Citation
Whitebark pine at treeline in the greater Yellowstone ecosystem : prevalence, facilitation, and biophysical characteristics of leeward microsites

Material Information

Title:
Whitebark pine at treeline in the greater Yellowstone ecosystem : prevalence, facilitation, and biophysical characteristics of leeward microsites
Creator:
Wagner, Aaron Christian
Place of Publication:
Denver, CO
Publisher:
University of Colorado Denver
Publication Date:
Language:
English

Thesis/Dissertation Information

Degree:
Master's ( Master of science)
Degree Grantor:
University of Colorado Denver
Degree Divisions:
Department of Integrative Biology, CU Denver
Degree Disciplines:
Biology
Committee Chair:
Tomback, Diana
Committee Members:
Das, Raibatak
Moreno-Sanchez, Rafael

Notes

Abstract:
Facilitative interactions are particularly important in the climatically stressful alpine treeline ecotone (ATE), the transition zone from closed canopy forest to alpine tundra. Within some ATEs on the harsh Eastern slope in the Rocky Mountains, whitebark pine (Pinus albicaulis)—a foundation and keystone species—plays a central role in tree island development by facilitating the survival and growth of leeward conifers more frequently than other species. However, the structure, composition, and dynamics of ATE formation are not well known for some geographic regions of western North America, including the Greater Yellowstone Ecosystem (GYE). Although facilitative interactions are considered important to treeline community development, comparisons of biophysical attributes among the leeward microsites of nurse objects and trees are also lacking. We conducted an exploratory study of four treeline communities in the GYE to investigate (1) the prevalence of whitebark pine, (2) the proportional occurrence of whitebark pine as a tree island initiator, (3) whether solitary whitebark pine abundance predicts its prevalence as a tree island initiator, (4) whether whitebark pine better ameliorates local biophysical conditions relative to other plants or objects, (5) the relationship between common nurse objects and solitary/tree island initiator establishment, and (6) the presence and severity of Cronartium ribicola, an invasive pathogen that attacks and kills whitebark pine, in these communities. We found that whitebark pine is the most prevalent solitary conifer and tree island component, and initiates tree island formation in direct proportion to its abundance as a solitary tree. We also found that whitebark pine leeward microsites did not consistently experience the most moderate microclimate compared to spruce, fir, rock and unprotected leeward microsites; differences in microclimatic conditions by microsite type depended substantially on the general climatic conditions observed at each study area. Among our study areas, C. ribicola infection rates ranged from less than 1% to 18%. Our findings suggest that whitebark pine is an important treeline species in the Greater Yellowstone Ecosystem. Losses of whitebark pine from infection by C. ribicola will lead to proportional changes in the composition and structure of ATE communities in the GYE, including decline in tree island formation on the landscape.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
Copyright Aaron Christian Wagner. Permission granted to University of Colorado Denver to digitize and display this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.

Downloads

This item is only available as the following downloads:


Full Text

PAGE 1

WHITEBARK PINE AT TREELINE IN THE GREATER YELLOWSTONE ECOSYSTEM: PREVALENCE, FACILITATION, AND BIOPHYSICAL CHARACTERISTICS OF LEEWARD MICROSITES by AARON CHRISTIAN WAGNER B.S., University of Colorado Denver, 2013 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Master of Science Biology Program 2017

PAGE 2

ii © 2017 AARON CHRISTIAN WAGNER ALL RIGHTS RESERVED

PAGE 3

iii This thesis for the Master of Science degree by Aaron Christian Wagner has been approved for the Biolog y Program by Diana Tomback, Chair Raibatak Das Raf ael Moreno Sanchez Date: May 13 , 2017

PAGE 4

iv Wagner, Aaron C hristian . (M.S. Biology Program ) WHITEBARK PINE AT TREELINE IN T HE GREATER YELLOWSTONE ECOSYSTEM: PREVALENCE, FACILITATION, AND BIOPHYSICAL CHARACTERISTICS OF LEEWARD MICROSITES Thesi s directed by Professor Diana Tomback Abstract Facilitative interactions are particularly important in the climatically stressful alpine treeline ecotone (ATE), the transition zone from closed canopy forest to alpine tundra. Within some ATEs on the harsh Eastern slope in the Rocky Mountains, whitebark pine ( Pinus albicaulis ) Ñ a foundation and keystone species Ñ plays a central role in tree isl and development by facilitating the survival and growth of leeward conifers more frequently than other species. However, the structure, composition, and dynamics of ATE formation are not well known for some geographic regions of western North America, incl uding the Greater Yellowstone Ecosystem (GYE). Although facilitative interactions are considered important to treeline community development, comparisons of biophysical attributes among the leeward microsites of nurse objects and trees are also lacking. We conducted an exploratory study of four treeline communities in the GYE to investigate (1) the prevalence of whitebark pine, (2) the proportional occurrence of whitebark pine as a tree island initiator, (3) whether solitary whitebark pine abundance predict s its prevalence as a tree island initiator, (4) whether whitebark pine better ameliorates local biophysical conditions relative to other plants or objects, (5) the relationship between common nurse objects and solitary/tree island initiator establishment , and (6) the presence and severity of Cronartium ribicola , an invasive pathogen that attacks and kills

PAGE 5

v whitebark pine, in these communities. We found that whitebark pine is the most prevalent solitary conifer and tree island component, and initiates tree island formation in direct proportion to its abundance as a solitary tree. We also found that whitebark pine leeward microsites did not consistently experience the most moderate microclimate compared to spruce, fir, rock and unprotected leeward microsites; differences in microclimatic conditions by microsite type depended substantially on the general climatic conditions observed at each study area. Among our study areas, C. ribicola infection rates ranged from less than 1% to 18%. Our findings suggest that whitebark pine is an important treeline species in the Greater Yellowstone Ecosystem. Losses of whitebark pine from infection by C. ribicola will lead to proportional changes in the composition and structure of ATE communities in the GYE, including decline in tree island formation on the landscape. The form and content of this abstract are approved. I recommend its publication Approved: Diana F. Tomback

PAGE 6

vi DEDICATION To my wife, Elizabeth Rose Pansin g, whose intelligence , grace, and adventurous spirit rou s e me to follow my heart . To my father , Edward James Wagner, whose passion for grand questions and profound contemplation always inspires me to think boldly. For my mother, Susan Lynn Knoll, whose boundless energy has always been contagious, and whose love and empathy make her the matriarch of our family. To my step father, Gary Evertt Knoll, whose patience and fortitude give me balance . Finally, to the giants who came before me, whose shoulders never tire. "If God were to hold all Truth concealed in his right hand, and in his left only the steady and diligent drive for Truth, albeit with the proviso that I would always and forever err in the process, and to offer me the choice, I would with all humility take the left hand." Ð Gotthold Ephraim Lessing (Anti Goese, 1778)

PAGE 7

vii ACKNOWLEDGEMENTS This product would not have been possible without Dr. Diana Tomback. Her determination , support, and guidance have been a sustained source of inspiration. My respect and appreciation for D r. Raib atak Das, whose statistical knowledge was invaluable to this project. My utmost appreciation for Dr. Rafael Moreno Sanchez, whose positivity was contagious, and whose background helped place this work in context. My sincerest gratitude to Dr. Michael Wunder for assisting in the deve lopment of my quantitative skills, and for being a person I aspire to emulate. I am profoundly grateful to Bill Lee, Jimmy Smail, Tanner Shuler, Daniel Byron, Scott Berkenfield, Ellen Jungck, and Carl Crittenden of the Unite d States Forest Service (USFS) f or assisting with logistical planning and execution . A special t hanks to Kent Houston (USFS) , whose professional support made this project possible . This project would not have been nearly as enjoyable without the help and humor of my friend, T. Ryan McCar ley. Finally, my sincerest respect and appreciation to my friends and colleagues in the Department of Integrative Biology at the University of Colorado Denver; their tenacity and desire to make this world a better place has profoundly influenced this proje ct , and my life .

PAGE 8

viii TABLE OF CONTENTS CHAPTER I. ! ECOLOGICAL FOUNDATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Background. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 The Alpine Treeline E coton e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 F acilitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Treeline dy namics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Whitebark Pine: A Keystone and Foundation S pecies . . . . . . . . . . . . . . . . . 6 Threats . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Mountain pine beetl e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Fire exclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 White pine blister rust . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Tables and F igures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 II. ! WHITEBARK PINE AT TREELINE IN THE GREATER YELLOWSTONE ECOSYSTEM: PREVALENCE, FACILITATION, AND POTENTIAL DISRUPTION BY AN I NTRODUCED PATHOGEN . . . . . . . . . . . . . . . . . . . . 1 2 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 Study areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 8 Community a ss essments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Mesoclimate assessments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

PAGE 9

ix Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Proportional abundance parameter estimates . . . . . . . . . . . . . . . . . . 21 Mean differences in proportional abundance parameter estimates . . 22 Growing seaso n temperature and precipitation parameter estimates . 23 Differences in mean growing season parameter estimates . . . . . . . . 24 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 4 Community structure and composition . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 Tibbs Butte . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Paintbrush Divide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Hurricane Pass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 Mutiny Ridge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Tree island development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Facilitation by nurse objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Tibbs Butte . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Paintbrush Divide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Hurricane Pass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 Mutiny Ridge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 Blister rust assessments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 Mesoclimate assessments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 Whitebark pine and tree island development in the ATE . . . . . . . . . . . . . . 32 The relationship of solitary trees to tree island initiators . . . . . . . . . . . . . 34 The impact of C. ribicola invasion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

PAGE 10

x Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 Tables and f igures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 III. ! FACILITATION IN TREELINE WHITEBARK PINE COMMUNITIES: BIOPHYSICAL DIFFERENCES AMONG LEEWARD MICRO SITES D EPEND ON LARGER SCALE CLIMATE INDICES . . . . . . . . . . . . . . . . . . . 63 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 Study areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 Study design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 M icroclimatic assessments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 Model selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 Bootstrapped linear regressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 Parameter estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 Assessi ng bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 Parameter estimates for mean differences . . . . . . . . . . . . . . . . . . . . 76 Parameter estimates for microclimatic variability . . . . . . . . . . . . . . . 77 Microclimatic variability mean difference parameter estimates. . . . 78 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 Air temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 Percent relative humidity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

PAGE 11

xi Soil temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 Soil moisture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 Wind and gust speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 Tables and f igures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 8 Appendix B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

PAGE 12

xii LIST OF TABLES TABLE 2 .1. Study areas across the Greater Yellowston e Ecosystem (GYE) . . . . . . . . . . . . . . . . . 38 2.2 . Descriptive stati stics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 A.1. Raw conifer species count data for all study areas . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 A.2. Raw conifer species count and proportional abundance data for Tibbs Butte . . . . . 50 A.3. Raw conifer species count and proportional abundance data for Paintbrush Divide. 51 A.4. Raw conifer species count and proportional abundance data for Hurricane Pass . . . 53 A.5. Raw conifer sp ecies count and proportional abundance data for Mutiny Ridge. . . . .55 A.6. Raw nurse object count data by plot for all study areas. . . . . . . . . . . . . . . . . . . . . . . 57 A.7. Raw nurse object proportional abundance data by plot for all study are as. . . . . . . . 59 A.8. Raw white pine blister rust count and proportional abundance data by plot. . . . . . . 61 B . 1. Air temperature bootstrapped parameter estimates. . . . . . . . . . . . . . . . . . . . . . . . . . 107 B.2. Percent relative humidity bootstrapped parameter estimates. . . . . . . . . . . . . . . . . . .108 B.3. Soil temperature bootstrapped parameter estimates. . . . . . . . . . . . . . . . . . . . . . . . . 109 B.4. Soil moisture bootstrapped parameter estimates. . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 B.5. Wind and gust speed bootstrapped parameter estimates. . . . . . . . . . . . . . . . . . . . . 111 B.6. Air temperature bootstrapped mean difference parameter estim ates. . . . . . . . . . . . .112 B.7 Percent relativ e humidity bootstrapped mean difference p arameter estimates. . . . . .113 B.8 . Soil temperature bootstrapped mean difference parameter esti mates. . . . . . . . . . . . 114 B.9 . Soil moisture bootstrapped mean difference parameter estimates. . . . . . . . . . . . . . .115 B.10 . Wind and gust speed bootstrapped mean difference parameter es timates. . . . . . . . 116 B.11 . Air temperature standard deviation bootstrapped parameter estimates. . . . . . . . . .117

PAGE 13

xiii B.12. Percent relative humidity standard deviation boots trapped parameter estimates. 118 B.13. Soil temperature standard deviation bootstrapped paramet er estimates. . . . . . . . .119 B.14. Soil moisture standard deviation bootstrapped parameter esti mates. . . . . . . . . . . . 120 B.15. Wind and gust spee d standard deviation bootstrapped p arameter estimates. . . . . .121 B.16. Air temperature SD bootstrapped mean difference parame ter e stimates. . . . . . . . 122 B.17. Percent relative humidity SD bootstrapped mean dif ference parameter estimates. 123 B.18. Soil temperature bootstrapped mean difference parameter e stimates. . . . . . . . . . .124 B.19. Soil moisture bootstrapped mean difference parameter estimat es. . . . . . . . . . . . . .125 B.20. Wind and gust speed bootstrapped mean difference parameter es timates. . . . . . . .126

PAGE 14

xiv LIST OF FIGURES FIGURE 1 .1. White pine blister rust infection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.1. Facilitation in action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 2.2. Study areas across the Greater Yellowstone Ecosystem . . . . . . . . . . . . . . . . . . . . . . . 41 2.3 a . Parameter estimates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 2.3b. Parameter estimates for mean differences. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 2.4. Mean difference parameter estimates: solitary minus initiator conifers . . . . . . . . . . 44 2.5a. Parameter estimates for nurse objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 2.5b. Mean difference parameter estimates for nurse objects. . . . . . . . . . . . . . . . . . . . . . 46 2.6a. Parameter estimates for climate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .47 2.6b. Difference in mean parameter estimates for climate. . . . . . . . . . . . . . . . . . . . . . . . . 48 3.1. Research study areas in the Greater Yellowstone Ecosystem. . . . . . . . . . . . . . . . . . .88 3.2. Microclimate weather station. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .89 3.3a. Time series plots for Tibbs Butte. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 3.3 b. Time series plots for Mutiny Ridge. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 3.4a. Best model fits for each response variable statistic at Tibbs Butte. . . . . . . . . . . . . . . 92 3.4b. Best model fits for each response variable statistic at Mutiny Ridge. . . . . . . . . . . . 93 3 .5a. Parameter estimates for mean differences in microclimate at Tibbs Butte . . . . . . . . . 94 3.5b. Parameter estimates for mean differences in microclimate at Mutiny R idge . . . . . . 95 3.6a. Microclimatic variability mean difference parameter e stimates for Tibbs Butte. . . 96 3.6b. Microclimatic variability mean difference parameter estimates for Mutiny Ridge. .97 B . 1. Nurse object widths, lengths, and heights. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

PAGE 15

xv B . 2. Density plots for raw biophysical variables at Tibbs Butte. . . . . . . . . . . . . . . . . . . . .99 B . 3. Density plots for raw biophysical variables at Mutiny Ridge. . . . . . . . . . . . . . . . . . .100 B . 3a. Parameter estimates for microclimate at Tibbs Butte . . . . . . . . . . . . . . . . . . . . . . . 101 B . 3b. Parameter estimates for microclimate at Mutiny Ridge. . . . . . . . . . . . . . . . . . . . . . 102 B . 4a. Parameter estimate s for bias es: Tibbs Butte. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .103 B.4b. Parameter estimates for biases: Mutiny Ridge. . . . . . . . . . . . . . . . . . . . . . . . . . . . .104 B.5a. Parameter estimates for microclimatic variability at Tibbs Butte . . . . . . . . . . . . . . .105 B.5b. Parameter estimates for microclimatic variability at Mutiny R idge. . . . . . . . . . . . .106

PAGE 16

xvi LIST OF ABBREVIATIONS 1. ATE Alpine Treeline Ecotone Ñ the region of transition from closed canopy upper su balpine forest to treeless alpine tundra. 2. GYE Greater Yellowstone Ecosystem Ñ a temperate zone area in the northern Rocky Mountains that encapsulates swaths of southwestern Montana, eastern Idaho, and northwestern Wyoming. It is defined by a large commun ity of interacting organisms. 3 . WPBR White Pine Blister Rust Ñ a disease caused by the invasive fungal pathogen Cronartium ribicola . 3 . HDI Highest Density Interval Ñ the Bayesian analog to frequentist confidence intervals. HDI boundaries are fixed, and the estimated parameter is a random variable. This is in contrast to frequentist confidence intervals, which have random variable boundaries and a fixed parameter value.

PAGE 17

1 Chapter I ECOLOGICAL FOUNDATIONS Overview Landscape pathology examines how pat hogens shape the distribution and dynamics of plant communities on a local scale (Holdenrieder et al., 2004). Across the range of five needle white pines, infection by white pine blister rust Ñ a disease caused by the exotic fungal pathogen, Cronartium ribic ola Ñ has caused extensive damage and mortality (Ellison et al., 2005; Schwandt et al., 2010; Tomback & Achuff, 2010; Bockino and Tinker, 2012). One of the most widely distributed of the white pines, whitebark pine ( Pinus albicaulis Engelm.) has experienced rapid decline in the western U.S. and Canada due to mortality from: 1) blister rust; 2) mountain pine beetle outbreaks; and 3) fire exclusion. A keystone and foundation species, whitebark pine fosters biodiversity and ecosystem stability (Tomback et al., 2 001; Ellison et al., 2005; Tomback & Achuff, 2010; Tomback et al., 2011). This ecological role results from whitebark pine's seed dispersal by Clark's nutcrackers ( Nucifraga columbiana ), hardy seedlings, and high tolerance for cold and windy sites (Tomback et. al., 2011). Under harsh conditions, whitebark pine may participate in positive interactions with neighboring plants (Callaway, 1998), in which leeward conifers may experience a moderated microclimate in the lee of whitebark pine, at no cos t to the wi ndward participant Ñ an interaction referred to as facilitation (Stachowicz, 2001; Butler et al., 2007). This interaction contributes to tree island initiation (Resler and Tomback 2008; Tomback et al. 2014), allowing treeline communities to develop under har sh environments. Because whitebark pine's ecosystem function may be diminishing due to

PAGE 18

2 mortality from white pine blister rust, community structure, whitebark pine's ecological role, and climatic variables must be used to assess treeline dynamics. Climate change is expected to alter forest tree distributions both latitudinally and in elevation. Depending on the species composition and landscape structure, some alpine treeline ecotones (ATEs) will have a greater ability to respond to climatic change, while others may be less equipped (Harsch et al., 2009). Upward movement of whitebark pine communities is predicted as a response to climate change (e.g., Schrag et al. 2007). However, damage and mortality to newly established recruits due to infection by white pine blister rust may counter the effects of warming in communities (Tomback and Resler, 2007). At this time, the combined effects of pathogen disturbance and increasing temperature on ATEs are largely unknown, and we need to examine the interactions among disease, vegetation, and climate change. Research previously conducted by: 1) D . F. Tomback and her students Ñ University of Colorado Denver; 2) L.M. Resler and her student Emily Smith McKenna Ñ Virginia Polytechnic Institute and State University; and 3) G. P . Malanson Ñ University of Iowa Ñ have examined the proportional abundance of whitebark pine, its role in tree island development, and the incidence of blister rust in selected areas across the Central and Northern Rocky Mountains, from 45¡ to about 54¡ latitu de. This work shows that geographic variation exists in the importance of facilitation by whitebark pine in tree island development. Whitebark pine has been found to be a majority tree island initiator and dominant conifer in the ATE at several study loc ations in the Rocky Mountains Ñ Stanley Glacier, Kootenay National Park, British Columbia; Lee Ridge and Divide Mountain, Glacier

PAGE 19

3 National Park, Montana; and Wyoming Creek, Custer National Forest, Wyoming (Resler and Tom back 2008, Tomback et al. 2014) . Howev er, whitebark pine is not a majority tree island initi ator in other ATE communities Ñ Tibbs Butte, Shos hone National Forest, Wyoming; Gibbon Pass, Banff National Park, Alberta; and Willmore Wilderness, Alberta Ñ although it is abundant in these areas (Tomback et al. 2014, Tomback and Resler, in preparation). The abundance and ecological role of whitebark pine in facilitation of tree islands in the ATE remains unknown in its southeastern most range (i.e., 42¡ latitude), the southern Wind River Mountains. Specifi c questions for this region are: What is the proporitonal abundance of whitebark pine? What is the incidence of blister rust? What is the extent of whitebark pine's ecological role? Do the microclimatic conditions differ from documented study locations in whitebark pine's central and northernmost range, and if so, how? Does whitebark pine provide better microsite protection than other conifers or unprotected microsites, and if so, which conditions? What do we predict in terms of the future movement of the ATE in the presence of blister rust and climate change? Background The Alpine Treeline Ecotone The alpine treeline ecotone (ATE) is the transition from closed canopy forest to treeless alpine tundra ( e.g., Holtmeier, 2009). Compared to subalpine forest, t he environment within the ATE is characterized by low temperatures and high winds, where species experience frozen soils, ele vated frost exposure , increased solar radiation, high rates of evapotranspiration, and amplified nutrient scarcity ( Kšrner and Paul sen, 2004, Holtmeier, 2009). In this zone, upright growth forms of trees are unable to develop due to

PAGE 20

4 wind blasting, and tissue formation is limited by low temperatures (Kšrner, 2008). When tree anatomy and physiology are impeded significantly by these cli matic extremes, dwarfed morphologies Ñ kn own as krummholz growth forms Ñ result. These stunted forms respond to slightly higher temperatures close to the ground (Kšrner & Paulsen 2004; Hoch & Kšrner, 2009), and often grow in the lee of biotic or abiotic struct ures Ñ termed nurse objects Ñ that provide shelter from the wind ( e.g., Resler and Tomback 2008 , Castro et al. 2011 ). As elevations increase, trees reach their upper distributional limit due to stress from these harsh conditions, giving way to herbaceous alpin e species (Malanson et al., 2011). Facilitation Within many ecosystems that exist under climatically stressful conditions, positive associations can occur among neighboring plants in which one individual experiences localized protection from climatic stre ss at no cost to others (Stachowicz, 2001; Butler et al., 2007). As conifers struggle to germinate, establish, and grow in the ATE, these facilitative interactions among neighbors become more important than competition for resources (Callaway et al., 2002) . Competitive interactions among spatially associated individuals at lower elevations are facilitative in this zone and often result in the formation of tree islands Ñ groups of two or more krummholz trees in close proximity (Marr, 1977). These emergent stru ctures occur when a conifer successfully establishes in the open or in the lee of a nurse object such as a rock, shrub, or terrace riser (Resler et al., 2005; Resler, 2006; Resler and Tomback, 2008). Other conifers will then germinate, establish, and grow in close leeward proximity to the initiating conifer, enjoying moderated climatic conditions (e.g. decreased wind speed ; Pyatt et al. 2016 ).

PAGE 21

5 These conifers then ameliorate the microclimate in their lee. This positive feedback of climate moderation and surv ival among neighbors occurs until tree island size is ultimately limited by topography or nutrient scarcity (Resler, 2006). The result is a landscape containing clusters of environmentally dwarfed trees above the continuous forest limit, described as islan d krummholz (Harsch and Bader, 2011). Treeline Dynamics The position and stability of the ATE is controlled by environmental and climatic influences on conifer regeneration, recruitment, and growth. Harsch and Bader (2011) sum up the mechanisms that affec t perform ance as: 1) growth limitation Ñ a result of impaired biosynthesis in response to low growing season temperature, short growing season duration, or nu trient deficiency; 2) dieback Ñ a result of physical damage due to high snow load, frost, frozen soil, excess radiation, and high wi nd; and 3) seedling mortality Ñ a result of physical damage caused by high snow load, frost, frozen soil, excess radiation, heat stress, high temperatures, and low precipitation. With all t hese environmental challenges Ñ coupled w ith geomorp hic barriers to establishment Ñ trees struggle to survive at the highest elevations. The prevailing theory is that the elevational boundaries of the ATE are governed by climatic conditi ons during the growing season Ñ July through September (Kšrner & Paulsen, 2004; Hoch & Kšrner, 2009; Harsch & Bader, 2011). Therefore, moderation of the environmental and climatic va riables of the growing season Ñ temporary or long term Ñ that limit growth and tree establishment may allow the ATE to move upward in latitude and elevation (Schrag, et al., 2007). In the next 100 years, global ambient temperatures are predicted to increase approximately 2.5¡C (Grace et al, 2002). Because ATE communities are among the first

PAGE 22

6 to display signs of climatic stress, it has been hypot hesized that many of them will serve as first indicators of climate change (Smith et al., 2009). It has already been demonstrated that 52% of all ATEs have advanced in elevation since 1900 A.D. Ñ with only 1% receding (Harsch et al., 2009). Research linking past temperature changes to ATE movement (Lloyd and Graumlich 1997), and ATE position to worldwide isotherms (Kšrner and Paulsen 2004) has provided evidence that the formation and maintenance of these ecosystems is largely temperature controlled. As the m inimum growing season temperature rises over time, growth and seedling establishment are expected to increase, and dieback due to freezing or frost at higher elevations should decrease (Kšrner, 2008). At the current rate of climate change, ATEs are expecte d to move upward in elevation an average of 140 Ñ 700 m (Grace et al., 2002). Bioclimatic envelope models also predict these distributional changes (McKenney et al., 2007; Warwell et al., 2007). At the highest elevations and latitudes, the environmental and climatic variables of the ATE are too extreme for the majority of conifer species (HŠttenschwil er and Smith, 1999). Pyatt et al. (2016 ) demonstrated that micosites in the lee of conifers provide ameliorated climatic conditions compared to rock and unprotec ted microsites. Successful establishment and community structure in harsh conditions is dependent upon facilitation via the formation of tree islands. Whitebark Pine: A keystone and Foundation Species One of the most widely distributed of the white pines in the U.S. and Canada, whitebark pine (Family Pinaceae, Subgenus Strobus .) has been designated as a keystone and foundation species in subalpine ecosystems and the ATE (Tomback et al., 2001). It's a keystone species because it fosters more biodiversity th an its proportional abundance

PAGE 23

7 would suggest ( e.g., Mills et al. 1993, Power et al. 1996, Soule ! et al. 2003 ). Additionally, whitebark pine often defines community structure, creating stable conditions for other species (Tomback et al., 2011). Due to its high tolerance for low temperatures, high wind speed s, and hardy seedlings, whitebark pine is able to play a key functional role as tree island initiator in many ATEs across the western United States (Tomback and Achuff 2010; McKinney and Tomback, 2011). Other important ecosystem services provided by whiteb ark pine include facilitating the growth of other conifers (Callaway, 1998), the regulation of runoff by slowing snow melt (Farnes, 1990), decreasing soil erosion via root systems (Arno and Hammerly, 1984), contributing to early successional regeneration following fires, resulting in community development (Tomback and Linhart, 1990; Tomback et al., 2001), and providing seeds that serve as an essential component of the diets of a number of birds and small mammals (Tomback and Kendall, 2001). The benefits of whitebark pine that pertain to watershed hydrology are also of the utmost importance to society. Farmers, ranchers, and Native Americans Ñ who inhabit many regions downstream of these mountain communities Ñ benefit from the protraction of snowmelt, which norm alizes t he levels of streams and creeks (Farnes 1990). Water levels in reservoirs utilized in urban communities are also maintained by the control of runoff (Smith et al., 2009). Among the many consequences of krummholz growth is a sharp decrease in cone and viable seed production (Grace et al., 2002). Species employing a wind dispersal strategy, such as Engelmann spruce ( Picea engelmannii Parry ex Engelm.) and subalpine fir ( Abies lasiocarpa (Hook.) Nutt.) experience a high probability of sun scorching an d predation of seeds by granivorous birds and mammals in the ATE, reducing the

PAGE 24

8 likelihood of germination. In contrast, whitebark pine is a n ornithochorous conifer Ñ meaning it relies on the assistance of bir ds for the dispersal of seeds Ñ and has evolved a spe cial mutualism with a corvid, the Clark's nutcracker ( Nucifraga columbiana ) (Tomback, D.F., 1982). To utilize whitebark pine seeds for food in the winter and spring, nutcrackers break off the tough scales of the cones with their beaks, place seeds in a sub lingual pouch, and transport them up to 32.6 kilometers (Lorenz et al. 2011 ) before burying small Ôcaches' Ñ g roups of seeds numbering 1 15 Ñ in the ground (Hutchins and Lanner, 1982; Tomback 1982). This process results in the transport of many whitebark pine seeds from subalpine forest to the ATE every year (Tomback, 1986; Tomback and Linhart, 1990). Unrecovered whitebark pine seeds may germinate and establish either in open or sheltered microsites and become the initiating conifer of tree islands. In a relati ve vigor study, Blakeslee (2012) was able to determine that whitebark pine has greater vigor at tre eline than other ATE conifers Ñ Engel mann spruce and subalpine fir Ñ by analyzing shoot length measurements. This means that whitebark pine may have superior har diness to harsh conditions and may regenerate, establish, and survive more frequently than other, less hardy species. Threats Twentieth century globalization and climate change have caused: 1) native mountain pine beetle ( Dendroctonus ponderosae ) outbreak s of unprecedented severity; 2) the spread of the invasive fungal pathogen C. ribicola , which causes the disease white pine blister rust ; and 3) wildfire suppression, resulting in the extensive loss of trees through successional replacement across the rang e of whitebark pine (Tomback and Achuff, 2010). In some areas, whitebark pine mortality rates approach one hundred

PAGE 25

9 percent (Smith et al., 2009). Additionally, these factors may diminish whitebark pine's role in affected ATE communities, resulting in unpred ictable changes in ecosystem function, stability, and biodiversity. Mountain Pine Beetle . Mountain pine beetles Ñ na tive to western North America Ñ are parasitic insects that attack pine trees periodically, killing mature trees within weeks of successful inf estation (Cole and Amman, 1969). Although pine species that comprise a large portion of lower elevation forest lodgepole ( Pinus contorta Douglas ex Loudon) and ponderosa ( Pinus ponderosa Lawson and C. Lawson) pine are primary hosts, all mature western pines qualify as targets (Tomback et al., 2011). During particularly severe outbreaks, these pests can advance to higher elevation forest communities (Perkins and Swetnam, 1996). Outbreaks are limited in severity and scope by climatic conditions that regul ate mountain pine beetle life cycles. Over the last 20 years, climate change has resulted in unprecedented mountain pine beetle proliferation (Logan et al., 2010), resulting in epidemic level mortality of mature whitebark pine in the central and northern R ocky Mountains (Gibson et al., 2008; Schwandt et al., 2010). Fire Exclusion. Whitebark pine is an early successional tree on productive sites, and is less shade tolerant than other subalpine and treeline conifers (Arno, 1986). Throughout the 20 th century, fire exclusion has inhibited the natural Ôsuccessional clock' and led to the replacement of whitebark pine by Engelmann Spruce and subalpine fir, fundamentally alter ing forest structure . In forests that have been Ôprotected' from fire, the result has been a decrease in biodiversity. White Pine Blister Rust. Whitebark pine trees are infected by C. ribicola when wind dispersed basidiospores enter needle stomata (McDonald and Hoff, 2001). Rust

PAGE 26

10 mycelia spread from the needles into the living wood and produce t he characteristic Ôcankers' and branch swelling associated with blister rust once aecial sacs break through the bark ( Fig. I.1; McDonald and Hoff, 2001). These sporulating cankers girdle branches and stems by inhibiting nutrient and water flow; the tree di es if cankers girdle the bole of the tree (McDonald and Hoff, 20 01). New spores are dispersed Ñ via wind Ñ as the sacs of aeciospores of the sporulating cankers rupture (McDonald and Hoff, 2001). Unlike mountain pine beetle and fire suppression, white pine bl ister rust affects regeneration, recruitment, and growth by attacking seedlings and saplings in addition to mature trees. Whitebark pine require up to 50 years to reach reproductive maturity (McCaughey and Schmidt, 1990), this means that once blister rust is introduced into a location, loss in cone production may last indefinitely. Additionally, the l oss of seedlings and saplings Ñ which die more rea dily than a fully mature tree Ñ drastically reduce the regeneration capacity of a community. Unlike mountain pine beetles, blister rust may also affect krummholz whitebark pine, altering whitebark pine's ecosystem services in the ATE. Because climate change is expected t o alter treeline dynamics, the e ffects of blister rust may affect treeline response to climate war ming. What could result is a diminished ecological function in whitebark pine across its range, or net forest loss.

PAGE 27

11 Tables and Figures Fig. 1 .1: White pine blister rust infection. Blisters containing aeciospores on the branches and bole of a whitebark pine.

PAGE 28

12 Chapter II WHITEBARK PINE AT TREELINE IN THE GREATER YELLOWSTONE ECOSYSTEM: PREVALENCE, FACILITATION, AND POTENTIAL DISRUPTION BY AN INTRODUCED PATHOGEN Abstract Facilitative interactions are particularly important in the climatically stressful al pine treeline ecotone (ATE), the transition zone from closed canopy forest to alpine tundra. Within some ATEs on the harsh eastern slope in the Rocky Mountains, whitebark pine ( Pinus albicaulis ) Ñ a foundation and keystone species Ñ plays a central role in tre e island development by facilitating the survival and growth of leeward conifers more frequently than other species. However, white pine blister rust (WPBR) Ñ a disease caused by an invasive fungal pathogen Cronartium ribicola Ñ is resulting in extensive morta lity across whitebark pine's range. The structure, composition, and dynamics of ATE formation are not well known for some geographic regions of western North America. The effect of disturbance to ATE whitebark pine is also understudied. We conducted an exp loratory study of four treeline communities in the Greater Yellowstone Ecosystem (GYE) to investigate (1) the prevalence of whitebark pine, (2) the proportional occurrence of whitebark pine as a tree island initiator, (3) whether solitary whitebark pine ab undance predicts its prevalence as a tree island initiator, (4) whether differences in climate correspond to geographic differences in whitebark pine's frequency of occurrence as a tree island initiator, (5) the relationship between common nurse objects an d solitary/tree island initiator establishment, and (6) the presence and severity of white pine blister rust in these communities. We found that whitebark pine is the most prevalent

PAGE 29

13 solitary conifer and tree island component, and initiates tree island form ation in direct proportion to its proportional community abundance as a solitary tree. Among our study areas, Cronartium ribicola infection rates ranged from less than 1% to 18%. Our findings suggest that losses of whitebark pine from infection by C. ribic ola will lead to profound changes in composition and structure of ATE communities in the GYE, including decline in tree island formation on the landscape. Introduction Temperate zone treeline communities provide important ecosystem services, such as snow retention and soil stabilization ( Farnes 1990, Geddes et al. 2005, Tomback et al. 2016 a ), and wildlife habitat, including refuge for disturbance sensitive species (e.g., Mattson 1997). The structure and composition, as well as dynamics of treeline formati on in the alpine treeline ecotone (ATE) Ñ the transition between open subalpine forest and tundra Ñ are not well known for some geographic regions of western North America. However, within a number of treeline communities in the Central and Northern Rocky Moun tains, and particularly those on the harsh, eastern Rocky Mountain Front, whitebark pine ( Pinus albicaulis Engelm, Family Pinaceae, Subgenus Strobus ) functions as a majority tree island initiator, facilitating the survival and growth of leeward conifers mo re frequently than other species (e.g., Resler and Tomback 2008, Tomback et al. 2014, Tomback et al. 2016a). There is now a general consensus that facilitation Ñ protective interactions between individuals of the same or different species (Callaway et al. 2 002) Ñ is an important biotic process shaping plant community structure and composition in abiotically stressful environments (Bertness and Callaway 1994, Antonsson et al. 2009).

PAGE 30

14 The ATE experiences extreme winds (Marr 1977), cold and variable temperatures ( Kšrner and Paulsen, 2004), poorly developed soils (Arno and Hammerly 1984), aridity (Holtmeier 2003), nutrient scarcity (Holtmeier 2003), high solar radiation (Maher et al. 2005), and truncated growing seasons (Arno and Hammerly 1984). ATE community develo pment relies on the initial establishment of solitary trees in unprotected microsites, or in protected microsites leeward of nurse objects such as rocks and topographic niches (Resler et al. 2005; Fig. 2. 1). Once established, these windward conifers create fa vorable leeward microsites by alleviating wind speeds (HŠttenschwiler and Smith 1999, Alftine and Malanson 2004, Pyatt et al. 2016), reducing solar radiation (Germino et al. 2002; Pyatt et al. 2016), decreasing soil temperature variation (Pyatt et al. 2016 ), and/or increasing soil moisture (Pyatt et al. 2016). Tree islands form when conifers become established in these leeward microsites, building a patch of krummholz trees (Marr 1977). Whitebark pine is often the most prevalent solitary conifer growing at treeline (Resler and Tomback 2008, Tomback et al. 2014, 2016); its prevalence as a solitary conifer predicts its frequency as a tree island initiator among different study areas (Tomback et al. 2016 a ). A moderately shade intolerant and slow growing conif er, whitebark pine competes poorly on productive sites, resulting in successional losses to faster growing, shade tolerant trees (Arno and Hoff 1990, Arno 2001). However, its slow growth and comparatively high carbon gain and water use efficiency relative to other conifers (Callaway et al. 2000, Bansal et al. 2011), coupled with hardy seedlings, contribute to its general tolerance of stressful conditions common to the ATE, including cold temperatures, aridity, nutrient poor soils, and high solar radiation ( Arno and Hoff 1990,

PAGE 31

15 Tomback et al. 2001b). Whitebark pine's recruitment success at high elevations may also be attributed to the seed storing behavior of its coevolved mutualist, Clark's nutcracker ( Nucifraga columbiana ; Tomback 1982). Nutcrackers bury whi tebark pine seeds in small caches near protective objects, such as rocks, tree trunks, and fallen trees (Tomback 1978); they also transport seeds from parent trees up to 32.6 kilometers for caching (Lorenz et al. 2011 ). This process results in the dispersa l of whitebark pine seeds from upper subalpine forest to treeline and tundra elevations (Tomback 1986). Whitebark pine is one of the most widely distributed five needle white pines in the U.S. and Canada (Tomback and Achuff 2010), and serves as a foundatio n and keystone species in many high elevation forest communities across the Central and Northern Rocky Mountains (Tomback et al. 2001a, Ellison et al. 2005). A foundation species is a locally abundant and regionally common ecosystem constituent that also c reates stable local conditions for others, thereby regulating ecosystem dynamics and disproportionately determining community spatial structure (Dayton 1972, Ellison et al. 2005, Ellison 2014). These F oundation species govern community assembly trajectorie s by modulating processes fundamental to the establishment and maintenance of associated flora and fauna (Ellison et al. 2014). Similarly, keystone species contribute disproportionately to biodiversity, influencing community composition more than their rel ative abundance would suggest (SoulÂŽ et al., 2003). Loss of foundation and keystone species through disease and other disturbance can destabilize ecosystems and have cascading effects on community structure and composition (Ellison et al. 2005). The keys tone and foundation functions of whitebark pine are discussed in detail elsewhere (Tomback et al. 2001a, Tomback and Kendall 2001, Tomback and Achuff

PAGE 32

16 2010, Tomback et al. 2011). They result from four major features of whitebark pine: large, nutritious see ds; seed dispersal by nutcrackers; hardy, robust seedlings; and tolerance of cold and windy sites (McKinney and Tomback 2011). The foundation functions include early establishment by whitebark pine after fire on poor seedbeds, or on harsh, exposed sites, leading to facilitation and establishment of other conifers; substrate stabilization against erosion at high elevations; snow retention and protraction of snow melt and regulation of downstream flows (e.g., Farnes 1990, Tomback et al. 2016 a ). The keystone function includes providing wildlife habitat at the highest elevations and providing a significant wildlife food source for a number of granivorous birds and mammals, including grizzly ( Ursus arctos ) and black bears ( Ursus americanus ) (e.g., Tomback and K endall 2001, Mattson et al. 2001). The extirpation of whitebark pine from subalpine and treeline communities is predicted to result in significant changes in community structure, function, and biodiversity (Tomback and Kendall 2001, Tomback and Resler 2007 , McKinney and Tomback 2011, Tomback et al. 2011, Tomback et al. 2016 a ). Throughout most of whitebark pine's range, the disease white pine blister rust , caused by the exotic fungal pathogen Cronartium ribicola , has resulted in extensive damage and mortali ty ( McDonald and Hoff 2001, Schwandt et al. 2010 ). White pine blister rust may infect all life stages of whitebark pine, from seedlings to mature trees. Stem infections kill trees; but branch infections reduce cone production and thus potential regeneratio n (McDonald and Hoff 2001). Cronartium ribicola infection threatens the foundation and keystone functions of whitebark pine (Tomback et al. 2001a, Tomback and Achuff 2010). When high infection rates from Cronartium ribicola as well as

PAGE 33

17 whitebark pine morta lity from blister rust were first detected in whitebark pine dominated ATE communities, the potential for disruption of structure and function and even altered response to climate warming were explored (Tomback and Resler 2007, Resler and Tomback 2008, Smi th McKenna et al. 2014). In the Greater Yellowstone Ecosystem (GYE) , Cronartium ribicola has infected an overall mean of 20 to 30% of the whitebark pine throughout the region (Jean et al. 2011). In this region, whitebark pine is widely represented in upper subalpine forests (Renkin and Despain 1992, McCaughey and Schmidt 2001), both as a successional component on productive sites and self replacing component on exposed sites with undeveloped soils. However, the prevalence and developmental role of whitebar k pine in ATE communities in the GYE, as well as the ATE structure and function, have been relatively unstudied, as is the prevalence of Cronartium ribicola . Elsewhere (Tomback et al. 2016 a ), we present information that whitebark pine is generally prevale nt in ATEs in the GYE but varies in occurrence and in its role as tree island initiator. We also provide information on variation in the occurrence of white pine blister rust in ATEs. Here, we provide more detailed information on community assessments and examine them in relation to mesoclimate conditions. For this study, we examined four ATE communities in the GYE from the Beartooth Plateau to the southeastern limits of whitebark pine's range in the Wind River Range to determine (1) the prevalence of whit ebark pine in treeline communities, (2) the importance of whitebark pine in facilitating tree island initiation, (3) whether solitary whitebark pine abundance predicts its prevalence as a tree island initiator, (4) whether differences in climate correspond to geographic differences in whitebark pine's frequency

PAGE 34

18 of occurrence as a tree island initiator, (5) the relationship between common nurse objects and solitary/tree island initiator establishment, and (6) whether blister rust is present, and, if so, the rate of infection in these communities. Methods Study A reas We performed community composition and climatic assessments in four ATEs in the GYE, two east and two west of the Continental Divide (Table. 2. 1, Fig. 2. 2). Within all study areas, conifer species inc luded whitebark pine, Engelmann spruce ( Picea engelmannii ), and subalpine fir ( Abies lasiocarpa ). At Tibbs Butte, the terrain followed a continuous upslope gradient with a few exposed rocks, and undeveloped soils stemming from a bedrock of Precambrian gran ite (Bevan 1923). Conifers and tree islands were scattered in occurrence. Dominant understory vegetation included Geum rossii (R. Br.) Ser. , Potentilla diversifolia Lehm., and Saxifraga spp. (Tomback et al. 2016a). Paintbrush Divide, Grand Teton National Park, Wyoming, consists of a steady rocky slope that becomes steeper with increasing elevation from SW to NE. Arctostaphylos uva ursi (L.) Spreng . , Myosotis asiatica (Vesterg.) Schischkin & Sergievskaja, Senecio spp., and Silene acaulis (L.) Jacq. are the understory dominants (Tomback et al. 2016a). The Hurricane Pass study area, Grand Teton National Park, Wyoming, has sparse rocks, and undeveloped soils supporting a diverse array of understory vegetation; dominants included included Dryas octopetala L., Se necio spp., Silene acaulis (L.) Jacq., and Graminoids species (Tomback et al. 2016a).

PAGE 35

19 Our southernmost study area was located on Mutiny Ridge in the Wind River Range, Shoshone National Forest, Wyoming, approximately 1.5 km west of Christina Lake. The stud y area comprised 117 ha on northeast, southeast, and southwest aspects from 3200 to 3400 m elevation. The terrain in the Mutiny Ridge ATE consists of a rock field with soils derived from Precambrian (Archean) crystalline bedrock, including granodiorite, po rphyritic quartz monzonite, and magmatic gneiss (e.g., Frost et al. 2000). The understory vegetation includes Geum rossii (R. Br.) Ser. , Pteryxia hendersonii (J.M. Coult. & Rose) Mathias & Constance , Phlox pulvinata (Wherry) Cronquist , and Silene acaulis ( L.) Jacq (Tomback et al. 2016a). Community A ssessments In July 2014 and 2015, we established assessment plots within each study area as follows: Tibbs Butte (n = 12), Paintbrush Divide (n = 20), Hurricane Pass (n = 20), and Mutiny Ridge (n = 30). Plots we re constructed at random locations generated using the splancs package in R (version 3.1.0, R Core Team 2014; Rowlingson and Diggle 2014). If random points were outside the ATE (i.e., in subalpine forest or alpine tundra), or resulted in plots without any conifers, we created the sampling plot in the nearest krummholz growth. We were cognizant that data collected within the same plot lack spatial independence. Each plot was the sampling unit; we used a plot size of 225 m 2 to minimize environmental variation within each plot , following the methods of Smith McKenna et al ( 2013). We assessed all conifers found completely or partly inside each plot. We identified each tree to species, and classified the ecological role of each as tree island initiator, leeward tree island component, or solitary tree. Tree islands comprised two or

PAGE 36

20 more contiguous conifers with spatially overlapping canopies. Conifers in tree islands were classified as tree island initiators or leeward conifers; initiators were the most windward c onifer, and leeward trees were all other trees within the tree island. For this study, there was only one initiating conifer per tree island, and thus the number of initiators equaled the number of tree islands for a given plot. To determine prevailing win d direction, and corresponding windward and leeward designations, we observed wind flagged branches of krummholz trees within the plot and in the immediate vicinity. Trees spatially isolated from other conifers (i.e., lacking overlap of branches, and often scattered throughout the plot) were classified as solitary. We identified nurse objects as any potentially protective object found directly windward of solitary trees or tree island initiators, including rocks, topographic depressions, and shrubs; we desi gnated conifers with the apparent absence of a nurse object as unprotected . Because each whitebark pine germinates from a cache of seeds buried by a Clark's nutcracker (Tomback 1982), some trees may occur as multi genet tree clusters (Tomback and Linhart 1 990). Additionally, Engelmann spruce and subalpine fir may have multiple stems of the same genotype due to clonal reproduction (Grime 1977). Following the methods of Resler and Tomback (2008), we characterized all multi stem growth forms as a single indivi dual. We recorded the following information for each plot: number of conifers by ecological role (i.e., solitary, initiator, or leeward), number of conifers by species, and nurse object for each solitary conifer and tree island initiator (i.e., rocks, top ographic depressions, shrubs, and unprotected). Additionally, we reco rded the number of blister rust infected whitebark pine by thoroughly examining all stems and canopies for evidence of infection. The criteria for a n infected blister rust designation inc luded at least

PAGE 37

21 one active canker, or three of the five following characteristics: 1) inactive cankers, 2) stem swelling, 3) branch flagging/dead foliage, 4) rodent gnawing, and 5) oozing sap (Burns et al. 2008). Mesoclimate A ssessments To estimate differe nces in climate variables among study areas, we used open source Precipitation Elevation Regressions on Independent Slopes Model (PRISM) data from the last 30 years (1985 2015; PRISM Climate Group, 2015). For each study site, we estimated the average daily growing season temperature and average growing season precipitation. Although literature definitions of growing season length differ, a global study by Kšrner and Paulsen (2004) found a narrow growing season temperature range in ATEs, and determined the l ower limit for plant tissue growth to be 5¡C. This temperature is also the threshold at which tissue growth is interrupted in cold adapted trees (Alvarez Uria and Kšrner, 2007) . Thus, we defined the growing season as the temporal period between average dai ly air temperatures of >5¡C for >5 consecutive days and average daily air temperatures of <5¡C for >5 consecutive days (Frich et al. 2002). Data A nalysis We converted raw counts for conifer ecological role (i.e., solitary, initiator, and leeward), species, nurse object association (rocks, topographic depressions, shrubs, and unprotected), and blister rust infection incidence (infected vs. uninfected) to proporitonal abundances for each assessment plot. We used R (version 3.2.4, R core team 2016) to conduct all statistical analyses. Proportional abundance parameter estimates. To estimate overall conifer proportional abundance parameters by ecological role, species, nurse object association,

PAGE 38

22 and blister rust infection incidence for each study area, we bootstr apped the raw proportional abundance data Ñ that is, sampled randomly with replacement over the size of the sample (e.g., 20 on Paintbrush Divide) Ñ for 5000 iterations to generate a sampling distribution of mean proportional abundances for each variable. We c alculated the mean of each sampling distribution to identify the proportional abundance parameter estimate ( ! ). Finally, we used the 0.025 and 0.975 quantiles of each sampling distribution to determine the lower and upper boundary, respectively, of the 95% high density interval (HDI). Mean differences in proportional abundance parameter estimates . We calculated pairwise differences in proportional abundance for conifer ecological role, species, nurse object association, and blister rust infection in cidence groups for each assessment plot. For example, comparing ecological roles at the species level, we calculated the difference in proporitonal abundance between whitebark pine and Engelmann spruce, whitebark pine and subalpine fir, and Engelmann spruc e and subalpine fir for each assessment plot. We also calculated the difference in proportional abundance between tree island initiators and solitary trees within species. Next, we bootstrapped the pairwise proportional abundance difference data Ñ that is, s ampled randomly with replacement over the size of the sample (e.g., 20 on Paintbrush Divide) Ñ for 5000 iterations to generate a sampling distribution of mean difference proportional abundances for each within group comparison. We calculated the mean of each mean difference in proportional abundance sampling distribution to identify the parameter estimate ( ! difference ). We made Bonferroni adjustments based on the number of pairwise

PAGE 39

23 comparisons to determine which sampling distribution quantiles to use as th e HDI lower and upper boundaries. The formula for these adjustments is the following: "#$ % & ' ( ) ' * + ' , (.. (1) T he numbe r of pairwise comparisons is denoted by n ; our * for all analyses in this study was 0.05. For three pairwise comparisons (i.e., species specific comparisons including whitebark pine, Engelmann spruce, and subalpine fir as well as ecological role comparisons including initiator, leeward, and solitary conifers), we estimated the 98.3% HDI by calculating the 0.0083 and 0.9917 quantil es of each sampling distribution. For six pairwise comparisons (i.e., conifer nurse object associations including r ocks, topographic depressions, shrubs, and unprotected; climactic variables across study areas ), we estimated the 99.17% HDI by calculating t he 0.0042 and 0.9958 quantiles of each sampling distribution as the lower and upper boundary. For all pairwise difference estimates, we concluded that proportional abundances did not differ (i.e., no mean difference) if the estimated HDI included zero. Growing season temperature and precipitat ion parameter estimates . We determined the duration of each growing season from 1985 to 2015 for all study sites by trimming the imported PRISM daily mean temperature and precipitation data according to the tempera ture parameters defined in methods. Then, we calculated the mean daily growing season air temperature and precipitation corresponding to each year. We bootstrapped Ñ that is, sampled randomly with replacement over the size of the sample (30 seasons; 1985 thr ough 2015 ) Ñ for 5000 iterations to generate a sampling distribution of means corresponding to each climatic parameter (i.e., average daily temperature and precipitation). We calculated the mean of each sampling distribution to estimate average

PAGE 40

24 daily tempera ture ( / ) and precipitation ( 0 ) across all growing seasons. Finally, we estimated the 95% HDI by calculating the 0.025 and 0.975 quantiles of each sampling distribution as the lower and upper boundary. Differences in mean growing se ason climatic parameter estimates . Once the average daily growing season temperature and precipitation sampling distributions had been generated, we calculated pairwise differences in means between all study areas. Then, we calculated the mean of eac h difference in mean growing season temperature ( / difference ) and precipitation ( 0 difference ) sampling distribution to identify the parameter estimate. For six pairwise comparisons (i.e. Tibbs Butte vs. Paintbrush Divide, Tibbs Butte vs. Hurricane Pass, Tibbs Butte vs. Mutiny Ridge, Paintbrush Divid e vs. Hurricane Pass, Paintbrush Divide vs. Mutiny Ridge, and Hurricane Pass vs. Mutiny Ridge), we estimated the 99.17% HDI by calculating the 0.0042 and 0.9958 quantiles of each sampling distribution as the lower and upper boundary, respectively. For all pairwise difference estimates, we concluded climatic conditions did not differ (i.e., no difference in means) if the estimated HDI included zero. Results All raw count and proportional abundance data for conifer ecological role, species, nurse object assoc iation, and blister rust infection incidence are available in Appendix A, Tables A . 1 through A . 8. For consistency with other studies assessing whitebark pine's prevalence and functional role at treeline (e.g., Resler and Tomback 2008, Tomback et al. 2014, To mback et al. 2016a), we report basic descriptive statistics on raw data in Table 2. 2. Community Structure and C omposition

PAGE 41

25 Tibbs Butte. On average, 0.956 (95% HDI: 0.880, 1.000) of all conifers in the Tibbs Butte study area we re solitary, and 0.044 (95% HDI : 0.000, 0.120) were tree island members (i.e., initiators or leeward conifers) . Average solitary tree density was 7.322 (95% HDI: 3.583, 12.583) trees per plot. Mean tree island density was 0.259 (95% HDI: 0.000, 0.667) tree islands per plot. Of all conife rs observed regardless of ecological role , whitebark pine comprised 0.914 (95% HDI: 0.843, 0.975), followed by Engelmann spruce at 0.076 (95% HDI: 0.014, 0.148) and subalpine fir at 0.011 (95% HDI: 0.000, 0.031; Fig. 2. 3a). Overall, whitebark pine was signif icantly more abundant than Engelmann spruce and subalpine fir (Fig. 2. 3b) Solitary trees were comprised of 0.926 (95% HDI: 0.855, 0.988) whitebark pine , followed by 0.074 (95% HDI: 0.012, 0.145) Engelmann spruce, and 0.000 (95% HDI: 0.000, 0.000) subalpine f ir. We found that whitebark pine was the most prevalent solitary conifer. Whitebark pine comprised 0.775 (95% HDI: 0.750, 0.800) of leeward tree island conifers; Engelmann spruce accounted for 0.123 (95% HDI: 0.000, 0.250), and subalpine fir comprised 0.10 1 (95% HDI: 0.000, 0.200). Whitebark pine was the most common leeward tree island conifer. The three tree island initiators we observed on Tibbs Butte included 2 whitebark pine, 1 subalpine fir, and 0 Engelmann spruce (A ppendix A, Table A. 1). Both whitebark pine tree island initiators were located within the same assessment plot. On average, tree island initiator species composition was 0.504 (95% HDI: 0.000, 1.000) whitebark pine, 0.496 (95% HDI: 0.000, 1.000) subalpine fir, and 0.000 (95% HDI: 0.000, 0.000 )

PAGE 42

26 Engelmann spruce. Pairwise mean differences between tree island initiator species abundances revealed no differences. Paintbrush Divide. At Paintbrush Divide, 0.620 (95% HDI: 0.482, 0.751) of all conifers were solitary, and 0.380 (95% HDI: 0.249, 0.518) were tree island members. Average solitary tree density was 14.030 (95% HDI: 10.100, 18.000) trees per plot. Mean tree island density was 2.144 (95% HDI: 1.450, 2.950) tree islands per plot. Whitebark pine comprised 0.579 (95% HDI: 0.441, 0.707) of all con ifers observed, followed by Engelmann spruce at 0.276 (95% HDI: 0.170, 0.395) and subalpine fir at 0.145 (95% HDI: 0.073, 0.229; Fig. 2. 3a). Whitebark pine was significantly more abundant than Engelmann spruce and subalpine fir (Fig. 2. 3b). Whitebark pine com prised 0.646 (95% HDI: 0.521, 0.753) of solitary trees, followed by 0.271 (95% HDI: 0.160, 0.405) Engelmann spruce, and 0.082 (95% HDI: 0.040, 0.130) subalpine fir. We found that whitebark pine was the majority solitary conifer, constituting more solitary trees than Engelmann spru ce and subalpine fir. We also found there were more solitary Engelmann spruce than solitary subalpine fir. For leeward conifers, whitebark pine comprised 0.666 (95% HDI: 0.474, 0.833), Engelman n spruce 0.201 (95% HDI: 0.0 67, 0.363) , and subalpine fir 0.132 (95% HDI: 0.037, 0.271). Our analysis revealed that whitebark pine was the majority leeward conifer. We observed 43 tree island initiators, which included 28 whitebark pine, 9 Engelmann spruce, and 6 subalpine fir (Appendix A, Ta ble A. 1). Average tree island initiator species proportional abundance was 0.561 (95% HDI: 0.353, 0.757) whitebark pine, 0.248 (95% HDI: 0.100, 0.419) Engelmann spruce, and 0.191 (95% HDI: 0.033,

PAGE 43

27 0.382) subalpine fir. We found no differences in species propo rtional abundance for tree island initiators. Hurricane Pass. On average, 0.537 (95% HDI: 0.404, 0.676) of the conifers at Hurricane Pass were solitary; 0.463 (95% HDI: 0.323, 0.596) were tree island members. Average solitary tree density was 6.987 (95% H DI: 4.900, 9.200) trees per plot. Mean tree island density was 2.406 (95% HDI: 1.500, 3.450) tree islands per plot. Whitebark pine comprised 0.419 (95% HDI: 0.298, 0.539) of all conifers observed, followed by Engelmann spruce at 0.326 (95% HDI: 0.231, 0.42 8) and subalpine fir at 0.255 (95% HDI: 0.173, 0.343; Fig. 2. 3a). We found no species to be an overall majority conifer at Hurricane Pass (Fig. 2. 3b). In our examination of the species composition of solitary trees, we found whitebark pine comprised 0.480 (95 % HDI: 0.334, 0.627), followed by 0.325 (95% HDI: 0.218, 0.437) for Engelmann spruce, and 0.194 (95% HDI: 0.092, 0.311) for subalpine fir. We found that no species comprised a higher proportion of solitary trees than any other species. We found that whiteb ark pine composed an average 0.419 (95% HDI: 0.247, 0.596) of leeward tree island conifers, followed by Engelmann spruce at 0.307 (95% HDI: 0.157, 0.489), and subalpine fir at 0.273 (95% HDI: 0.119, 0.453). No species comprised a higher proportion of leewa rd tree island conifers than any other species. Our sample at Hurricane Pass included 48 tree island initiators: 16 whitebark pine, 17 Engelmann spruce, and 15 subalpine fir (Appendix A, Table A. 1). On average, tree island initiator species composition was 0.263 (95% HDI: 0.127, 0.417) whitebark pine, 0.415 (95% HDI: 0.237, 0.598) subalpine fir, and 0.322 (95% HDI: 0.151, 0.561)

PAGE 44

28 Engelmann spruce. Pairwise mean differences among tree island initiator species revealed no differences. Mutiny Ridge. A t Mutiny R idge, we found that 0.745 (95% HDI: 0.618, 0.860) of all conifers were solitary, and 0.255 (95% HDI: 0.140, 0.382) were tree island members. Average solitary tree density was 8.256 (95% HDI: 5.167, 12.067) trees per plot. Mean tree island density was 1.367 (95% HDI: 0.667, 2.267) tree islands per plot. In total, whitebark pine comprised 0.904 (95% HDI: 0.856, 0.945) of all conifers observed, followed by subalpine fir at 0.064 (95% HDI: 0.028, 0.109) and Engelmann spruce at 0.032 (95% HDI: 0.015, 0.052; Fig. 2. 3a). We found that whitebark pine was the overall majority conifer (Fig. 2. 3b). Of the solitary trees in the Christina Lake ATE, whitebark pine comprised 0.924 (95% HDI: 0.882, 0.962), followed by 0.041 (95% HDI: 0.014, 0.074) subalpine fir, and 0.035 (95% HDI: 0.014, 0.059) Engelmann spruce. Whitebark pine was the majority solitary tree. We also found whitebark pine constituted an average 0.955 (95% HDI: 0.900, 0.995) of leeward tree island conifers; Engelmann spruce accounted for 0.031 (95% HDI: 0.000, 0 .086), and subalpine fir comprised 0.014 (95% HDI: 0.000, 0.037). We found whitebark pine was the majority leeward tree island constituent. The 41 tree island initiators we observed at Mutiny Ridge included 33 whitebark pine, 4 subalpine fir, and 4 Engelma nn spruce (Appendix A , Table A. 1). Average tree island initiator species composition was 0.652 (95% HDI: 0.409, 0.873) whitebark pine, 0.283 (95% HDI: 0.071, 0.538) subalpine fir, and 0.065 (95% HDI: 0.000, 0.163)

PAGE 45

29 Engelmann spruce. We found there were more w hitebark pine tree island initiators than Engelmann spruce. Tree Island D evelopment To determine whether solitary whitebark pine proportional abundance predicted its prevalence as a tree island initiator, we examined the difference in proportional abundan ce between tree island initiators and solitary trees within species. At Tibbs Butte, Paintbrush Divide, and Mutiny Ridge, we found there were no differences between the estimated proportion of tree island initiators and solitary trees for any species (Fig. 2. 4). At Hurricane Pass, however, our results indicated whitebark pine initiated fewer tree islands than would be expected given its proportional abundance as a solitary tree; subalpine fir initiated more tree islands than would be expected given its propor tional abundance as a solitary tree. Facilitation by Nurse O bjects Tibbs Butte . Our sample collected at Tibbs Butte contained 91 conifers that were either solitary trees or tree island initiators (Appendix A, Table A. 1). Of these, 0.545 (95% HDI: 0.383, 0.7 03) were in geomorphic depressions, 0.237 (95% HDI: 0.094, 0.417) were leeward of woody plants, 0.174 (95% HDI: 0.080, 0.273) were protected by rocks, and 0.045 (95% HDI: 0.000, 0.098) were unprotected (Fig. 2. 5a). There were significantly more conifers prot ected by geomorphic depressions than rocks, and significantly more conifers protected by geomorphic depressions than those that were unprotected (Fig. 2. 5b). Paintbrush Divide. There were 324 conifers that were either solitary trees or tree island initiator s at Paintbrush Divide (Appendix A, Table A. 1). Of these, 0.532 (95%

PAGE 46

30 HDI: 0.381, 0.684) were protected by rocks, 0.429 (95% HDI: 0.284, 0.579) were unprotected, 0.029 (95% HDI: 0.000, 0.068) were in geomorphic depressions, and 0.010 (95% HDI: 0.000, 0.027) w ere in the lee of woody plants (Fig. 2. 5a). We found that rocks protected more conifers than woody vegetation and geomorphic depressions. We also found there were more unprotected conifers than those protected by woody plants or geomorphic depression (Fig. 2. 5 b). Hurricane Pass. The Hurricane pass sample encompassed 188 trees that were either solitary or tree island initiators (Appendix A, Table A. 1). Of these, 0.413 (95% HDI: 0.268, 0.562) were unprotected, 0.338 (95% HDI: 0.207, 0.489) were protected by rocks, 0.193 (95% HDI: 0.068, 0.342) were in geomorphic depressions, and 0.055 (95% HDI: 0.008, 0.124) were in the lee of woody plants (Fig. 2. 5a). We found there were more conifers protected by rocks than geomorphic shelters, and more unprotected conifers than th ose protected by geomorphic depressions (Fig. 2. 5b). Mutiny Ridge . At Mutiny Ridge, 288 trees were either solitary or tree island initiators (Appendix A, Table A. 1). Of these, 0.689 (95% HDI: 0.570, 0.800) were protected by rocks, 0.174 (95% HDI: 0.092, 0.274) were unprotected, 0.086 (95% HDI: 0.038, 0.142) were in the lee of woody plants, and 0.051 (95% HDI: 0.023, 0.082) were in geomorphic depressions (Fig. 2. 5a). We found there were more conifers protected by rocks than those that were unprotected, or protecte d by geomorph ic depressions or woody plants (Fig. 2. 5b). Blister Rust A ssessments M ean blister rust infection was 0.006 (95% HDI: 0.000, 0.018) at Tibbs Butte, 0.180 (95% HDI: 0.099, 0.273) at Paintbrush Divide, 0.143 (95% HDI: 0.062, 0.242) at

PAGE 47

31 Hurricane Pa ss, and 0.019 (95% HDI: 0.005, 0.038) at Mutiny Ridge Ñ a range of less than 1%, to 18%. There were no differences between white pine blister rust infection in solitary and tree island whitebark pine at any study area. Mesoclimate A ssessments For our study locations, the mean daily growing season temperature from 1985 to 2015 was 8.51¡C (95% HDI: 8.05¡C, 8.97¡C) at Tibbs Butte, 9.90¡C (95% HDI: 9.49¡C, 10.28¡C) at Paintbrush Divide, 8.95¡C (95% HDI: 8.50¡C, 9.38¡C) at Hurricane Pass, and 9.13¡C (95% HDI: 8.7 0¡C, 9.54¡C) at Mutiny Ridge (Fig. 2. 6a). When examining pairwise differences in mean daily growing season temperature between study locations, we found Paintbrush Divide was warmer than Hurricane Pass and Tibbs Butte (Fig. 2. 6b). Average daily growing season precipitation was 1.65mm (95% HDI: 1.44mm, 1.87mm) on Tibbs Butte, 2.02mm (95% HDI: 1.78mm, 2.27mm) on Paintbrush Divide, 2.14mm (95% HDI: 1.85mm, 2.43mm) on Hurricane Pass, and 1.42mm (95% HDI: 1.21mm, 1.62mm) on Mutiny Ridge (Fig. 2. 6a). We found Paintbru sh Divide and Hurricane Pass received more precipitation than Mutiny Ridge (Fig. 2. 6b). Discussion We designed this exploratory study of community structure and composition at four ATE communities in the GYE to examine specifically (1) the prevalence of whi tebark pine, (2) the proportional occurrence of whitebark pine as a tree island initiator, (3) whether solitary whitebark pine abundance predicts its prevalence as a tree island initiator, (4) whether differences in climate correspond to geographic differe nces in whitebark pine's frequency of occurrence as a tree island initiator, (5) the relationship

PAGE 48

32 between common nurse objects and solitary/tree island initiator establishment, and (6) the presence and severity of white pine bli ster rust in these communiti es. Whitebark Pine and Tree Island D evelopment in the GYE Results of this study confirm whitebark pine's importance in ATE community structure and composition in the Greater Yellowstone Ecosystem. We found that whitebark pine was the overall majority conif er at Tibbs Butte ( ! ' = 0.914), Paintbrush Divide ( ! ' = 0.579), and Mutiny Ridge ( ! ' = 0.904). It was also the most abundant solitary conifer at all study areas except Hurricane Pass, and by a wide margin for study areas outside Grand Teton National Park ( ! ' ' > 0.853 more than any other conifer). Additionally, whitebark pine comprised the majority of leeward tree island conifers at Tibbs Butte ( ! ' = 0.775 ), Paintbrush Divide ( ! ' = 0.665 ), and Mutiny Ridge ( ! ' = 0.955 ). The preceding evid ence support s the notion that the prevalence of whitebark pine in the ATE is increased relative to other conifer species , particularly on the arid and windy sites east of the Continental Divide. It also highlights the importance of mechanisms that may driv e whitebark pine's elevated proportional abundance, namely, efficient seed dispersal by Clark's nutcrackers to protected sites (Tomback 1978) and whitebark pine's relatively high tolerance of abiotic stress (e.g., A rno and Hoff 1990). We also found that solitary trees and conifers that serve as tree island initiators in ATEs within the GYE often establish leeward of nurse objects (i.e. rocks, topog raphic depressions, and shrubs) , and the proportional occurrence of common nurse objects types found windward of solitary conifers and tree island initiators differs by study area. This study did not assess whether the proportional occurrence of nurse objects as windward facilitators differs from the proportion of nurse objects available on the landscape. We

PAGE 49

33 reco mmend future studies account for the proportional abundance of nurse objects types available when assessing the relative importance of certain nurse objects to solitary conifer establishment. Given that whitebark pine functions as the majority tree island initiator in many ATEs across the Central and Northern Rocky Mountains (Tomback et al. 2016a), we hypothesized that it would serve this role in the southeastern most limit of its range. Yet, our results indicated that whitebark pine did not serve as major ity tree island initiator at any study area. If whitebark pine is the majority conifer in regards to overall, solitary, and leeward compositon at Tibbs Butte, Paintbrush Divide, and Mutiny Ridge, why is it not the majority tree island initiator at these lo cations? One explanation may lie in the high proportional abundance of solitary whitebark pine, which Resler et al. (2014) hypothesized as indicative of an early stage in ATE development or a steady age structure with rapid new establishment. High proporti onal abundance of leeward whitebark pine supports the conjecture that whitebark pine readily establishes leeward of solitary conifers, most likely because nutcrackers place seeds near objects disproportionately (Tomback 1978), and whitebark pine early life stage survival is increased near rocks and trees relative to unprotected microsites in some ATEs (Pansing et al. in review). Another answer may be that other ATE conifers, namely Engelmann spruce and subalpine fir, are unable to establish leeward of white bark pine due to seed dispersal limitations, or lower tolerances to stressful conditions (Bansal et al. 2011), even with windward facilitation Finally, use of a study design that applies assessment plots may have led to a low number of tree islands sample d (n = 3, n = 43, n = 48, and n = 41 at Tibbs Butte, Paintbrush Divide, Hurricane Pass, and Mutiny Ridge, respectively), decreasing the

PAGE 50

34 precision of our estimates to a degree that precluded our ability to detect mean differences in proportional abundance b etween species. Previous plot or transect based studies have overlooked this issue of intra plot spatial interdependence and combined tree island totals across replicates (e.g., Resler and Tomback 2008, Tomback et al. 2014) Ñ a form of pseudoreplication (Hu rlbert 1984); this work invoked a more rigorous approach. In deed , had we combined tree island totals across assessment plots and conducted multinomial goodness of fit tests under the null hypothesis that tree island initiator species proportional abundance s were equal ( ! ' 1 2 3456789 ' :3;5 = ! ' <=67>:3;5 ' ?38 = ! ' @;A5>B7;; ' <:8=C5 = 0.333; e.g. Resler and Tomback 2008, Tomback et al. 2014), we would have designated whitebark pine a majority tree island initi ator at Paintbrush Divide ( X 2 = 19.86, df = 2, p " 0.000), and Mutiny Ridge ( X 2 = 41.02, df = 2, p " 0.000). In future examinations of tree island initiator s pecies proportional abundances, we recom mend two possible approaches: (1) forego assessment plots and transects in favor of a simple random sampling strategy that employs tree islands as replicates, or (2) use hierarchical models that account for spatial dependence among observations to increase the precision . The Relationship of Solitary Trees to T re e Island I nitiators Our second objective was to determine whether solitary conifer species proportional abundances would predict tree island initiator species proporitonal abundances, as they did previously in Tomback et al. (2016a). Our results support th ose findings. With the exception of whitebark pine and subalpine fir at Hurricane Pass, the mean difference effect size between solitary and tree island initiator proportional abundances did not deviate from zero for any species at any study area. These re sults

PAGE 51

35 provide support for the general notion that ATE conifer species facilitate subsequent conifer establishment (Marr 1977), and thereby initiate tree island development at a rate proportional to their proportional abundance as solitary trees. Mean diffe rences in proportional abundance among species, therefore, relate primarily to differences in establishment and survival (Malanson et al. 2007). At Hurricane Pass, however, whitebark pine initiated 0.207 (95% HDI: 0.028, 0.393) fewer tree islands than wou ld be expected given its proportional abundance as a solitary tree; subalpine fir initiated 0.175 (95% HDI: 0.025, 0.334) more. These results may be explained by long known physiological differences between whitebark pine and subalpine fir. Griggs (1938) n oted that subalpine fir survive and better grow in moist, snowy environments in relation to other timberline species. Little (1994) found subalpine fir establishment and survival to be limited by low soil moisture, especially on southern aspects. Indeed, o ur pairwise comparisons of climactic conditions between study sites revealed that Hurricane Pass received, on average, 0.489 mm and 0.720 mm per day more precipitation than Tibbs Butte and Mutiny Ridge, respectively . Our findings suggest that local climact ic conditions, particularly mean daily growing season precipitation, may influence species' functional ecology and tree island composition at high elevations in some communities in the Greater Yellowstone Ecosystem. The I mpact of C. ribicola I nvasion The s pread of C. ribicola and the resulting incidence of white pine blister rust in ATEs within the Shoshone National Forest and Grand Teton National Park have the potential to disr upt tree island development and community structure and composition. As a stress tolerant foundation species, whitebark pine establishes on cold, arid sites, and

PAGE 52

36 facilitates the leeward establishment of other conifers in locations less conducive to the survival of other species (Resler and Tomback 2008 , Tomback et al. 2016b ). Resler a nd T omback (2008), working south of the Canadian border on Divide Mountain and Lee Ridge on the eastern slope of Glacier National Park, documented whitebark pine mortality from blister rust in the ATE; they reported an overall infection level of 33.7%. In fact, this northern region along with the southern Canadian Rocky Mountains have the highest blister rust infection levels (83%) recorded for whitebark pine overall (Smith et al. 2013). In this study, we found that ATE white pine blister rust incidences ra nged from <1% at Tibbs Butte in Shoshone National Forest to 18% at Paintbrush Divide in Bridger Teton National Forest . The blister rust incidence of 83% among subalpine whitebark pine communities in Grand Teton National Park was as high as the northern sur veys (Bockino and Tinker 2012), and explains the high blister rust inciden ce at treeline in this region. The progressive loss of solitary whitebark pine to white pine blister rust, therefore, may ultimately limit tree island development on the ATE landscap e, significantly affecting community structure and biodiversity. Given whitebark pine's ecological role as majority solitary tree and leeward tree island component in three out of four study areas, ATE conifer communities in the Greater Yellowstone Ecosyst em will likely experience severe changes in structure and composition resulting from mortality by white pine blister rust, as described by Tomb ack and Resler (2007) : First, the seed sources for ATE conifer regeneration and community development are trees f rom lower elevations (Malanson et al. 2007). As C. ribicola spreads through stands in the upper subalpine, trees succumbing to white pine blister rust infection will produce fewer seeds (e.g., Hoff et al. 2001). As a result, Clark's

PAGE 53

37 nutcrackers will have f ewer opportunities to disperse to the ATE (e.g., Tomback, 1986). McKinney et al. (2009) and Barringer et al. (2012) demonstrated that the probability of nutcracker seed dispersal or stand visitation decreases with decreased cone production in subalpine whi tebark pine stands; there may also be a threshold below which the probability of nutcracker visitation declines steeply. Recent studies in the GYE have estimated subalpine blister rust incidences range from 20% to 30%. Second, white pine blister rust kills whitebark pine in the ATE, and mortality of solitary whitebark pine will reduce the number of trees available to facilitate the establishment of other conifers, precluding tree island development (Tomback and Resler 2007). Proportionally, whitebark pine's ecosystem function within these ecosystems may also be reduced, altering its functional ecology, impacting community assembly trajectories, and reducing the ecosystem services provided by treeline communities. Conclusions Here, for the study areas sampled in the GYE, we found evidence that whitebark pine is generally the most prevalent solitary conifer and tree island component in treeline communities, and whitebark pine initiates tree island development at a rate directly proportional to its proportional abundance as a solitary tree. Our findings suggest that declines in ATE whitebark pine are imminent, stemming from community invasion by the pathogen C. ribicola . Whitebark pine mortality will likely lead to a serious decline in ATE conifer abundance, a pr oportional loss of tree islands on the landscape, and reduction in whitebark pine's ecological role and general ecosystem services in ATE communities in the Greater Yellowstone Ecosystem .

PAGE 54

38 Tables and figures Table 2.1 Study areas across the Greater Yello wston e Ecosystem (GYE). The GYE comprises the southernmost distribution of whiteba rk pine in the Rocky Mountains. Study Area National Forest Latitude & Longitude Aspects Elevation [m] Area Sampled [ha] Tibbs Butte Shoshone 44¡56'46.48" N 109¡26'39.7 6" W NE 2,983 Ð 3,238 4.2 Paintbrush Divide Bridger Teton 43¡47'32.90" N 110¡49'17.96" W NE, NW, SW 3,055 Ð 3,289 121 Hurricane Pass Bridger Teton 43¡43'55.06" N 110¡50'31.45" W NE, SE, SW 3,045 Ð 3 , 078 92 Mutiny Ridge Shoshone 42¡35'33.56" N 108¡5 8'07.44" W NE, SE, SW 3 , 200 Ð 3 , 400 117

PAGE 55

39 Table 2. 2 D escriptive statistics . Minimum, median, and maximum statistics for per plot conifer counts by ecological role and species for all study areas. ABLA = A. lasiocarpa , PIEN = P. Engelmannii , PIAL = P. al bicaulis . Study area Ecological role Descriptive statistics ABLA PIEN PIAL Total Min Med Max Min Med Max Min Med Max Min Med Max Tibbs Butte initiator 0 0.5 1 0 0.0 0 0 1.0 2 1 1.5 2 leeward 0 0.5 1 0 0.5 1 3 3.5 4 4 4.5 5 solitary 0 0.0 0 0 0. 0 3 1 4.0 32 1 5 32 Paintbrush Divide initiator 0 0.0 3 0 0.0 2 0 1.0 5 1 2 7 leeward 0 0.0 3 0 1.0 2 0 4.0 17 1 4 20 solitary 0 1.0 3 0 4.0 17 0 8.5 23 3 14 34 Hurricane Pass initiator 0 1.0 3 0 1.0 3 0 1.0 5 1 2 8 leeward 0 1.0 6 0 1.0 3 0 2.0 14 1 4 17 solitary 0 1.0 6 0 2.0 11 0 1.0 9 1 6 18 Mutiny Ridge initiator 0 0.0 1 0 0.0 2 0 2.0 10 1 2 10 leeward 0 0.0 1 0 0.0 1 1 5.0 55 1 5.5 55 solitary 0 0.0 6 0 0.0 6 1 5.5 33 1 6 45

PAGE 56

40 Fig. 2.1 Facilitation in action. Krummholz whitebark pin e ( Pinus albicaulis ) established leeward of a rock at Tibbs Butte, Shoshone National Forest, Wyomi ng. Photograph credit: E.R. Pansing.

PAGE 57

41 Fig. 2. 2. Study areas across the Greater Yellowstone Ecosystem . Study areas are in the southernmost distribution of wh itebark pine in the Rocky Mountains. Map created using ggmap package in R (Kahle and Wickham 2013).

PAGE 58

42 Fig. 2. 3a. Parameter estimates. Bootstrapped parameter estimates and 95% HDIs for the proportional abundance of solitary, initiator, and leeward conifer s by species for each study area.

PAGE 59

43 Fig. 2. 3b. Parameter estimates for mean differences. Bootstrapped estimates and 98.3% HDIs for mean differences in proportional abundance of solitary, initiator, and leeward conifers by species for each study area. PIA L = P. Albicaulis , ABLA = A. lasiocarpa , and PIEN = P. engelmannii . We concluded no mean difference if the HDI included zero; fields in which there was a difference are marked with an asterisks (*).

PAGE 60

44 Fig. 2. 4. Mean difference parameter estimates: solita ry minus initiator conifers. Bootstrapped mean differences in proportional abundance between tree island initiators and solitary trees by species at each study area. Estimates are indicated by open circles, and lines delineate 95% HDIs. We concluded that n o mean difference occurred between distributions if the HDI included zero. F ields in which there was a difference are marked with an asterisks (*).

PAGE 61

45 Fig. 2. 5a. Parameter estimates for nurse objects. Bootstrapped proportional abundance of nurse objects f ound facilitating solitary and tree island initiator conifers, by study site. Estimates are indicated by open circles, and lines delineate 95% HDIs.

PAGE 62

46 Fig. 2. 5b. Mean difference parameter estimates for nurse objects. Bootstrapped mean differences in prop ortional abundance of nurse objects found protecting solitary conifers and tree island initiators, by study area. R = Rock, U = Unprotected, D = Depression, and S = Shrubs. Estimates are indicated by open circles, and lines delineate 99.17% HDIs. We conclu ded that no mean difference occurred if the HDI included zero. F ields in which there was a difference are marked with an asterisks (*).

PAGE 63

47 Fig. 2. 6a. Parameter estimates for climate. Bootstrapped mean daily growing season temperature (top) and precipitati on (bottom), by study area . We estimated climactic parameters using PRISM data from the last 30 years (1985 2015). Estimates are indicated by open circles, and lines delineate 95% HDIs.

PAGE 64

48 Fig. 2. 6b. Difference in mean parameter estimates for climate . Bootst rapped pairwise differences in mean growing season temperature (top) and precipitation (bottom) between study area . Estimates are indicated by open circles, and lines delineate 99.17% HDIs. We concluded that no difference in means occurred if the HDI inclu ded zero. F ields in which there was a difference are marked with an asterisks (*). TB = Tibbs Butte, PD = Paintbrush Divide, HP = Hurricane Pass, MR = Mutiny Ridge.

PAGE 65

49 Appendix A Table A . 1. Raw conifer species count data for all study areas. ABLA = A. Lasi ocarpa , PIEN = P. Engelmannii , PIAL = P. Albicaulis . Study Area Ecological Role ABLA PIEN PIAL Total Tibbs Butte Initiator 1 0 2 3 Leeward 1 1 7 9 Solitary 0 6 82 88 Total 2 7 91 100 Paintbrush Divide Initiator 6 9 28 43 Leeward 11 11 81 103 So litary 17 89 175 281 Total 34 109 284 427 Hurricane Pass Initiator 15 17 16 48 Leeward 22 20 54 96 Solitary 32 47 61 140 Total 69 84 131 284 Mutiny Ridge Initiator 4 4 33 41 Leeward 2 3 143 148 Solitary 14 13 220 247 Total 20 20 396 436

PAGE 66

50 Table A . 2. Raw conifer species count and proportional abundance data for Tibbs Butte.

PAGE 67

51 Table A . 3. Raw conifer species count and proportional abundance data for Paintbrush Divide.

PAGE 68

52

PAGE 69

53 Table A . 4. Raw conifer species count and proportional abundance data f or Hurricane Pass.

PAGE 70

54

PAGE 71

55 Table A . 5. Raw conifer species count and proportional abundance data for Mutiny Ridge.

PAGE 72

56

PAGE 73

57 Table A . 6. Raw nurse object count data by plot for all study areas.

PAGE 74

58

PAGE 75

59 Table A . 7. Raw nurse object proportional abundance data by plot for all study areas.

PAGE 76

60

PAGE 77

61 Table A . 8. Raw white pine blister rust count and proportional abundance data by plot, all study areas.

PAGE 78

62

PAGE 79

63 C hapter III FACILITATION IN TREELINE WHITEBARK PINE COMMUNITIES: BIOPHYSICAL DIFFERENCES AMONG LEEWARD MICROSITES DE PEND ON LARGER SCALE CLIMATE INDICES Abstract In many treeline communities across the Northern and Central Rocky Mountains, whitebark pine ( Pinus albicaulis Engelm, Family Pinaceae, Subgenus Strobus .) Ñ a widely distributed five needle white pine in the U.S . and Canada Ñ is the most common tree island initiator, facilitating the survival and growth of leeward conifers more frequently than other species. Recent comparisons of biophysical characteristics of leeward microsites suggest that conifers offer a more m oderated microclimate than rocks or exposed microsites, but that that whitebark pine's leeward microclimate may be similar or slightly harsher than spruce microsites, except for moisture and sky exposure. We assessed differences in leeward microclimate in two treeline communities on the harsh Rocky Mountain eastern front to test whether whitebark pine better ameliorates local conditions relative to other nurse plants or objects. Using a simple random sampling design, as opposed to a block study, we found th at whitebark pine leeward microsites did not consistently experience the most moderate microclimate. Differences in microclimatic conditions by microsite type depended substantially on the general climatic conditions observed at each study area. W hitebark pine microsites experienced the lowest minimum, mean, and maximum daily soil moisture compared to all other microsite types at moister locations, but among the highest soil moisture levels in more arid conditions. We also found an inverse relationship betw een wind/gust speed severity and differences

PAGE 80

64 in wind protection by microsite type. Where wind and gust speeds were lower, microsites sheltered by conifers experienced lower wind and gust speeds compared to rocks and unprotected sites. Extreme wind and gust speeds, however, precluded ameliorative effects based on microsite type. Introduction The link between local processes and community spatial structure is one of the central principles of landscape ecology (Callaway 1995, Turner et al. 2001). In plant co mmunities, decades of cross sectional observations and experiments have revealed that ecosystem structure emerges from a complex combination of negative and positive associations (Callaway and Walker 1997). Although many factors tilt the balance among thes e interactions, including the life stages and physiologies of constituent individuals (Callaway and Walker 1997), there is a general consensus that facilitation helps shape population and community structure in environments experiencing harsh abiotic stres s (e.g., Bertness and Callaway 1994, Choler et al. 2001, Callaway et al. 2002, Antonsson et al. 2009). Facilitative interactions among plants (Callaway 1995) Ñ nurse plant syndromes Ñ have been observed in many stressful environments, including deserts (Arriag a et al. 1993), salt marshes (Bertness and Hacker 1994) and alpine tundra (Whitehead, 1951). Similarly, where abiotic conditions reduce the likelihood of seed germination and seedling growth and survival, common geomorphic objects such as rocks and topogra phic niches Ñ nurse objects or conditions Ñ may also increase establishment and survival success (Castro et al. 2011). The magnitude of effect that facilitation has in governing community assembly in harsh environments relies on the moderation of climatic c onditions at the local scale

PAGE 81

65 (Maher et al. 2005). The term "microclimate" refer s to the site specific, fine scale abiotic conditions that are subject to variation attributable to small differences in air and soil temperature, light intensity, humidity, moi sture, and wind/gust speeds (Spittlehouse and Sathers 1990). The modus operandi of sheltering individuals in nurse plant and nurse object syndromes is the amelioration of microclimates to a degree that promotes plant germination and seedling survival (Pyat t et al. 2016). Mechanisms by which plants can be facilitative in harsh environments include solar radiation reduction (Germino et al. 2002), wind speed moderation (Baumeister and Callaway 2006), soil temperature amelioration (Chambers 2001), and soil nutr ient provision (Callaway et al. 1991); McIntire et al. (2016) found solar radiation and wind speed to be most important limitations on seedling establishment above treeline. Facilitative interactions are particularly important in the climatically stressfu l alpine treeline ecotone (ATE), the transition zone from closed canopy forest to alpine tundra (Callaway et al. 2002, Resler and Tomback 2008, Tomback et al. 2016 a ). Compared to lower elevations, conditions in the ATE typify abiotic stress for plants (e.g ., Grime 1977). Extreme winds (Marr 1977, McIntire et al. 2016), cold and variable temperatures (Kšrner and Paulsen, 2004), poorly developed soils (Arno and Hammerly 1984), aridity (Holtmeier 2009), nutrient scarcity (Stevens and Fox 1991), high solar rad iation and sky exposure (Stevens and Fox 1991, Maher et al. 2005), and truncated growing seasons (Arno and Hammerly 1984) decrease the likelihood of conifer seed germination and early seedling survival (e.g., Pansing et al. 2017). Harsch and Bader (2011) s ummarize the mechanisms that impede conifer survival in the ATE as: (1) growth limitation Ñ impaired biosynthesis in response to low growing season

PAGE 82

66 temperature, short growing season duration, or nutrient deficiency, (2) dieback Ñ physical damage due to high sn ow load, frost, frozen soil, excess radiation, and high wind; and 3) seedling mortality Ñ a result of physical damage caused by high snow load, frost, frozen soil, excess radiation, heat stress, high temperatures, and low precipitation. ATE conifer community development relies on the initial establishment of solitary trees in unprotected microsites, or leeward of nurse objects (Resler et al. 2005). Once established, these conifers create favorable leeward microsites by alleviating wind speeds (HŠttenschwiler and Smith 1999), reducing solar radiation and sky exposure (Germino et al. 2002, Maher et al. 2005), decreasing soil temperature variation (Pyatt et al. 2016), and/or increasing water availability (Callaway 1998). Tree islands form when conifers become est ablished in the leeward microsites, building a patch of krummholz trees (Marr 1977). The capacity to survive, establish, and grow as a solitary tree allows whitebark pine to contribute significantly to community structure and biodiversity in these ecosyste ms (Resler and Tomback 2008). Seed germination and early life stage survival patterns give rise to ATE conifer community structure over time (e.g., Maher and Germino 2006, Pansing et al. 2017), and are governed largely by local climatic conditi ons during t he growing season Ñ July through September (e.g., Kšrner and Paulsen, 2004; Hoch and Kšrner, 2009; Harsch and Bader, 2011). Air and soil temperatures do not limit tissue growth, but rather the investment of photo assimilates into structural growth (i.e., sin k activity; Germino and Smith 1999); low air and soil temperatures slow physiological processes, including root growth (LandhŠusser et al. 1996). Elevated sky exposure, radiation, and wind speeds decrease humidity and soil moisture (Holtmeier, 2009). High winds desiccate and

PAGE 83

67 damage branches (Marr 1977). Although much is known about the climatic conditions governing treeline dynamcs (e.g., Harsch and Bader 2011) and nurse plant object syndromes (e.g., Resler and Tomback 2008), studies comparing the differenc es in microclimate provided by nurse objects and/or nurse plants (such as Pyatt et al 2016) are lacking. In many ATEs across the Northern and Central Rocky Mountains, whitebark pine ( Pinus albicaulis Engelm, Family Pinaceae, Subgenus Strobus .) Ñ one of the m ost widely distributed five needle white pines in the U.S. and Canada Ñ serves as a foundation species , facilitating the survival and growth of leeward conifers more frequently than other species (e.g., Resler and Tomback 2008; Tomback et al. 2016 a ). Due to its slow growth, high water use efficiency (Callaway et al. 2000) and elevated carbon gain relative to other conifers (Bansal et al. 2011), whitebark pine tolerates stressful conditions in the ATE. Whitebark pine seedlings are also robust and tolerant of a variety of conditions, including poor soils and harsh seedbeds (Arno and Hoff 1990, Tomback et al. 2001, McCaughey and Tomback 2001). Whitebark pine's survival and recruitment success at high elevations may also be attributable to a coevolved mutualism wi th the corvid Clark's nutcracker ( Nucifraga columbiana ; Tomback 1982). Nutcrackers often bury whitebark pine seeds in small caches near protective objects, such as rocks, tree trunks, and fallen trees (Tomback 1978); they may also transport seeds from pare nt trees up to 32.6 kilometers in distance (Lorenz et al. 2011 ). This process results in the dispersal of whitebark pine seeds from upper subalpine forest to the highest elevations (Tomback 1986). Unrecovered seed caches may germinate and establish in open or sheltered microsites, facilitate leeward conifer recruitment, and eventually become the

PAGE 84

68 initiating conifer of tree islands. Additionally, nutcrackers cache near nurse objects disproportionately; seed caches in geomorphic niches or leeward of rocks boos t whitebark pine survival in the ATE (Tomback 1978, Tomback 1982, Maher et al. 2005, Resler et al. 2005, Resler and Tomback 2008, Pyatt et al 2016). Pyatt et al. (2016) postulated two explanations for whitebark pine's prevalence as a tree island initiator: (1) whitebark pine may present more frequent opportunities for microclimate amelioration, given its landscape abundance, or (2) it may provide better quality microclimate amelioration compared to other nurse plants, or nurse objects. Using a block design , they compared biophysical attributes in microsites leeward of whitebark pine, Engelmann spruce ( Picea engelmannii ), rocks, and in unprotected (exposed) microsites. They found whitebark pine and Engelmann spruce microsites experienced higher daily soil te mperature minima, lower daily soil temperature maxima, lower daily photosynthetically active radiation (PAR), and lower daily mean wind speeds than rock and unprotected microsites. Here, we examined growing season microclimate for two ATE communities on t he harsh Rocky Mountain eastern front to test whether microsites leeward of conifers, and particularly P. albicaulis , experience less extreme and variable microclimates than those of similarly sized rocks, or those in open, unprotected microsites (e.g., Py att et al. 2016), but using a simple random sampling design. In both these treeline communities, whitebark pine is the predominant treeline conifer, and a locally abundant tree island initi ator (Wagner, chapt. 2). In both communities examined by Pyatt et a l. (2016) , whitebark pine was the most frequent tree island initiator (Tomback et al. 2016a). In this study, we hypothesized that whitebark pine leeward microsites, in comparison to other

PAGE 85

69 microsite types, would be characterized by (1) higher minimum and lo wer maximum growing season air and soil temperatures (2) higher levels of percent relative humidity and soil moisture, and (3) lower wind and gust speeds. To test these hypotheses, we compared differences in biophysical variables among four microsite types at both study areas, for the growing season, following the methods of Pyatt et al. (2016). Our focus was comparative performance of four microsite types at opposite ends of the Greater Yellowstone Ecosystem over part of the growing season in each study ar ea. Methods Study A reas We performed microclimate assessments in two ATEs at northernmost and southernmost locations in the Greater Yellowstone Area (GYE, Fig. 3. 1), which is situated in northwestern Wyoming. In both study areas, conifer species included w hitebark pine, Engelmann spruce, and subalpine fir ( Abies lasiocarpa ), which occurred either solitarily or as members of tree islands (Wagner, chapt. 2) . At Tibbs Butte, Shoshone National Forest, the study area comprised approximately 4.2 ha on the northea st aspect, from 2,983 to 3,238 m elevation. The topography consisted of a continuous upslope gradient, and the terrain featured large, scattered rocks, and undeveloped soils stemming from a bedrock of Precambrian granite (Bevan 1923). Conifers were scarce and primarily solitary. Dominant understory vegetation included Geum rossii (R. Br.) Ser. , Potentilla diversifolia Lehm., and Saxifraga spp. (Tomback et al. 2016a). Our southernmost study site was located 250 km south of Tibbs Butte on Mutiny Ridge in the Wind River Range, Shoshone National Forest, approximately 1.5 km west of

PAGE 86

70 Christina Lake. The Mutiny Ridge ATE terrain is strewn with numerous large rocks and scattered pockets of soil; the bedrock consists of Precambrian (Archean) crystalline minerals, in cluding granodiorite, porphyritic quartz monzonite, and magmatic gneiss (e.g., Frost et al. 2000). Dominant vegetation includes Geum rossii (R. Br.) Ser. , Pteryxia hendersonii (J.M. Coult. and Rose) Mathias and Constance , Phlox pulvinata (Wherry) Cronquist , and Silene acaulis (L.) Jacq (Tomback et al. 2016a). The study area included 117 ha on northeast, southeast, and southwest aspects (3,200 Ñ 3,400 m elevation). W agner (chapt 2) performed mesoclimate comparisons between these two study areas using open sou rce Precipitation Elevation Regressions on Independent Slopes Model (PRISM) data for the years 1985 through 2015 (PRISM Climate Group, 2015). He found no significant differences in mean daily growing season temperature and precipitation between the study a reas. Study Design We defined a microsite as a cylindrical space 20 cm in diameter and 15 cm in height directly leeward of a conifer or rock, or on open ground with no windward protection (Pyatt et al. 2016). At each study area, we recorded microclimatic conditions directly leeward of the two most prevalent conifer species (Tibbs Butte: whitebark pine and subalpine fir; Mutiny Ridge: whitebark pine and Engelmann spruce), rocks, and in unprotected (exposed) microsites. The heights, lengths, and widths of n urse trees and rocks depended on local availability and ar e depicted in Appendix B , Fig. B . 1. We found no mean difference in nurse object height between Tibbs Butte and Mutiny Ridge. There were also no mean differences in nurse object height by microsite t ype.

PAGE 87

71 In 2014, we established twenty microclimatic weather stations at Tibbs Butte and in 2015 at Mutiny Ridge. E ach microsite type (i.e., P. albicaulis , A. lasiocarpa , rock, and unprotected on Tibbs Butte; P. albicaulis , P. Enelmannii , rock, and unprotecte d on Mutiny Ridge) had five replicates, resulting in a sample population of 20 leeward microsites, stratified by type. We employed a simple random sampling design, selecting microsite locations based on random positions generated using the splancs package in R (version 3.1.0, R Core Team 2014; Rowlingson and Diggle 2014). We also randomized the assignment of each nurse object to its corresponding random location. We determined leeward microsite position by the direction of wind flagging on branches of nearb y conifers. We defined the leeward microsite as directly opposite the source of wind causing the most prevalent branch flagging. At both study sites, we selected microsites not protected by objects other than the nurse object specified; those microsites de signated to be unprotected were not protected by any nearby object. Microsites at Tibbs Butte typically supported a higher percent ground cover (75% Ð 100%) than Mutiny Ridge (50% Ð 75%). Microclimatic A ssessments Using Onset Computer HOBO sensors and Micro S tation Data Loggers, we measured the following microclimate characteristics at each microsite for part of the growing season, from August 2 nd to September 5 th (2014 for Tibbs Butte and 2015 for Mutiny Ridge): air temperature (¡C) 10 cm above the ground (S TMB M002), percent relative humidity 10 cm above the ground (S TMB M002), soil temperature (¡C) 3 cm below ground (S TMB M017), soil moisture (m 3 m 3 ) 6 cm below ground (S SMC M00), wind speed (m s 1 ) 10 cm above ground (S WSA M003), and gust speeds (m s 1 ) 10 cm

PAGE 88

72 above ground (S WSA M003; Pyatt et al. 2016). The wind and gust speed sensor was an anemometer. We constructed protective housing using PVP piping and vinyl gutters to prevent damage to the sensors and cables during the recording interval (Pyatt et al. 2016; Fig. 3. 2). For placement of the air temperature and percent relative humidity sensor, we used a protective radiation shield mounted on a 7.6 cm wooden block; the apparatus was attached to a metal stake for stability (Pyatt et al. 2016). The soil temperature and soil moisture sensors were placed directly in the ground on opposite sides of the microsite (approximately 20 cm apart) and at least 4 cm from other sensors to avoid interference. The anemometer was staked near the center of the microsite. All the sensors delivered data to a Micro Station Data Logger established 1 m away from the microsite; the data logger was mounted on a vinyl gutter pipe 30 cm above the ground (Pyatt et al. 2016). To record all microclimate variables, we launched the Mic ro Station Data Loggers using HOBOware Pro software (version 3.3.0), and set each to record in 15 minute intervals. For wind speed, data loggers averaged automatically over each 15 minute interval from one minute observations. Gust speeds were designated a s the highest wind speed over any 3 second period within each 15 minute logging interval (Pyatt et al. 2016). On Tibbs Butte, all 20 Micro Station Data Loggers began recording on August 2, 2014 at 12:00am MST and were stopped on September 5, 2014 at 11:45p m MST. On Mutiny Ridge, all 20 Micro Station Data Loggers began recording on August 2, 2015 at 12:00am MST and were stopped on September 5, 2015 at 11:45pm MST. The recording duration was 35 days at both study areas for each year . Data Analysis

PAGE 89

73 We used R (version 3.2.4, R core team 2016) to conduct all statistical analyses. For response variables (i.e., air temperature, percent relative humidity, soil temperature, soil moisture, wind speed, and gust speed) at each microclimate weather station, we computed the following statistics prior to data analysis: (1) daily minimum, mean, and maximum for air temperature, percent relative humidity, soil temperature, and soil moisture, (2) daily mean and maximum for wind speed, and (3) daily maximum for gust speed. Mod el selection. We began the process of estimating growing season microclimatic parameters by microsite type at each study area by fitting intercept ( D % ' E F G ' H ), linear ( D % ' E F G ' E I J G ' H ), and quadratic ( D % ' E F G ' E I J G ' E K J K G ' H ) models to each response variable statistic at each station. In these models, y corresponds to the response variable statistic, x corresponds to time (in days), E 's are parameter estimates, and H are residuals (i.e., model errors; the distances between each data point and the model in the y direction). To determine which model type best explained the general behavior of each response variable statistic throughout the recording interval, we calculated Akaike's Information Criterion (AIC) scores for each m odel at each station, and selected the one with the lowest AIC score (i.e., the most parsimonious model type) over the majority of stations. By study area, we used the best fitting model for its associated response variable statistic for the remainder of t he analysis. Bootstrapped linear regressions. At this point, we deviated from the statistical methods of Pyatt et al. (2016), and selected an approach that precludes a priori assumptions about each response variable's underlying distribution. Once a model type had been selected for a given response variable statistic, we performed a bootstrapped

PAGE 90

74 linear regression of each response variable statistic ( D ) by time ( J , in days) for each microsite type at each study area Ñ that is, we (1) fit a model to each response variable statistic for each microsite type; (2) harvested the predicted values (the y predictions by day for each response variable statistic) and residuals (model errors) by microsite type; (3) for each of 5000 iterations, sampled randomly with replacement from the microsite specific residuals and added them to the predicted values to generate 20,000 bootstrapped samples, 5000 for each microsit e type; and (4) fit a model to each bootstrapped sample (n = 5000 models per microsite type). Parameter estimates. The bootstrap regression resulted in 5000 models per microsite type per response variable statistic, each predicting its corresponding resp onse variable statistic in terms of time (date from Aug. 2 ( L F ) to Sept 5 ( L ? )). For these models, the total area under the curve ( A ) over the duration of the recording interval is equivalent to the predicted seasonal average ( M <57 ) (Eq. 1). M <57 % ' N L ? ) ' L O ' (1) To estimate the microclimatic conditions at a given microsite type over the recording interval, we integrated each model with respect to time, and divided the integral by the total chang e in time (35 days). The calculation for estimating seasonal microclimatic conditions from an intercept model takes the following form: N % ' & E F , PL 4 Q 4 R intercept (2) N % S ' E F L ? ) L F ' T intercept (3)

PAGE 91

75 M <57< F ;7> % ' ( L ? ) ' L O ' S ' E F L ? ) L F ' T intercept (4) M <57 % ' E F intercept (5) The calculation for estimating seasonal microclimatic conditions form a linear model takes the following form: N % ' & E F G ' E I L , PL 4 Q 4 R linear (6 ) N % S ' E F L ? ) L F G E I L ? K ) L F K U ' T linear (7 ) M <57 % ( L ? ) ' L O ' S ' E F L ? ) L F G E I L ? K ) L F K U ' T linear (8 ) M <57 % ' E F G ' E I L ? G L F U ' linear (9) The calculation for estimating seasonal microclimatic conditions from a quadratic model takes the following form: N % ' & E F G ' E I L ' G ' E K L K ' , PL 4 Q 4 R quadratic (10 ) N % S ' E F L ? ) L F G E I L ? K ) L F K U G E K L ? V ) L F V W T quadratic (11 ) M <57 % ( L ? ) ' L O ' S ' E F L ? ) L F G E I L ? K ) L F K U G E K L ? V ) L F V W T quadratic (12 ) M <57 % ' E F G E I L ? G L F U G E K L ? K G L ? L F G L F K W quadratic (13) This process generated four sampling distributions for each response variable statistic, one corresponding to each microsite type. Then, we calculated the mean of each sampling distribution to identify the parameter estimate for each microclimatic condition at each

PAGE 92

76 microsite type. Finally, we used the 0.025 and 0.9 75 quantiles of each sampling distribution to determine the lower and upper boundary, respectively, of the 95% high density interval (HDI). Assessing bias . We employed the model selection, bootstrap regression, and integration to maximize the likelihood of achieving accurate parameter estimates for microclimate in each microsite type. To assess how our estimates differed from simply calculating the crude global mean (i.e., taking the mean of the raw data for each microsite over the entire recording interv al), we subtracted the crude mean from our estimated seasonal average ( M <57 ) by microsite during each iteration of the bootstrap regression. This process resulted in a sampling distribution of 5000 biases for each microsite type. We calculated the mean of each sampling distribution to identify the parameter e stimate for the bias in each microclimatic condition at each microsite type. We used the 0.025 and 0.975 quantiles of each sampling distribution to determine the lower and upper boundary, respectively, of the 95% high density interval (HDI). Parameter est imates for mean differences . We calculated pairwise mean differences in seasonal microclimatic conditions between microsite types by subtracting the bootstrap regression derived sampling distributions from one another. For example, we calculated the differ ence in daily minimum air temperature from August 2 nd through September 5 th between whitebark pine and Engelmann spruce on Tibbs Butte by subtracting the sampling distribution corresponding to whitebark pine from the sampling distribution of Engelmann spru ce. This calculation resulted in a distribution of mean differences between whitebark pine and Engelmann spruce. Once this had been done for each pairwise comparison at each study area, we calculated the mean of each mean

PAGE 93

77 difference in seasonal microclimat ic condition sampling distribution to identify its corresponding parameter estimate. We made Bonferroni adjustments based on the number of pairwise comparisons to determine which sampling distribution quantiles to use as the HDI lower and upper boundaries. The formula for these adjustments is: "#$ % ' * X + , where + is the number of pairwise comparisons; our * for all analyses in this study was 0.05. Comparisons among our four microsite types resulted in six pairwise comparisons, so we estimated the 99.17% HDI by calculating the 0.0042 and 0.9 958 quantiles of each sampling distribution as the lower and upper boundary. For all pairwise difference estimates, we concluded that seasonal microclimatic conditions did not differ (i.e., no mean difference) if the estimated HDI included zero. Parameter estimates for microclimatic variability. To assess variability in each response variable statistic by microsite, we calculated the standard deviation (SD) of each of the 20,000 sets of residuals derived from the bootstrap regression. For this, we followed the following process: We (1) fit a model to each response variable statistic for each microsite type, (2) harvested the residuals (model errors) by microsite type, (3) for each of 5000 iterations, sampled randomly with replacement from the microsite spec ific residuals, and (4) calculated the standard deviation of each set of residuals ( n = 5000 per microsite type). This process resulted in four sampling distributions corresponding to the variability in each response variable statistic, one corresponding t o each microsite type. Then, we calculated the mean of each sampling distribution to identify the parameter estimate for the variability of each microclimatic condition at each microsite type. We used the 0.025 and 0.975 quantiles of each sampling distribu tion to determine the lower and upper boundary, respectively, of the 95% high density interval (HDI).

PAGE 94

78 Microclimatic variability mean difference parameter est imates . As we had done mean difference parameter estimates between microsite types, we cal culated mean differences in seasonal microclimatic variability by subtracting the bootstrap regression derived sampling distributions of standard deviations from one another. In doing so, we achieved a sampling distribution of mean differences in variabili ty for each response variable statistic by microsite type. Once this had been done for each pairwise comparison for each study area, we calculated the mean of each mean difference in seasonal microclimatic condition variability sampling distribution to ide ntify its corresponding parameter estimate. As with previous pairwise difference HDIs, we made Bonferroni adjustments based on six pairwise comparisons. We estimated the 99.17% HDI by calculating the 0.0042 and 0.9958 quantiles of each sampling distributio n as the lower and upper boundary. For all pairwise difference estimates, we concluded that the variation in seasonal microclimatic conditions did not differ (i.e., no mean difference) if the estimated HDI included zero. Results Raw data for all daily resp onse variable statistics are displayed in Figs. 3. 3a and 3. 3b. Figs 3. 4a and 3. 4b display model selection results for those data. Additionally, density plots for all raw daily response variable statis tics are displayed in Appendix B , Figs. B. 2a and B. 2b. F or stand alone parameter estimates and 95% HDIs regarding seasonal microcli mat ic conditions, see Appendix B, F igs. B. 3a and B.3b, and Appendix B , Tables B. 1 through B. 5 for tabulated values. Biases for those parameter estimat es are represented in Appendix B, figs. B. 4a and B. 4b. The primary results of this study, pairwise mean difference parameter estimates for microclimatic conditions (effect sizes),

PAGE 95

79 are displayed in figures 3. 5a and 3. 5b, and their corresponding va lues are tabulated in Appendix B , Tables. B. 6 through B. 10. For parameter estimates and 95% HDIs regarding the variability of seasonal microcli matic conditions, see Appendix B , figures B. 5a and B.5b, or Appendix B , Tables. B. 11 through B. 15 for tabulated values. Finally, pairwise mean difference parameter estimates for microclimatic variability (effect sizes) can be found in figures 3. 6a and 3. 6b; their corresponding values are also tabulated in Appendix B , Tables. B. 16 through B. 20. Air T emperature The minimum, mean, and maximum daily air temper atures at Tibbs Butte decreased over the course of the growing season, but not linearly; air temperature declines were amplified later in the season following a quadratic curve (Figs. 3. 3a and 3. 4a). At Mutiny Ridge, the minimum, mean, and maximum daily ai r temperature were consistent over time following an intercept model (Figs. 3. 3b and 3. 4b). In general, Mutiny Ridge experienced higher minimum daily air temperatures than Tibbs Butte (Figs 3. 3a, 3. 3b, 3. 5a, 3.5b; Appendix B : B. 3a, B. 3b, and Tables B. 1 and B. 6). We found no differences in mean daily air temperature by microsite at either study area (Figs. 3. 5a, 3.5b; Appendix B : Figs. B. 3a, B. 3b, and Tables B. 1 and B. 6). At Tibbs Butte, we found that whitebark pine and Engelmann spruce microsites had warmer maximum daily air temperatures than rock microsites. We also found whitebark pine microsites had warmer maximum daily air temperatures than unprotected microsites. For Mutiny Ridge, we found significant differences between rock and unprotected microsites; rocks had a lower minimum and higher maximum daily air temperature than that of unprotected microsites.

PAGE 96

80 Our assessments of variability in seasonal microclimatic conditions revealed that whitebark pine microsites experienced significantly greater variabil ity in maximum daily air temperature than rock microsites at both study areas (Fig. 3. 6a and 3. 6b ; Appendix B : F ig. B. 4a and B. 4b , and Tables B. 11 and B. 16). Percent Relative H umidity For daily percent relative humidity on Tibbs Butte, the minimum and me an decreased over the course of the growing season following a quadratic curve; maximum daily percent relative humidity held relatively constant (Figs. 3. 3a and 3. 4a). The general trends for those response variable statistics were identical on Mutiny Ridge (Figs. 3. 3b and 3. 4b). In general, Tibbs Butte experienced higher minimum and mean daily percent relative humidity than Tibbs Butte (Figs 3. 3a, 3. 3b, 3. 5a, 3.5b; Appendix B : B. 3a, B. 3b, and Tables B. 2 and B. 7). Pairwise differences for minimum, mean, and maximum daily percent relative humidity on Tibbs Butte revealed several significant differences (Fig. 3.5a; Appendix B : Fig. B.3a, and Tables B. 2 and B. 7). First, we found that whitebark pine microsites had a higher minimum and lower maximum daily percent relative humidity than that of rock microsites and Engelmann spruce microsites. Second , we found unprotected microsites had higher minima and lower maxima for daily percent relative humidity compared to rock and Engelmann spruce microsites. Third, Engelman n spruce microsites had a higher mean daily percent relative humidity than whitebark pine and unprotected microsites . Finally, we found Engelmann spruce microsites had the highest maximum daily percent relative humidity compared to all other microsites.

PAGE 97

81 W e found no differences in minimum and mean daily percent relative humidity by microsite on Mutiny Ridge (Fig. 3. 5b; Appendix B : Fig. B. 3b, and Tables B. 2 and B. 7). H owever, whitebark pine microsites had the lowest daily maximum percent relative humidity co mpared to all other microsite types. Our assessments at Tibbs Butte revealed that whitebark pine microsites experienced significantly less variability in mean and maximum daily percent relative humidity than rock and Engelmann spruce microsites, which in turn had significantly less variability in maximum daily percent relative humidity compared to unprotected microsites (Fig. 3. 6a; App endix B : Fig. B. 4a, and Tables. B. 12 and B. 17). At Mutiny Ridge, we found that whitebark pine microsites experienced signif icantly more variability in maximum daily percent relative humidity than rock microsites (Fig. 3.6b; Appendix B : Fig. B. 4b, and Tables. B. 12 and B. 17). Soil Temperature Mirroring air temperature trends, the minimum, mean, and maximum daily soil temperatur e at Tibbs Butte decreased over the course of the growing season, but not linearly; declines were amplified later in the season following a quadratic curve (Figs. 3. 3a and 3. 4a). For soil temperature on Mutiny Ridge, the minimum and mean were relatively co nsistent over the course of the growing season, but the maximum daily percent relative humidity decreased following a quadratic curve (Figs. 3. 3b and 3. 4b). In general, Mutiny Ridge experienced higher minimum daily soil temperatures than Tibbs Butte (Figs 3. 3a, 3. 3b, 3. 5a, 3.5b; Appendix B : B. 3a, B. 3b, and Tables B. 3 and B. 8). At Tibbs Butte, the mean daily soil temperature in whitebark p ine microsites was lower than that of Engelmann spruce microsites (Fig. 3.5a; Appendix B : Fig. B. 3a, and

PAGE 98

82 Tables B. 3 and B . 8). We also found that the daily minimum soil temperature in Engelmann spruce and rock microsites was higher than that of unprotected microsites. For daily maximum soil temperature, we found Engelmann spruce microsites to be warmer than rock microsites. The minimum and mean daily soil temperature at Mutiny Ridge was significantly higher for whitebark pine microsites than for subalpine fir and unprotected microsites (Fig. 3. 5b; App endix B : Fig. B.3b, and Tables B.3 and B. 8). We also found that the minimum and mean daily soil temperatures in rock microsites were warmer than that of unprotected microsites. We found the minimum daily soil temperature was greater in rock microsites than that of subalpine fir microsites. Our assessments of seasonal microclimatic variability at Tibbs Butte revealed no significant di fferences by microsite type for mean daily soil temperature. However, we found the minimum daily soil temperature in whitebark pine microsites varied more than unprotected microsites. W e also found the maximum daily soil temperature in whitebark pine microsites varied more than any other microsite type throughout the season (Fig. 3.6a; Appendix B : Fig. B. 4a, and Tables B. 13 and B. 18). We also found the maximum daily soil temperatures varied more in Engel mann spruce microsites than in rock microsites. At Mutiny Ridge, unprotected microsites varied significantly more than rock microsites for minimum, mean, and maximum soil temperature (Fig. 3. 6b; Appendix B : Fig. B. 4b, and Tables B. 13 and B. 18). For mean a nd maximum daily soil temperature, we found that whitebark pine microsites experienced greater variability than rock microsites. W e also found whitebark pine microsites varied less than subalpine fir and unprotected

PAGE 99

83 microsites for minimum daily soil temper ature. Finally, we found that subalpine fir microsites were more variable in minimum daily soil temperature than rock microsites. Soil Moisture Throughout the course of the growing season on Tibbs Butte, the minimum and mean daily soil moisture increased following a quadratic curve; the maximum daily soil moisture held constant (Figs. 3. 3a and 3. 4a). Conversely, on Mutiny Ridge the minimum, mean, and maximum daily soil moisture decreased linearly over the course of the season (Figs. 3. 3b and 3. 4b). Every pairwise comparison for minimum, mean, and maximum soil moisture on Tibbs Butte revealed a significant difference, establishing a clear hierarchy (Fig. 3.5a; Appendix B : Fig. B. 3a, and Tables B. 4 and B. 9). Whitebark pine microsites experienced the lowest minimum, mean, and maximum daily soil moisture, followed by rock, unprotected, and Engelmann spruce microsites. At Mutiny Ridge, however, whitebark pine microsites did not experience the driest conditions. In fact, whiteb ark pine microsites experienced hig her soil moisture levels than subalpine fir microsites (Fig. 3.5b; appendix B : Fig. B.3b, and Tables B. 4 and B. 9). Additionally, whitebark pine microsites experienced higher daily minimum soil moisture than unprotected microsites. W e also found rock micros ites had higher minimum and mean daily soil moisture than that of subalpine fir microsites. We found that whitebark pine microsites experienced the most variable maximum daily soil moisture at Tibbs Butte, followed by Engelmann spruce, unprotected, and ro ck microsites, respectively (Fig. 3. 6a; Appendix B : Fig. B. 4a, and Tables B. 14 and B. 19). Results for minimum and mean daily soil moisture showed that whitebark pine and

PAGE 100

84 Engelmann spruce microsites had greater variability than rock and unprotected microsites. For Mutiny Ridge, results were the same for minimum, mean, and maximum soil moisture variability, with whitebark pine microsites experiencing more variability than any other microsite type assessed (Fig. 3. 6b; Appendix B : Fig. B. 4b, and Tables B. 14 and B. 19). Rock mic rosites experienced more variability than subalpine fir and unprotected microsites. Wind and Gust S peed The minimum, mean, and maximum daily wind and gust speeds at Tibbs Butte increased linearly over the course of the growing season (Figs. 3. 3a and 3. 4a). At Mutiny Ridge, minimum, mean, and maximum wind and gust speeds increased, but the increases were amplified later in the season following a quadratic curve (Figs. 3. 3b and 3. 4b). In general, Mutiny Ridge experienced greater mean and maximum wind speeds, and grea ter maximum gust speeds, than Tibbs Butte (Figs 3. 3a, 3. 3b, 3. 5a, 3. 5b; Appendix B : B. 3a, B. 3b, and Tables B. 5 and B. 10). On Tibbs Butte, we observed whitebark pine and Engelmann spruce microsites experienced the lowest wind and gust speeds across every statistic assessed (Fig. 3. 5a; Appendix B : Fig. B. 3a, and Tables B. 5 and B. 10). Rock microsites experienced lower mean and maximum daily wind speeds compared to unprotected microsites. At Mutiny Ridge, our analysis revealed no significant differences for daily mean wind speed or m aximum daily gust speed (Fig. 3. 5b; Appendix B : Fig. B. 3b, and Tables B. 5 and B. 10). Interestingly, we found maximum daily wind speeds were significantly

PAGE 101

85 greater in subalpine fir microsites compared to whitebark pine, rock, and unprotected microsites. Our assessm ents of variability of microclimatic conditions at Tibbs Butte revealed that Engelmann spruce experienced the least variability in daily mean and maximum wind speed s relative to any other microsite type (Fig. 3. 6a; Appendix B : Fig. B. 4a, and Tables B. 15 and B. 20). For daily maximu m gust speed , we found that whitebark pine experienced greater variability than all other microsite types. At Mutiny Ridge, we found that unprotected microsites had more variable mean daily wind speeds than whitebark pine, subalpine fir, and rocks (Fig. 3. 6b; Appendix B : Fig. B. 4b, and Tables B. 15 and B. 20). We also found that microsites protected by subalpine fir experienced greater maximum daily wind speed variability than rock sheltered microsites. For maximum daily gust speed, we found that su balpine fir microsites had the most variable conditions. Finally, we found whitebark pine microsites had less variable maximum daily gust speeds than unprotected microsites. Discussion Our primary objective for this study was to determine whether whiteba rk pine's prevalence as a tree island initiator in many Rocky Mountain ATEs was attributable to a leeward microclimate that ameliorated the extreme ATE environment. We hypothesized that microsites leeward of whitebark pine would experience (1) higher minim um and lower maximum growing season air and soil temperatures, (2) higher percent relative humidity and soil moisture, and (3) lower wind and gust speeds compared to other conifers and nurse objects.

PAGE 102

86 We found that whitebark pine leeward microsites did not consistently provide the most moderate leeward microclimate. Rather, differences in microclimatic conditions by microsite type depend substantially on general climatic conditions at each study area. At Tibbs Butte, where soil moisture was higher and incre ased later in the season, we found that w hitebark pine microsites experienced the lowest minimum, mean, and maximum daily soil moisture compared to all other microsite types. Conversely, whitebark pine microsites were generally more moist than other micros ite types at Mutiny Ridge, where soil moisture conditions were lower and potentially more limiting for tree survival. We also found an inverse relationship between wind and gust speeds and microclimate amelioration differences by microsite type. At Tibbs Butte, where wind and gust speeds were lower than that on Mutiny Ridge, we found that microsites leeward of conifers experienced lower wind and gust speeds compared to rocks and unprotected sites. At Mutiny Ridge, however, extreme wind and gust speeds prec luded microsite ameliorative effects. When comparing biophysical attributes in microsites leeward of whitebark pine, Engelmann spruce (Picea engelmannii), rocks, and in unprotected (exposed) microsites, Pyatt et al. (2016) found whitebark pine and Engelman n spruce microsites experienced higher daily soil temperature minima, lower daily soil temperature maxima, lower daily photosynthetically active radiation (PAR), and lower daily mean wind speeds than rock or unprotected microsites. Our findings were in lin e with Pyatt et al. (2016) in that the minimum daily soil temperatures were higher leeward of Engelmann spruce, but did not find evidence for minimum soil temperature moderation by whitebark pine. In fact, we observed generally higher levels of soil temper ature variability in whitebark pine

PAGE 103

87 microsites compared to other microsites. For wind/gust speeds, our findings at Tibbs Butte are in line with those observed by Pyatt et al. (2016); conifers significantly ameliorated mean and maximum daily wind speeds, as well as maximum daily wind speeds. However, at Mutiny Ridge, where the ATE experienced significantly greater wind/gust speeds than Tibbs Butte, our results deviated from their general findings. Finally, our results also agree with Pyatt et al. (2016) in t hat Engelmann spruce may produce a somewhat more favorable microclimate than whitebark pine in regards to mean daily wind speeds.

PAGE 104

88 Tables and Figures Figure 3. 1. Research study areas in the Greater Yellowstone Ecosystem. Study areas include , Tibbs Butte , Shoshone National Forest, WY (44¡56'46.48" N, 109¡26'39.76" W), and Mutiny Ridge, Shoshone National Forest, WY (42¡35'33.56" N, 108¡58'07.44" W). Map created using ggmap package in R (Kahle and Wickham 2013).

PAGE 105

89 Figure 3. 2. Microclimate weather station . The weather station is established in an unprotected (exposed) microsite on Mutiny Ridge. Photocredit: Elizabeth Pansing.

PAGE 106

90 Figure 3. 3a. Time series plot s for Tibbs Butte. Plots display raw response variable statistics for all biophysical variables at Ti bbs Butte. Circles are daily averages for each weather station, and lines are daily averages for each microsite type.

PAGE 107

91 Figure 3. 3b. Time series plot s for Tibbs Butte. Plots display raw response variable statistics for all biophysical variables at Mutiny Ridge. Circles are daily averages for each weather station, and lines are daily averages for each microsite type.

PAGE 108

92 Figure 3. 4a. Best model fits for each response variable statistic at Tibbs Butte.

PAGE 109

93 Figure 3. 4b. Best model fits for each response vari able statistic at Mutiny Ridge.

PAGE 110

94 Figure 3. 5a. Parameter estimates for mean differences in microclimate at Tibbs Butte . Violin plot displaying bootstrapped sampling distributions (curves), estimates (open circles), and 99.17% HDIs (lines) for pairwise me an differences in seasonal microclimatic conditions by microsite type at Tibbs Butte. R = Rock, P = Pine, U = Unprotected, and S = Spruce. Asterisks (*) denote significant differences.

PAGE 111

95 Figure 3. 5b. Parameter estimates for mean differences in microclima te at Mutiny Ridge . Violin plot displaying bootstrapped sampling distributions (curves), estimates (open circles), and 99.17% HDIs (lines) for pairwise mean differences in seasonal microclimatic conditions by microsite type at Mutiny Ridge. R = Rock, P = P ine, U = Unprotected, and F = Fir. Asterisks (*) denote significant differences.

PAGE 112

9 6 Figure 3. 6a . Microclimatic variability mean difference parameter estimates for Tibbs Butte . Violin plot displaying bootstrapped sampling distributions (curves), estimates (open circles ), and 99.17% HDIs (lines) for pairwise mean differences in seasonal microclimatic standard deviations by microsite type at Tibbs Butte. R = Rock, P = Pine, U = Unprotected, and S = Spruce. Asterisks (*) denote significant differences.

PAGE 113

97 Fig ure 3. 6b. Microclimatic variability mean difference parameter estimates for Mutiny Ridge. Violin plot displaying bootstrapped sampling distributions (curves), estimates (open circles), and 99.17% HDIs (lines) for pairwise mean differences in seasonal micro climatic standard deviations by microsite type at Mutiny Ridge. R = Rock, P = Pine, U = Unprotected, and F = Fir. Asterisks (*) denote significant differences.

PAGE 114

98 Appendix B Figure B. 1. Nurse object widths, lengths, and heights. To estimate each, we boots trapped for 5000 iterations to generate a sampling distribution of mean widths, lengths, and heights. We calculated the mean of each sampling distribution to identify its parameter estimate. We used the 0.025 and 0.975 quantiles of each sampling distributi on to determine the lower and upper boundary, respectively, of the 95% high density interval (HDI). Estimates are indicated by open circles, and lines delineate 95% HDIs.

PAGE 115

99 Figure B. 2a. Density plots for raw biophysical variables at Tibbs Butte.

PAGE 116

100 Figur e B. 2b. Density plots for raw biophysical variables at Mutiny Ridge.

PAGE 117

101 Figure B. 3a. Parameter estimates for microclimate at Tibbs Butte . Violin plot displaying bootstrapped sampling distributions (curves), estimates (open circles), and 95% HDIs (lines) o f seasonal microclimatic conditions by microsite type at Tibbs Butte.

PAGE 118

102 Figure B. 3b. Parameter estimates for microclimate at Mutiny Ridge. Violin plot displaying bootstrapped sampling distributions (curves), estimates (open circles), and 95% HDIs (lines) of seasonal microclimatic conditions by microsite type at Mutiny Ridge.

PAGE 119

103 Figure B. 4a. Parameter estimates for bias es: Tibbs Butte. Violin plot displaying bootstrapped sampling distributions (curves), estimates (open circles), and 95% HDIs (lines) of bi as by microsite type at Tibbs Butte. Bias is the difference between the seasonal climactic estimates derived from modelling and the aggregate temporal mean.

PAGE 120

104 Figure B. 4b. Parameter estimates for biases: Mutiny Ridge. Violin plot displaying bootstrapped sampling distributions (curves), estimates (open circles), and 95% HDIs (lines) of bias by microsite type at Mutiny Ridge. Bias is the difference between the seasonal climactic estimates derived from modelling and the aggregate temporal mean.

PAGE 121

105 Figure B . 5a. Parameter estimates for microclimatic variability at Tibbs Butte. Violin plot displaying bootstrapped sampling distributions (curves), estimates (open circles), and 99.17% HDIs (lines) of standard deviations in seasonal microclimatic conditions by mic rosite type at Tibbs Butte.

PAGE 122

106 Figure B. 5b. Parameter estimates for microclimatic variability at Mutiny Ridge. Violin plot displaying bootstrapped sampling distributions (curves), estimates (open circles), and 99.17% HDIs (lines) of standard deviations in seasonal microclimatic conditions by microsite type at Mutiny Ridge.

PAGE 123

107 Table B. 1. Air temperature bootstrapped parameter estimates and 95% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock , and UN = Unprotected.

PAGE 124

108 Table B. 2. Percent relative humidity bootstrapped parameter estimates and 95% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 125

109 Table B. 3. Soil temperature bootstrapped parameter estimates and 95% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 126

110 Table B. 4. Soil moisture bootstrapped parameter e stimates and 95% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 127

111 Table B. 5. Wind and gust speed bootstrapped parameter estimates and 95% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 128

112 Table B. 6. Air temperature bootstrapped mean difference parameter estimates and 99.17% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 129

113 Table B. 7. Percent relative humidity bootstrapped mean difference parameter estimates and 99.17% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 130

114 Table B. 8. Soil temperature bootstrapped mean difference parameter estimates and 99.17% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 131

115 Table B. 9. Soil moisture bootstrapped mean difference parameter estimates and 99.17% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 132

116 Table B. 10. Wind and gust speed bootstrapped mean difference parameter estimates and 99.17% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected .

PAGE 133

117 Table B. 11. Air temperature standard deviation bootstrapped parameter estimates and 95% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 134

118 Table B. 12. Percent relative humidity standard deviation bootstrapped parameter estimates and 95% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 135

119 Table B. 13. Soil temperature stan dard deviation bootstrapped parameter estimates and 95% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 136

120 Table B. 14. Soil moisture standard deviation bootstrappe d parameter estimates and 95% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 137

121 Table B. 15. Wind and gust speed standard deviation bootstrapped parameter estimate s and 95% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 138

122 Table B. 16. Air temperature standard deviation bootstrapped mean difference parameter estimates and 99 .17% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 139

123 Table B. 17. Percent relative humidity standard deviation bootstrapped mean difference parameter estimates an d 99.17% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 140

124 Table B. 18. Soil temperature standard deviation bootstrapped mean difference parameter estimates and 99. 17% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 141

125 Table B. 19. Soil moisture standard deviation bootstrapped mean difference parameter estimates and 99.17% HDIs . TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 142

126 Table B. 20. Wind and gust speed standard deviation bootstrapped mean difference parameter estimates and 99.17% HDIs. TB = Tibbs Butte, MR = Mutiny Ridge; WB = Whitebark pine, ES = Engelmann spruce, SF = Subalpine fir, RK = Rock, and UN = Unprotected.

PAGE 143

127 REFERENCES Alvarez ! Uria, P., & Kšrner, C. (2007). Low temperature limits of root growth in deciduous and evergreen temperate tree species. Functional Ecology , 21.2 , 211 218. Amman, G.D., McGregor, M.D., Cahill, D.B., & Klein, William H. (1977). Guidelines for reducing loss es of lodgepole pine to the mountain pine beetle in unmanaged stands in the Rocky Mountains. Gen. Tech. Report INT 36. Ogden, UT: USDA Forest Service, Intermountain Forest & Range Experiment Station. 1:19. Antonsson, H., Bjšrk, R. G., and Molau, U. (2009 ). Nurse plant effect of the cushion plant Silene acaulis (L.) Jacq. in an alpine environment in the subarctic Scandes, Sweden. Plant Ecology & Diversity , 2: 1 , 17 25. Arno, S.F., & Hammerly, R.P. (1984). Timberline: mountain and arctic forest frontiers . T he Mountaineers . Seattle, WA. Arno, S. F., & Hoff, R. J. (1990). Pinus albicaulis Engelm. whitebark pine. Silvics of North America , 1 , 268 279. Arriaga, L., Maya, Y., Diaz, S., & Cancino, J. (1993). Association between cacti and nurse perennials in a het erogeneous tropical dry forest in northwestern Mexico. Journal of Vegetation Science , 4 :3, 349 356. Bansal, S., Reinhardt, K., & Germino, M. J. (2011). Linking carbon balance to establishment patters: comparison of whitebark pine and Engelmann spruce seedl ings along an herb cover exposure gradient at treeline. Plant Ecol . 212 , 219 Ð 228. Barringer, L.E., Tomback, D.F., Wunder, M.B., & McKinney, S.T. (2012). Whitebark pine stand condition, tree abundance, and cone production as predictors of visitation b y Clark's Nutcracker. PloS ONE , 7 : e37663, 1 11. Baumeister, D., & Callaway, R. M. (2006). Facilitation by Pinus flexilis during succession: a hierarchy of mechanisms benefits other plant species. Ecology , 87 :7 , 1816 1830. Bertness, M. D., & Callaway, R. (1994). Positive interactions in communities. Trends in Ecology & Evolution , 9.5 , 191 193. Bertness, M. D., and Hacker, S. D. (1994). Physical stress and positive associations among marsh plants. The American Naturalist , 144 :3 , 363 372. Bevan, A. (1929). Rocky Mountain Front in Montana. Geological Society of America Bulletin , 40.2 , 427 456.

PAGE 144

128 Blakeslee, S.C. (2012) Assessing whitebark pine vigor and facilitation in the alpine treeline ecotone. Diss. University of Colorado Denver. Bockino, N. K., & Tinker , D. B. (2012). Interactions of white pine blister rust and mountain pine beetle in whitebark pine ecosystems in the southern Greater Yellowstone Area. Natural Areas Journal , 32 (1), 31 40. Burns, K. S., Schoettle, A. W., Jacobi, W. R., & Mahalovich, M. F. (2008). Options for the management of white pine blister rust in the Rocky Mountain Region. Butler, D.R., Malanson, G.P., Walsh, S.J. & Fagre, D.B. (20 07) Influences of geomorphology and geology on alpine treeline in the American West Ð more important t han climatic influences? Physical Geography , 28.5 , 434 450. Caldwell, M. M. (1971). Solar UV irradiation and the growth and development of higher plants. Photophysiology , 6 , 131 177. Chicago Callaway, R. M., Nadkarni, N. M., and Mahall, B. E., (1991) Fac ilitation and interference of Quercus douglasii on understory productivity in central California. Ecology, 72 , 1484 Ð 1499. Callaway, R. M. (1995). Positive interactions among plants. The Botanical Review , 61:4 , 306 349. Callaway, R. M., and Walker, L. R. (1997). Competition and facilitation: a synthetic approach to interactions in plant communities. Ecology , 78 (7), 1958 1965. Callaway, R.M. (1998). Competition and facilitation on elevation gradients in subalpine forests of the northern Rocky Mountains, US A. Oikos , 82.3 , 561 573. Callaway, R. M., Sala, A., & Keane, R.E. (2000). Succession may maintain high leaf area: sapwood ratios and productivity in old subalpine forests. Ecosystems , 3 , 254 Ð 268. Callaway, R. M., Brooker, R. W., Choler, P., Kikvidze, Z. , Lortie, C. J., Michalet, R., ... & Armas, C. (2002). Positive interactions among alpine plants increase with stress. Nature , 417 :6891 , 844 848. Castro, J., Allen, C. D., Molina ! Morales, M., Mara–—n ! JimŽnez, S., S‡nchez ! Miranda, ç., and Zamora, R. (2011 ). Salvage Logging Versus the Use of Burnt Wood as a Nurse Object to Promote Post ! Fire Tree Seedling Establishment. Restoration Ecology , 19 (4), 537 544. Chambers, J. C., 2001: Pinus monophylla establishment in an expanding Pinus ! Juniperus woodland: Environ mental conditions, facilitation and interacting factors. Journal of Vegetation Science, 12 , 27 Ð 40.

PAGE 145

129 Choler, P., Michalet, R., and Callaway, R. M. (2001). Facilitation and competition on gradients in alpine plant communities. Ecology , 82 :12 , 3295 3308. Co le, W. E., & Amman, G. D. (1969). Mountain pine beetle infestations in relation to lodgepole pine diameter , 95 , US Dept. of Agriculture, Forest Service, Intermountain Forest & Range Experiment Station. Cui, M., and Smith, W. K. (1991). Photosynthesis, wate r relations and mortality in Abies lasiocarpa seedlings during natural establishment. Tree Physiology, 8 , 37 Ð 46. Dayton, P.K. (1972). Toward an understanding of community resilience and the potential effects of enrichments to the benthos at McMurdo Sound , Antarctica. Proceedings of the colloquium on conservation problems in Antarctica , Allen Press, Lawrence, Kansas, USA. Ellison, A. M., & Farnsworth, E. J. (2001). Mangrove communities. Marine community ecology , 16 , 423 442. Sinauer Associates Sunderland, MA, USA. Ellison, A.M., Bank, M.S., Clinton, B.D., Colburn, E.A., Elliott, K., Ford, C.R., Foster, D.R., Kloeppel, B.D., Knoepp, J.D., Lovett, G.M., Mohan, J., Orwig, D.A., Rodenhouse, N.L., Sobczak, W.V., Stinson, K.A., Stone, J.K., Swan, C.M., Thomps on, J., Von Holle, B. & Webster, J.R. (2005). Loss of foundation species: consequences for the structure and dynamics of forested ecosystems. Frontiers in Ecology and the Environment , 3.9 , 479 486. Ellison, A. M., Barker ! Plotkin, A. A., Foster, D. R., & Orwig, D. A. (2010). Experimentally testing the role of foundation species in forests: the Harvard Forest Hemlock Removal Experiment. Methods in Ecology and Evolution , 1.2 , 168 179. Ellison, A. M. (2014). Experiments are revealing a foundation species: a case study of eastern hemlock ( Tsuga canadensis ). Advances in Ecology . Farnes, P. E. (1990). SNOTEL and snow course data: describing the hydrology of whitebark pine ecosystems. General technical report INT (USA). Frich, P., Alexander, L. V., Della Marta, P., Gleason, B., Haylock, M., Tank, A. K., & Peterson, T. (2002). Observed coherent changes in climatic extremes during the second half of the twentieth century. Climate Research , 19.3 , 193 212. Frost, B.R.; Chamberlain, K.R.; Swapp, S.; Frost, C.D.; Hulsebosch, T.P. (2000). Late Archean structural and metamorphic history of the Wind River Range: Evidence for a long lived active margin on the Archean Wyoming craton. Geological Society of America Bulletin. 112.4 , 564 Ð 578. Germino, M. J., and Smith, W. K. (1999). Sky exposure, crown architecture, and low ! temperature photoinhibition in conifer seedlings at alpine treeline. Plant, Cell & Environment , 22.4 , 407 415.

PAGE 146

130 Germino, M. J., Smith, W. K., & Resor, A. C. (2002). Conifer seedling distribution and survival in an alpine treeline ecotone. Plant Ecology , 1 62.2 , 157 168. Grace, J., Berninger, F. & Nagy, Laszlo. (2002). Impacts of climate change on the tree line. Annals of Botany , 90.4 , 537 544. Greater Yellowstone Whitebark Pine Monitoring Working Group (GYWPMWG). (2015). Monitoring whitebark pine in the Greater Yellowstone ecosystem ecosystem: 2015 annual report. Natural Resource Data SeriesNPS/GRYN/NRDS Ñ 2015/278. National Park Service, Fort Collins, Colorado Griggs, R. F. (1938). Timberlines in the northern Rocky Mountains. Ecology , 19 .4 , 548 564. Gri me, J. P. (1977). Evidence for the existence of three primary strategies in plants and its relevance to ecological and evolutionary theory. The American Naturalist , 111.982 , 1169 1194. Hanski, I. (1982). Dynamics of regional distribution: the core and sate llite species hypothesis. Oikos . 1 :210 221. Harsch, M.A., Hulme, P.E., McGlone, M.S. & Duncan, R.P. (2009) Are treelines advancing? A global meta analysis of treeline response to climate warming. Ecology Letters , 12 , 1040 1049. Harsch, M. A., & Bader, M . Y. (2011). Treeline form Ð a potential key to understanding treeline dynamics. Global Ecology and Biogeography , 20 (4), 582 596. Harvell, C. D., Mitchell, C. E., Ward, J. R., Altizer, S., Dobson, A. P., Ostfeld, R. S., & Samuel, M. D. (2002) Climate warmin g and disease risks for terrestrial and marine biota. Science , 296 :5576 , 2158 2162. HŠttenschwiler, S. and Smith, W.K. (1999). Seedling occurrence in alpine treeline conifers: a case study from the central Rocky Mountains, USA. Acta Oecologica , 20.3 , 219 224. Hoch, G., & Kšrner, C. (2009). Growth and carbon relations of tree line forming conifers at constant vs. variable low temperatures. Journal of Ecology , 97 :1 , 57 66. Hoff, R. J., Ferguson, D. E., McDonald, G. I., & Keane, R. E. (2001). Strategies for m anaging whitebark pine in the presence of white pine blister rust. In Tomback, D. F., Arno, S. F., & Keane, R. E. (eds.), Whitebark Pine Communities: Ecology and Restoration. Washington, D.C.: Island Press, 346 Ð 366.

PAGE 147

131 Holdenrieder, O., Pautasso, M., Weisb erg, P. J., & Lonsdale, D. (2004). Tree diseases and landscape processes: the challenge of landscape pathology. Trends in Ecology & Evolution , 19:8 , 446 452. Holtmeier, F. (2009). Mountain timberlines: ecology, patchiness, and dynamics . Vol. 36, Germany. Hurlbert, S. H. (1984). Pseudoreplication and the design of ecological field experiments. Ecological monographs , 54 .2 , 187 211. Hutchins, H.E. & Lanner , R.M . (1982) The central role of Clark's nutcracker in the dispersal and establi shment of whitebark pi ne. Oecologia , 55.2 , 192 201. Jean, C., Shanahan, E., Daley, R., DeNitto, G., Reinhart, D., & Schwartz, C. (2011). Monitoring white pine blister rust infection and mortality in whitebark pine in the Greater Yellowstone Ecosystem. Kahle, D and Wickham, H. Ggmap: Spatial Visualization with ggplot2. The R Journal, 5:1 , 144 161. URL http://journal.r project.org/archive/2013 1/khale wickhap.pdf . Kšrner, C., & Paulsen, J. (2004). A wo rld ! wide study of high altitude treeline temperatures. Journal of Biogeography , 31.5 , 713 732. Kšrner, C . (2008) Winter crop growth at low temperature may hold the answer for alpine treeline formation. Plant Ecology & Diversity , 1.1 , 3 11. Krugman, S. L., & Jenkinson, J. L. (1974). Pinus L. pine. Seeds of shrubs in the United States , 598 638. LandhŠusser, S. M., Wein, R. W., and Lange, P., 1996: Gas exchange and growth of three Arctic tree line tree species under different soil temperature and drought prec onditioning regimes. Canadian Journal of Botany , 74 : 686 Ð 693. Little, R. L., Peterson, D. L., & Conquest, L. L. (1994). Regeneration of subalpine fir (Abies lasiocarpa) following fire: effects of climate and other factors. Canadian Journal of Forest Resea rch , 24 .5 , 934 944. Lloyd, A.H., & G raumlich, L.J . (1997) Holocene dynamics of treeline forests in the Sierra Nevada. Ecology , 78.4 , 1199 1210. Lorenz, T. J., and Sullivan, K. A. (2009). Seasonal differences in space use by Clark's Nutcrackers in the Cas cade Range. The Condor , 111 .2 , 326 340. Lorenz, T. J., Sullivan, K. A., Bakian, A. V., & Aubry, C. A. (2011). Cache site selection in Clark's Nutcracker (Nucifraga columbiana). The Auk , 128 (2), 237 247.

PAGE 148

132 Maher, E. L., Germino, M. J., & Hasselquist, N. J. (2005). Interactive effects of tree and herb cover on survivorship, physiology, and microclimate of conifer seedlings at the alpine tree line ecotone. Canadian Journal of Forest Research , 35.3 , 567 574. Maher, E. L., & Germino, M. J. (2006). Microsite dif ferentiation among conifer species during seedling establishment at alpine treeline. Ecoscience , 13 (3), 334 341. Malanson, G.P., Butler, D.R., Fagre, D.B., Walsh, S.J., Tomback, D.F., Daniels, L.D., Resler, L.M., Smith, W.K., Weiss, D.J., Peterson, D.L., Bunn, A.G., Hiemstra, C.A., Liptzin, D., Bourgeron, P.S., Shen, Z. & Millar, C. (2007). Alpine treeline in western North America: linking organism to landscape dynamics. Invited contribution for special issue on alpine treeline. Physical Geography , 28 , 378 396. Malanson, G.P., Butler, D.R., Fagre, D.B., Walsh, S.J., Tomback, D.F., Daniels, L.D., Resler, L.M., Smith, W.K., Weiss, D.J., Peterson, D.L., Bunn, A.G., Hiemstra, C.A., Liptzin, D., Bourgeron, P.S., Shen, Z. & M illar, C. (2007) Alpine treeline in western North America: linking organism to landscape dynamics. Invited contribution for special issue on alpine treeline. Physical Geography , 28 , 378 396. Marr, J.W. (1977). The development and movement of tree islands near the upper limit of Tree grow th in the southern Rocky Mountains. Ecology , 58.5 , 1159 1164. Mattson, D. (1997). Wilderness dependent wildlife. International Journal of Wilderness , 3 , 34 38. McCaughey, W. W., & Tomback, D. F. (2001). The natural regeneration process (pp. 105 120). Isla nd Press, Washington DC. McDonald, G.I., & Hoff, R.J. (2001). Blister rust: an introduced plague. Whitebark pine communities: ecology and restoration. Edited by DF Tomback, SF Arno, and RE Keane. Island Press, Washington , DC , 193 220. McKinney, S. T., F iedler, C. E., & Tomback, D. F. (2009). Invasive pathogen threatens bird Ð pine mutualism: implications for sustaining a high ! elevation ecosystem. Ecological Applications , 19 .3 , 597 607 McIntire, E. J., Piper, F. I., and Fajardo, A. (2016). Wind exposure and light exposure, more than elevation ! related temperature, limit tree line seedling abundance on three continents. Journal of Ecology , 104.5 , 1379 1390. Mills, L. S., Soule ! , M. E., & Doak, D. F., (1993) The keystone species concept in ecology and conservation. Bioscience 43 , 219 Ð 224. Pansing, E.R., Tomback, D.F., Wunder, M.B., French, J.P. & Wagner, A.C. (2017). Simulated s eed dispersal reveals microsite and abiotic stress driven recruitment discordance in a bird dispersed pine, Ecology and Evolution , in review.

PAGE 149

133 Perkins, D. L., & Swetnam, T. W. (1996). A dendroecological assessment of whitebark pine in the Sawtooth Ð Salmon River region, Idaho. Canadian Journal of Forest Research , 26.12 , 2123 2133. Power, M. E.; Tilman, D.; Estes, J. A.; Menge, B. A.; Bond, W. J.; Mills, L. S.; Daily, G.; Castilla, J. C.; Lubchenco, J.; Paine, R. T., (1996) Challenges in the quest for keysto nes. Bioscience 46, 609 Ð 620. PrevŽy, J. S., Germino, M. J., & Huntly, N. J. (2010). Loss of foundation species increases population growth of exotic forbs in sagebrush steppe. Ecological Applications , 20 .7 , 1890 1902. PRISM Climate Group, Oregon State Un iversity, http://prism.oregonstate.edu, created 10 Dec 2015. Pyatt, J. C., Tomback, D. F., Blakeslee, S. C., Wunder, M. B., Resler, L. M., Boggs, L. A., and Bevency, H. D. (2016). The importance of conifers for facilitation at treeline: Comparing biophys ical characteristics of leeward microsites in whitebark pine communities. Arctic, Antarctic, and Alpine Research , 48.2 , 427 444. R Core Team (2014). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Au stria. URL https://www.R project.org/ . R Core Team (2016). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R project.org/. Pyatt , J. C., Tomback, D. F., Blakeslee, S. C., Wunder, M. B., Resler, L. M., Boggs, L. A., & Bevency, H. D. (2016). The importance of conifers for facilitation at treeline: Comparing biophysical characteristics of leeward microsites in whitebark pine communiti es. Arctic, Antarctic, and Alpine Research , 48.2 , 427 444. Resler, L.M., Butler, D.R. & Malanson, G.P. (2005). Topographic shelter and conifer establishment and mortality in an alpine environment, Glacier National Park, Montana. Physical Geography , 26.2 , 112 125. Resler, L. M. (2006) Geomorphic Controls of Spatial Pattern and Process at Alpine Treeline Y' The Professional Geographer , 58.2 , 124 138. Resler, L. M. & Tomback, D.F. (2008). Blister rust prevalence in krummholz whitebark pine: implications for t reeline dynamics. Arctic, Antarctic, and Alpine Research , 40 , 161 170.

PAGE 150

134 Resler, L. M., Shao, Y., Tomback, D. F., & Malanson, G. P. (2014). Predicting functional role and occurrence of Whitebark Pine (Pinus albicaulis) at Alpine Treelines: Model accuracy and variable importance. Annals of the Association of American Geographers , 104.4 , 703 722. Rowlingson, B., Diggle, P. (2014). Splancs: Spatial and Space Time Point Pattern Analysis. R package version 2.01 38. https://CRAN.R project.org/package=splancs S afranyik, L. (1978). Effects of climate and weather on mountain pine beetle populations. Chicago. Schšb, C., Butterfield, B. J., & Pugnaire, F. I. (2012). Foundation species influence trait ! based community assembly. New Phytologist , 196.3 , 824 834. Schoe ttle, A. W., & Sniezko, R. A. (2007). Proactive intervention to sustain high elevation pine ecosystems threatened by white pine blister rust. Journal of Forest Research , 12.5 , 327 336. Schrag, A. M., Bunn, A. G. & Graumlich, L. J. (2007) Influence of bioc limatic variables on treeline conifer distribution in the Greater Yellowstone Ecosystem: implications for species of conservation concern. Journal of Biogeography , 35 , 698 710. Schwandt, J. W., Lockman, I. B., Kliejunas, J. T., & Muir, J. A. (2010). Cur rent health issues and management strategies for white pines in the western United States and Canada. Forest Pathology , 40:3 4 , 226 250. Soule ! , M. E., Estes, J. A., Berger, J., & Del Rio, C. M. (2003). Ecological effectiveness: Conservation goals for in teractive species . Conservation Biology , 17 , 1238 Ð 1250. Smith, C. M., Wilson, B., Rasheed, S., Walker, R. C., Carolin, T., & Shepherd, B. (2008). Whitebark pine and white pine blister rust in the Rocky Mountains of Canada and northern Montana. Canadian J ournal of Forest Research , 38.5 , 982 995. Smith, W.K., Germino, M.J., Johnson, D.M. & Reinhardt, K. (2009) The altitude of alpine treeline: a bellwether of climate change effects. The Botanical Review , 75.2 , 163 190. Smith, E.K., Resler, L.M., Vance, E. A., Carstensen Jr, L.W. & Kolivras, K.N. (2011). Blister rust incidence in treeline whitebark pine, Glacier National Park, USA: environmental and topographic influences. Arctic, Antarctic, and Alpine Research , 43.1 , 107 117. Smith, C. M., Langor, D. W ., Myrholm, C., Weber, J., Gillies, C., & Stuart Smith, J. (2013). Changes in white pine blister rust infection and mortality in limber pine over time. Canadian Journal of Forest Research , 43 (10), 919 928.

PAGE 151

135 Smith McKenna, E.K., Resler, L.M., Tomback, D.F., Zhang, H. & Malanson, G.P. (2013). Topographic influences on the distribution of white pine blister rust in Pinus albicaulis treeline communities. Ecoscience , 20:3 , 215 229. Spittlehouse, D. L., and Stathers, R. J. (1990). Seedling microclimate , No. 65 . Stachowicz, J. J. (2001) Mutualism, Facilitation, and the Structure of Ecological Communities Bioscience , 51.3 , 235 246. Stevens, G. C., and Fox, J. F. (1991). The causes of treeline. Annual review of ecology and systematics , 22 :1 , 177 191. Tomback, D. F . (1978). Foraging strategies of Clark's nutcracker. Living Bird , 16 , 123 160. Tomback, D.F. (1982). Dispersal of whitebark pine seeds by Clark's nutcracker: a mutualism hypothesis. The Journal of Animal Ecology , 51 , 451 467. Tomback, D.F. (1986). Post fire regeneration of krummholz whitebark pine: a consequence of nutcracker seed caching. Madro–o , 33 , 100 110 Tomback, D.F. & Linhart, Y.B. (1990). The evolution of bird dispersed pines. Evolutionary Ecology , 4.3 , 185 219. Tomback, D. F., Arno, S. F., & Keane, R. E. (2001). Whitebark pine communities: ecology and restoration . Island Press. Tomback, D.F. & Kendall, K.C. (2001) Biodiversity losses: the downward spiral. Whitebark pine communities: ecology and restoration. Edited by DF Tomback, SF Arno, a nd RE Keane. Island Press, Washington, DC , 243 262. Tomback, D. F., Anderies, A. J., Carsey, K. S., Powell, M. L., & Mellmann Brown, S. (2001). Delayed seed germination in whitebark pine and regeneration patterns following the Yellowstone fires. Ecology , 8 2.9 , 2587 2600. Tomback, D.F. & Resler, L.M. (2007). Invasive pathogens at alpine treeline: Consequences for treeline dynamics. Physical. Geography. 28 , 397 Ð 418. Tomback, D.F. & Achuff, P. (2010) Blister rust and western forest biodiversity: Ecology, va lues and outlook for white pines. Invited contribution for spec ial issue on white pine blister rust. Forest Pathology , 40 , 186:225.

PAGE 152

136 Tomback, D.F., Achuff, P., Schoettle, A.W., Schwandt, J.W. & Mastrogiuseppe, R.J . (2011) The magnificent high elevation five needle white pines: ecological roles and future outlook. The future of high elevation, five needle white pines in Western North America: Proceedings of the High Five Symposium. US Department of Agriculture, Forest S ervice, Rocky Mountain Research Station, Missoula, MT . Tomback, D. F., Chipman, K. G., Resler, L. M., Smith McKenna, E. K., & Smith, C. M. (2014). Relative abundance and functional role of whitebark pine at treeline in the northern Rocky Mountains. Arctic, Antarctic, and Alpine Research , 46.2 , 407 418. Tomback, D. F., Resler, L. M., Keane, R. E., Pansing, E. R., Andrade, A. J., & Wagner, A. C. (2016). Community structure, biodiversity, and ecosystem services in treeline whitebark pine communities: potential impacts from a non native pat hogen. Forests , 7 .1 , 21. Tomback, D. F., Blakeslee, S. C., Wagner, A. C., Wunder, M. B., Resler, L. M., Pyatt, J. C., & Diaz, S. (2016). Whitebark pine facilitation at treeline: potential interactions for disruption by an invasive pathogen. Ecology and Ev olution , 6.15 , 5144 5157. Tovar S‡nchez, E., L. Valencia Cuevas, E. Castillo Mendoza, P. Mussali Galante, R. V. PŽrez Ruiz, and A. Mendoza. (2013). Association between individual genetic diversity of two oak host species and canopy arthropod community st ructure. European Journal of Forest Research , 132 , 165 Ð 179. Van Miegroet, H., Hysell, M. T., & Johnson, A. D. (2000). Soil microclimate and chemistry of spruce Ð fir tree islands in northern Utah. Soil Science Society of America Journal , 64 :4 , 1515 1525 V ander Wall, S. B., & Balda, R. P. (1977). Coadaptations of the Clark's nutcracker and the pinon pine for efficient seed harvest and dispersal. Ecological monographs , 47.1 , 89 111. Warwell, M.V., Rehfeldt, G. E. & Crookston, N.L. (2007) Modeling contemporar y climate profiles of whitebark pine (Pinus albicaulis) and predicti ng responses to global warming. Proceedings of the Conference Whitebark Pin e: A Pacific Coast Perspective. USDA Forest Service R6 Ð NR Ð FHP 2007 Ð 01 . Whitehead, F. H. (1951). Ecology of the altipiano of Monte Maiella, Italy. Journal of Ecology , 39 :2 , 330 355. Yakovis, E. L., A. V. Artemieva, N. N. Shunatova, and M. A. Varfolomeeva. (2008). Multiple foundation species shape benthic habitat islands. Oecologia , 155 , 785 Ð 795.