Dynamics 39: 1823-1840.
Ray, A., J. Barsugli, K. Averyt, K. Wolter, M. Hoerling, N. Doesken, B. Udall, and R.S. Webb
2008. Climate change in Colorado. A synthesis to support water resources management and
adaptation. A report for the Colorado Water Conservation Board, edited, Colorado Water
Conservation Board, Denver. Western Water Assessment and University of Colorado,
Boulder.
Ray, C., E.A. Beever, and S.R. Loarie. 2012. Retreat of the American pika: up the mountain or
into the void? In: Conserving wildlife populations in a changing climate. University of
Chicago Press.
Rhodes, J. R. and N. Jonzen. 2011. Monitoring temporal trends in spatially structured
populations: how should sampling effort be allocated between space and time? Ecography
62
34: 1040-1048.
Rodhouse, T.J., E.A. Beever, L.K. Garrett, K.M. Irvine, M.R. Jeffress, M. Munts, and C. Ray.
2010. Distribution of American pikas in a low-elevation lava landscape: conservation
implications from the range periphery. Journal of Mammalogy 91: 1287-1299.
Salek, M., J. Kreisinger, F. Sedlacek, and T. Albrecht. 2010. Do prey densities determine
preferences of mammalian predators for habitat edges in an agricultural landscape? Landscape
and Urban Planning 98: 86-91.
Sheafor, B.A. 2003. Metabolic enzyme activities across an altitudinal gradient: an examination
of pikas (genus Ochotona). Journal of Experimental Biology 206: 1241-1249.
Sibly, R.M. and J. Hone. 2002. Population growth rate and its determinants: an overview.
Philosophical Transactions of the Royal Society of London Series B-Biological Sciences
357: 1153-1170.
Smith, A.T. 1974a. The distribution and dispersal of pikas: consequences of insular population
structure. Ecology 55: 1112-1119.
Smith, A.T. 1974b. The distribution and dispersal of pikas: Influences of behavior and climate.
Ecology 55: 1368-1376.
Smith, A.T. and M.E. Gilpin. 1997. Spatially correlated dynamics in a pika metapopulation. In
Metapopulation Biology. Ed. I.A. Hanski and M.E. Gilpin: 407-428. Academic Press, San
Diego, CA.
Smith, A.T. and M.L. Weston. 1990. Ochotona princeps. Mammalian Species 352: 1-8.
Smith, J.A. and L.P. Erb. 2013. Patterns of selective caching behavior of a generalist herbivore,
the American pika (Ochotona princeps). Arctic, Antarctic, and Alpine Research 45: 396-403.
Tapper, S.C. 1973. The spatial organization of pikas (Ochotona), and its effect on population
63
recruitment. Ph.D. dissertation, University of Alberta, Edmonton, 154 pp.
Thomas, C.D., A. Cameron, R.E. Green, M. Bakkenes, L.J. Beaumont, Y.C. Collingham, B.F.N.
Erasmus, M.F. de Siqueira, A. Grainger, L. Hannah, L. Hughes, B. Huntley, A.S. van
Jaarsveld, G.F. Midgley, L. Miles, M.A. Ortega-Huerta, A.T. Peterson, O.L. Phillips, S.E.
Williams. 2004. Extinction risk from climate change. Nature 427:145-148.
Venables, W.N. and Ripley, B.D. 2002. Modern Applied Statistics with S. 4th Ed. Springer, NY.
Venn, S., C. Pickering, K. Green. 2012. Short-term variation in species richness across an
altitudinal gradient of alpine summits. Biodiversity and Conservation 21:3157-3186.
Waltari, E. and R. Guralnick. 2009. Ecological niche modeling of Great Basin montane
mammals: examining past and present connectivity of species across basin and ranges.
Journal of Biogeography 36: 148-161.
Wilkening, J.L., C. Ray, E.A. Beever, and P.F. Brussard. 2011. Modeling contemporary range
retraction in Great Basin pikas (Ochotona princeps) using data on microclimate and
microhabitat. Quaternary International 235:77–88.
64
Appendix I
Climate histograms. Panels A and B represent 1908-2007 mean annual precipitation values and mean summer (June-August)
maximum temperatures, respectively, for the 69 sites analyzed for 2008 pika occupancy in Chapter 1. A subset of 19 of these sites was
selected for further analysis of mean occupancy 2008-2011; panels C and D represent 1908-2007 mean annual precipitation and mean
summer maximum temperatures, respectively, for this subset of 19 sites.
A B 65
C D
Appendix II
Predictor coefficients for potential predictors of mean pika occupancy in the Southern Rocky Mountains. Predictors (described in
Appendix V) were standardized as Zi = (X i − μ(X i ))/ σ(X i ) , allowing for intuitive interpretation of model coefficients: each
coefficient (βi) measures the effect of Xi in units of standard deviation in Xi. 66
Appendix III
Comparison of climatic predictors among 18 study sites and the 69 historically occupied sites from which they were selected via
random sampling stratified by latitude, longitude, and elevation. 67
Appendix IV
Pika latrine density (latrines/200 m2) at 18 sites in the Southern Rocky Mountains.
Latrine Density Site Name (latrines/200 m2)
Nambe Lake, NM 4
Valles Caldera, NM 9
Kennebec Pass, CO 3
Wolf Creek Pass, CO 8
Del Norte Peak, CO 5
Crystal Lake, CO 6
Cochetopa Dome, CO 0
Mount Gothic, CO 13
Grand Mesa, CO 3
Halfmoon Creek, CO 5
Papoose Basin, CO 16
Pagoda Peak, CO 57
Grand Lake, CO 0
Bighorn Peak, CO 9
Trap Lake, CO 3
Bridger Peak, WY 3
Silver Lake, WY 35
Highway 130, WY 1
68
Appendix V
Models predicting pika density with ΔAIC<2, along with the null model. Key to predictor codes provided below model table.
2 Model AIC ΔAIC R N
Neg10 + Area 119.99 0 0.47
ForbSpp + Area 120.24 0.25 0.46
GF + Neg10 + Area 120.27 0.28 0.56
ForbCover + Area 120.36 0.37 0.45
GF + Area 120.37 0.38 0.45
MeanPrecip + GF + Area 120.42 0.43 0.56
ForbCover + Neg10 + Area 120.46 0.47 0.56
ForbSpp + Area + Rock.Veg 120.47 0.48 0.56
MeanPrecip + ForbCover + Area 120.95 0.96 0.55
MaxMean + GF + Area 121.05 1.06 0.54
ForbSpp + GF + Area 121.35 1.36 0.54
GF + JJATemp + Area 121.41 1.42 0.54
Area 121.42 1.43 0.30
ForbSpp + Neg10 + Area 121.71 1.72 0.53
…
Null (y~1) 124.97 4.98
69
Predictor Code Predictor Description
Area Patch area
ForbCover Percent of vegetation consisting of forbs
ForbSpp Forb species richness
GF Gram:forb
JJATemp Sub-surface summer average temperature
MaxMean Surface summer maximum temperature
MeanPrecip Mean annual precipitation
Neg10 Proportion of recorded sub-surface temperatures below -10° C
Rock.Veg % rock cover multiplied by % vegetation cover
70