<<

Red pine (Pinus resinosa Ait) Dynamics under Changing Climate in Northern ,

by

Muhammad Waseem Ashiq

A Thesis presented to The University of

In partial fulfillment of requirements for the degree of Doctor of Philosophy in Environmental Sciences

Guelph, Ontario, Canada

© Muhammad Waseem Ashiq, May, 2015

ABSTRACT

RED PINE (PINUS RESINOSA AIT) DYNAMICS UNDER CHANGING CLIMATE IN , CANADA

Muhammad Waseem Ashiq Advisor: Dr. Madhur Anand University of Guelph, 2015 Advisory Committee: Dr. Andrew. M. Gordon Dr. Roger Suffling

This thesis is an investigation of climate effects on radial growth of red pine (Pinus resinosa Ait) in northern Ontario over multiple spatio-temporal scales. In the context of climate change, these investigations provide insight about species survival, growth and range shift potential. In this thesis I investigate these three aspects of red pine dynamics using tree-ring width (TRW) data from 54 sites across northern Ontario. I first identified leading edge populations (16 No.) of red pine using climate data in hierarchical cluster analysis (HCA). I used various combinations of climatic variables as predictors in

HCA, and evaluated their performance based on the compactness of formed clusters. These analyses reveal that leading edge populations of red pine in northern Ontario can be determined using three monthly climatic variables: mean minimum temperature (Tmin), mean maximum temperature (Tmax) and climate moisture index (CMI). I performed correlation and response function analyses to identify climatic controls on growth in these leading edge red pine populations. My results show that the effect of seasonal climate during prior summer was significant for growth in leading edge populations. Specifically, growth response was positive to prior summer precipitation (Prec) and negative to prior summer Tmax. This combined effect suggests the potential role of drought in controlling red pine northern range limit. My results question the validity of studies that predict large scale range shift potential of red pine under warming scenarios. My growth – climate analyses for the remaining red pine populations reveal a complex growth response of red pine in northern Ontario. In general, I observed four main results: (i) red pine populations in northwestern Ontario are more sensitive to climate than populations in northeastern

Ontario, (ii) the growth-limiting effect of Prec is more significant than Tmax, (iii) growth – climate relationships follow a longitudinal gradient, and (iv) a shift in climatic controls of red pine temperature during first half of the 20th century to precipitation during recent decades. I also investigated age effects on red pine growth – climate relationships in an old-growth red pine forest at Wolf Lake Forest Reserve in northern Ontario. My analyses reveal that growth response to climate in young red pines is different from old red pines. Winter season Tmin positively influences growth in young red pine trees, whereas summer Prec (specifically July) positively influences growth in old red pine trees. I finally conclude that climate effects on red pine growth vary across space and time, and such variations must be considered in any decision-making process.

iii

AUTHOR'S DECLARATION

I hereby declare that I am the sole author of this thesis. This is a true copy of the thesis, including any required final revisions, as accepted by my examiners. I understand that my thesis may be made electronically available to the public.

iv

PREFACE

This thesis was written in manuscript format with the intent of publishing individual chapters as original research papers in peer reviewed journals. I tried my best to avoid any redundancy. However, some repetition was unavoidable considering each chapter (Chapter‎ 2 - 4) as an individual manuscript.

For this thesis, I took the lead in conceptualizing research ideas, developing research design, conducting fieldwork, preparing samples, recording/collecting data, analyzing data and producing manuscripts, including figures, tables, appendices, for all chapters in this thesis. However, I could not have accomplished these milestones without highly valuable guidance and support from my research advisor, Professor Dr. Madhur Anand, and advisory committee members: Professor Dr. Andrew M.

Gordon and Professor Dr. Roger Suffling.

In addition to continual feedback on research conceptualization and design, Dr. Anand advised me heavily on quantitative analysis approaches that were used in Chapter‎ 2 to identify ‘leading edge populations’ of red pine in northern Ontario. She was also instrumental in evaluating/selecting analysis techniques used in Chapter‎ 3, and refining the research questions for Chapter‎ 4. Her advice also helped to improve figures and tables in Chapter‎ 2 - 4. As such, she is included as co-author on the following manuscripts;

Ashiq, M.W., Anand, M., (Chapter‎ 2). Using cluster analysis to determine growth - climate relationships of leading edge red pine (Pinus resinosa Ait) populations in northern Ontario, Canada.

Ashiq, M.W., Anand, M. (Chapter‎ 3). Spatial and temporal variability in dendroclimatic response of red pine (Pinus resinosa Ait) to climate in northern Ontario, Canada. Forest Ecology and Management (submitted).

Ashiq, M.W., Anand, M. (Chapter‎ 4). Age dependent climate sensitivity of radial growth in red pine at Wolf Lake old-growth forest in . PLOS ONE (submitted).

v

ACKNOWLEDGEMENTS

In the name of Almighty ALLAH (God), the Gracious, the Merciful. I am most grateful to Almighty ALLAH for blessing me with knowledge, and making me capable to accomplish this PhD thesis. ALHAMD-U-LILLAH.

I feel pleasure to express my profound admiration and sincere gratitude to my PhD advisor Professor Dr. Madhur Anand for her guidance, valuable suggestions and thoughtful critical comments. Her consistent mentoring throughout my PhD made it possible to conceptualize research questions and hypotheses, design research, and apply appropriate statistical methods. I am deeply impressed with her enthusiasm and vision of academic research. The provision of full financial support for field work, and state of the art lab facilities for data preparation and analysis are also her contributions towards the successful completion of my research. She continuously provided critical and constructive feedback on my research progress and thesis write up. In addition, she always informed and encouraged me to attend conferences and summer schools, and also provided financial support to attend these events.

I am equally grateful and express my warm appreciation to my advisory committee members Professor Dr. Andrew M. Gordon (University of Guelph) and Professor Dr. Roger Suffling (University of Waterloo) for their mentoring during my PhD studies. I always find them approachable and supportive. Their constructive feedback on my research, information about useful resources and valuable suggestions for thesis write up provided me with a strong foundation for my research work.

I feel pleasure to put on record my heartfelt thanks to my thesis examiners Dr. Margot Kaye (Penn State University)‎and‎Dr.‎Ze’ev‎Gedalof‎(University‎of‎Guelph)‎for‎their‎valuable‎time‎and‎constructive‎ feedback to improve my thesis. Dr. Gedalof was also member of my PhD qualifying examination committee. He also provided me with thoughtful insights to elaborate my research.

Special thanks to Alice Cecile and Elisabeth Shapiro for assistance in data collection and fieldwork; Colin Bowling, Dan Duckert and many field officers of Ontario Ministry of Natural Resources and Forestry for fieldwork logistics; Chris Wagner and Amanda Shamas for data preparation; Dan McKenney and Pia Papadopol for climate data; Martin Girardin, Scott St. George and Steven Gore for additional tree-ring data; Franco Biondi for DendroClim software; and Melissa Tong for thesis proofreading. Alice Cecile also helped me by writing R codes and preparing graph.

I am indebted to Professor Dr. Paul K. Sibley (former graduate officer) and SES office staff (Marie Vickery, Linda Bissell, Virginia Warren, Jo-Anne Scarrow, Linda Wing, and Debbie A. MacDonald) for vi their valuable guidance and tremendous help in administration related issues. Dr. Sibley also chaired committees for my PhD qualifying and final oral examinations.

I feel pleasure to put on record my acknowledgements for financial support from Dr. Anand, Registrar Office University of Guelph, School of Environmental Sciences, NSERC (GGSF travel grant), Forest Complexity Modelling (NSERC CREATE), Ontario Graduate Scholarship, Doug and Esther Ormord Scholarship, and Arthur D. Latornell Scholarship.

I owe my eternal gratitude to my parents, parents-in-law, wife (Rozina), children (Meema, Hamiz and Behzad) and other family members for their encouragement and moral support to pursue this endeavour.

I am also grateful to my lab mates, friends and many other people who helped me in studies or otherwise during my time at University of Guelph. Thank you all!

vii

DEDICATION

I dedicate this thesis to my father-in-law, Allauddin Khan (may ALLAH bless his soul and award him

Paradise) for his continual support and inspiration.

viii

TABLE OF CONTENTS ABSTRACT ...... II AUTHOR'S DECLARATION ...... IV PREFACE ...... V ACKNOWLEDGEMENTS ...... VI DEDICATION ...... VIII TABLE OF CONTENTS ...... IX LIST OF ACRONYMS AND ABBREVIATIONS ...... XII LIST OF SPECIES ...... XIII LIST OF TABLES ...... XIV LIST OF FIGURES ...... XV CHAPTER 1 INTRODUCTION ...... 1 1.1 Role of Climate in Tree Growth ...... 1 1.2 Our Changing Climate ...... 3 1.2.1 Climate Change during 20th Century...... 3 1.2.2 Climate Projections for 21st Century ...... 9 1.3 Climate Change Effects on Tree Growth ...... 12 1.4 Research Objectives ...... 14 1.5 Thesis Structure ...... 14 CHAPTER 2 USING CLUSTER ANALYSIS TO DETERMINE GROWTH - CLIMATE RELATIONSHIPS OF LEADING EDGE RED PINE (PINUS RESINOSA AIT) POPULATIONS IN NORTHERN ONTARIO, CANADA ...... 16 Abstract ...... 16 2.1 Introduction ...... 17 2.2 Materials and Methods ...... 20 2.2.1 Study Area and Data ...... 20 2.2.2 Hierarchical Cluster Analysis ...... 23 2.2.3 Red Pine Chronologies...... 24 2.2.4 Red Pine Growth – Climate Analyses ...... 28 2.3 Results and Discussion ...... 28 2.3.1 Cluster Analysis ...... 28 2.3.2 Growth – Climate Relationships ...... 32 ix

2.4 Conclusions ...... 41 CHAPTER 3 SPATIAL AND TEMPORAL VARIABILITY IN DENDROCLIMATIC RESPONSE OF RED PINE (PINUS RESINOSA AIT) TO CLIMATE IN NORTHERN ONTARIO, CANADA ...... 42 Abstract ...... 42 3.1 Introduction ...... 44 3.2 Materials and Methods ...... 47 3.2.1 Study Area ...... 47 3.2.2 Data Collection and Preparation ...... 50 3.2.3 Chronology Development ...... 54 3.2.4 Spatial and Temporal Growth – Climate Analyses ...... 58 3.3 Results ...... 60 3.3.1 Red Pine Chronologies...... 60 3.3.2 Growth – Climate Associations ...... 61 3.3.3 Spatial Variability in Growth – Climate Associations ...... 63 3.3.4 Temporal Variability in Growth _ Climate Associations...... 66 3.4 Discussion ...... 73 3.5 Conclusions ...... 78 CHAPTER 4 AGE DEPENDENT CLIMATE SENSITIVITY OF RADIAL GROWTH IN RED PINE AT WOLF LAKE OLD-GROWTH FOREST IN ONTARIO CANADA ...... 80 Abstract ...... 80 4.1 Introduction ...... 81 4.2 Materials and Methods ...... 85 4.2.1 Field Sampling and Data Preparation ...... 85 4.2.2 Chronology Development ...... 85 4.2.3 Growth – Climate Analysis ...... 88 4.3 Results ...... 90 4.4 Discussion ...... 98 4.5 Conclusions ...... 102 CHAPTER 5 CONCLUSIONS AND RECOMMENDATIONS ...... 103 5.1 Conclusions ...... 103 5.2 Recommendations ...... 105 x

Literature Cited ...... 107 APPENDIX A RED PINE CHRONOLOGIES WITH CORRESPONDING SAMPLE SIZE. BOLD RED LINE INDICATES THE RUNNING AVERAGE OF CHRONOLOGY INDEX VALUES ...... 133 APPENDIX B DISSIMILARITY MATRIX ...... 137 APPENDIX C SPSS OUTPUT FOR HCA AFTER CHANGING THE ORDER OF SITES ENTERING THE ANALYSIS ...... 147 APPENDIX D SPSS OUTPUT FOR HCA USING A MATRIX OF 54 X 28. ANALYSIS WAS RUN BY REMOVING RANDOMLY SELECTED 8 (~20%) CLIMATIC VARIABLES 150 APPENDIX E SPSS OUTPUT FOR HCA VALIDATION USING SPLIT HALF METHOD ...... 153 APPENDIX F CHARACTERISTICS OF 37 RED PINE SITE CHRONOLOGIES ...... 157 APPENDIX G RED PINE CHRONOLOGIES AND THE CORRESPONDING SAMPLING SIZE. BOLD RED LINE INDICATES THE RUNNING AVERAGE ...... 159 APPENDIX H RESPONSE COEFFICIENTS OF RED PINE GROWTH – SEASONAL PRECIPITATION RELATIONSHIPS. SIGNIFICANT COEFFICIENTS, AT 95% CONFIDENCE INTERVAL, ARE SHOWN IN BOLD ITALIC...... 169 APPENDIX I RESPONSE COEFFICIENTS OF RED PINE GROWTH – SEASONAL TMIN RELATIONSHIPS. SIGNIFICANT COEFFICIENTS, AT 95% CONFIDENCE INTERVAL, ARE SHOWN IN BOLD ITALIC...... 171 APPENDIX J RESPONSE COEFFICIENTS OF RED PINE GROWTH – SEASONAL TMAX RELATIONSHIPS. SIGNIFICANT COEFFICIENTS, AT 95% CONFIDENCE INTERVAL, ARE SHOWN IN BOLD ITALIC...... 173

xi

LIST OF ACRONYMS AND ABBREVIATIONS AIK Akaike Information Criterion AR Autoregressive Modelling CCA Canonical Correspondence Analysis Cef Centre‎ďétude‎de‎la‎forêt CMI Climate Moisture Index Elv Elevation EPS Expressed Population Signal FRI Forest Resource Inventory GCM General Circulation Model HCA Hierarchical Cluster Analysis ITRDB International Tree-Ring Database Lat Latitude Lon Longitude m Meter MAP Mean Annual Precipitation MAT Mean Annual Temperature MME Multi-Model Ensemble MS Mean Sensitivity NEO NPP Net Primary Productivity NWO Northwestern Ontario PPE Perturbed Physical Ensemble P-ps Prior Summer Precipitation Prec Total Precipitation P-sm Summer Precipitation SC Serial Correlation Tmax Mean Maximum Temperature Tmin Mean Minimum Temperature Tn-ps Prior Summer Minimum Temperature TRW Tree-ring Width Tx-ps Prior Summer Maximum Temperature xii

LIST OF SPECIES Balsam Fir Abies balsamea (L.) Mill. Black Spruce Picea mariana Mill B.S.P. Douglas-Fir Pseudotsuga menziesii Eastern Hemlock Tsuga canadensis L. Eastern Larch Larix laricina (Du Roi) K. Koch. European Larch Larix decidua Jack Pine Pinus banksiana Lamb Lodgepole Pine Pinus contorta var. latifoloia Engelm Longleaf Pine Pinus palustris P. Mill. Mountain Hemlock Tsuga mertensiana Norway spruce Picea abies Paper Birch Betula papyrifera Marshall Pedunculate Oak Quercus robur Quilian Juniper Sabina przewalskii Kom. Red Maple Acer rubrum L. Red Oak Quercus rubra L. Red Pine Pinus resinosa Ait Scott Pine Pinus sylvestris L. Spanish Juniper Juniperus thurifera Stone Pine Pinus pinea Subalpine Fir Abies lasiocarpa Sugar Maple Acer saccharum Marsh. Swiss Stone Pine Pinus cembra Trembling Aspen Populus tremuloides Michx White Birch Betula pendula White Oak Quercus alba L. White Pine Pinus strobus L. White Spruce Picea glauca (Moench) Voss

xiii

LIST OF TABLES

Table 2.1‎ : Location characteristics and chronology statistics of red pine leading edge populations in northern Ontario, Canada...... 27

Table 2.2‎ : Correlation coefficients between annual growth (in NEO) and seasonal climate variables during the 20th century...... 35

Table 2.3‎ : Correlation coefficients between annual growth (in NWO) and seasonal climate variables during the 20th century...... 36

Table 2.4‎ : Response coefficients between annual growth (in NEO) and seasonal climate variables during the 20th century ...... 37

Table 2.5‎ : Response coefficients between annual growth (in NWO) and seasonal climate variables during the 20th century...... 38

Table 3.1‎ : General statistics of 37 red pine site chronologies...... 56

Table 4.1‎ : Chronology statistics of young (< 100 years) and old (>100 years) red pines at Wolf Lake Forest in northern Ontario, Canada. Values in parenthesis are for the analysis period 1950 – 2010...... 87

xiv

LIST OF FIGURES

Figure 1.1:‎ Spatial variability of mean annual temperature (MAT) and mean annual precipitation (MAP) in Ontario during 1951 - 2002. Shown maps are 52-year average of MAT (A1) and MAP (B1); average rate of change during 52 years for MAT (A2) and MAP (B2); statistical confidence in change in MAT (A3) and MAP (B3). Statistical confidence was determined from least square p-values. These maps were generated using data from Girvetz et al. (2009). Also shown in A1 are major lakes in blue and three Ontario regions (SO – , NEO – northeastern Ontario, and NWO – northwestern Ontario) demarcated by white lines. Grey line polygons in each map represent the distribution range of red pine (USGS 1999)...... 7

Figure 1.2‎ : Change in various temperature indices in Ontario during 1950 – 2003. Shown maps are (A) warm days (number of days with Tmax > 90th percentile), (B) warm nights (number of days with Tmin > 90th percentile), (C) diurnal temperature range (mean of the difference between Tmax and Tmin), and (D) summer days (number of days with Tmax > 25°C). Significant trends at 5% level are shown by dots and non-significant trends are shown by crosses. Dot size is proportional to magnitude of the trend. Maps are adopted from Vincent and Mekis (2006)...... 8

Figure 1.3‎ : Climate projections of mean annual temperature (MAT) and mean annual precipitation (MAP) in Ontario for 2050s (2040-69). Maps show change in MAT (°C) and MAP (%) with reference to baseline period 1961-90. These projections are based on ensemble average of 16 climate models (General Circulation Models - GCMs) for a balanced emission scenario A1B. These maps were generated using data from Girvetz et al. (2009). Grey line polygons in each map represent the distribution range of red pine (USGS 1999)...... 10

Figure 1.4‎ : A hypothetical model of species growth response to climate change-induced warming. Under warming conditions, growth in species growing in warmer environments (red) will decrease and growth in species growing in colder environments (blue) will, in most cases, increase. Despite warming, species from colder environments can also experience decrease in growth due to limiting effects of water and nutrients. Figure adopted from Way and Oren (2010).13

Figure 2.1‎ : Extent of the study area in northern Ontario. Areas in pink represent red pine distribution range‎in‎northern‎Ontario‎(USGS‎1999).‎Symbols‎(▲‎and‎■)‎show‎the‎locations‎of‎red‎pine‎ sampling sites: Colour of symbols differentiate data source: red (this study), blue (Girardin et

xv

al. 2006) and green (St. George et al. 2008). Hierarchical Cluster Analysis (HCA) was performed‎to‎differentiate‎sites‎as‎leading‎edge‎(▲)‎and‎non-leading‎edge‎(■)...... 22

Figure 2.2‎ : Dendrogram of hierarchical cluster analysis (HCA) to determine leading edge populations. Along Y-axis are the sites that successively merge into clusters. X-axis is the rescaled (0 - 25) proximity matrix distances. Dotted vertical line shows the distance for two clusters...... 29

Figure 2.3‎ : Plot of distance statistic differences between every two successive stages of hierarchical cluster analysis (HCA). Highest value at stage 52 indicates the possible clustering solution (2 clusters)...... 31

Figure 2.4‎ : Summary of significant functions between red pine chronologies of leading edge populations and monthly climatic variables: total precipitation (Prec), mean minimum temperature (Tmin) and mean maximum temperature (Tmax). Negative numbers at y-axis show number of significant‎negative‎functions.‎Prior‎year’s‎months‎(abbreviated‎in‎capital‎letters‎at‎x-axis) were included to account‎for‎effect‎of‎prior‎year’s‎climate‎on‎growth.‎Chronologies‎are‎ separated in two groups: NEO (northeastern Ontario) and NWO (northwestern Ontario).34

Figure 3.1‎ : Study area in northern Ontario. Locations of sampling sites are shown with square symbol (■).‎ Data sources are differentiated by colour of symbols: red (this study), blue (Girardin et al. 2006) and green (St. George et al. 2008). Ecoregions are delineated with blue dotted lines Ontario regions are delineated with black lines...... 49

Figure 3.2‎ : Interpolated data of mean minimum temperature (Tmin in oC) for various seasons. Mean Tmin for the period 1901-2010 are shown in plot A (winter), C (spring), and E (prior summer). Temporal variability for the period 1901-2010 (standard deviations) are shown in plots B (winter), D (spring), and F (prior summer). Interpolation method and seasons are described in section 3.2.2.1.‎ ...... 51

Figure 3.3‎ : Interpolated data of mean maximum temperature (Tmax in oC) for various seasons. Mean Tmax for the period 1901-2010 are shown in plot A (winter), C (spring), and E (prior summer). Temporal variability for the period 1901-2010 (standard deviations) are shown in plots B (winter), D (spring), and F (prior summer). Interpolation method and seasons are described in section 3.2.2.1.‎ ...... 52

Figure 3.4‎ : Interpolated data of total precipitation (Prec in mm) for various seasons. Mean Prec for the period 1901-2010 are shown in plot A (winter), C (spring), and E (prior summer). Temporal xvi

variability for the period 1901-2010 (standard deviations) are shown in plots B (winter), D (spring), and F (prior summer). Interpolation method and seasons are described in section 3.2.2.1.‎ ...... 53

Figure 3.5‎ : Scatter plot showing geographic distribution of significant seasonal climatic variables influencing‎red‎pine‎growth‎in‎northern‎Ontario:‎▲‎(precipitation),‎■‎(maximum‎ temperature),‎●‎(minimum‎temperature),‎♦‎(precipitation‎and‎minimum/maximum‎ temperature) and + (precipitation, minimum and maximum temperature)...... 62

Figure 3.6‎ : Results of Canonical Correspondence Analysis of significant/non-significant response functions. These response functions were derived from red pine growth – climate analyses. Triplot is amplified by factor of 5 to improve the visuals. The eigenvalues of axis 1 (horizontal) and axis 2 (vertical) are 0.313 and 0.186 respectively. Together these axes account for about 50% of the variation. Sites are shown in black triangles. Three sites (out of 37) and one climatic variable (out of 12) were excluded from the analysis due to absence of any significant response function. Less points of sites are visible due to their overlap in 2-D. Three explanatory (geographic) variables, namely latitude (Lat), longitude (Lon) and altitude (Alt) are represented by green lines. Names of response functions (blue) are abbreviated in two parts separated by hyphen ( - ): first part represents climatic variable [precipitation (P), minimum temperature (Tn), maximum temperature (Tx)] and second part is the season [prior summer (ps), winter (wn), spring (sp) and summer (sm)]...... 65

Figure 3.7‎ : Ecoregion 4E – Moving response function analyses (with a 30-year window) using significant climatic variables for red pine growth. At the top of each graph are site ID (left) and seasonal climatic variable (right). X-axis shows the last year of each 30-year window. Y-axis shows moving response coefficients. Black lines show standardized coefficients of moving response functions. Grey areas show 95% confidence interval. All sites show non-stable temporal dynamics of growth-climate relationships...... 67

Figure 3.8‎ : Ecoregion 3W - Moving response function analyses (with a 30-year window) using significant climatic variables for red pine growth. At the top of each graph are site ID (left) and seasonal climatic variable (right). X-axis shows the last year of each 30-year window. Y-axis shows moving response coefficients. Black lines show standardized coefficients of moving response functions. Grey areas show 95% confidence interval. All sites show non-stable temporal dynamics of growth-climate relationships...... 68 xvii

Figure 3.9‎ : Ecoregion 4S - Moving response function analyses (with a 30-year window) using significant climatic variables for red pine growth. At the top of each graph are site ID (left) and seasonal climatic variable (right). X-axis shows the last year of each 30-year window. Y-axis shows moving response coefficients. Black lines show standardized coefficients of moving response functions. Grey areas show 95% confidence interval. All sites show non-stable temporal dynamics of growth-climate relationships...... 70

Figure 3.10‎ : Ecoregion 5S - Moving response function analyses (with a 30-year window) using significant climatic variables for red pine growth. At the top of each graph are site ID (left) and seasonal climatic variable (right). X-axis shows the last year of each 30-year window. Y- axis shows moving response coefficients. Black lines show standardized coefficients of moving response functions. Grey areas show 95% confidence interval. All sites show non- stable temporal dynamics of growth-climate relationships...... 71

Figure 3.11‎ : Ecoregion 4W - Moving response function analyses (with a 30-year window) using significant climatic variables for red pine growth. At the top of each graph are site ID (left) and seasonal climatic variable (right). X-axis shows the last year of each 30-year window. Y- axis shows moving response coefficients. Black lines show standardized coefficients of moving response functions. Grey areas show 95% confidence interval. All sites show non- stable temporal dynamics of growth-climate relationships...... 72

Figure 4.1‎ : Map of Wolf Lake Forest. This forest contains the largest remaining old-growth red pine dominated forests in North America. Forest age is shown in two age classes: young (<100 years) and old (>100 years). Size of the tree symbols shows red pine composition (%) in various stands. Water bodies are shown in blue. Location of Wolf Lake Forest in northeastern Ontario is shown by red dot on Ontario map (top left)...... 83

Figure 4.2:‎ Time series (1950 – 2010) of seasonal climate data used in this study for red pine growth - climate analysis. Climatic variables (top to bottom) are mean maximum temperature (Tmax), mean minimum temperature (Tmin) and total precipitation (Prec). Seasons (left to right) are prior summer (prior June – prior September), summer (July - September), spring (April - Jun) and winter (prior October - March). Dotted red lines are 10-year averaged trend lines. . 89

Figure 4.3‎ : Red pine chronologies for old (black) and young (red) red pines at Wolf Lake Forest, northern Ontario, Canada. (A) time series of each chronology index values (mean correlation 0.75, p <

xviii

0.001), (B) year to year variability (sensitivity) of index values in each chronology (mean correlation 0.40, p < 0.001), and (C) number of crossdated tree-ring width series used in each chronology over time...... 92

Figure 4.4‎ : Results of response function analyses for old (black) and young (grey) red pines at Wolf Lake Forest, northern Ontario, Canada. Analyses were performed using three climatic variables (Tmax - mean maximum temperature, Tmin - mean minimum temperature and Prec - total precipitation) for four seasons (prior summer, summer, spring and winter). Seasons are explained in section 4.2.3.‎ Dotted red bars are the 1000 times bootstrap confidence intervals at 95 percentile. Growth-limiting factors for old and young red pines are summer precipitation and winter minimum temperature respectively...... 93

Figure 4.5‎ : Results of moving response function analysis of growth-limiting seasonal climatic variables for red pine at Wolf Lake Forest, northern Ontario, Canada. Analyses were performed using a 30-year window with a1-year moving interval. Gray areas are the 1000 times bootstrap confidence intervals at 95 percentile. Upper – response coefficients of old red pines growth – summer precipitation moving response analysis. Lower – response coefficients of young red pines growth – winter minimum temperature moving response analysis...... 95

Figure 4.6‎ : Results of response function analysis showing growth-limiting monthly climate for old red pines at Wolf Lake Forest, northern Ontario, Canada. Upper – response coefficients of old red pines growth – precipitation response analysis for summer months. Dotted red bars are the 1000 times bootstrap confidence intervals at 95 percentile. Lower – response coefficients of old red pines growth – July precipitation moving response analysis. Gray areas are the 1000 times bootstrap confidence intervals at 95 percentile...... 96

Figure 4.7‎ : Results of response function analysis showing growth-limiting monthly climate for young red pines at Wolf Lake Forest, northern Ontario, Canada. Top – response coefficients of young red pines growth – minimum temperature response analysis for winter months. Dotted red bars are the 1000 times bootstrap confidence intervals at 95 percentile. Lowers – response coefficients of young red pines growth – minimum temperature moving response analysis for prior November, February and March. Gray areas are the 1000 times bootstrap confidence intervals at 95 percentile...... 97

xix CHAPTER 1 INTRODUCTION

1.1 Role of Climate in Tree Growth

Energy and moisture are two essential requirements for growth in trees. Climate regulates energy and moisture, and as such influences forest dynamics in terms of species distribution, abundance and composition (Bonan and Shugart 1989, Allen et al. 2002). Among various climatic factors, temperature and precipitation are of particular importance. These two climatic factors strongly influence energy and moisture availability for tree growth (Bonan et al. 1990, Saxe et al. 2001).

Temperature regulates photosynthetic rate and biomass production (Rustad et al. 2001, Tanja et al. 2003, Hall et al. 2009, Yamori et al. 2014). The produced biomass is allocated to various components of trees including radial growth (growth). Biomass allocation for growth depends on factors such as species, age, site condition and local populations, and it varies between 20 – 50% of net primary productivity (NPP) (Pereira 1995, Pretzsch 2009). At higher temperatures, increased photorespiration reduces photosynthesis efficiency (Doughty and Goulden 2008, Lukac et al. 2010), and hence affects biomass allocation for growth. Complexity in determining actual biomass allocation for growth makes it difficult to accurately assess the effects of temperature on growth. Species growth – climate relationships are therefore determined through experimental (e.g; control growth chambers) and field (e.g; tree-ring based) studies. Such studies show that growth response to temperature is species specific, and it also varies with conditions such as water and nutrient availability, growth stage and site conditions, and species ability to acclimate to warming etc. (Slatyer 1977, Hill et al. 1988, Cunningham and Read 2002,

Tanja et al. 2003, Sage and Kubien 2007, Sigurdsson et al. 2013). Wilmking et al. (2004) studied growth response of white spruce (Picea glauca (Moench) Voss) to temperature at the tree line in Alaska. Their results show that temperature effect on growth was positive in 36% of the sampled trees, negative in 40% and non-significant in 24%. Studies also show that growth response to temperature is non-linear, and

1 growth declines after a certain threshold temperature (Way and Oren 2010, Beck et al. 2011, Newman et al. 2011). McLane et al. (2011) studied the growth response of lodgepole pine (Pinus contorta var. latifolia Engelm) grown in a large provenance experiment in British Columbia and Yukon, Canada. Their results show that, depending on site and provenance, growth in lodgepole pine maximizes between 3.9°C and 5.1°C.

The role of temperature thresholds in limiting species distribution has been recognized for a long time (Parmesan 2006). Species growth at the upper (latitudinal/elevational) boundary of their distribution range is generally limited by lower temperatures (Jump et al. 2007, Colwell et al. 2008), whereas species growth at their lower boundary is limited by high temperature (Jump et al. 2006). Non-climatic factors such as species competition also play important role in determining the lower boundary of species distribution range (Colombo 2008).

Precipitation regulates the availability of water that is necessary for various biochemical (such as photosynthesis) and physiological (such as cell turgor) processes in plant growth. Reduced water supply causes metabolic impairment or stomata closure that affects photosynthesis and carbon fixation efficiency

(Pallardy 2008, Newman et al. 2011). Reduced availability of water under drought conditions significantly affects growth in many species including red pine (Pinus resinosa Ait) (St. George et al.

2008).

The combined effect of high temperature and low precipitation also creates drought-like conditions that affect growth. Such drought-induced growth reduction has been reported for a number of species such as white spruce and lodgepole pine (Barber et al. 2000, Chhin et al. 2008). Drought affected growth‎can‎increase‎species’ susceptibility to insect/pathogen infestation and thus affect species composition and distribution. McDowell et al. (2008) postulated that drought effects growth via three different mechanisms: hydraulic failure, carbon starvation and biotic agent demographics. Hydraulic

2 failure occurs during intense droughts when reduced soil moisture availability and increased evaporative demand causes xylem conduits to cavitate. Plants avoid hydraulic failure through stomatal closure.

Stomatal closure also decreases photosynthetic carbon uptake. Continued stomatal closure during prolonged drought may lead to carbon starvation due to continued metabolic demand of carbohydrates.

Hydraulic failure or carbon starvation decreases production of carbon-based defensive compounds (resin), increases emission of volatiles that may attract insects/pathogens and alteration of food quality for insects/pathogens. These conditions lead to increased susceptibility to insect/pathogen attacks. Droughts associated with warmer weather can further amplify biotic infestation, for example, as a result of increased growth rate and number of generations per year.

1.2 Our Changing Climate

1.2.1 Climate Change during 20th Century The 5th assessment report of the Intergovernmental Panel on Climate Change (IPCC) highlights the footprint of anthropogenic activity on global climate (IPCC 2014). According to this report, earth is experiencing noticeable changes in its climate since the 1880s. Typically, many changes in climate since the 1950s are unprecedented over decades to millennia. During the period 1880 – 2012, global temperature rose 0.85°C with substantial decadal and inter-annual variability. The global warming trend during the 20th century increased at a rate of 0.075°C per decade to 0.31°C per decade during 1901-2001 and 1971-2002 respectively (Girvetz et al. 2009). Temperature changes in the northern hemisphere are more pronounced in recent decades. In the northern hemisphere, the period from 1983- 2012 marked the warmest period in 1400 years of recorded history. Variations in seasonal temperatures are also non- proportional. Mitchell and Jones (2005) reported an increase of 1.1°C and 0.8°C in seasonal temperature of spring and autumn respectively in the northern hemisphere during the last two decades. Climate analyses also suggest that annual warming is mainly due to a temperature rise in winter and spring

3 seasons (Robeson 2004, Girvetz et al. 2009). In eastern North America temperature variations exhibited a complex pattern of warming and cooling during the latter half of the 20th century. Warming during the winter season is shown to be more intense than other seasons. Warming during the spring is not uniform, while some areas in eastern North America are also experiencing a cooling trend (Schwartz et al. 2006).

Precipitation change during the 20th century also showed considerable spatio-temporal variation.

Although, at the global scale, precipitation change was non-significant during the latter half of the 20th century (Girvetz et al. 2009), mid-latitudes in the northern hemisphere (~ 23.5° - 66.5°N) received significantly higher precipitation during the same period (IPCC 2014).

Climate in Canada is characterized by high spatio-temporal variability, and the changes in climate observed during the 20th century are higher than global averages (Lemmen et al. 2008). Analyses of instrumental climate data (period 1990 – 1998) for southern Canada (south of 60°N) shows an increase of

0.9°C and 12% in mean annual temperature (MAT) and mean annual precipitation (MAP) respectively

(Zhang et al. 2000). This increase in MAT during the 20th century is mainly due to (i) rising temperatures prior to 1940s and after 1970s, and (ii) rising temperatures in spring followed by winter. Also, increase in night (minimum) temperature (Tmin) is greater as opposed to increase in day (maximum) temperature

(Tmax). This has resulted in a significant decrease in seasonal daily temperature ranges. Increase in MAP in southern Canada during the 20th century is mainly due to increased precipitation during the period from

1920s to 1970s (Zhang et al. 2000, Bonsal et al. 2001).

Climate in Ontario exhibits variability, both at temporal (seasonal/annual) and spatial scales.

Spatial variability of MAT and MAP in Ontario is shown in Figure 1‎ .1. During the latter half of the 20th century MAT in Ontario increased as much as 1.5°C. Southern Ontario experienced the least warming, whereas western areas in northern Ontario experienced the highest warming. Some areas west of Lake

Superior also exhibit high increases in temperature. Temperature increases in most of northern Ontario was highly significant (p<0.05). Seasonal climate analyses also show that warming is more pronounced 4 during spring and winter (Chiotti and Lavender 2008). Vincent and Mekis (2006) analyzed temperature data to compare warming trends between Tmax and Tmin. Their results, shown in Figure 1‎ .2, reveal that during the latter half of the 20th century the magnitude and significance of increase in Tmin was higher than Tmax in Ontario. This has caused significant reduction in diurnal temperature range in eastern and western Ontario. Their analyses also reveal an increase in the number of summer days in Ontario.

However, the magnitude and significance of this increase in number of summer days is not uniform throughout Ontario. Increase in summer days is mostly significant in northern Ontario, whereas increase in summer days in non-significant in southern Ontario and southern areas of northwestern Ontario.

Precipitation in Ontario is spatially more variable than temperature. In general, southern Ontario receives more precipitation than northern Ontario. In northern Ontario, the eastern region receives more precipitation than the western region. Figure 1‎ .1 shows some increase in precipitation in southern and most of northern Ontario during the latter half of the 20th century. In contrast, most of the eastern region of northern Ontario received as much as 20% less precipitation. Precipitation decrease in this region was marginally (p<0.10) to highly significant (p<0.05). Areas west of Lake Superior that exhibited the highest temperature during the same period also experienced a decreasing trend in precipitation during the latter half of the 20th century. Areas in Figure 1‎ .1 showing higher rates of precipitation change with lower statistical confidence have greater year to year variability in precipitation (Girvetz et al. 2009). During the

20th century, the ratio of snow to total precipitation during spring has also decreased in Ontario except in areas that are under the influence of Lake Climate (Zhang et al. 2000). Areas under the influence of Lake

Climate also receive higher precipitation during autumn and winter (Chiotti and Lavender 2008).

These variations in climate can have implications for forests in Ontario. For example, evapotranspiration increases due to warmer temperature and increased number of summer days. This, in turn, affects tree growth. Moisture availability can further be depleted due to less precipitation and/or low accumulation of snow pack (Manabe et al. 2004, Vincent and Mekis 2006). Climatic variations in 5 northern Ontario during the 20th century suggest a decreasing moisture budget that could affect growth in some species. As variations in seasonal climate are greater than the annual climatic variations, a better insight to climate effects on species growth can be achieved through seasonal climate analyses.

6

Figure ‎1.1: Spatial variability of mean annual temperature (MAT) and mean annual precipitation (MAP) in Ontario during 1951 - 2002. Shown maps are 52-year average of MAT (A1) and MAP (B1); average rate of change during 52 years for MAT (A2) and MAP (B2); statistical confidence in change in MAT (A3) and MAP (B3). Statistical confidence was determined from least square p-values. These maps were generated using data from Girvetz et al. (2009). Also shown in A1 are major lakes in blue and three Ontario regions (SO – southern Ontario, NEO – northeastern Ontario, and NWO – northwestern Ontario) demarcated by white lines. Grey line polygons in each map represent the distribution range of red pine (USGS 1999).

7

Figure ‎1.2: Change in various temperature indices in Ontario during 1950 – 2003. Shown maps are (A) warm days (number of days with Tmax > 90th percentile), (B) warm nights (number of days with Tmin > 90th percentile), (C) diurnal temperature range (mean of the difference between Tmax and Tmin), and (D) summer days (number of days with Tmax > 25°C). Significant trends at 5% level are shown by dots and non-significant trends are shown by crosses. Dot size is proportional to magnitude of the trend. Maps are adopted from Vincent and Mekis (2006).

8

1.2.2 Climate Projections for 21st Century We have an incomplete understanding of climate systems in terms of the complexity of physical processes involved and the feedback loop of various driving forces. Implementation of this limited knowledge, through climate models and emission scenarios, results in uncertain climate projections.

Emission scenarios (IPCC SRES scenarios) used in climate projections are based on a wide range, from demographic to technological and economic development, of main driving forces of future emissions.

These IPCC SERES scenarios however do not include any future climate change policy (Nakićenović‎et‎ al. 2000). Nonetheless, commonly used IPCC SERES scenarios are A1B, A2 and B1. A1B is a balanced emission scenario that describes a future world of technological development with a balanced consumption of energy sources (fossil vs non-fossil). A2 is a high emission scenario that describes a heterogeneous world with fragmented and relatively slower technological development. B1 is a low emission scenario that describes a convergent world with emphasis on service and information economy

(Nakićenović‎et‎al.‎2000).

For any IPCC SERES scenario, robust estimates of climate projections are usually achieved through perturbed physical ensemble (PPE) or multi-model ensemble (MME). In PPE technique, multiple outputs from a single climate model with perturbed parameters are analyzed together (Yokohata et al.

2010). In MME approach data from various climate models are analyzed using various techniques (Araújo and New 2006, Girvetz et al. 2009, Taylor et al. 2012). This approach accounts for predictions uncertainty due to structural differences in climate models.

Climate projections from a regional climate model PRECIS1 for A1B emission scenario suggest that MAT in northern Ontario will increase as much as 7°C (-3 to 4°C) by the end of the 21st century

(Wang et al. 2014b). In most parts of Ontario, temperature will increase between 3-5°C by 2050s (2040-

69). During the same period Tmax and Tmin in Ontario will increase 4-5°C. A spatial gradient in

1 Providing REginoal Climate for Impact Studies 9 temperature change from northeastern Ontario (higher change) to (lower change) has also been reported for the 21st century (Wang et al. 2014a). Girvetz et al. (2009) analyzed ensemble of

16 global climate models (General Circulation Models - GCMs) to project global climate for the 21st century under A1B emission scenario. Their analysis show a projected increase in MAT up to 3.8°C for

Ontario by 2050s (Figure 1‎ .3). Nonetheless, climate projections using either of the for-mentioned analysis techniques (PPE and MME) suggest that temperature in Ontario will continue rising during the 21st century. It is likely that during the 21st century temperature rise in Ontario will be greater in winter than summer (Colombo 2008).

Figure ‎1.3: Climate projections of mean annual temperature (MAT) and mean annual precipitation (MAP) in Ontario for 2050s (2040-69). Maps show change in MAT (°C) and MAP (%) with reference to baseline period 1961-90. These projections are based on ensemble average of 16 climate models (General Circulation Models - GCMs) for a balanced emission scenario A1B. These maps were generated using data from Girvetz et al. (2009). Grey line polygons in each map represent the distribution range of red pine (USGS 1999).

10

Precipitation projections are generally much more uncertain than temperature, and the same is true for Ontario. Chiotti and Lavender (2008) provided a good overview of this uncertainty based on their analysis for seven emission scenarios from seven GCMs. Although, majority of their simulations for

2050s projected increased MAP in Ontario, few projected decrease in MAP up to 2.5%. Colombo (2008) reported that MAP in western part of northwestern Ontario, under A2 scenario, will decrease up to 10% by 2070. These results are consistent with Chiotti and Lavender (2008). Girvetz et al. (2009) analyzed climate data from 16 GCMs to project 21st century global precipitation. An extract of their projections for

Ontario under A1B scenario is shown in Figure 1‎ .3. These projections show that Ontario will receive increased precipitation (10-20%) during the 2050s. The spatial variability in projected precipitation is however less when compared with spatial variability of precipitation during the 20th century (Figure 1‎ .2).

This is likely due to the coarse resolution (50 km) of the GCMs used for projections. Wang et al. (2014b) generated fine scale (10 km) precipitation projections for Ontario. Their results show higher MAP in some areas in the west and southeast Ontario, and lower MAP in northern Ontario during 2050s. The discrepancy between two projections (Girvetz et al. 2009 vs Wang et al. 2014b) reiterates the fact that precipitation predictions are highly uncertain. Analyses of seasonal precipitation suggest that, although

MAP is likely to increase, precipitation in southern Ontario during summer and fall may decrease up to

10% by 2050s (Chiotti and Lavender 2008).

Overall, temperature increase in Ontario during the 20th century is significant and a temperature increase is likely to continue during the 21st century. Contrarily, precipitation in Ontario exhibits large scale spatio-temporal variability both in magnitude and trend. Although climate predictions for the 21st century suggest increased precipitation in Ontario, these predictions are associated with high levels of uncertainty. Nevertheless, warmer temperatures and longer growing seasons have affected the moisture budget in Ontario (Gabriel and Kreutzwiser 1993), and such effects are likely to increase, or at least continue, during the 21st century (Chiotti and Lavender 2008).

11

1.3 Climate Change Effects on Tree Growth

Forests play an important role in the global carbon cycle, and climate change is likely to have a profound impact on this role by inducing changes in growth (Barber et al. 2000). It is generally assumed that climate change-induced warming will enhance growth (Hanson et al. 1989) and increase in CO2 uptake (Black et al. 2000, Schimel et al. 2000). Rising temperatures also affect growth positively by increasing growing season length and enhancing nutrient mineralization (Peltola et al. 2002, Strömgren and Linder 2002, Linderholm 2006, Girardin et al. 2008, Pretzsch 2009). The recent increase in forest productivity in the northern hemisphere has been attributed to warming during the last two decades (Zhou et al. 2001, Piao et al. 2011). However, the positive effects of warming on productivity are likely to decrease under continuing climate change (Piao et al. 2006). This is typically true for trees that are growing in warmer environments where temperature is already close to the threshold for photosynthesis

(Doughty and Goulden 2008, Newman et al. 2011). With climate change, other growth controlling factors, such as temperature-induced stresses (Barber et al. 2000, Wilmking et al. 2004, Ciais et al. 2005), competition ((Leithead et al. 2012) and nutrients (Saxe et al. 2001, Colombo 2008) are becoming disproportionately important. Way and Oren (2010) proposed a hypothetical model of growth response to increasing temperature. This model as shown in Figure 1‎ .4 suggests that, due to climate change-induced warming, growth in trees growing at warmer regions will decrease, whereas growth in trees growing at colder regions may increase or decrease. In many cases, trees from colder regions will benefit in growth due to rising temperatures. However, such benefits may diminish due to other growth-limiting factors such as water and nutrients. A study on multiple species (black spruce – Picea mariana Mill B.S.P., red oak – Quercus rubra L., red maple – Acer rubrum L., and red pine) across temperate and boreal forests in

Ontario, Canada suggests that recent decline in growth, despite warming trends, is due to climatic and non-climatic stresses such as nutrient limitation (Silva et al. 2010). Recent IPCC report (the 5th

12 assessment) also acknowledges the increasing role of multiple and interacting climate stresses on forest growth under climate change (Romero-Lankao et al. 2014).

Figure ‎1.4: A hypothetical model of species growth response to climate change-induced warming. Under warming conditions, growth in species growing in warmer environments (red) will decrease and growth in species growing in colder environments (blue) will, in most cases, increase. Despite warming, species from colder environments can also experience decrease in growth due to limiting effects of water and nutrients. Figure adopted from Way and Oren (2010).

In fact, quantifying growth response of trees to climate change is remarkably complex and challenging (Williamson et al. 2009, Ohse et al. 2012). Studies have shown that not only climate controls of growth vary over space and time (Lloyd and Fastie 2002, Driscoll et al. 2005, Lloyd et al. 2011,

Chavardès et al. 2013), but coexisting species and their provenances also have differential adaptive capacity to changes in their climate controls (Colombo 2008, McLane et al. 2011). Yeh and Wensel

(2000) reported that the effects of climate change on growth in coniferous species vary among species and their densities. Goldblum and Rigg (2005) reported a differential growth response of three co-existing species (sugar maple – Acer saccharum Marsh., balsam fir – Abies balsamea (L.) Mill., and white spruce) in a deciduous – boreal ecotone in Ontario. According to their results, climate change will increase

13 growth in sugar maple followed by white spruce, whereas growth in balsam fir will decrease. The differential growth response to climate change alters species composition and distribution that may result in new forest types (Reed and Desanker 1992, Mittler 2006). For example, warming-induced moisture deficiency in boreal Alaska can shift spruce forest to dry aspen forests (Bonan et al. 1990). Thus species specific studies on growth response to climate can also improve our understanding of forest dynamics.

1.4 Research Objectives

It is very likely that changing climate is affecting forests in Ontario. Such effects are complicated and depend on species, site conditions, and interaction among various growth-influencing factors (Matthews et al. 2011). Indeed our knowledge about the role of species specific growth-influencing climatic factors is limited (Chiotti and Lavender 2008). This lack of knowledge has implications for the management of forests in Ontario. For example, contemporary forest management relies heavily on models. Many growth models use uniform climate variables that lead to erroneous results (Loehle and Leblanc 1996). Studies have shown that same species can have different climatic controls over larger areas (Schumacher and Day

1939, Foster and Brooks 2001), and species climatic controls also vary over time (Devall et al. 1991). To understand climatic control of a species, growth – climate investigations should span over a multitude of sites and time periods (Henderson 2006). In this thesis, I fill some of these knowledge gaps concerning the growth response of red pine to various climatic variables. This research encompasses, for the first time, a range of spatial and temporal scales that provide a broader perspective of growth response of red pine to climatic variability during the 20th century.

1.5 Thesis Structure

The remainder of this thesis consists of four chapters. Chapter‎ 2 deals with the leading edge populations of red pine. In this chapter I have put equal emphasis on using objective methods to identify leading edge populations and the growth response of these populations to climate. Chapter‎ 3 provides an

14 in-depth examination of spatio-temporal variation in red pine growth – climate relationships in northern

Ontario. In Chapter‎ 4 I have compared growth – climate relationships between two age classes in the largest remaining red pine old-growth forests at Wolf Lake Forest Reserve. Chapter‎ 5 summarizes the findings of this thesis and some recommendations for future research.

15

CHAPTER 2 USING CLUSTER ANALYSIS TO DETERMINE GROWTH - CLIMATE RELATIONSHIPS OF LEADING EDGE RED PINE (PINUS RESINOSA AIT) POPULATIONS IN NORTHERN ONTARIO, CANADA

Abstract

Bioclimatic studies suggest a northward range shift potential of red pine (Pinus resinosa Ait) under warming climate. However, growth dynamics of leading edge red pine populations under changing climate is unknown. Here I investigate growth – climate relationships of leading edge populations of red pine in northern Ontario to assess its range shift potential under changing climate. I identified red pine leading edge populations from a larger set of its northern populations by applying hierarchical cluster analysis (HCA). HCA analyses suggest that leading edge populations of red pine in northern Ontario follow a combined gradient of three climatic variables: minimum temperature, maximum temperature and climate moisture index. I performed correlation and response function analyses using tree-ring width data from leading edge populations to identify climatic controls for the northern range limit of red pine. For these analyses, I used monthly and seasonal data of three climatic variables: precipitation, minimum temperature and maximum temperature. My results show that growth in leading edge red pine populations was significantly influenced by prior summer temperatures (negatively) and precipitation (positively) during the 20th century. Combined effects of temperature and precipitation suggest that drought is a dominant factor in red pine historical growth dynamics. My results do not support the hypothesis that warming temperatures are favouring growth in leading edge populations of red pine.

Key Words climate change, dendrochronology, hierarchical cluster analysis, leading edge populations, northern

Ontario, red pine, species range shift.

16

2.1 Introduction

Growth in many trees declines at their northern range limit due to unfavourable growth conditions created by factors such as climate (Mather and Yoshioka 1968, Kremenetski et al. 1998, Lenoir et al. 2008, Chen et al. 2011), disturbance (Flannigan and Bergeron 1998, Asselin et al. 2003, Leithead et al. 2010, 2012), species competition (Moorcroft et al. 2006, Ettinger and HilleRisLambers 2013) and edaphic factors

(Lafleur et al. 2010). It is well recognized that tree growth at northern range limits is mainly affected by climate (Woodward 1987), but to date, there is no consensus on the most growth-limiting climatic variable (Stephenson 1990, Brown et al. 1996). The role of temperature in limiting growth at northern range limit is emphasized (Hocker 1956, Fang and Lechowicz 2006), and studies show that growth is particularly affected by low temperature during growing season (Woodward et al. 1990, Flannigan and

Woodward 1994) and winter minimum temperature (Sakai and Weiser 1973, Arris and Eagleson 1989).

Projected climate warming should diminish the growth-limiting effects of temperature, and trees are expected to show increases in growth at their northern range limit (Reich and Oleksyn 2008).

Moisture-related climatic variables, such as precipitation, also play important role in limiting growth at northern range limits (Girardin et al. 2001, Bogino et al. 2009). The limiting effects of moisture-related climatic variables on growth are likely to increase due to climate change-induced moisture stress. As a result, it is possible that tree growth in many species will not increase despite warming at their northern range limit (Gavin and Hu 2006). This dynamic nature of growth-limiting climatic factors suggests that studies on growth response to climate should use a combination of temperature and moisture-condition variables (Shao and Halpin 1995, McKenney et al. 2007).

Climate influences tree-ring growth, and as such, tree-ring data are useful to study species growth

– climate relationships across a range of temporal and spatial scales (Swetnam et al. 1999, Gedalof and

Smith 2001, Hughes 2002, Silva et al. 2010, Wettstein et al. 2011). In general, studies on growth – climate‎relationships‎of‎species’‎leading‎edge‎populations‎are‎confined‎to‎small‎areas‎(Bergeron and 17

Gagon 1987, Tardif et al. 2006, Bhuta et al. 2009, Liang et al. 2011, Paul et al. 2014). Such studies provide a good insight into growth – climate relationships at a local scale, but fail to capture large scale spatial variability (Griesbauer and Green 2010a). As a result, it is not appropriate to generalize findings from such studies for species with wide geographic distribution (Vila et al. 2008). In addition, leading edge populations in such studies are selected either following a latitudinal gradient (e.g; Paul et al. 2014) or an isotherm (Rudolf 1990, Rudolph and Laidly 1990). These approaches to delineate leading edge populations are also questionable. Latitudinal gradients are used as surrogate for bioclimatic zones to differentiate northern and central populations. However, these zones rarely follow a simple latitudinal gradient (Hills 1969, Rowe 1972). Likewise, present northern limits of species in eastern Canada do not coincide with isotherms (Asselin et al. 2003).

One way to deal with growth – climate variability of populations among spatial locations is to try to group together areas with similar responses using data properties (Hartigan 1975). Hierarchical Cluster

Analysis (HCA) is one of such techniques, and is commonly used in various ecological applications

(Pielou 1984, Xu and Wunsch 2005, Legendre and Legendre 2012), typically for plant/animal classification (Kaufman and Rousseeuw 1990). In HCA, partitioning of data into non-overlapping clusters is based on a distance matrix, and without any a priori hypothesis or training samples (Kaufman and

Rousseeuw 1990, Viattchenin 2013). Examples of HCA applications in growth – climate studies are limited. HCA has been used to group response functions in tree-ring studies to interpret growth – climate relationships. Fritts (1974) used HCA to group large number of response functions into four major clusters; each cluster described a unique set of monthly/seasonal temperature and precipitation influencing growth of arid site conifers in western North America. Similarly, Tessier (1989) applied HCA to response functions to determine the role of habitat in growth – climate relationships. Other applications include classifying anatomical variables for their potential to record climate signals (Wimmer and

Grabner 2000), reconstructing distinct climate episodes from growth ring time series (Bunn et al. 2005),

18 grouping chronologies along an elevational gradient (Wilson and Hopfmueller 2001, Piovesan et al.

2005), identifying growth increment patterns (García Gonzálea et al. 1999, Koprowski and Zielski 2006,

García-González 2008), and grouping tree-ring chronologies based on growth response to extreme climatic events (Oberhuber et al. 1998). Climate is a strong driver of species distribution, and specific climatic conditions within a species distribution range can be used to explain variations in its occurrence

(Toledo et al. 2012). However, climate, as a predictor variable in HCA, has rarely been used to identify groups of species populations within its distribution range.

Red pine (Pinus resinosa Ait) in eastern Canada has wide geographic distribution ranging from the east coast to southeastern (Rudolf 1990). It has potential for large-scale northern range limit growth – climate investigations, but studies so far are mainly from lake islands in northwestern Quebec.

In these lake islands, various climatic factors, such as snow cover and drought, affect red pine regeneration (Bergeron and Brisson 1994). Confinement of red pine in these islands is mainly attributed to changing fire regimes (Bergeron and Gagon 1987, Bergeron and Brisson 1990), and the role of climate in determining this northern range limit has been ruled out (Flannigan and Bergeron 1998). On the other hand, a bioclimatic study suggests that the northern range limit of red pine is mainly controlled by insufficient warmth during the growing season (Flannigan and Woodward 1994). Based on such climatic controls, red pine is projected to move about 500 – 800 km towards higher latitudes under future warming scenarios (Overpeck et al. 1991, Flannigan and Woodward 1994, Clark and Perera 1995). However, such large-scale potential range shifts are largely criticized due to the faulty assumptions in the models such as ignoring effects of species dispersal rate, species interaction, and landscape fragmentation (Schenk 1996,

Davis et al. 1998) and the contradicting evidence from empirical studies (Moore 2003, Liang et al. 2011,

Zhu et al. 2012).

In fact, the climate driven range shift potential of red pine is unknown due to lack of empirical studies on its growth dynamics under changing climate. This is typically true for northern Ontario that 19 hosts the largest extent of northern populations of red pine. These populations extend from 50° N in northeastern Ontario (NEO) to about 51° N in northwestern Ontario (Little 1971). Here I study growth – climate relationships of leading edge of northern populations of red pine in northern Ontario, Canada. I apply HCA, with climatic variables as predictors, to identify leading edge of northern populations of red pine.

The overall objective of this study is to establish more reliable climatic indicators of red pine growth. More specific objectives are (i) to identify leading edge populations of red pine in northern

Ontario using an unbiased structural approach, and (ii) to examine the growth – climate relationships of these leading edge populations during the 20th century. In the context of species range shift potential under climate change, I also test the hypothesis that growth – temperature relationships in leading edge populations of red pine are positive.

2.2 Materials and Methods

2.2.1 Study Area and Data During summer 2012, tree-ring samples of red pine were collected from 28 locations throughout northern

Ontario. Potential sampling sites were identified a priori from Ontario Forest Resource Inventory database

(FRI 2009), and their field status were confirmed with the district foresters. The criteria for site selection were naturally regenerated red pine dominant stands without any recent disturbances such as fire, wind thrown, and insect infestation.

At each site 30 – 45 dominant red pine trees were sampled for tree-ring data. Core samples were extracted using increment borers following standard procedures (Maeglin 1979, Speer 2011). In general, a single core was extracted from each sampling tree at breast height. Each sample was then visually inspected to ensure its quality for further processing and analyses. A second sample was taken from a tree if the first one was not in good condition due to a knot or rotten part. Samples were then mounted on

20 plywood mounts and sanded to sharpen the ring boundaries. These prepared samples (growth series) were then scanned and tree-ring widths (TRW) were measured using a semi-automatic image analysis program

WinDendro (WinDENDRO 2012). In each growth series every TRW was also assigned a growth year.

These TRW data were supplemented with additional data from 26 sites available at ITRDB (International

Tree-Ring Database)‎and‎cef‎(Centre‎ďétude‎de‎la‎forêt).‎These‎additional TRW data included 9 sites of

St. George et al. (2008) and 17 sites of Girardin et al. (2006). Locations of all TRW data sites are shown in

Figure 2‎ .1.

Monthly climate data were obtained from Natural Resource Canada for the period of 1901 – 2010

(Price et al. 2000, McKenney et al. 2013). The data included total precipitation (water equivalent of all kinds of precipitations - Prec), mean minimum temperature (Tmin), mean maximum temperature (Tmax), and climate moisture index (CMI). CMI is a drought index calculated as difference between precipitation and potential evapotranspiration.

21

Figure ‎2.1: Extent of the study area in northern Ontario. Areas in pink represent red pine distribution range in northern Ontario (USGS 1999). Symbols‎(▲‎and‎■)‎show‎the‎locations‎of‎red‎pine‎sampling‎sites:‎Colour‎of‎symbols differentiate data source: red (this study), blue (Girardin et al. 2006) and green (St. George et al. 2008). Hierarchical Cluster Analysis (HCA) was performed‎to‎differentiate‎sites‎as‎leading‎edge‎(▲)‎and‎ non-leading‎edge‎(■).

22 2.2.2 Hierarchical Cluster Analysis

I applied HCA as an objective mean to identify groups of homogeneous sites based on their similarity in climatic conditions. This identified the representative sites for red pine leading edge populations. A typical HCA procedure involves the following steps: (i) calculate proximity (similarity/dissimilarity) matrix between objects (ii) identify the most similar objects from this matrix (iii) combine these objects into a new composite object - cluster (iv) calculate the proximity matrix of composite objects and the remaining objects (v) iterate the process until there are no more original/composite objects to combine or some stopping criteria is achieved (Revelle 1979, Mooi and Sarstedt 2011). Essentially, a proximity matrix 퐷 = (푑푖푗) of a data with 푛 items is a symmetrical 푛 by 푛 matrix where 푛 (푛 − 1)⁄2 upper diagonal entries are distinct and non-negative. The nature of proximity matrix is dependent on the calculation method used. Similarity matrix is based on correlations and dissimilarity matrix is based on

Euclidian distances.

In the present study, climatic variables in an n by p matrix were used in HCA to identify leading edge populations of red pine. Here n represented number of sites and p the number of climatic variables. I adopted agglomerative method whereby each site is treated as a nucleus, and the nuclei are progressively merged into larger clusters until all sites are merged into a single cluster (Hepsen and Vatansever 2012).

Decisions for successive merger of clusters were based on proximity matrix of climatic variables and not on their actual values (Bailey and Dubes 1982). Proximity matrix was calculated using squared Euclidean distance function (Tibshirani et al. 2001) as shown in equation (2.1). Here, 푑 represents the proximity between two objects 푥 and 푦, and is calculated based on j number of variables.

퐽 2 푑푥,푦 = ∑(푥푗 − 푦푗) (2.1) 푗=1

23

I used farthest neighbour rule to avoid any chaining effect (drawn-out clusters) during the clustering process (Legendre and Legendre 2012). The mathematical expression of this rule using three objects 푎, 푏 and 푒 is shown in equation (2.2). Here 푑 represents the proximity of an object 푎 to the composite object 푏푐, and is estimated from its distance to the farthest member of the 푏푐. Composite object 푏푐 is a cluster of 푏 and c formed at step (iii) of the procedure mentioned above. I also standardized data to eliminate any possible effects of different measurement units (Kaufman and Rousseeuw 1990).

푑(푏푐,푎) = 푚푎푥{푑(푏,푎), 푑(푐,푎)} (2.2)

In HCA I entered temperature and moisture variables in various combinations (monthly, seasonal and annual), and used spatial cohesion of formed clusters to evaluate their performance. Finally, I selected

HCA of three monthly climatic variables: Tmin, Tmax and CMI. This HCA provided spatially the most distinct clusters as shown in Figure 2‎ .1. I used IBM® SPSS® Statistics to perform HCA.

2.2.3 Red Pine Chronologies I developed chronologies for leading edge red pine populations using their respective TRW data.

Chronology development included the processes of crossdating and removing non-climatic signal from

TRW series.

Crossdating is the matching of TRW patterns between different trees across a stand. This is necessary to ensure that correct year is assigned to each TRW across growth series (Fritts 2001). In the absence of crossdating, a locally absent or false ring will produce error in growth series and may lead to erroneous results in subsequent analysis with climate data (Speer 2011). Dated growth series for each site were then tested for crossdating using COFECHA2 program (Grissino-Mayer 2001). By default,

COFECHA uses a 32-year cubic smoothing spline to remove low frequency variance from TRW series.

The remaining high frequency variance is then used for crossdating TRW series. I followed these default

2 An‎inverted‎Spanish‎word‎meaning‎‘co-date’‎or‎‘cross-date’‎(Grissino-Mayer 2001). 24 settings, and evaluated the quality of crossdating from TRW series correlations at the 99% confidence interval. For TRW series with low correlations, I re-inspected their core samples for any possible dating/measurement errors. After correcting such errors, I again included these series into COFECHA run to check their improved correlation. I retained only those TRW series for further chronology development that had high correlation in crossdating.

Theoretically, a TRW series is a combined representation of growth-influencing climatic (e.g; temperature and moisture) and non-climatic (e.g; age, competition and disturbances) factors (Cook 1990).

I removed the non- climatic component from TRW series by detrending using ARSTAN program (Cook

1985). This process included (i) fitting each series by a 40-year smoothing spline, with a 50% frequency response, and (ii) standardizing series into indices by dividing TRW with fitted values. I used same detrending to all growth series to fulfil the statistical requirement of uniform treatment (St. George et al.

2008). I also performed autoregressive modeling to remove within series correlation

(autocorrelation/persistence) and to fulfil the statistical requirement of independent observations (Speer

2011, Legendre and Legendre 2012). I created residual chronologies by using biweight robust mean of their respective indices series and developed final chronologies by adding their respective coefficients of pooled autoregression. Cook (1985) suggested this step to preserve any climatic signal that may have been lost during autoregressive modeling. I used these final chronologies for growth – climate analysis.

These chronologies and their respective sample depths are shown in Appendix‎ A.

Here I use a set of statistics to assess chronologies. These include (i) mean sensitivity (MS), (ii) serial correlation (SC), and (iii) two measures of signal strength (r-bar and EPS). These statistics are shown in Table 2‎ .1. MS is a statistical measure (scale 0-1) of mean difference in ring width from year to year (Speer 2011). It is suggested that MS sensitivity reflects higher sensitivity to climate signal (Fritts and Shatz 1975). However, lower MS does not mean the absence of climatic signal (Hughes et al. 1982).

In fact, chronologies with MS around 0.2 are sensitive enough for climatic signal (Speer 2011). SC is the 25 correlation‎of‎index‎value‎with‎the‎previous‎year’s‎index‎value‎(Fritts 1974), and the higher values reflect more‎dependence‎of‎growth‎on‎previous‎year’s‎growth‎(Campelo et al. 2006). The r-bar (average correlation among all series in a chronology) and the EPS (expressed population signal) are running measures of common signal strength (percent common variance). The achieved r-bar values (0.2 – 0.4) indicate the suitability of these chronologies for growth - climate analysis (Cook et al. 2000). EPS, common variability in the chronology, is the correlation between developed chronology and a theoretical chronology of infinite samples (Wigley et al. 1984, Cook et al. 2000). The obtained values for the analysis period (1901-2010) are well within the range of acceptable threshold values i.e; 0.80 – 0.85

(Cook et al. 2000, Speer 2011).

26 Table ‎2.1: Location characteristics and chronology statistics of red pine leading edge populations in northern Ontario, Canada. Site ID± Region Lon Lat Ele Data Source Crossdated MS SC rbar* EPS* (°W) (°N) (m) Period Series Redpr94 NWO -93.82 51.08 365 Girardin et al. 2006 1818-2001 34 0.15 0.387 0.440 0.964 PAKPRC98 NWO -93.43 50.75 349 Girardin et al. 2006 1744-2002 34 0.16 0.428 0.253 0.920 NWO25 NWO -91.003 50.602 398 This Study 1709-2011 40 0.122 0.649 0.292 0.943 CANA259 NWO -93.111 50.388 363 St. George et al. 2008 1807-2004 30 0.125 0.401 0.272 0.918 SEUPR95 NWO -92.28 50.32 379 Girardin et al. 2006 1837-2001 40 0.102 0.26 0.191 0.902 SIOPRA88 NWO -91.92 50.07 378 Girardin et al. 2006 1766-2002 28 0.144 0.34 0.291 0.909 NEO07 NEO -84.166 50.001 182 This Study 1843-2011 30 0.133 0.52 0.376 0.948 NEO09 NEO -84.583 49.557 307 This Study 1884-2011 27 0.114 0.429 0.363 0.939 NWO12 NWO -88.101 49.466 353 This Study 1869-2011 10 0.168 0.208 0.263 0.781 NWO11 NWO -88.121 49.322 399 This Study 1866-2011 26 0.143 0.569 0.395 0.944 NEO05 NEO -79.907 49.247 304 This Study 1785-2011 21 0.158 0.351 0.348 0.914 NEO06 NEO -81.203 49.049 275 This Study. 1836-2011 18 0.149 0.464 0.407 0.925 NEO08 NEO -84.276 48.964 331 This Study. 1719-2011 26 0.124 0.489 0.289 0.914 BLUPRC55 NEO -81.72 48.58 302 Girardin et al. 2006 1606-2002 46 0.141 0.551 0.307 0.953 NEO29 NEO -89.239 48.422 197 This Study. 1857-2011 42 0.201 0.459 0.362 0.959 IVAPRC52 NEO -82.5 48.15 363 Girardin et al. 2006 1828-2002 36 0.141 0.576 0.432 0.962 The abbreviations are: Lon (longitude), Lat (latitude), Ele (elevation), m (meter), Series (number of growth series), MS (mean sensitivity), SC (serial correlation), EPS (expressed population signal). Statistical measures (MS, SC, rbar and EPS) are explained in Section ‎2.2.2. ± Site‎IDs‎for‎“this‎study”‎refer‎to‎the‎location‎(NEO‎– northeastern Ontario / NWO – northwestern Ontario) and numerical order of sampling. For description of Site IDs of archived data please refer to the respective data source. * Running value of a 20- year window for the year 1990. This year represents the latest common year in all chronologies.

27 2.2.4 Red Pine Growth – Climate Analyses I‎performed‎correlation‎(Pearson’s‎product‎moment)‎and‎response‎functions‎analyses‎to‎assess‎growth – climate relationships of leading edge red pine populations. Response functions (Fritts et al. 1971, Blasing et al. 1984) are used in tree-ring studies to account for collinearity of climatic predictors (Briffa and Cook

1990). These analyses were done on a monthly and seasonal basis using three climatic variables (Prec,

Tmin and Tmax) for the period of 1901 – 2010. For seasonal analyses I used the following four seasons: prior summer (prior June – prior September), winter (prior October – current March), spring (current

April – current June) and summer (current July – current September). I preferred extreme temperatures

(Tmin and Tmax) over the mean temperature to better capture the limiting effect of temperature (Tessier

1989). I performed these analyses using DendroClim 2000. This program estimates coefficients for correlation and response functions from a 1000 times bootstrap; and the median coefficients are considered significant if they exceed half the difference between 97.5 and 2.5 percentile of 1000 estimates

(Biondi and Waikul 2004).

2.3 Results and Discussion

2.3.1 Cluster Analysis The HCA was performed in a 54 x 36 matrix representing sites and climatic variables respectively. The dendrogram of the clustering structure (Figure 2‎ .2) shows the successive merger of sites/clusters into larger clusters with increasing distance. The distances are rescaled (0-25) from a dissimilarity matrix

(Appendix‎ B).

28

Figure ‎2.2: Dendrogram of hierarchical cluster analysis (HCA) to determine leading edge populations. Along Y-axis are the sites that successively merge into clusters. X-axis is the rescaled (0 - 25) proximity matrix distances. Dotted vertical line shows the distance for two clusters.

29

A caveat with respect to HCA is that it tends to produce clusters that keep merging at higher distances (Dubes and Jain 1979). To identify/decide on optimal number of clusters in HCA, a number of methods have been proposed and applied. Examples include variance ratio criteria (Calinski and Harabasz

1974), gap statistics (Tibshirani et al. 2001, Yin et al. 2008), prediction based resampling (Dudoit and

Fridlyand 2002), clustering stability (Ben-Hur et al. 2002), Bayesian model (Gao et al. 2002), jump method (Sugar and James 2003), prediction strength (Tibshirani and Walther 2005), maximum clustering similarity (Albatineh and Niewiadomska-Bugaj 2011), and variance contrast statistic (Lyakh et al. 2012).

These methods vary in complexity and have inherent merits and limitations, and their applicability depends on data characteristics and the objectives of the investigation. For example, the variance ratio criteria can only be used if data have potentially three or more clusters (Mooi and Sarstedt 2011). As the objective of present study was to identify the cluster that represents leading edge populations, the issue of

“optimal‎number‎of‎clusters”‎was not of paramount importance. I used the distinctive break method to differentiate leading edge populations from non – leading edge populations. This method is based on proximity matrix distance statistics (coefficients); whereby the highest difference value of coefficients is used to determine optimal number of clusters (Mooi and Sarstedt 2011). I calculated coefficients differences between every two successive steps in cluster formations and found highest coefficient difference between clustering stage 51 and 52. This corresponds to the two cluster solution: leading edge and non-leading edge. Results are shown in Figure 2‎ .3.

30

Number of Clusters 52 48 44 40 36 32 28 24 20 16 12 8 4

110

90

70

50

30

10

Distance Statistic Difference Statistic Distance -10 2 6 10 14 18 22 26 30 34 38 42 46 50 Clustering Stage

Figure ‎2.3: Plot of distance statistic differences between every two successive stages of hierarchical cluster analysis (HCA). Highest value at stage 52 indicates the possible clustering solution (2 clusters).

Assessment of the stability of identified clusters, in term of their memberships, further requires validation of the clustering structure (Smith and Dubes 1980). However, applying any statistical test for validation is inappropriate as the number of clusters cannot be determined prior to applying a clustering algorithm (Bailey and Dubes 1982). A number of alternate methods/indices have thus been proposed in literature for this purpose (Strauss et al. 1973, Baker and Hubert 1976, Smith and Dubes 1980, Halkidi et al. 2001, Ben-Hur et al. 2002, Tibshirani and Walther 2005, Fränti et al. 2014). None of these has universal applicability and the choice of any method depends mainly on data characteristics and purpose of clustering (Milligan and Cooper 1985, Halkidi et al. 2001). I assessed the stability of developed clustering structure by introducing perturbations in the dataset using the following three approaches: (i) changing the order of sites entering the HCA algorithm (ii) randomly removing some climatic variables, and (iii) randomly partitioning half of the samples (Strauss et al. 1973, Mooi and Sarstedt 2011). In general, changing the order should not affect the output but presence of outliers in the dataset can result in changes in cluster membership (Mooi and Sarstedt 2011). A separate HCA was performed for each

31 perturbed dataset using the same parameters and results were compared with the original dataset. It is suggested that in stable clusters the membership should not change much due to such perturbations

(Strauss et al. 1973). The resultant clustering structures from the first two approaches were not different from the original. SPSS outputs of these tests are in Appendix‎ C and Appendix‎ D. In the split method, the dataset was randomly divided into two subsets (subset 1 and subset 2) and HCA was performed for each subset using similarity measure (Pearson Correlation). For subset 1, all leading edge populations were clustered together in a 3-cluster solution. For subset 2, all leading edge populations were also clustered together. In addition, five non-leading edge populations were also clustered with leading edge populations in a 2-cluster solution. These findings are in agreement with Strauss et al. (1973) who introduced this method to validate a 5-cluster solution for psychiatric patients. According to Strauss et al. (1973) a clustering solution is relatively stable if one of the half datasets yields the membership in the same clusters‎as‎that‎of‎the‎full‎dataset.‎Further,‎the‎strength‎of‎this‎method‎is‎to‎identify‎“discreteness‎of‎the‎ subgroups”.‎The‎obtained‎subgroups‎of‎leading‎edge‎populations,‎using‎this‎split‎half‎method, are also discrete from the non-leading edge populations (Appendix‎ E). These perturbation based validity methods suggest that the cluster of red pine leading edge population is relatively stable.

2.3.2 Growth – Climate Relationships Results of correlation and response function analyses of red pine growth in leading edge populations with monthly climate are summarized in Figure 2‎ .4. The number of significant relations (growth – climate) captured by response function are relatively less than the correlation analyses. The conservative behaviour of response function was due to its inherent property of uncorrelating the predictor (climatic) variables

(Fritts et al. 1971, Fritts 1974). Overall, these results suggest that summer months (June, July and August) are important for the growth of leading edge populations. During the 20th century, growth was more influenced by Tmax (negatively) and Prec (positively) than Tmin. Growth response varied among sites; and the limiting effect of climate at the regional level was more in NWO than NEO. Results also show 32 that limiting‎effect‎of‎prior‎year’s‎Tmax was‎more‎significant‎than‎growth‎year’s‎Tmax. Out of 16 populations,‎growth‎in‎10,‎6‎and‎8‎populations‎was‎negatively‎correlated‎with‎prior‎year’s‎Tmax during

August, July and June respectively. The Tmax of June and July during the growth year also limited growth in 7 and 8 populations respectively. Growth in populations at the highest latitude (Redpr94,

PAKPRC98 and CANA259) was also limited by September Tmax. Growth response to Prec during the summer months (prior and current year) was positive. Five populations showed positive correlation with current‎year’s‎July‎Prec followed by 4 sites for Prec during current June and prior September. At few sites growth – precipitation relationships were negative during winter months.

Further analyses (correlation and response function) were performed using seasonal climate data for each climatic variable and the results are shown in Table 2‎ .2 - 2‎ .5. In conformity with the monthly analyses, growth was least influenced by seasonal Tmin. Overall growth in most populations was correlated with seasonal Prec than Tmax. For most of the populations, growth during the 20th century was positively correlated with prior summer Prec. There were, however, regional differences. In NEO growth was positively correlated with spring Prec, whereas in NWO winter Prec was more significant and had negative effect on growth. Growth response to spring Prec was significant for populations at drier sites with relatively steep gradient (NEO08, NEO09, NEO29 and NWO12). Overall, growth response was spatially more consistent for prior summer Tmax. Growth of all populations in NWO and 5 (out of 8) in

NEO had negative correlation with prior summer Tmax. Response function coefficients also confirmed these relationships. Growth response (negative) to prior summer was more consistent in NWO than NEO.

33 Correlation Functions Response Functions 6 6 NEO NWO NEO NWO

4 4

2 2 Prec 0 0

-2 -2 4 4

0 0

-4 -4 Tmin -8 -8

-12 -12 4 4 0 0

-4 -4

Count of significant (positive/negative) functions functions significant Count(positive/negative) of -8 -8 Tmax -12

-12

Jul

Jan

Jun

JUL

Oct

Apr

SEP

Feb Sep

Aug

Jul

JUN

DEC

OCT Mar

May

Jan

AUG

Jun NOV

JUL

Oct

Apr

SEP

Feb Sep

Aug

JUN

DEC

OCT Mar

May AUG NOV Figure ‎2.4: Summary of significant functions between red pine chronologies of leading edge populations and monthly climatic variables: total precipitation (Prec), mean minimum temperature (Tmin) and mean maximum temperature (Tmax). Negative numbers at y-axis show number of significant negative functions. Prior‎year’s‎months‎(abbreviated in capital letters at x-axis) were included to‎account‎for‎effect‎of‎prior‎year’s‎ climate on growth. Chronologies are separated in two groups: NEO (northeastern Ontario) and NWO (northwestern Ontario).

34

Table ‎2.2: Correlation coefficients between annual growth (in NEO) and seasonal climate variables during the 20th century.

NEO07 NEO09 NEO05 NEO08 NEO06 NEO29 IVAPRC52 BLUPRC55 Prec Pr Summer 0.1126 0.2975 0.1210 0.2307 0.1351 0.1949 0.2482 0.1809 Winter -0.0822 -0.0493 0.0206 0.0336 0.0192 -0.1019 -0.0313 0.0459 Spring 0.0981 0.2063 0.1576 0.2944 0.0998 0.2052 0.0311 0.1333 Summer 0.0112 0.2644 0.0962 0.1373 0.0587 0.0282 0.0540 0.0534 Tmin Pr Summer -0.0636 -0.1762 -0.2008 -0.1925 -0.1380 -0.1923 -0.1803 -0.2145 Winter 0.1329 0.1778 0.0357 0.1413 0.1009 0.0454 0.1687 0.1179 Spring 0.0785 0.1120 -0.0077 0.0526 0.1392 0.0591 0.0884 0.0792 Summer 0.1821 0.0834 -0.1905 -0.0466 0.0318 -0.0280 0.0430 0.0396 Tmax Pr Summer -0.1469 -0.2071 -0.1979 -0.2115 -0.2587 -0.1254 -0.2596 -0.2777 Winter 0.0827 0.1604 -0.0311 0.1414 0.0210 0.1251 0.1490 0.0981 Spring 0.0300 0.0676 -0.0491 0.0017 -0.0002 0.0048 0.0644 0.0129 Summer 0.0816 -0.1382 -0.2568 -0.2189 -0.1063 -0.0151 0.0316 -0.0630

The abbreviations are: Tmin (mean minimum temperature), Tmax (mean maximum temperature), Prec (mean Precipitation), Pr Summer (prior Jun – Sep), Winter (prior Oct – current March), Spring (current Apr – Jun), Summer (current Jul – Sep), Significant values shown in bold italic are based on confidence interval between 97.5 and 2.5 percentile.

35

Table ‎2.3: Correlation coefficients between annual growth (in NWO) and seasonal climate variables during the 20th century.

NWO11 NWO12 NWO25 SIOPRA88 Redpr94 SEUPR95 PAKPRC98 CANA259 Prec Pr Summer 0.1924 0.0886 0.2615 0.3118 0.2958 0.2420 0.1211 0.1573 Winter -0.2066 -0.2372 -0.0124 -0.2651 -0.0408 -0.0716 -0.1054 -0.1142 Spring 0.0461 0.2571 0.0344 0.1863 0.1259 0.0941 -0.0605 0.1575 Summer 0.1170 0.1507 0.1593 0.2080 0.1523 0.0240 0.0412 0.0025 Tmin Pr Summer -0.1884 -0.1668 -0.1525 -0.2232 -0.0854 -0.1038 -0.3362 -0.2332 Winter 0.1001 0.0935 -0.0283 -0.0076 0.0004 0.1268 -0.0564 0.0293 Spring 0.1219 0.0392 0.0324 0.0466 0.0581 0.2173 0.0325 0.0563 Summer -0.0537 -0.0257 0.0011 -0.0686 0.0676 0.0734 -0.1240 -0.0300 Tmax Pr Summer -0.2407 -0.3466 -0.3209 -0.3608 -0.3914 -0.2530 -0.4163 -0.3838 Winter 0.1869 0.0731 -0.0600 0.0150 -0.0904 0.1431 -0.0161 0.0888 Spring 0.0803 -0.1505 -0.0249 0.0200 -0.0464 0.2085 -0.0078 -0.0018 Summer -0.1137 -0.2021 -0.1478 -0.1326 -0.1549 0.0226 -0.1603 -0.1137

The abbreviations are: Tmin (mean minimum temperature), Tmax (mean maximum temperature), Prec (mean Precipitation), Pr Summer (prior Jun – Sep), Winter (prior Oct – current March), Spring (current Apr – Jun), Summer (current Jul – Sep), Significant values shown in bold italic are based on confidence interval between 97.5 and 2.5 percentile.

36

Table ‎2.4: Response coefficients between annual growth (in NEO) and seasonal climate variables during the 20th century

NEO05 NEO06 NEO07 NEO08 NEO09 NEO29 IVAPRC52 BLUPRC55 Prec Pr Summer 0.0243 -0.0045 0.0591 0.1712 0.2381 0.1904 0.1760 0.0523 Winter 0.0849 0.0250 -0.0681 0.0773 0.0122 -0.0741 0.0511 0.1272 Spring 0.0980 0.0752 0.0911 0.2857 0.1918 0.2162 0.0152 0.1176 Summer -0.0133 -0.0590 -0.0169 0.0597 0.1909 0.0130 0.0555 -0.0064 Tmin Pr Summer -0.0431 0.0394 -0.0360 -0.2091 -0.2043 -0.1685 -0.1381 -0.1211 Winter 0.0921 0.1324 0.0513 0.0861 0.0584 0.0292 0.0534 0.0846 Spring 0.0624 0.1746 0.0398 0.0475 0.0117 -0.0014 -0.0008 0.0535 Summer -0.0552 0.1132 0.0826 -0.1734 -0.0568 -0.1107 -0.0304 0.0015 Tmax Pr Summer -0.1554 -0.2804 -0.1134 -0.0747 -0.0905 -0.0265 -0.1293 -0.1941 Winter -0.0063 -0.0609 0.0219 0.1826 0.1597 0.1553 0.0754 0.0515 Spring -0.0453 -0.0743 -0.0229 0.0546 0.0913 0.0590 0.0405 -0.0318 Summer -0.2156 -0.2505 0.0384 -0.1350 -0.0556 0.0107 0.0299 -0.0882

The abbreviations are: Tmin (mean minimum temperature), Tmax (mean maximum temperature), Prec (mean Precipitation), Pr Summer (prior year Jun – Sep), Winter (prior Oct – current March), Spring (current Apr – Jun), Summer (current Jul – Sep), Significant values shown in bold italic are based on confidence interval between 97.5 and 2.5 percentile.

37

Table ‎2.5: Response coefficients between annual growth (in NWO) and seasonal climate variables during the 20th century.

NWO11 NWO12 NWO25 SIOPRA88 Redpr94 SEUPR95 PAKPRC98 CANA259 Prec Pr Summer 0.1583 -0.0321 0.2136 0.2127 0.1867 0.1878 0.0827 0.0859 Winter -0.1491 -0.1298 0.0626 -0.1950 0.0271 0.0589 0.0591 -0.0007 Spring 0.0485 0.2532 0.0377 0.2310 0.0569 0.0784 -0.0177 0.1762 Summer 0.0448 0.0682 0.1124 0.1177 0.0998 -0.1131 0.0095 -0.0270 Tmin Pr Summer -0.1323 -0.1227 -0.0842 -0.1233 -0.0107 -0.1251 -0.1878 -0.1582 Winter 0.0929 0.1096 0.0020 -0.0386 0.0249 0.0377 0.0388 0.0512 Spring 0.0995 0.0921 0.0672 0.0700 0.0826 0.1760 0.0881 0.0917 Summer -0.1059 -0.0055 0.0321 -0.0005 0.1330 -0.0256 -0.0104 0.0116 Tmax Pr Summer -0.1562 -0.2759 -0.2643 -0.2222 -0.3469 -0.2494 -0.3122 -0.3021 Winter 0.1503 0.0970 -0.0322 -0.0072 -0.0813 0.0419 0.0167 0.0429 Spring 0.0691 -0.1032 0.0472 0.0285 -0.0293 0.1721 0.0702 0.0327 Summer -0.1332 -0.1961 -0.0786 -0.0386 -0.1305 -0.0321 -0.0586 -0.0485

The abbreviations are: Tmin (mean minimum temperature), Tmax (mean maximum temperature), Prec (mean Precipitation), Pr Summer (prior year Jun – Sep), Winter (prior Oct – current March), Spring (current Apr – Jun), Summer (current Jul – Sep), Significant values shown in bold italic are based on confidence interval between 97.5 and 2.5 percentile.

38 Growth response‎to‎prior‎year’s‎climatic‎conditions‎is‎not‎surprising‎(Jump et al. 2007), and can be explained by carbohydrate reserves and bud formation (Fritts 2001, Tardif et al. 2001a). Increased respiration due to high temperatures affects net photosynthesis and thereby reduces the accumulation of carbohydrates as reserve food. In such cases, when cambial activity commences in the following growing season, the growth is limited by the availability of reserved food. Further, limited water availability results in fewer and smaller buds. This affects leaves area and reduces photosynthetic area in the following year resulting in reduced growth. These results are consistent with studies on northern populations of some other species in Canada. Griesbauer and Green (2010a) found that growth of

Douglas-fir (Pseudotsuga menziesii) along its northern range margins in British Columbia was negatively correlated with prior summer temperature and positively correlated with prior summer and spring precipitation. Henderson and Grissino-Mayer (2009) reported a negative relationship between longleaf pine (Pinus palustris P. Mill.) and warm summer temperature in the South-eastern coastal plain, USA.

Results are also in conformity of Jacobi and Tainter (1988) who found negative correlation with prior summer temperature for white oak (Quercus alba L.) in South Carolina, USA.

The combined effect of temperature (negative correlation) and precipitation (positive correlation) on red pine growth also suggests importance of soil water balance; and the sensitivity of this species’‎ growth to drought during the prior year. The limiting effect of drought is relatively stronger in NWO than

NEO. In NWO, long spells of dry weather are common due to upper level high pressure ridge and the northwesterly flow aloft (Flannigan and Harrington 1988). Importance of drought as a limiting factor in species distribution has been reported for other species as well (Sakai and Weiser 1973, Tardif et al.

2006).

Studies suggest that at higher latitudes low temperatures affect species growth and survival and limit species distribution (Parker 1963, Fang and Lechowicz 2006). Increase in temperature under climate change is assumed to increase species growth, and their potential for survival at higher latitudes. This

39 assumption is commonly used to predict range shifts of species including red pine (Flannigan and

Woodward 1994, Newman et al. 2011). In the context of red pine range shift (migration) under climate change, I hypothesized a positive growth – temperature relationships for leading edge red pine populations. I also assumed that increased growth, due to rising temperature, will yield more cones

(Sutton et al. 2002), and more seedlings if favourable site conditions are met (Rudolf 1990, Carleton et al.

1996, Leithead et al. 2010). Further, the chances of survival and establishment of these seedlings at higher latitudes will increase due to warmer temperatures (Kozlowski and Borger 1971). However, my results indicate red pine growth – temperature relationships are not positive at leading edge. This suggests that northern range limit of red pine is not controlled by low temperatures. An earlier study by Sakai and

Weiser (1973) supports this observation. They tested freezing resistance of one year old red pines twigs and reported that the species is resistant to severe winter desiccation. My results also support the findings of studies on eastern (northwestern Quebec) and western (Manitoba) populations that red pine range limit is not controlled by cold climate (Flannigan and Woodward 1993, Tardif et al. 2001b, Sutton et al. 2002).

These findings also conform the observation that northern range limit of many other species in Canada are not controlled by cold climate (Meilleur et al. 1997, Tardif and Stevenson 2001, Asselin et al. 2003,

Tardif et al. 2006). These results further support the argument that empirical evidence of species specific growth – climate relationships, such as from tree-ring studies, should be considered in modeling for reliable range shift predictions.

Studies also show that species populations along their northern limit are more fragmented

(Meilleur et al. 1997) and perhaps habitat availability and disturbance regime overrides the physiological response to climate (Asselin et al. 2003). This is typically true for northern populations of red pine in

Quebec, Canada due to supportive fire regime for red pine germination and establishment, and harsh edaphic conditions for associate species (Bergeron and Brisson 1990, Flannigan and Woodward 1994,

Flannigan and Bergeron 1998). Red pine is dependent on fire for successful regeneration (Horton and

40

Brown 1960, Drever et al. 2006). Conducive conditions created by fire for successful regeneration include a competition free opening with a mineral seedbed or lightly covered with duff (Rudolf, 1990). Generally, a summer surface fire of intensity 200 – 500 Btu/sec-ft is sufficient to achieve these conditions (van

Wagner 1970). Occasional crown fires may also help by creating open patches for increased light penetration and releasing nutrients into the soil. Climate change-induced drought however, can alter such favourable fire regimes into more hostile crown fires that will further limit survival of red pine leading edge populations. Certainly, investigating such indirect effects of drought will help to formulate effective management strategies and reliable predictions for climate change impacts.

2.4 Conclusions

An unbiased structural approach, HCA, was applied to identify leading edge red pine populations over a larger spatial domain in northern Ontario. It is clear that leading edge populations cannot be defined by a simple latitudinal gradient. Further, growth – climate relationships during the 20th century were studied for these leading edge populations. Results‎show‎that‎red‎pine‎growth‎was‎related‎to‎prior‎year’s‎summer‎ temperatures (negatively) and precipitation (positively) suggesting that drought is a growth-limiting factor. Further, this study does not support the assumption that growth in red pine along its northern range limit is limited by low temperatures (Flannigan and Woodward 1994) and warming temperatures are favouring growth in leading edge populations. Results of this study question the validity of forest dynamic models based on the general assumption that temperature is a determining factor in controlling the northern limits of species.

41

CHAPTER 3 SPATIAL AND TEMPORAL VARIABILITY IN DENDROCLIMATIC RESPONSE OF RED PINE (PINUS RESINOSA AIT) TO CLIMATE IN NORTHERN ONTARIO, CANADA

Abstract

Growth – climate relationships are complex in trees. Investigating growth response to climate over large geographic areas and climatic gradients can provide insights into spatio-temporal complexity of species growth – climate relationships and the potential implications of climate change on forest dynamics. Here,

I study red pine (Pinus resinosa Ait) by analyzing tree-ring width data from 37 sites across northern

Ontario, Canada for the period 1901-2010. I performed response function and moving response function analyses using seasonal data of three climatic variables: precipitation, minimum temperature and maximum temperature. My analyses show considerable spatio-temporal variability in red pine growth, and growth response to climate during the 20th century. Regional droughts affected red pine growth during 1910s, 1940s and 1970-80s. In general, red pine growth response to climate was more prevalent in northwestern Ontario than in northeastern Ontario. Overall, precipitation was the most significant climatic control of red pine, with seasonal variations across sites. Unlike many other studies, I did not find any significant growth response to summer temperature. Instead, at many sites in northwestern Ontario red pine had a significant negative growth response to prior summer temperatures. The results of canonical correspondence analysis show that red pine growth response to climate follows more a longitudinal gradient than a latitudinal gradient. Precipitation patterns in northern Ontario also vary along a longitudinal gradient. Temporal stability analyses revealed that growth response of red pine at any site was not stable during the 20th century. I observed a general shift in significant growth response to temperature during the first half of the 20th century to precipitation during recent decades. My results suggest that effects of climate on red pine vary across northern Ontario.

42

Key Words canonical correspondence analysis, climate change, dendrochronology, drought, gradient analysis, growth

– climate relationships, northeastern Ontario, northwestern Ontario, Ontario ecoregions, red pine, response functions.

43

3.1 Introduction

Investigations of tree growth responses to climate change are required for vulnerability assessment, predictive modelling and adaptive management of forests (Graumlich 1989, Swetnam et al. 1999,

Spittlehouse 2005, Griesbauer and Green 2012). Climate variability influences radial growth (Fritts

2001), and thus large scale investigations on growth – climate relationships can help to predict species response to climate change (Gedalof and Smith 2001, Huang et al. 2010). The need to study individual species is also emphasized due to diversity of responses among species despite identical climatic changes

(Aber and Federer 1992, Hofgaard et al. 1999).

In general, a static growth – climate relationship for individual species is used to simulate forest dynamics (Hanson et al. 1989). Studies however suggest that static relationships are over simplifications

(Loehle and Leblanc 1996, Tardif et al. 2001b), as these overlook the variability that exists in growth – climate relationships (Cook‎and‎Cole‎1991,‎O’Neill‎et‎al.‎2008). Indeed, individual species have distinct growth – climate relationships that vary across spatial scales (Benavides et al. 2013). Dendroclimatic analyses of various species have confirmed spatial variability in growth – climate relationships. For example, the climate – growth relationships of Douglas-fir [Pesudotsuga menziesii var. glauca (Mirb.)] in

British Columbia are stronger for within range populations than those at the northern range margins

(Griesbauer and Green 2010a). In addition, various climatic variables have been reported to influence the growth of this species depending on the site characteristics such as annual precipitation at warm and dry sites, and annual temperature at wet and cold sites (Griesbauer and Green 2010b). Spatial variability in growth – climate relationships has also been reported for many other species in Canada such as black spruce [Picea mariana (Mill.) BSP.], jack pine (Pinus banksiana Lamb), lodgepole pine (Pinus contorta var. latifolia), mountain hemlock (Tsuga mertensiana), paper birch (Betula papyrifera Marshall), subalpine fir (Abies lasiocarpa) , trembling aspen (Populus tremuloides Michx), and white spruce [Picea glauca (Moench) Voss] (Gedalof and Smith 2001, Miyamoto et al. 2010, Huang et al. 2010, Griesbauer

44 and Green 2012). Huang et al. (2010) also postulated differential responses of black spruce and jack pine under future warming in eastern Canada. They predicted that, under future warming, northern populations

(north of 47°N) of these species will benefit in growth, whereas the southern populations will experience decline in growth.

In the face of climate change, investigations on temporal stability of growth – climate relationships are also necessary for proper model parameterization and validation (Graumlich 1989, Cook and Cole 1991).‎Dendroclimatic‎studies‎show‎that‎individual‎species’‎growth‎- climate relationships changed during the 20th century. Reduced sensitivity of growth to temperature and precipitation has been reported during recent decades in the northern hemisphere (Briffa et al. 1998a, 1998b, Wilson and Elling

2004,‎D’Arrigo‎et‎al.‎2008,‎Silva‎et‎al.‎2010). Biondi (2000) reported temporal inconsistency in growth response of Douglas-fir to May precipitation in North-Central Idaho, USA during the 20th century. The relationship was relatively weaker during the middle of the century compared to the earlier and latter parts of the century. Similar results have been observed for northern populations of this species in British

Columbia, Canada. Growth – annual precipitation relationships were significant during the earlier and later part of the 20th century but non-significant during the middle period (1935-65) (Griesbauer and

Green 2010b). A shift from positive growth response to summer temperatures to negative growth response to annual drought has been reported for high elevation white spruce in Yukon since the 1960s

(D’Arrigo‎et‎al.‎2004,‎Griesbauer‎and‎Green‎2012). Temporal instability of species specific growth – climate relationships has been reported in many other studies as well (Rolland et al. 1998, Hofgaard et al.

1999, Lloyd and Fastie 2002, Tardif et al. 2003, Huang et al. 2010, DeSoto et al. 2012).

Red pine (Pinus resinosa Ait) is one of the most genetically uniform tree species native to northeastern North America. It has a wide geographic distribution in Canada extending from southeastern

Manitoba eastward to Cape Breton Island (Nova Scotia) and in the south as far as West Virginia (USA).

Within its range, red pine mostly occurs in boreal forest/Great Lakes – St. Lawrence forest ecotone – an 45 area characterized by steep climatic gradients (Hills 1969, Rowe 1972, Rudolf 1990). Characteristic climatic requirements of red pine are low to moderate rainfall, cold winters and cool to warm summers

(OMNR 1998). Macrofossil records suggest that the red pine distribution reached into the eastern part of northern Ontario (NEO) around 7400 BP during the Hypsithermal when boreal forests transformed into

Great Lakes – St. Lawrence forests due to warmer and drier climate (Liu 1990). In the western part of northern Ontario (NWO) red pine reached its northern limit as far as Indian Lake (~ 51°N) around 8800

BP (Björck 1985). Red pine is a shade intolerant species, and depends on low intensity fires that provide favourable conditions for its regeneration and establishment (Bergeron and Brisson 1990). Red pine is a drought resistant species and is most commonly found on coarse to fine well-drained sandy soils with low nutrients. On poor sites it usually grows in pure even-aged stands, whereas at sites with moderate nutrients it grows with mixedwood (white birch and trembling aspen). Other associate species are white pine (Pinus strobus L.), balsam fir [Abies balsamea (L.)], black spruce and jack pine (Sims et al. 1990).

Red pine preserves climatic signal in annual growth (Kipfmueller et al. 2010), and its radial growth data has been used in combination with other species for various regional dendroclimatic analyses. These studies include investigations on synoptic scale atmospheric circulation and summer drought in boreal Canada (Girardin et al. 2006), summer climate variability over 300 years in Winnipeg

River basin (St. George et al. 2008), summer drought severity in northwestern Ontario (St. George et al.

2009), and‎growth‎response‎of‎species,‎in‎Ontario’s‎temperate‎and‎boreal‎forests,‎to increasing atmospheric CO2 (Silva et al. 2010). Red pine tree-ring data have also been used to investigate fire regimes (Bergeron and Brisson 1990), post fire regeneration (Bergeron and Brisson 1994) and age structure (Bergeron and Gagon 1987) in northwestern Quebec, Canada. However, little is known about the broad scale associations of red pine radial growth with seasonal climate, and the spatio-temporal variability of these associations. Studies done so far are for southern populations in upper Great Lakes region (Graumlich 1993) and northern Minnesota (Kipfmueller et al. 2010). This information is lacking

46 for northern populations in northern Ontario, Canada. The purpose of this study is to investigate spatio- temporal patterns of red pine growth – climate associations in northern Ontario. These associations can help to create plausible scenarios for this species under global warming conditions and to formulate adaptive management practices. The specific objectives of the study are (i) to identify red pine growth – climate associations in northern Ontario, (ii) to identify spatial patterns of red pine growth – climate associations in northern Ontario, and (iii) to assess the temporal stability of red pine growth – climate associations in northern Ontario. I also test the null hypotheses that (i) red pine growth - climate associations in northern Ontario exhibit no spatial variability, and (ii) red pine growth – climate associations in northern Ontario were stable during the 20th century.

3.2 Materials and Methods

3.2.1 Study Area The study area, in northern Ontario, consists of the northern parts of the Great Lakes – St. Lawrence forests and some southern sections of the boreal forest. The study area is further differentiated into two regions: NEO – northeastern Ontario and NWO – northwestern Ontario (Figure 3‎ .1). The study area can be‎described‎using‎the‎information‎provided‎by‎the‎Hills’‎classification of terrestrial ecosystems in

Ontario. The study area contains five ecoregions [4E (NEO); 4S, 4W, 5S and 3W (NWO)] which have been described based on distinct climatic patterns of temperature, precipitation and humidity (Hills 1959,

1969, Burger 1993, Crins and Uhlig 2000, Crins et al. 2009). In general, moisture in the study area decreases from East to West. Ecoregion 4E, also known as the Lake Temagami ecoregion, mainly consists of Great Lakes – St. Lawrence forests with numerous lakes and rivers. Climate in this ecoregion is humid and cold with mean annual temperature (MAT) of 0.8 – 4.3°C and growing season of 171 – 200 days. Climate in the ecoregion 4S is influenced by prairie climate and is relatively dry and cool with

MAT 0.1 – 2.6°C and growing season of 174 – 188 days. Low summer rainfall (245 – 291 mm) in this

47 ecoregion often leads to droughts. A steep climatic gradient in the 4S results a rapid ecological transition from Great Lakes – St. Lawrence forests in the warmer and drier south to the boreal forests in the north.

Forest productivity particularly in the western part is low due to shallow substrates (Mackey et al. 1996a,

1996b, Crins et al. 2009). The 4W (Pigeon river ecoregion) is in the Quetico section of Great Lakes – St.

Lawrence forests (Rowe 1972). In this ecoregion, MAT ranges between 0.2 – 2.7°C and growing season is from 168 to 188 days. Red pine in this ecoregion grows on well-drained sites. Climate in the 5S is characterized by warm, moist (rainfall 243 – 287 mm) summers and cold winters. MAT is 1.4 – 2.8°C and growing season is 182 – 190 days. The 3W is a typical boreal forest region with shorter growing season

(161-182) and moist and colder climate than the adjacent areas of similar latitude. Scattered stands of red pine also occur in warmer areas of the southern 3W (Mackey et al. 1996a, 1996b, Crins et al. 2009).

48

Figure ‎3.1: Study area in northern Ontario. Locations of sampling sites are shown with square symbol (■).‎Data‎sources‎are‎differentiated‎by‎ colour of symbols: red (this study), blue (Girardin et al. 2006) and green (St. George et al. 2008). Ecoregions are delineated with blue dotted lines Ontario regions are delineated with black lines.

49 3.2.2 Data Collection and Preparation

3.2.2.1 Climate Data The extrapolated climate data for three variables namely mean minimum temperature (Tmin), mean maximum temperature (Tmax) and total precipitation (Prec), for the period 1901 – 2010, were obtained from Natural Resource Canada. These extrapolated data were generated, from point (observatory) climate data, using thin plate smoothing splines, as surface fitting, in ANUSPLIN package (McKenney et al.

2001, 2013, Price et al. 2011). Surface fitting process essentially used latitude and longitude as independent variables; and occasionally used altitude as a third independent variable or as an independent covariate (Hutchinson 1995). ANUSPLIN generated climate data have already been used in dendroclimatic studies in Canada (Chhin et al. 2008, Silva et al. 2010, Huang et al. 2010). The obtained monthly climate data were then used to generate the following seasonal data for each climatic variable: prior summer (prior June – prior September), winter (prior October – current March), spring (current

April – current June) and summer (current July – current September). I preferred extreme temperatures, over mean temperature, due to their influences on growing season length (Tmin) and net productivity - gross photosynthesis minus leaf respiration (Tmax) (Aber and Federer 1992). To illustrate spatio-temporal variability of seasonal Tmin, Tmax and Prec, I interpolated these data for 1901 – 2010 with ordinary kriging using a spherical semi-variogram model (Figure ‎3.2 - ‎3.4). These figures show that seasonal climatic variables not only vary across the study area (as shown by mean values plotted in A, C, and E) but also exhibit temporal variability (as shown by standard deviations plotted in B, D, and E). In general, temporal variability in seasonal climate is high in the northwestern part of the study area.

50

(A)

(B)

(C)

(D)

(E)

(F)

Figure ‎3.2: Interpolated data of mean minimum temperature (Tmin in oC) for various seasons. Mean Tmin for the period 1901-2010 are shown in plot A (winter), C (spring), and E (prior summer). Temporal variability for the period 1901-2010 (standard deviations) are shown in plots B (winter), D (spring), and F (prior summer). Interpolation method and seasons are described in section ‎3.2.2.1.

51

(A)

(B)

(C)

(D)

(E)

(F)

Figure ‎3.3: Interpolated data of mean maximum temperature (Tmax in oC) for various seasons. Mean Tmax for the period 1901-2010 are shown in plot A (winter), C (spring), and E (prior summer). Temporal variability for the period 1901-2010 (standard deviations) are shown in plots B (winter), D (spring), and F (prior summer). Interpolation method and seasons are described in section ‎3.2.2.1.

52

(A)

(B)

(C)

(D)

(E)

(F)

Figure ‎3.4: Interpolated data of total precipitation (Prec in mm) for various seasons. Mean Prec for the period 1901-2010 are shown in plot A (winter), C (spring), and E (prior summer). Temporal variability for the period 1901-2010 (standard deviations) are shown in plots B (winter), D (spring), and F (prior summer). Interpolation method and seasons are described in section ‎3.2.2.1.

53

3.2.2.2 Tree-Ring Data Tree-ring samples of red pine were collected from 18 sampling sites during summer 2012. Sampling site selection criteria included naturally regenerated stands, red pine as a dominant species, and an absence of recent disturbances (fire, wind throw, insect/disease infestation etc.). At each sampling site, core samples were extracted at breast height using a 0.5 mm increment borer from 30 – 45 trees for tree-ring width

(TRW) data. All core samples were prepared for TRW measurement. The preparation procedure included sample mounting, sanding and scanning. Scanned images were then imported to WinDendro® to delineate annual TRW using light intensity parameter (WinDENDRO, 2012). Annual TRW measurements were recorded with 0.001 mm precision, and every TRW series was crossdated by assigning a calendar year to each annual TRW. These crossdated TRW series were checked for their accuracy using COFECHA program. This program is designed to assist in testing the accuracy/validity of the crossdating using correlation statistics (Grissino-Mayer 2001). A 32-year cubic smoothing spline with 50% wavelength cut- off was applied to each crossdated TRW series for an individual site. Pearson correlation coefficients with master dating series were estimated and their significance was determined based on 99% confidence interval level (Grissino-Mayer 2001, Speer 2011). Those series or section of series that yielded correlation coefficients less than critical values, were re-examined for possible dating errors. Where possible, dating errors were removed and COFECHA was re-run to verify the improvements in correlations.

To increase the spatial resolution of sampling sites, archived data of red pine TRW from 19 other sites (Girardin et al. 2006, St. George et al. 2008) were also used in this study. Quality of these data was assessed using COFECHA, and some TRW series were removed due to their poor correlation with their master dating series. Distribution of all data sites and their sources are shown in Figure 3‎ .1.

3.2.3 Chronology Development To relate ring patterns with yearly climatic variations, crossdated TRW series were standardized by dividing the actual growth values with the estimated values from a fitted curve. This process essentially 54 removes age and size related growth difference in growth series (Fritts 1966, 2001). Cubic spline was used for curve fitting (Cook and Peters 1981, Cook et al. 1990a). Growth ring series were fitted with various wavelengths of cubic spline to compare their performance. A 40-year cubic spline with 50% frequency response was finally selected to standardize (detrend) the TRW series. This preserved 99% of the individual TRW series variation at the wavelength of approximately 13 years (Speer 2011). A 40-year cubic spline has also been used to develop chronologies for other species such as white spruce, lodgepole pine, and subalpine fir in Canada (Miyamoto et al. 2010), and stone pine (Pinus pinea) in dry

Mediterranean Portugal (Campelo et al. 2006).

TRW series indices were further pre-whitened, through autoregressive modeling (AR), to remove persistence (correlation over successive years). Minimum Akaike Information Criterion (AIK) was used to select the best AR model (Akaike 1974). Pre-whitened TRW series for each site were then pooled using a biweight robust mean to develop residual site chronology (Cook et al. 1990b). Finally, the pooled coefficient of autoregression was added back to each residual chronology to preserve low frequency climatic signal (Cook 1985). These chronologies were used in subsequent analyses for growth – climate associations.

A few descriptive statistics were used to assess the quality of chronologies and their suitability for further growth – climate analyses. These include expressed population signal (EPS), mean sensitivity

(MS), serial correlation (SC) and standard deviation (SD). The respective values of these statistics for each chronology are shown in Table 3‎ .1.

55

Table ‎3.1: General statistics of 37 red pine site chronologies. Site ID Lon Lat EPS* MW SD MS SC (°W) (°N) (mm) NEO02 79.4795 47.1430 0.913ₐ 0.995 0.199 0.187 0.280 NEO03 79.6741 47.8596 0.862 0.988 0.247 0.183 0.479 NEO01 79.9769 46.9703 0.870 0.998 0.176 0.136 0.481 NEO04 80.1959 47.6161 0.928ₐ 0.999 0.212 0.157 0.590 556PRB 83.4500 46.8700 0.931 0.994 0.162 0.108 0.653 MONPRE 84.6500 47.2200 0.982 0.994 0.258 0.229 0.386 SGWPR 88.7300 48.4500 0.966 0.996 0.180 0.149 0.486 NWO13 88.7862 48.4012 0.969 0.993 0.196 0.185 0.346 SOWPRA 91.1700 49.5300 0.948 0.995 0.175 0.131 0.548 CANA264 91.2310 48.6064 0.958 0.992 0.139 0.110 0.558 NWO28 91.3192 48.7094 0.951ₑ 0.990 0.133 0.130 0.290 SABPRA 91.5500 49.4500 0.882 0.993 0.202 0.140 0.630 NWO27 91.7038 48.8978 0.939 0.992 0.160 0.141 0.413 NWO14 91.8133 48.9090 0.957 0.997 0.153 0.135 0.416 STOPRB 92.2300 49.3500 0.864 0.993 0.162 0.111 0.643 NWO15 92.8079 49.6525 0.962ₑ 0.994 0.269 0.170 0.683 NWO26 93.0746 48.7980 0.881 0.984 0.206 0.149 0.610 BIBPR 93.3300 49.7800 0.940 0.992 0.168 0.155 0.333 NWO16 93.3581 49.8515 0.965ₐ 1.006 0.175 0.159 0.338 CANA246 93.4870 50.0347 0.975 0.996 0.170 0.164 0.329 NWO17 93.5000 49.9253 0.975ₐ 0.990 0.207 0.179 0.418 NWO19 93.5059 50.0596 0.966 0.996 0.188 0.175 0.352 CANA263 93.6700 49.6444 0.899 0.998 0.190 0.154 0.456 CANA252 93.7390 49.8446 0.911 0.993 0.196 0.122 0.597 CANA255 93.8800 49.6444 0.911 0.999 0.152 0.138 0.347 CALPR 93.9200 49.0500 0.919 0.998 0.177 0.158 0.422

56

CANA245 94.0099 49.0484 0.971 0.996 0.188 0.168 0.411 LOOPR2 94.0500 49.4200 0.970 0.994 0.199 0.190 0.303 NWO24 94.0883 49.6410 0.891 0.991 0.139 0.115 0.488 JOHPRB 94.1200 49.9200 0.933 0.996 0.139 0.133 0.303 CANA265 94.1610 49.1441 0.943 0.994 0.197 0.204 0.252 LONPR3 94.2800 49.7200 0.967 0.992 0.197 0.198 0.231 NWO20 94.4069 49.6394 0.960ₑ 1.000 0.148 0.129 0.421 NWO21 94.4999 49.7795 0.969 0.994 0.169 0.150 0.430 NWO23 94.5731 49.7247 0.963ₑ 0.995 0.178 0.160 0.299 NWO22 94.5919 50.1393 0.953 0.998 0.223 0.177 0.520 CANA253 94.8700 49.6444 0.968 0.994 0.215 0.212 0.304 Abbreviations used are MW (indices mean width), SD (standard deviation), MS (mean sensitivity), SC (serial correlation), and EPS (expressed population signal). * Residual chronology values for common period 1900 – 2000 ₑ Common period is 1920 – 2010. ₐ Common period is 1960 – 2010.

57

EPS is a measure of common variability in a chronology, and it depends on sample depth (Wigley et al. 1984). The mathematical expression of EPS is shown in equation (3.1) where 푁 is the total number of series, 푟̅ is the average value of correlations between series, and 푅̅푁 is the expected correlation between

푁- series averages.

푁푟̅ 퐸푃푆 = (푅̅ )2 ≈ (3.1) 푁 1 + (푁 − 1)푟̅

MS is a statistical measure of year to year variation in chronology (Speer 2011). SC, in a chronology,‎is‎the‎correlation‎of‎index‎value‎with‎the‎previous‎year’s‎index‎value‎(Fritts 1974). It is suggested that for dendroclimatic analysis, a chronology should have high EPS, MS, and SD, but low SC

(DeWitt and Ames 1978, Wigley et al. 1984). An EPS threshold value (0.80) is desirable to capture common climate variance in a chronology (Cook et al. 2000, Speer 2011). All chronologies developed for this study have higher EPS values (0.86 – 0.98) than the suggested threshold.

3.2.4 Spatial and Temporal Growth – Climate Analyses I performed response function analyses, using three climatic variables (Tmin, Tmax and Prec), to determine growth – climate associations. Response functions coefficients are the multivariate estimates from a principal component regression model, and are commonly used in tree-ring studies to deal with collinearity of climatic predictors (Fritts et al. 1971, Blasing et al. 1984, Briffa and Cook 1990, Biondi

2000, Gedalof and Smith 2001, Huang et al. 2010). I performed these analyses using DendroClim 2002

(Biondi and Waikul 2004).

I examined spatial variability using gradient analysis. I used Canonical Correspondence Analysis

(CCA) to understand the patterns of significant response functions with respect to three geographic variables (latitude, longitude, altitude). CCA is an asymmetric ordination method in which ordination is constrained to represent the variation in response variable that is due to explanatory variables. Since its

58 first application in community ecology (ter Braak 1986), CCA has been widely used in various disciplines of ecology (Palmer 1993, Birks et al. 1998, Legendre and Birks 2012, Legendre and Legendre 2012). The output of CCA is an ordination diagram (triplot) that displays response variables, sites and explanatory variables along ordination axes. Locations of response variables in the ordination diagram represent their relative position in two dimensional niches essentially located at the centroid of the site points at which they occur. Explanatory variables are shown with lines originating from the centre (0,0). The coordinates of line head for quantitative explanatory variables, like in this study, are determined by their correlations with the axes. The length of line for any explanatory variable, in general, reflects the relative importance of that variable with respect to other explanatory variables (ter Braak and Verdonschot 1995). Ranking of response variables with respect to any explanatory variable is assessed by weighted averages. These weighted averages are approximated by projecting the response variable on an explanatory variable line

(Legendre and Legendre 2012).

I examined temporal variability by measuring stability of growth – climate associations using a

30-year moving response function starting with the period 1902- 31. The 30-year window was progressively moved with 1-year step till it reached the last 30 years of the chronology. Stability of response /moving response coefficients was tested with 1000 times bootstrap (Guiot 1990). A median value from these 1000 times bootstrap was used as final coefficient for each response function/moving response function. Statistical significance of each coefficient was assessed using confidence interval at

95% limit (between 97.5 and 2.5 percentile).

I also performed correlation analyses between significant moving response coefficients and standard deviations (temporal variability) of seasonal climatic variables at multiple spatial scales. The purpose of these analyses was to test if the stability of red pine growth – climate relationships is influenced by variability in any/all climatic variables.

59

3.3 Results

3.3.1 Red Pine Chronologies

Site characteristics of red pine study sites are summarized in Appendix‎ F. A total of 37 site chronologies were developed using 1291 dated TRW series. These site chronologies with their respective sample depths are shown in Appendix‎ G. The number of crossdated TRW series per site varied from 14 (NEO01) to 75

(CANA264) with an average of 34.9 ± 10.4. This variation in number of TRW series was mainly due to the number of samples collected at respective sites. Length of individual growth series varied from 79 to

362 years. The shortest one was from 1933 AD, whereas the longest one dated to 1640 AD.

Generally, the higher MS reflects a higher sensitivity of growth to climate (Fritts and Shatz 1975) especially drought stress (Fritts et al. 1965). MS of eastern species in North America is relatively lower than the western species. I compared MS of various species in western North America (Holmes et al.

1986) and eastern North America (DeWitt and Ames 1978), and found that difference in MS of species between both regions was significant at 95% confidence level. Lower MS of eastern species is attributed to temperate climate and higher stand competition (DeWitt and Ames 1978). Studies also show that despite low MS chronologies can still have strong climatic signals (Hughes et al. 1982), and a value of 0.2 is reasonable (Speer 2011). My site chronologies’ MS ranged from 0.11 – 0.23 (0.16 ± 0.03). This matched the MS (0.175 ± 0.041) of eastern species including red pine (DeWitt and Ames 1978). Average values of SC and SD for all chronologies were 0.43±0.13 and 0.18±0.03 respectively; and these values closely match with other studies on multiple species in eastern North America (DeWitt and Ames 1978,

Huang et al. 2010).

My results show that red pine growth patterns were synchronous during 20th century. Three episodes (1910s, 1940s and 1970-80s) of synchronous reduced growth occurred. During the 1910s almost all sites showed reduced growth. During the 1940s, most of the eastern sites up to -93.5° longitude

60 showed reduced growth. This reduced growth was not common in the most western populations. Reduced growth of red pine during the 1910s and 1940s has also been reported in the Winnipeg River region, which overlaps with this study area (St. George et al. 2008). The third reduced growth episode happened during the 1970s and it shifted to the 1980s in most of the populations west of 93.5° longitude. These regional level reduced growth patterns indicate the possible influence of some large scale climatic patterns. It is likely that such reduced growth is due to droughts that have been reported in most parts of this study area during these periods (Girardin et al. 2006, Girardin and Wotton 2009).

3.3.2 Growth – Climate Associations

Results of response function analyses are summarized in Figure 3‎ .5 and Appendix‎ H - J. Red pine growth during the 20th century was influenced more by seasonal Prec and Tmax than Tmin. Although, overall

Prec was the most important factor for growth, prior summer Tmax had the most consistent negative effect on red pine growth throughout the study area. In general five growth – climate associations were observed in the study area: (i) negative growth response to prior summer Tmax, (ii) positive growth response to spring Prec, (iii) positive growth response to summer Prec, (iv) negative growth response to prior summer Tmin, and (v) positive growth response to prior summer Prec. A few sites showed negative response to winter precipitation. Results show that prior summer climate plays an important role in red pine growth in northern Ontario.

61 50.50 Legend ▲ - Precipitation 50.00 ■ - Max Temperature ● - Min Temperature ♦ - Precipitation and Min/Max Temperature 49.50 + - Precipitation, Min and Max Temperature

49.00

N) °

48.50 Latitude( 48.00

47.50

47.00

46.50 95 94 93 92 91 90 89 88 87 86 85 84 83 82 81 80 79 Longitude (°W)

Figure ‎3.5: Scatter plot showing geographic distribution of significant seasonal climatic variables influencing red pine growth in northern Ontario: ▲‎(precipitation),‎■‎(maximum‎temperature),‎●‎(minimum‎temperature),‎♦‎(precipitation‎and‎minimum/maximum‎temperature) and + (precipitation, minimum and maximum temperature).

62 3.3.3 Spatial Variability in Growth – Climate Associations A shift of growth-influencing climatic factors along latitudinal and altitudinal gradients has been reported for many species in Canada such as white spruce, black spruce, red pine, red oak, paper birch and trembling aspen (Silva et al. 2010, Huang et al. 2010, Griesbauer and Green 2012). I did not find any patterns of spatial variability in red pine growth – climate relationships along altitudinal and latitudinal gradients. However, spatial variability of the growth – precipitation association was observed along a longitudinal gradient. A shift in positive response to seasonal precipitation along 92° was observed. In general, red pine growth east of 92° is positively related to prior summer Prec, whereas red pine growth west of 92° is positively related to spring and summer Prec.

Overall growth in NEO was responsive to prior summer and summer climate. Growth at a few sites also had positive response to winter Tmin Spatial patterns were more prominent in NWO. Response function‎analyses‎show‎that‎growth’s‎negative‎response‎to‎prior summer Tmax was almost persistent throughout NWO. There was also a transition zone along 93°W where additional influencing factors were winter Prec (negative response), summer Prec (positive response) and preceding summer Tmin (negative response). Gedalof and Smith (2001) also reported a transition zone for additional growth-influencing climatic variables for mountain hemlock in western North America.

At the ecoregion level, red pine growth response to climate also varied. Natural stands of red pine are abundant in ecoregion 4S. In this ecoregion growth response to spring and summer Prec was positive.

This is in addition to a negative response to preceding summer Tmax. In 5S growth mainly responded to prior summer Tmax (negative) and spring Prec (positive). Contrarily, growth – climate associations in 4W were not significant. In this ecoregion data were analysed from six populations: two in each eastern, central and western part of the ecoregion. In fact, only a week negative response to prior summer Tmax was noticed in the western part of 4W. In 3W (boreal forest region) natural stands of red pine are not very common. Results show that red pine growth in this ecoregion was associated with prior summer

63 temperature (negative) and precipitation (positive). These results are however based on only two sampling sites and therefore should be interpreted with caution.

Red pine growth response to climate is more prominent along longitude than latitude (Figure 3‎ .6).

Two precipitation based response functions (i.e; prior summer and summer) are more related to longitude, whereas prior summer minimum temperature response function is closely related to latitude. Almost equal projection distance of prior summer maximum temperature for latitude and longitude suggest that these explanatory variables are equally important for red pine growth response to prior summer maximum temperature.

The remaining response variables show a weaker relationship with latitude, longitude and altitude as is evident from their respective larger projection distances. The canonical axes account together for about 50% of the variation in growth response (Axis – 1: 31% of the variation – p = 0.009 after 999 permutations, Axis – 2: 19% of variation – p = 0.004 after 999 permutations). Axis – 1 shows strong gradient as its eigenvalue is greater than 0.30 (ter Braak and Verdonschot 1995).

64

Figure ‎3.6: Results of Canonical Correspondence Analysis of significant/non-significant response functions. These response functions were derived from red pine growth – climate analyses. Triplot is amplified by factor of 5 to improve the visuals. The eigenvalues of axis 1 (horizontal) and axis 2 (vertical) are 0.313 and 0.186 respectively. Together these axes account for about 50% of the variation. Sites are shown in black triangles. Three sites (out of 37) and one climatic variable (out of 12) were excluded from the analysis due to absence of any significant response function. Less points of sites are visible due to their overlap in 2-D. Three explanatory (geographic) variables, namely latitude (Lat), longitude (Lon) and altitude (Alt) are represented by green lines. Names of response functions (blue) are abbreviated in two parts separated by hyphen ( - ): first part represents climatic variable [precipitation (P), minimum temperature (Tn), maximum temperature (Tx)] and second part is the season [prior summer (ps), winter (wn), spring (sp) and summer (sm)].

65 3.3.4 Temporal Variability in Growth _ Climate Associations Results of moving response function analyses suggest that growth response to any climatic variable was not stable during the 20th century. Negative responses to summer/prior summer temperatures were more stable during first half of the 20th century, whereas positive responses to summer and spring precipitation were prominent during recent decades. In this section, temporal stability (or variability) of only those growth – climate associations is discussed that pertain to spatially significant growth – climate associations at various sites. In general I found that temporal responses to climate varied spatially with some areas showing positive response to prior summer/summer Prec and others showing negative response to prior summer Tmax. The detailed results of temporal response for each ecoregion are as follow.

In ecoregion 4E the most stable growth response was with summer Prec. Overall radial growth was positively related to summer Prec for 30-year periods ending 1967 and onward. This signal however diminished towards the end of the 20th century. The positive response to winter Tmin started during the first quarter of the 20th century (30-year ending on 1947) and lasted until the 1980s. Radial growth at two sites (556PRB and NEO03) benefitted for the longest period from warm winters. In addition growth at higher latitude (NEO03) was also limited due to rising summer Tmax and Tmin during the earlier periods

(up to 1959). This limiting effect was however non-significant at lower latitude (556PRB). These results are shown in Figure 3‎ .7.

66

Figure ‎3.7: Ecoregion 4E – Moving response function analyses (with a 30-year window) using significant climatic variables for red pine growth. At the top of each graph are site ID (left) and seasonal climatic variable (right). X-axis shows the last year of each 30-year window. Y-axis shows moving response coefficients. Black lines show standardized coefficients of moving response functions. Grey areas show 95% confidence interval. All sites show non-stable temporal dynamics of growth-climate relationships.

67

In 3W the positive responses to prior summer Prec were significant for very short durations (2-6 consecutive 30-year windows) mostly ending during 1980s. However, negative effects of prior summer

Tmax were persistent for relatively longer periods: 1957-75 (SOWPRA) and 1979-88 (SABPRA). These results are shown in Figure 3‎ .8.

Figure ‎3.8: Ecoregion 3W - Moving response function analyses (with a 30-year window) using significant climatic variables for red pine growth. At the top of each graph are site ID (left) and seasonal climatic variable (right). X-axis shows the last year of each 30-year window. Y-axis shows moving response coefficients. Black lines show standardized coefficients of moving response functions. Grey areas show 95% confidence interval. All sites show non-stable temporal dynamics of growth-climate relationships.

In ecoregion 4S growth – climate associations were not stable during the 20th century

(Figure 3‎ .9). In general, there were more periods of significant positive responses of growth to spring

Prec compared to responses with summer Prec; however, the latter lasted for longer periods. Likewise, the negative response to prior summer Tmax were neither stable spatially nor across the sites. However, these negative responses are fewer in recent decades. The negative responses to prior summer Tmin and winter Prec were more stable during the earlier to middle period of the 20th century than recent decades.

68

69

Figure ‎3.9: Ecoregion 4S - Moving response function analyses (with a 30-year window) using significant climatic variables for red pine growth. At the top of each graph are site ID (left) and seasonal climatic variable (right). X-axis shows the last year of each 30-year window. Y-axis shows moving response coefficients. Black lines show standardized coefficients of moving response functions. Grey areas show 95% confidence interval. All sites show non-stable temporal dynamics of growth-climate relationships.

In 5S the majority of the populations (8 out of 10) showed significant growth response to seasonal Prec. Six populations also had temporally infrequent significant response to seasonal Tmax.

Over the course of study period, infrequent positive growth response to summer Prec shifted to spring

Prec. For seven out of ten populations, spring Prec was a limiting factor for 30-year periods ending in last two decades. However, at most of the sites these responses were present for very short durations. Growth at the two most southern populations (CANA245 and CALPR) also benefitted from rising spring Tmin during last two decades (Figure 3‎ .10).

In 4W climate was not a limiting factor to red pine growth during the first half of the 20th century.

However, some growth – climate associations were revealed in temporal analyses that were not captured in spatial analyses. In recent decades Prec during all seasons (except winter) positively affected red pine growth. Red pine growth in the eastern and western parts of the ecoregion showed negative response to prior summer/summer Tmax during 30-year periods ending between 1989 and 2007. During the same period growth in the central part was negatively correlated to only summer Tmax. The once positive response to spring Tmin (periods ending between 1956 and 1975) in the central part was no more significant in recent decades. These results are shown in Figure 3‎ .11.

70

Figure ‎3.10: Ecoregion 5S - Moving response function analyses (with a 30-year window) using significant climatic variables for red pine growth. At the top of each graph are site ID (left) and seasonal climatic variable (right). X-axis shows the last year of each 30-year window. Y-axis shows moving response coefficients. Black lines show standardized coefficients of moving response functions. Grey areas show 95% confidence interval. All sites show non-stable temporal dynamics of growth-climate relationships.

71

Figure ‎3.11: Ecoregion 4W - Moving response function analyses (with a 30-year window) using significant climatic variables for red pine growth. At the top of each graph are site ID (left) and seasonal climatic variable (right). X-axis shows the last year of each 30-year window. Y-axis shows moving response coefficients. Black lines show standardized coefficients of moving response functions. Grey areas show 95% confidence interval. All sites show non-stable temporal dynamics of growth-climate relationships.

72

Results of correlation analyses between significant moving response functions and temporal variability of climate variables for various seasons during the study period reveal that all correlations were non-significant except for prior summer climate (r = 0.414, significant at ∝ 0.05). This shows that stability of red pine growth – climate relationships was influenced by climatic (Tmin, Tmax and Prec) variability of prior summer season during the 20th century.

3.4 Discussion

Climate during prior year and/or prior to growing season plays an important role in red pine radial growth. These results are in agreement with others who have examined growth – climate associations of various tree species. Huang et al. (2010) reported a consistent influence of prior summer temperature

(negative) and prior summer precipitation (positive) on radial growth of four boreal species (trembling aspen, paper birch, black spruce and jack pine) at the same latitude along the Ontario - Quebec border.

Hofgaard et al. (1999) also reported negative influence of prior summer temperature and positive influence of prior summer precipitation on black spruce and jack pine along in adjoining areas, in western

Quebec, at the same latitudes. Significance of prior summer climate in radial growth has also been reported for other North American tree species such as white spruce (Chhin et al. 2004), mountain hemlock (Gedalof and Smith 2001), lodgepole pine (Chhin et al. 2008), Douglas-fir (Griesbauer and

Green 2010b) and boreal species in Manitoba (Girardin and Tardif 2005). Most of these studies used mean monthly or mean seasonal temperatures in their growth – climate analyses. Studies also show that growth is mostly correlated with prior summer maximum temperature compared with minimum temperatures (Chhin et al. 2004, Huang et al. 2010). I also observed negative influence of prior summer maximum temperature at the majority of sites. This implies that prior summer maximum temperature is more important for red pine growth than minimum temperature and precipitation.

73

Growing season spring/summer temperature is not significant for red pine radial growth in the study area. These results are in contradiction to many other studies that find growing season summer temperature affecting radial growth for other tree species (Hofgaard et al. 1999, Gedalof and Smith 2001,

Girardin et al. 2005, Huang et al. 2010). Radial growth of red pine along its western margin in northern

Minnesota is also negatively correlated with growing season summer temperatures (Kipfmueller et al.

2010). Studies on other southern populations of red pine show a negative response to growing season summer temperature (Graumlich 1993) and positive response to spring temperature (Kilgore and

Telewski 2004). Comparison of my results with earlier studies on southern populations of red pine shows the variability of climatic control of southern and northern populations of red pine. My results do not confirm the general perception of increasing growth due to rising temperatures (Hanson et al. 1989).

Growth – climate associations of red pine identified in this study are based on statistical correlations, and thus should not be misinterpreted as cause and effect. However, some physiological processes affecting growth provide plausible biological explanations of the identified growth - climate associations in this study. The lag effect of prior summer/prior to growing season climate on radial growth in conifers has long been recognized (Fritts 1966). Climate-induced preconditions influence radial growth during growing season. It is postulated that high temperatures and low precipitation causes water stress through reduced soil moisture and increased evapotranspiration. This water stress reduces photosynthetic activity, lowers food reserves, reduces rate of cambial activity: and subsequently results in reduced growth (Fritts 1966). This pathway has been reported for shade-intolerant upper canopy coniferous species such as jack pine (Huang et al. 2010). Red pine is also an upper canopy species that usually grows on poor soils in northern Ontario. In these areas, high temperatures increase evapotranspiration and lower soil water reserves. The situation is further aggravated by lower precipitation, more prevalent in northwestern Ontario, leading to drought like conditions and potential water stress effect on radial growth. My results, namely the negative response to maximum temperature

74 and positive response to prior summer precipitation, confirm this water stress effect on red pine growth.

However, water stress cannot be considered as the only cause of reduced growth of red pine in the study area. At many sites, growth response (negative) to prior summer maximum temperature was significant, whereas growth response to prior summer precipitation was non-significant. This implies that, at these sites, water stress was not a limiting factor for red pine radial growth. Reduced growth due to high temperature without any apparent water stress has also been reported (Fritts 1966).

It is postulated that higher temperatures may lead to increased respiration resulting in lower food reserves for the following year. This pathway has been suggested for reduced radial growth in lodgepole pine (Chhin et al. 2008). However, the role of high temperature in disturbing photosynthetic respiratory balance in red pine has not been observed (Decker 1944). In fact, high temperatures result in reduced net primary productivity in red pine (Decker 1944). It is likely that reduced radial growth in red pine is due to the negative effect of this reduced productivity on apical activity (Larson 1960). Studies on resource allocation in trees also suggest a trade-off mechanism in which increased resource allocation to a certain process (such as cone production) is at the expense of resource allocation to other processes (Puritch

1977). In conifers, cone production increases with warmer conditions. Resultantly, more resources are allocated to these cones, in subsequent year, leaving less for radial growth. The lag effect of warmer climate through cone production pathway has been reported for mountain hemlock and subalpine fir

(Woodward et al. 1994, Gedalof and Smith 2001). A surge in cone and seed production in red pine has been documented to occur once every 3-7 years (Godman and Mattson 1976). Warmer temperature stimulates more cone production. In the subsequent year these cones utilize most of the reserve food resources (Dickmann and Kozlowski 1968, 1970, Wetzel and Burgess 1994) and hence affect radial growth.

My results show positive growth response to precipitation during all seasons. These results confirm findings in an experimental study that precipitation influences red pine daily 75 radial growth (Braekke et al. 1978). Precipitation effects soil moisture and the water balance of the trees (Fritts 1966), and water stress affects radial growth in red pine (Becker et al. 1987).

Growth at a few sites showed a negative response to winter precipitation. Negative influence of winter precipitation has also been observed in other species such as jack pine (Hofgaard et al. 1999, Huang et al.

2010) and lodgepole pine (Chhin et al. 2008). This may be due to termination of cambial activity as a result of increased winter snowfall (Vaganov et al. 1999). However, a probe in climatic data for these sites and the neighbouring sites did not reveal any significant differences. It is likely that delay in snow melt, due to some local conditions, cause low soil temperature for longer periods that result in late onset of the growing season (Evans and Fonda 1990). This result however warrants further investigation.

The results of this study are also in conformity with other studies showing spatial variability in growth – climate associations (Mäkinen et al. 2002, Miyamoto et al. 2010). Some studies attribute spatial variability of growth – climate associations to local environmental conditions (Peterson et al. 2002), whereas others consider such variability as a function of latitudinal gradient (Silva et al. 2010, Huang et al. 2010). I identified variability in growth – climate associations along longitudinal (east – west) gradient. Such observations are not common in the region. This is perhaps due to the fact that studies on growth – climate associations mainly focus on latitudinal gradient. Although major forest types are differentiated based on climatic variations along a latitudinal gradient (Rowe 1972), the within forest/species productivity (and response to climate) varies with local climate, which does not necessarily follow a latitudinal gradient (Hills 1959, 1969). This is typically true for species like red pine that are prone to moisture-related stress. Indeed, moisture gradient in the study area has more variation along longitude than latitude (Hills 1969), and warrants the need to investigate growth – climate associations along latitudinal gradients. Spatial variability in red pine growth – climate associations also suggests that local site conditions are important

76

Temporal analyses show that red pine growth – climate associations in northern Ontario are unstable and complex. Given the spatial variability (due to larger study area) and inherent temporal fluctuations in climate over a century period, the complexity of red pine growth – climate associations is not surprising (Brubaker 1986). These results conform findings of Briffa et al. (1998a, 1998b) who reported decreased sensitivity of growth to summer temperatures during second half of the 20th century.

Red pine forests are adapted to seasonal cycles and undergo dormancy due to colder temperatures in winter (Sims et al. 1990). Warmer winters however, favour growth by extending the growing season.

These results are in conformity of findings in other studies such as Andreu et al. (2007).

A shift in red pine growth response from summer to spring precipitation may be associated with a warmer spring that provides conditions conducive to optimal growth and hence spring precipitation becomes significant (Biondi 2000). This is typical for conifers that take advantage of warmth during spring due to their evergreen foliage (Graumlich 1993). Tardif et al. (2001b) also reported increased red pine growth due to warmer springs. Springs in northern Ontario experience thawing and freezing cycles.

These abrupt changes in temperature affect red pine growth as red pine is prone to frost damage (Woolsey and Chapman 1914, Sims et al. 1990). This is likely the explanation for a shifting response (positive and negative) to spring temperatures during different periods. For certain periods, growth responded to summer temperatures (negatively) and summer precipitation (positively). This combination suggests the role of moisture stress during those periods (Biondi 2000, Andreu et al. 2007). Positive responses to precipitation during recent decades suggest that increased precipitation offsets the negative effects of rising temperatures (drought) by the replenishment of soil moisture (Jump et al. 2007).

Temporally instable red pine growth – climate relationships also suggest the possible role of underlying climatic thresholds (Way and Oren 2010, Newman et al. 2011). Such effects have also been reported in other studies (Wilmking et al. 2004, Carrer et al. 2007, McLane et al. 2011). My results show that variability of prior summer climate influences the stability of red pine growth response to prior 77 summer climate. Such results have also been observed for some pine species in Iberian Peninsula (Andreu et al. 2007). However, due to overall non-significant correlations, I will argue that instability in climate is not the main contributing factor for instability in growth – climate relationships of red pine. Eco- physiological studies considering large scale environmental changes, such as changes in synoptic scale climate, atmospheric CO2, stratospheric O3 and anthropogenic nitrogen deposition, can perhaps provide some further insight on temporal instability of growth – climate relationships (Girardin and Tardif 2005,

Carrer et al. 2007).

3.5 Conclusions

For effective forest management in the midst of climate change, a better understanding of species specific responses to climate change across large regions is required. I addressed this question by examining spatio-temporal patterns of red pine growth – climate associations over large areas in northern Ontario,

Canada. This study shows that red pine populations exhibit multiple growth – climate associations, and an overall significant effect of precipitation (moisture). These associations not only varied across the geo- climatic ranges in the study area, but were also unstable over the course of past 100 years. It is generally assumed that growth – climate relationships change along latitudinal gradients due to changes in climate.

In the case of red pine, this does not hold true in northern Ontario. Although broader patterns of these relationships can be captured on an east – west gradient, better understanding can be achieved by investigating at ecoregion or sub-ecoregion levels. These results also support the notion that effective adaptation of forest management in the wake of climate change requires site specific solutions (Millar et al. 2007). These results have implications for growth models like PnET (Aber and Federer 1992) that overlook the role of precipitation. This study also (i) provides important information to develop strategies for managing complex forests where individual species climatic controls vary across space and time (ii) gives sufficient evidence that assuming spatially uniform response to climate is an over simplification, and (iii) suggests the need to improve existing modeling techniques to capture realistic patterns for 78 various management scenarios. In addition, temporal instability of red pine growth – climate relationships also suggest that reconstruction of past climatic conditions based on static growth – climate relationships could be biased.

79

CHAPTER 4 AGE DEPENDENT CLIMATE SENSITIVITY OF RADIAL GROWTH IN RED PINE AT WOLF LAKE OLD-GROWTH FOREST IN ONTARIO CANADA

Abstract

Many studies suggest that species growth – climate relationships are age dependent. Such information is lacking about multi-cohort red pine (Pinus resinosa Ait) forests. Here I investigate growth – climate relationships using tree-ring width (TRW) data of red pine from Wolf Lake Old-Growth Forest Reserve in northeastern Ontario. I developed TRW chronologies of red pine for two age classes: young (<100 years) and old (>100 years). I performed response function analyses using seasonal climate data (precipitation, minimum temperature and maximum temperature) for the period 1950 – 2010. My analyses show that difference in growth between young and old red pines is statistically non-significant, however the difference between their mean sensitivities is statistically significant. My results show that radial growth in young red pines was limited by minimum temperature during winter season, while radial growth in old red pines was limited by summer season, mainly July, precipitation. I further performed moving response function analyses to assess the stability of identified growth – climate relationships in each class. These analyses reveal that growth response of red pine in both age classes to their respective climatic control was not stable during the study period. Growth in young trees was limited by winter season minimum temperature during 1960 – 80s, while limiting effect of summer precipitation on old trees was significant during 1965 – 90s. Overall, these results suggest that growth – climate relationships of red pine at the study site are age dependent. My results have implications for predicting red pine old-growth forest dynamics to changing climate.

Key Words climate change, dendrochronology, growth – climate relationships, multi-cohort forest, northern Ontario, old-growth, red pine, response functions, Wolf Lake forest. 80

4.1 Introduction

Tree-ring width (TRW) data are commonly used to investigate species growth response to climate. These data are usually standardized to maximize climatic signal and remove any non-climatic signal. Further analyses, based on these standardized data, have an underlying assumption that age has no effect on growth – climate relationships. Although few studies have validated this assumption (Colenutt and

Luckman 1991, Fritts 2001), many studies suggest that growth – climate relationships are age dependent.

Szeicz and MacDonald (1994) studied growth response of Picea glauca to climate for two age-classes and observed that response of older trees (> 200 years) was different from younger trees (< 200 years).

Carrer and Urbinati (2004) studied growth – climate relationships in multi-cohort Larix decidua and

Pinus cembra in eastern Italian Alps ecotone under various age classes. They observed that growth response to climate varied with age. In Larix decidua growth sensitivity to significant climatic variables increased with age, whereas in Pinus cembra growth response did not vary much with age. Linderholm and Linderholm (2004) observed varying growth response in different age classes of Pinus sylvestris L. in

Scandinavian mountains. They reported that during the latter part of the 20th century (1967 – 96) growth response to July temperature was only significant (negative) in younger (< 100 years) Pinus sylvestris, whereas during the same period growth response of all older classes (>100 years) was positive to April temperature. Yu et al. (2008) reported significant differences in growth response to climate among various age classes of Quilian Juniper (Sabina przewalskii Kom.) in Qinghai – Tibetan Plateau. Rozas et al. (2009) observed a decreasing sensitivity to Jun – July precipitation with increasing age in Juniperus thurifera. Schuster and Oberhuber (2013) observed that, in a drought-prone mixed forest, older (> 121 years) Picea abies were sensitive to May-June precipitation, whereas young Picea abies were insensitive to any climatic factors. These studies suggest that climate sensitivity in many species is age dependent, and ignoring such age effects in growth – climate investigations for multi-cohort forests can lead to biased growth dynamics (Brienen et al. 2012).

81

In this study I performed in-depth analyses of age specific growth – climate relationships of a multi-cohort red pine (Pinus resinosa Ait) forest in Wolf Lake Forest Reserve (46.85 N, 80.64 E) in northern Ontario, Canada (Figure 4‎ .1). Wolf Lake Forest extends to an area of about 1600 ha and is the largest remaining red pine old-growth forest in North America (Iles 1990, Quinby 1996, Henry and

Quinby 2009, Anand et al. 2013). The very existence of this old-growth‎“Forest‎Reserve”‎is threatened due to existing mining tenure (OMNR 2003). A recent attempt to revoke its protective status

(Government of Ontario 2012) has‎raised‎concern‎for‎the‎protection‎of‎this‎‘living‎legacy”‎(Quinby et al.

2013). In fact, our scientific knowledge about this old-growth forest is very limited despite its high scientific value (Anand et al. 2013).

82

Figure ‎4.1: Map of Wolf Lake Forest. This forest contains the largest remaining old-growth red pine dominated forests in North America. Forest age is shown in two age classes: young (<100 years) and old (>100 years). Size of the tree symbols shows red pine composition (%) in various stands. Water bodies are shown in blue. Location of Wolf Lake Forest in northeastern Ontario is shown by red dot on Ontario map (top left). .

83

At Wolf Lake, red pine is a dominant tree species. In an earlier study on species recruitment dynamics at Wolf Lake Forest, Leithead et al. (2010) reported a competitive disadvantage of red pine over other species in large and old treefall gaps. This may have implications for red pine recruitment and overall dynamics of this forest under climate change. The very existence of Wolf Lake Forest in a transition zone between deciduous and boreal forests suggests that climate change may have pronounced effects (Goldblum and Rigg 2010, Anand et al. 2013). However, very little attention has been given to the potential impacts of changing climate on growth dynamics in this forest. Negative effects of climate on growth predispose a forest to disturbances that ultimately contribute to increased mortality (Chmura et al.

2011). If such is the case at Wolf Lake then it is likely that frequency of large treefall gaps will increase.

As a result increased establishment of species other than red pine (Leithead et al. 2010) will have profound effects on species dynamics in this red pine-dominated old-growth forest.

Silva et al. (2010) presented some results on growth response to climate for young and old red pines at Wolf Lake Forest in another context. They reported a negative correlation between growth in young red pines and mean annual precipitation (MAP); and an insignificant influence of temperature/precipitation on older red pine. The results of Silva et al. (2010) suggest that (i) climate sensitivity of young and old pine varies at Wolf Lake, and (ii) it is unlikely that large treefall gaps will increase under climate change. However, the generalizability of the results of Silva et al. (2010) is limited due to reasons such as using few growth samples (5 individual trees for each of the two age class), a lack of crossdating and, using annual climatic data (instead of monthly/seasonal data) for growth – climate analysis. This warrants further detailed investigation of age dependent growth – climate relationships of red pine at Wolf Lake Forest. Such investigations are pivotal to improve our understanding of red pines in this old-growth forest.

Here I use TRW data from Wolf Lake to test the following null hypotheses: (i) there is no difference in growth between young and old red pine trees, (ii) red pine growth response to climate is 84 independent of age classes, and (iii) the influence of each climatic factor on red pine growth is stable over time.

4.2 Materials and Methods

4.2.1 Field Sampling and Data Preparation Fifty nine red pine core samples were collected from Wolf Lake Forest during summer 2011 and 2013. In general, one core, at breast height, was collected from each sampled tree using increment borer. Cores were prepared for tree-ring width measurements. Each tree-ring was measured and subsequently assigned a calendar year (crossdating) using WinDendro (WinDENDRO 2012). Quality of crossdating was checked‎through‎Pearson’s‎correlations‎at‎99%‎confidence‎level‎as‎follows.‎First, each growth series was converted into an index series, a standardization process, using a 32-year cubic smoothing spline with

50% wavelength cutoff. A master dating series was then developed from the mean values of these index series.‎Pearson’s‎correlation‎coefficient‎of‎each‎index‎series‎was‎estimated‎with‎master‎dating‎series.‎For‎ each correlation, the respective index series was first removed from the master dating series. These analyses were performed using a program COFECHA (Grissino-Mayer 2001). This program flagged series that had lower correlations (critical threshold 0.3281) with the master dating series. For such series, any potential measurement/dating errors were corrected by re-inspecting their core samples. After error correction, their improved correlations were verified through correlation analysis in COFECHA.

4.2.2 Chronology Development A total of 59 TRW series, from 59 red pine trees, were crossdated and used to develop chronologies.

These TRW series were divided in two groups based on their age at breast height: young (less than 100 years) and old (more than 100 years). The rationale for selecting these age groups is the availability of data for this study; and these are by no means standard age classes of red pine. Generally, red pine over

80 years is classed as mature, however trees of age 100 – 120 years characteristically form old-growth

85 stands (Gilmore and Palik 2006, Quinby et al. 2013). Chronologies for old and young red pines were developed from 37 and 22 crossdated TRW series respectively. The study site is at a transition zone between temperate and boreal forests. Radial growth at such sites is likely to be influenced to a greater extent by non-climatic factors including stand dynamics (Conkey 1986). This warrants the need to apply a flexible detrending technique like cubic spline (Cook and Peters 1981). I thus fitted all individual TRW series with a 40-year cubic smoothing spline. Each series was then standardized by dividing TRW values with their respective fitted values. This process essentially creates indices that have a mean of 1.00 and a variance independent from the effects of tree age, sampling position on tree and mean growth of the tree

(Ferguson 1970). Index chronology of each age group was developed from the biweight robust mean of standardized series in that age group (Cook et al. 1990b). Each index chronology was prewhitened, using autoregressive modeling, to remove persistence. It is likely that prewhitening may remove some common climatic signal as well. To preserve this common climatic signal, pooled autoregression for each age group was added back to its prewhitened chronology. All these analyses were performed using a specialized program ARSTAN (Cook 1985). Summary statistics of these chronologies are shown in

Table 4‎ .1.

86

Table ‎4.1: Chronology statistics of young (< 100 years) and old (>100 years) red pines at Wolf Lake Forest in northern Ontario, Canada. Values in parenthesis are for the analysis period 1950 – 2010. Old Red Pine Young Red Pine Chronology Characteristics (> 100 Yearsa) (<100 Yearsa) Number of core samples / growth series 37 22 Number of tree-rings 5585 1410 Time Span 1767-2012 1915-2012 Common Period 1915-2012 1915-2012 Time Span with at least 10 growth series 1835-2012 1950-2012 Growth - Climate Analysis Period 1950 - 2010 1950 - 2010 Mean Index 0.993 (0.989b) 0.990 (0.993b) Standard Deviation 0.234 (0.175) 0.171 (0.168) Mean Sensitivity 0.179 (0.149) 0.140 (0.136) Serial Correlation 0.486 (0.465) 0.461 (0.494) Expressed Population Signal 0.90-0.94 0.85-0.86 a Age is considered at breast height. b Mean index value after standardization for the analysis period 1950 – 2010.

87

4.2.3 Growth – Climate Analysis I used response function analysis to statistically determine the growth-influencing climatic factors for young and old red pines at Wolf Lake Forest. Response function analysis is a multiple regression technique that uses principal components of predicting variables to estimate the index value of predictant variable. The principal components themselves are calculated from the eigenvectors of normalized data of predicting variables (Fritts et al. 1971). In this study, I performed response function analysis on young and old red pine chronologies using three predicting variables: total precipitation (Prec), mean minimum temperature (Tmin), and mean maximum temperature (Tmax). Interpolated monthly data for these climatic variables were obtained from Natural Resource Canada (McKenney et al. 2013). Monthly climate data were converted to seasonal data: prior summer (prior June – prior September), winter (prior

October – current March), spring (current April – current June) and summer (current July – current

September). Time series of these data are shown in Figure 4‎ .2. Response function analyses were performed for the period 1950 – 2010 using program DendroClim 2000 (Biondi and Waikul 2004).

Statistical significance of response coefficients was assessed using 1000 times bootstrap.

88

Prior Summer Summer Spring Winter

(°C)

Max T

(°C)

Min T

(mm)

Prec

Figure ‎4.2: Time series (1950 – 2010) of seasonal climate data used in this study for red pine growth - climate analysis. Climatic variables (top to bottom) are mean maximum temperature (Tmax), mean minimum temperature (Tmin) and total precipitation (Prec). Seasons (left to right) are prior summer (prior June – prior September), summer (July - September), spring (April - Jun) and winter (prior October - March). Dotted red lines are 10-year averaged trend lines.

89 4.3 Results

Descriptive statistics of old and young red pines chronologies are shown in Table 4‎ .1. Four statistics are mentioned to describe chronologies. These statistics are measures of (i) variation in growth ring index about the mean index value (standard deviation), (ii) relative year to year variation in growth ring index

(mean sensitivity),‎(iii)‎degree‎of‎dependence‎of‎a‎growth‎ring‎index‎upon‎the‎index‎of‎preceding‎year’s‎ growth (serial correlation or 1st order autocorrelation) and expressed population signal (EPS). Old red pine chronology has relatively higher values than the young for standard deviation (0.234 vs 0.171), mean sensitivity (0.179 vs 0.140) and serial correlation (0.486 vs 0.461). We used z-score‎from‎Fisher’s‎r-to-z transformation (test of the difference between two independent coefficients) to check the statistical significance of differences in standard deviations and serial correlations. Results show that differences in standard deviation (z = 0.541, p = 0.5887) and serial correlations (z = 0.264, p = 0.7918) between young and old tree chronologies are non-significant. However, t-test shows that difference in mean sensitivity is statistically significant (p = 0.0078, ∝ 0.05).

Figure 4‎ .3 shows time series of old and young red pines for (A) chronology index values, (B) year to year variation (sensitivity) in chronology index values, and (C) chronology sample depths. I correlated chronology index values for the overlapping period of old and young trees (1915 – 2012), and the correlation was highly significant (r = 0.75, p < 0.001) (Figure 4.3A). The correlation between sensitivity of both chronologies was relatively weak, but also significant (r = 0.40, p < 0.001; Figure

4.3B).

Results of response function analysis show that red pine growth – climate relationships at Wolf

Lake Forest are age dependent (Figure 4‎ .4). Overall, Prec is the most growth-influencing climatic variable followed by Tmax and Tmin. In both age classes, trends in growth responses to most of the seasonal climate are similar (either positive or negative) with the exception for Tmin during prior summer and summer. In the case of prior summer/summer Tmin, growth response in old red pines is negative and

90 growth response in young red pines is positive. However, these effects of climate on growth, in either age class, are non-significant. In the case of old red pines, positive growth response to summer precipitation is significant; but in the case of young red pines positive growth response to winter Tmin is significant.

These significant climate controls of growth vary in magnitude. The positive effect of summer Prec on growth in old red pines is greater compared to the positive effect of winter Tmin on growth in young red pines

91 2.5 (A)

2.0

1.5

1.0

Chronology Index Chronology 0.5

0.0

1.0 (B) 0.8 0.6 0.4

0.2 Mean Sensitivity Mean 0.0

40 (C) 30 20 10 0

Number of Series of Number 1760 1810 1860 1910 1960 2010 Year

Figure ‎4.3: Red pine chronologies for old (black) and young (red) red pines at Wolf Lake Forest, northern Ontario, Canada. (A) time series of each chronology index values (mean correlation 0.75, p < 0.001), (B) year to year variability (sensitivity) of index values in each chronology (mean correlation 0.40, p < 0.001), and (C) number of crossdated tree-ring width series used in each chronology over time.

92

0.8

0.6

0.4

0.2

0

-0.2 Response Coefficients Response

-0.4

-0.6 Prior Summer Spring Winter Prior Summer Spring Winter Prior Summer Spring Winter Summer Summer Summer Tmax Tmin Precipitation

Figure ‎4.4: Results of response function analyses for old (black) and young (grey) red pines at Wolf Lake Forest, northern Ontario, Canada. Analyses were performed using three climatic variables (Tmax - mean maximum temperature, Tmin - mean minimum temperature and Prec - total precipitation) for four seasons (prior summer, summer, spring and winter). Seasons are explained in section ‎4.2.3. Dotted red bars are the 1000 times bootstrap confidence intervals at 95 percentile. Growth-limiting factors for old and young red pines are summer precipitation and winter minimum temperature respectively.

93 The limiting effect of summer precipitation on old red pines was not significant throughout the analysis period (1950 – 2010) (Figure 4‎ .5). Positive growth response in old red pines to summer precipitation was significant during the period 1965 – 1990. In recent decades summer precipitation is no longer significant for growth in old red pines. Likewise, the limiting effect of winter Tmin on growth in young red pines was not stable throughout the analysis period. The positive growth response of young red pines to winter Tmin was significant during 1960 – 1980, and in recent periods this positive effect was not observed. These results also reveal that the positive effect of summer Prec on old red pines was stronger and for a longer period than the positive effect of winter Tmin on young red pines.

Further analyses show that in old red pines the significant influence of summer Prec was mainly due to July Prec during 1960s – 1990s (Figure 4‎ .6). This is the same time period when seasonal (summer)

Prec significantly influenced growth in old trees. However, in the case of young red pines, the effect of

Tmin in any individual month during winter was not significant (Figure 4‎ .7). Growth response in young red pines to Tmin was relatively higher for prior November, February and March. I performed moving response analyses for these three months. Results of these analyses (Figure 4‎ .7) reveal that, at the most, the effect of February Tmin was significant, but infrequent, during 1960 – 1980. These results suggest that the effect of Tmin on growth in young red pines was due to overall winter season rather than any individual month.

94 1

Old Red Pine Prec_ Summer

0.5

0

Response Coefficients Response -0.5

-1

1 Young Red Pines Tmin_Winter

0.5

0

-0.5 Response Coefficients Response

-1 1950 1960 1970 1980 1990 2000 2010

Last year of 30-year moving window Figure ‎4.5: Results of moving response function analysis of growth-limiting seasonal climatic variables for red pine at Wolf Lake Forest, northern Ontario, Canada. Analyses were performed using a 30-year window with a1-year moving interval. Gray areas are the 1000 times bootstrap confidence intervals at 95 percentile. Upper – response coefficients of old red pines growth – summer precipitation moving response analysis. Lower – response coefficients of young red pines growth – winter minimum temperature moving response analysis.

95

0.5

0.25

0

-0.25 Response Coefficients Response

-0.5 Jul Aug Sep

Precipitation 1

0.5

0

-0.5 Response Coefficients Response

-1 1950 1960 1970 1980 1990 2000 2010

Last year of 30-year moving window Figure ‎4.6: Results of response function analysis showing growth-limiting monthly climate for old red pines at Wolf Lake Forest, northern Ontario, Canada. Upper – response coefficients of old red pines growth – precipitation response analysis for summer months. Dotted red bars are the 1000 times bootstrap confidence intervals at 95 percentile. Lower – response coefficients of old red pines growth – July precipitation moving response analysis. Gray areas are the 1000 times bootstrap confidence intervals at 95 percentile.

96

0.5

0.25 0 -0.25 -0.5

OCT NOV DEC Jan Feb Mar Response Response Coefficients

Tmin 1 Tmin_Nov 0.5

0

-0.5

Response Response Coefficients -1

1 Tmin_Feb 0.5

0

-0.5 Response Response Coefficients -1

1 Tmin_Mar 0.5

0

-0.5

Response Response Coefficients -1 1950 1960 1970 1980 1990 2000 2010 Last year of 30-year moving window Figure ‎4.7: Results of response function analysis showing growth-limiting monthly climate for young red pines at Wolf Lake Forest, northern Ontario, Canada. Top – response coefficients of young red pines growth – minimum temperature response analysis for winter months. Dotted red bars are the 1000 times bootstrap confidence intervals at 95 percentile. Lowers – response coefficients of young red pines growth – minimum temperature moving response analysis for prior November, February and March. Gray areas are the 1000 times bootstrap confidence intervals at 95 percentile.

97 4.4 Discussion

In this study, I performed dendroclimatic analyses for the red pine population at Wolf Lake Forest. The focus of this research was to determine growth-influencing climatic factors in this red pine old-growth forest, and to see whether or not the growth-influencing climatic factors are different for old vs. young trees. I developed chronologies for old and young red pines from 59 TRW series and performed response function analysis to assess the effects of climate change during the latter half of the 20th century. The main study objectives were (i) to compare radial growth in old and young red pines at Wolf Lake Forest,

(ii) to identify climatic variables that significantly affect growth in old and young red pines at Wolf Lake

Forest, and (iii) to assess temporal sensitivity of old and young red pines to their respective growth- limiting climatic variables.

Although the mean sensitivity of developed chronologies for red pine at Wolf Lake Forest was relatively low, I argue that these chronologies are suitable for the purpose of this study because (i) eastern

North American species tend to exhibit lower mean sensitivity due to influence of non-climatic growth- limiting factors (DeWitt and Ames 1978), and (ii) mean sensitivity of these chronologies are comparable with that of other red pine chronologies in northern Ontario (Chapter‎ 3) and elsewhere in eastern North

America (DeWitt and Ames 1978, Kilgore and Telewski 2004). Standard deviation, serial correlation and

EPS of my chronologies are also comparable with red pine chronologies from other sites (Chapter‎ 3,

DeWitt and Ames 1978).

During the analysis period (1950 – 2010), young red pines exhibit higher serial correlation than old trees. Similar results have been observed for Quercus robur L. where young trees had higher serial correlation compared to mature and old trees (Rozas 2005). Rozas (2005) postulates that difference in serial correlation may be due to varying ratios of earlywood and latewood. Eilmann et al. (2011) observed shortening of growth (wood formation) period in pines under drought conditions. Such reduction in

98 growth period may lead to less wood formation. Drought also causes early initiation of latewood formation (late July vs late September) in red pine (Zahner et al. 1964).

My results show that old red pines at Wolf Lake Forest are sensitive to summer precipitation.

Decreasing precipitation along with rising temperatures at Wolf Lake Forest may create moisture deficit resulting in reduced growth. On the other hand, growth in young red pines at Wolf Lake Forest is affected by winter minimum temperature. Evergreen foliage in pines is helpful in earlier growth initiation during warmer winter and resultantly, the growing season is usually longer (Lebourgeois 2000, Fritts 2001). It is likely that larger proportions of earlywood, as a result of the longer growing season, create higher serial correlation in young red pines at Wolf Lake Forest.

I found that there is no difference in growth between young and old red pines at Wolf Lake

Forest. These results support the findings of Silva et al. (2010) who reported a consistency in basal area increment patterns between old and young red pines at Wolf Lake Forest. However, my analyses reveal different and contrasting growth – climate relationships for old and young red pines at Wolf Lake Forest.

Age dependent growth – climate relationships have also been observed for old-growth forests of other species (Rozas 2005). These results are in concordance with physiological studies that suggest that growth related functional processes are age dependent (Greenwood 1995, Bond 2000, Day et al. 2001,

Spicer and Gartner 2001). Radial growth, a complex physiological process, is essentially influenced by lag effects of growth itself integrated with the effects of various environmental factors (Fritts 2001).

Under the influence of these factors, growth response to climate is likely to vary among trees of different ages (Szeicz and MacDonald 1994). Growth in old red pines was positively related to summer precipitation and influence of July precipitation was significant. Growth in young red pines was related to overall winter minimum temperature. However, analyses based on monthly climate data do not reveal a significant influence of any individual winter month minimum temperature on young red pines. A positive influence of summer precipitation on red pine growth has been observed in other red pine 99 populations in northern Ontario, Canada (Section 3.3‎ ), northern Minnesota, USA (Kipfmueller et al.

2010) and upper Great Lakes Region, USA (Graumlich 1993), whereas literature on positive influence of warmer winters on red pine does not exist. The studies that reported positive influence of summer precipitations on red pine mainly collected growth data (chronology) from older (> 100 years) red pine trees. Hence, these studies support my findings for old red pines at Wolf Lake. Kilgore and Telewski

(2004) performed growth – climate analyses for the period 1931-98 using a chronology from about 100 year old red pine plantation trees in Michigan, USA, and found significant growth response to warmer temperature in April (late winter/early growing season). The difference between the two studies about the timing of significant growth response (winter season vs April) is likely due to the use of different temperature variables (minimum vs mean).

Silva et al. (2010) reported negative correlation between growth in young red pines and mean annual precipitation at Wolf Lake. They also reported that growth response in old red pine to temperature/precipitation is non-significant. I argue that a number of factors contribute to this disagreement between the results of Silva et al. (2010) and this study. One is that in this study I had a much greater sample depth, which is more representative of the red pine population at Wolf Lake Forest.

Silva et al. (2010) used annual mean temperature and precipitation data to determine red pine growth climate relationships at Wolf Lake Forest. Studies show that growth is mainly influenced by climate during the growing season (Fritts 1974). Averaging data from the non-growing season months skews climate data negatively (in case of temperature) or positively (in case of precipitation) and ultimately may affect the analyses output. This is particularly true for Wolf Lake Forest where a lower proportion of total annual precipitation is available during the growing season. Precipitation at Wolf Lake during the growing season is less than in the dormant season (winter). In addition, Wolf Lake Forest is characterized by bare rock outcrops and thin organic soil that has poor water holding capacity. As a result, a larger proportion of the precipitation falling during the dormant season is not actually retained for the growing

100 season. Cambial cell activity in pines is linked with water availability during the growth season and thus increased radial growth is expected during summer with high precipitation (Fritts 2001).

Characteristically, old red pines at Wolf Lake form the upper canopy. These are also exposed to more solar radiation and wind compared to young trees. During summers with low precipitation, such exposure affects their water budget through higher evapotranspiration; and ultimately less growth activity. On the other hand, young red pines are mostly confined to lower canopy and are thus not affected by water stress during summer months. This relative canopy position explains the differential response of old and young red pines at Wolf Lake.

Moving response function analyses indicate that growth response of each age group to its significant climatic variable varies through time. Studies on red pine and other species have shown that growth – climate relationships vary through time. Possible reasons could be changes in growth- influencing environmental factors (LeBlanc 1993), stand dynamics such as aging (Rozas 2005), competition intensity (Tardif et al. 2001a), and canopy position (van den Brakel and Visser 1996, Orwig and Abrams 1997). However, complete explanation of the results in this study is difficult due to our limited understanding of physiological processes in red pines (Wetzel and Burgess 1994). In young red pines, positive growth response to warmer winters was significant only during 1960 – 1978. After that this relationship was non-significant. This is likely due to the aging of young trees. The literature suggests that early initiation of physiological processes (including growth) due to warmer winters is not stable throughout the life span of trees. Trees benefit from warmer winter only during their early years and after

20-25 years the initiation of physiological processes in trees become more or less independent of rising temperature during late winter / early summer months (Plas and Kassen in van den Brakel and Visser

1996). This is a likely explanation of the results of significant growth response to winter minimum temperature during 1960 – 78, as the majority of sampled young trees were younger than 30 years during that period.

101

4.5 Conclusions

I studied age dependent growth climate relationships of red pine for two age classes at Wolf Lake Forest in northern Ontario, Canada. The results presented here have implications for predicting the response of red pine at Wolf Lake to climate change. First, climatic variations affect red pine growth at Wolf Lake even though these forests are well within distribution range of this species. Second, growth response to climate is differential: old red pines are influenced mainly by monthly climate (July precipitation), whereas young red pines are influenced by seasonal climate (winter minimum temperature). Third, growth response in each age class to its respective growth-influencing significant climatic factor is not stable over time. Influence of summer precipitation on old red pines was more stable than the influence of winter minimum temperature on young red pines. The overall influence of these climatic factors was significant during 1960s – 90s, but in recent decades the influence of these climatic factors is non – significant. The literature suggests that warmer winters provide conditions conducive for early growth initiation, and hence positively affect radial growth. The influence of warm winters on young red pines is no longer significant despite recent warming trend. The results presented here also show the inability of existing models, which use relatively simple static growth – climate relationships, to predict forest dynamics under changing climate.

102

CHAPTER 5 CONCLUSIONS AND RECOMMENDATIONS

5.1 Conclusions

Climate is one of many factors that affect radial growth of trees. Scientific investigations on radial growth response to climate provide insights that allow us to reconstruct climate, validate climate/growth models, document disturbance history, assess forest productivity, predict species range shift potential under climate change, evaluate forest management options, devise restoration strategies, and conserve endangered species/ecosystems.

In this thesis I examined growth - climate relationships of red pine in northern Ontario under changing climate during the 20th century. I studied these relationships at multiple spatio-temporal levels to get a better insight about the complexity of these relationships. Through this research I sought to inform forestry professionals who often rely on the assumption of stable climate (Colombo 2008). I also aim to provide data and scientific evidence to the modeling community for their model improvements and validation. For example, my results show that precipitation is a major growth controlling factor for red pine in northern Ontario species. However, precipitation data has rarely been used to validate red pine growth models. Growth models validated with precipitation can provide reliable growth projections under climate change.

There is growing literature on growth – climate relationships of leading edge populations of species. However, it was difficult to find any study that has used objective methods to select leading edge populations over larger areas. Knowing the fact that species climatic control may vary among locations, selecting and analyzing non-representative populations as leading edge populations may lead to biased results. Such results could be misleading if the study objective is to assess species range shift potential under climate change. I find that hierarchical cluster analysis (HCA) is helpful in identifying clusters of red pine populations based on common properties (i.e; climate in my case). As well as being an objective 103 method, HCA is an exploratory approach and can help to identify predictor variables for clusters based on the study objectives.

It is perhaps worth noting that working with red pine TRW data is relatively easy compared to other coniferous species. I hardly found any missing or absent ring in my growth series. Clear ring boundaries make crossdating easy. However, in older trees some very narrow growth rings with inconspicuous boundary make crossdating difficult.

My results show that precipitation is the most significant climatic control of growth at majority of sites. These results are in confirmation of other studies that suggest that species in eastern North America are more responsive to moisture-related factors than temperature (Tryon et al. 1957, Grissino-Mayer and

Butler 1993, Orwig and Abrams 1997, Pan et al. 1997). It is generally observed that climatic controls of species leading edge populations vary from non-leading edge populations. However, in the case of red pine, I found precipitation as a growth-influencing factor both in leading edge and non-leading edge populations. The combined effect of maximum temperature and precipitation on leading edge red pine suggests the influence of drought on the northern range limit of red pine. This is an important finding and no earlier study has reported a drought effect on red pine leading edge populations. This finding questions the validity of range shift predictions of red pine, and warrants the need to improve bioclimatic models for reliable projections.

Another interesting result from my research is that growth - climate relationships of red pine in northern Ontario do not follow latitudinal gradient. Instead, they closely follow the precipitation patterns in longitudinal direction. This result further confirms that many leading edge and non-leading edge populations have precipitation as a common climatic control.

Comparison of two age classes from an old-growth forest for growth response to climate is rare from an eastern North American perspective. I found different climatic controls for both age classes, and

104 these results contradicted findings of an earlier study by Silva et al. (2010) in the same study area. I argue that my results are more reliable as I performed detailed analysis using sufficient number of tree-ring width samples and monthly/seasonal climate data. These results are useful to understand growth dynamics in an old-growth forest. Similar studies from other old-growth forests of red pine can provide insights about their conservation/management potential.

Overall results in my thesis suggest that diversity in red pine growth response to climate across northern Ontario is due to local conditions, and as such local scale variations in growth – climate relationships should be incorporated in decision-making tools. What makes it difficult is the instability of growth – climate relationships over time.

5.2 Recommendations

My research shows that growth – climate relationships of red pine in northern Ontario vary at multiple spatial scales. These results can be helpful to achieve various objectives for adaptive forest management under climate change. The results from this research could be used to develop and improve existing growth/forest dynamics models.

My results show that growth – climate relationships of red pine in northern Ontario are not stable over time. Similar results have been reported by many other studies. Research suggests the possible role of‎“thresholds”‎in‎growth‎– climate relationships. Investigating these thresholds for various populations can further improve our‎understanding‎of‎forest‎dynamics.‎Additionally,‎incorporating‎such‎“thresholds”‎ in growth models will help in more reliable projections under climate change.

In this research I focused on red pine growth response to monthly and seasonal climatic variables.

Studies have suggested that extreme climate events may have a strong influence on radial growth, and their carry-over effects may last longer. With the recent changes in climate, we are experiencing more

105 extreme events than before. Such events can affect growth potential and productivity of species. I suggest that future studies examine extreme climatic events in determining growth response to climate.

106

Literature Cited

Aber, J. D., and C. A. Federer. 1992. A generalized, lumped-parameter model of photosynthesis, evapotranspiration and net primary production in temperate and boreal forest ecosystems. Oecologia 92:463–474.

Akaike, H. 1974. A new look at the statistical model identification. IEEE Transactionson Automatic Control 19:716–723.

Albatineh, A. N., and M. Niewiadomska-Bugaj.‎2011.‎MCS :‎a‎method‎for‎finding‎the‎number‎of‎ clusters. Journal of Classification 28:184–209.

Allen, A. P., J. H. Brown, and J. F. Gillooly. 2002. Global biodiversity, biochemical kinetics , and the energetic-equivalence rule. Science 297:1545–1548.

Anand, M., M. Leithead, L. C. R. Silva, C. Wagner, M. W. Ashiq, J. Cecile, I. Drobyshev, Y. Bergeron, A. Das, and C. Bulger. 2013. The scientific value of the largest remaining old- growth red pine forests in North America. Biodiversity and Conservation 22:1847–1861.

Andreu, L., E. Gutiérrez, M. Macias, M. Ribas, O. Bosch, and J. Camarero. 2007. Climate increases regional tree-growth variability in Iberian pine forests. Global Change Biology 13:804–815.

Araújo, M. B., and M. New. 2006. Ensemble forecasting of species distributions. Trends in Ecology and Evolution 22:42–47.

Arris, L. L., and P. S. Eagleson. 1989. Evidence of a physiological basis for the boreal-deciduous forest ecotone in North America. Vegetatio 82:55–58.

Asselin, H., S. Payette, M.-J. Fortin, and S. Vallee. 2003. The northern limit of Pinus banksiana Lamb. in Canada: explaining the difference between the eastern and western distributions. Journal of Biogeography 30:1709–1718.

Bailey, T. A., and R. Dubes. 1982. Cluster validity profiles. Pattern Recognition 15:61–83.

Baker, F. B., and L. J. Hubert. 1976. A graph-theoretic approach to goodness-of-fit in complete- link hierarchical clustering. Journal of the American Statistical Association 71:870–878.

Barber, V. A., G. P. Juday, and B. P. Finney. 2000. Reduced growth of Alaskan white spruce in the twentieth century from temperature-induced drought stress. Nature 405:668–673.

107

Beck, P. S. A., G. P. Juday, C. Alix, V. A. Barber, S. E. Winslow, E. E. Sousa, P. Heiser, J. D. Herriges, and S. J. Goetz. 2011. Changes in forest productivity across Alaska consistent with biome shift. Ecology Letters 14:373–379.

Becker, C. A., G. D. Mroz, and L. G. Fuller. 1987. The effects of plant moisture stress on red pine (Pinus resinosa) seedling growth and establishment. Canadian Journal of Forest Research 17:813–820.

Benavides, R., S. G. Rabasa, E. Granda, A. Escudero, J. A. Hódar, J. Martinez-Vilalta, A. M. Rincón, R. Zamora, and F. Valladares. 2013. Direct and indirect effects of climate on demography and early growth of Pinus sylvestris at the rear edge: changing roles of biotic and abiotic factors. PLoS ONE 8:e59824.

Ben-Hur, A., A. Elisseeff, and I. Guyon. 2002. A stability based method for discovering structure in clustered data. Pacific Symposium on Biocomputing 7:6–17.

Bergeron, Y., and J. Brisson. 1990. Fire regime in red pine stands at the northern limit of the species’‎range.‎Ecology‎71:1352–1364.

Bergeron, Y., and J. Brisson. 1994. Effect of climatic fluctuations on post-fire regeneration of two jack pine and red pine populations during the twebtieth century. Geographie Physique et Quaternaire 48:145–149.

Bergeron, Y., and D. Gagon. 1987. Age structure of red pine (Pinus resinosa Ait.) at its northern limit in Quebec. Canadian Journal of Forest Research 17:129–137.

Bhuta, A. A. R., L. M. Kennedy, and N. Pederson. 2009. Climate-radial growth relationships of northern latitudinal range margin longleaf pine (Pinus palustris P. Mill.) in the Atlantic Coastal Plain of Southeastern Virginia. Tree-Ring Research 65:105–115.

Biondi, F. 2000. Are climate - tree growth relationships changing in North - Central Idaho, U.S.A.? Arctic, Antarctic, and Alpine Research 32:111–116.

Biondi, F., and K. Waikul. 2004. DENDROCLIM2002: A C++ program for statistical calibration of climate signals in tree-ring chronologies. Computers and Geosciences 30:303–311.

Birks, H. J. B., H. A. Austin, N. E. Indrevaer, S. M. Peglar, and C. Rygh. 1998. An annotated bibliography of canonical correspondence analysis and related constrained ordination methods 1986 – 1996. http://adn.biol.umontreal.ca/~numericalecology/cca_bib/index.html.

Björck, S. 1985. Deglaciation chronology and revegetation in northwestern Ontario. Canadian Journal of Earth Sciences 22:850–871.

108

Black, T. A., W. J. Chen, A. G. Barr, M. A. Arain, Z. Chen, Z. Nesic, E. H. Hogg, H. H. Neumann, and P. C. Yang. 2000. Increased carbon sequestration by a boreal deciduous forest in years with a warm spring. Geophysical Research Letters 27:1271–1274.

Blasing, T. J., A. M. Solomon, and D. N. Duvick. 1984. Response functions revisited. Tree-Ring Bulletin 44:1–15.

Bogino, S., M. J. F. Nieto, and F. Bravo. 2009. Climate effect on radial growth of Pinus sylvestris at its southern and western distribution limits. Silva Fennica 43:609–623.

Bonan, G. B., and H. H. Shugart. 1989. Environmental factors and ecological processes in boreal forests. Annual Review of Ecology and Systematics 20:1–28.

Bonan, G. B., H. H. Shugart, and D. L. Urban. 1990. The sensitivity of some high latitude boreal forests to climatic parameters. Climatic Change 16:9–29.

Bond, B. J. 2000. Age-related changes in photosynthesis of woody plants. Trends in Plant Science 5:349–353.

Bonsal, B. R., X. Zhang, L. A. Vincent, and W. D. Hogg. 2001. Characteristics of daily and extreme temperatures over Canada. Journal of Climate 14:1959–1976.

Ter Braak, C. J. F. 1986. Canonical correspondence analysis: a new eigenvector technique for multivariate direct gradient analysis. Ecology 67:1167–1179.

Ter Braak, C. J. F., and P. F. M. Verdonschot. 1995. Canonical correspondence analysis and related multivariate methods in aquatic ecology. Aquatic Sciences 57:255–289.

Braekke, F. H., T. T. Kozlowski, and T. Skröpra. 1978. Effects of environmental factors on estimated daily radial growth of Pinus resinosa and Betula papyrifera. Plant and Soil 49:491–504.

Van den Brakel, J. A., and H. Visser. 1996. The influence of environmental conditions on tree- ring series of norway spruce for different canopy and vitality classes. Forest Science 42:206–219.

Brienen, R. J. W., E. Gloor, and P. A. Zuidema. 2012. Detecting evidence for CO2 fertilization from tree ring studies: the potential role of sampling biases. Global Biogeochemical Cycles 26:GB1025(1–13).

Briffa, K., and E. R. Cook. 1990. Methods of response function analysis. Pages 240–247 in E. R. Cook and L. A. Kairiukstis, editors. Methods of Dendrochronology: Applications in the Environmental Sciences. Kluwer Academic Publishers, Dordrecht, The Netherlands.

109

Briffa, K. R., F. H. Schweingruber, P. D. Jones, T. J. Osborn, I. C. Harris, S. G. Shiyatov, E. A. Vaganov, and H. Grudd. 1998a. Trees tell of past climates: but are they speaking less clearly today? Philosophical Transcations of the Royal Society of London. Series B, Biological Sciences 353:65–73.

Briffa, K. R., F. H. Schweingruber, P. D. Jones, T. J. Osborn, S. G. Shiyatov, and E. A. Vaganov. 1998b. Reduced sensitivity of recent tree-growth to temperature at high northern latitudes. Nature 391:678–682.

Brown, J. H., G. C. Stevens, and D. M. Kaufman. 1996. The geographic range: size, shape, boundaries, and internal structure. Annual Review of Ecology and Systematics 27:597–623.

Brubaker, L. B. 1986. Responses of tree populations to climatic change. Vegetatio 67:119–130.

Bunn, A. G., L. J. Graumlich, and D. L. Urban. 2005. Trends in twentieth-century tree growth at high elevations in the Sierra Nevada and White Mountains, USA. The Holocene 15:481– 488.

Burger, D. 1993. Revised site regions of Ontario: concepts, methodology and utility. Forest Research‎Report‎No.‎129.‎Queen’s‎Printer‎for‎Ontario,‎Sault‎Ste‎Marie,‎ON.

Calinski, T., and J. Harabasz. 1974. A dendrite method for cluster analysis. Communications in Statistics 3:1–27.

Campelo, F., C. Nabais, H. Freitas, and E. Gutiérrez. 2006. Climatic significance of tree-ring width and intra-annual density fluctuations in Pinus pinea from a dry Mediterranean area in Portugal. Annals of Forest Science 64:229–238.

Carleton, T. J., P. F. Maycock, R. Arnup, and A. M. Gordon. 1996. In situ regeneration of Pinus strobus and P. resinosa in the Great Lakes forest communities of Canada. Journal of Vegetation Science 7:431–444.

Carrer, M., P. Nola, J. L. Eduard, R. Motta, and C. Urbinati. 2007. Regional variability of climate-growth relationships in Pinus cembra high elevation forests in the Alps. Journal of Ecology 95:1072–1083.

Carrer, M., and C. Urbinati. 2004. Age-dependent tree ring growth responses to climate in Larix Decidua and Pinus Cembra. Ecology 85:730–740.

Chavardès, R. D., L. D. Daniels, P. O. Waeber, J. L. Innes, and C. R. Nitschke. 2013. Unstable climate-growth relations for white spruce in southwest Yukon, Canada. Climatic Change 116:593–611.

110

Chen, I., J. K. Hill, R. Ohlemüller, D. B. Roy, and C. D. Thomas. 2011. Rapid range shifts of species associated with high levels of climate warming. Science 333:1024–1026.

Chhin, S., E. H. (Ted) Hogg, V. J. Lieffers, and S. Huang. 2008. Potential effects of climate change on the growth of lodgepole pine across diameter size classes and ecological regions. Forest Ecology and Management 256:1692–1703.

Chhin, S., G. G. Wang, and J. Tardif. 2004. Dendroclimatic analysis of white spruce at its southern limit of distribution in the Spruce Woods Provincial Park, Manitoba, Canada. Tree-Ring Research 60:31–43.

Chiotti, Q., and B. Lavender. 2008. Ontario. Pages 227–274 in D. S. Lemmen, F. J. Warren, J. Lacroix, and E. Bush, editors. From Impacts to Adaptation: Canada in a Changing Climate 2007. Natural Resources Canada, Government of Canada, , ON.

Chmura, D. J., P. D. Anderson, G. T. Howe, C. A. Harrington, J. E. Halofsky, D. L. Peterson, D. C. Shaw, and J. B. St. Clair. 2011. Forest responses to climate change in the northwestern United States: ecophysiological foundations for adaptive management. Forest Ecology and Management 261:1121–1142.

Ciais, P., M. Reichstein, N. Viovy, A. Granier, J. Ogée, V. Allard, M. Aubinet, N. Buchmann, C. Bernhofer, A. Carrara, F. Chevallier, N. De Noblet, A. D. Friend, P. Friedlingstein, T. Grünwald, B. Heinesch, P. Keronen, A. Knohl, G. Krinner, D. Loustau, G. Manca, G. Matteucci, F. Miglietta, J. M. Ourcival, D. Papale, K. Pilegaard, S. Rambal, G. Seufert, J. F. Soussana, M. J. Sanz, E. D. Schulze, T. Vesala, and R. Valentini. 2005. Europe-wide reduction in primary productivity caused by the heat and drought in 2003. Nature 437:529– 533.

Clark, T. P., and A. H. Perera. 1995. An overview of ecology of red and white pine old-growth forests in Ontario. Forest Fragmentation and Biodiversity Project Report No. 18. Sault Ste Marie, ON.

Colenutt, M. E., and B. H. Luckman. 1991. Dendrochronological investigation of Larix lyallii at Larch Valley, Alberta. Canadian Journal of Forest Research 21:1222–1233.

Colombo,‎S.‎J.‎2008.‎Ontario’s‎forests and forestry in a changing climate. Climate Change Research Report CCRR-12. Peterborough, ON.

Colwell, R. K., G. Brehm, C. L. Cardelús, A. C. Gilman, and J. T. Longino. 2008. Global warming, elevational range shifts, and lowland biotic attrition in the wet tropics. Science 322:258–261.

Conkey, L. E. 1986. Red spruce tree-ring widths and densities in eastern North America as indicators of past climate. Quaternary Research 26:232–243. 111

Cook, E. 1990. A conceptual linear aggregate model for tree rings. Pages 98–104 in E. R. Cook and L. A. Kairiukstis, editors. Methods of Dendrochronology: Applications in the Environmental Sciences. Kluwer Academic Publishers, Dordrecht, The Netherlands.

Cook, E. R. 1985. A time series analysis approach to tree ring standardisation. PhD Thesis. The University of Arizona, Tucson, AZ.

Cook, E. R., K. Briffa, S. Shiyatov, and V. Mazepa. 1990a. Tree ring standardization and growth ring estimation. Pages 104–123 in E. R. Cook and L. A. Kairiukstis, editors. Methods of Dendrochronology: Applications in the Environmental Sciences. Kluwer Academic Publishers, Dordrecht, The Netherlands.

Cook,‎E.‎R.,‎B.‎M.‎Buckley,‎R.‎D.‎D’Arrigo,‎and‎M.‎J.‎Peterson.‎2000.‎Warm-season temperatures since 1600 BC reconstructed from Tasmanian tree rings and their relationship to large-scale sea surface temperature anomalies. Climate Dynamics 16:79–91.

Cook, E. R., and J. Cole. 1991. On predicting the response of forests in eastern North America to future climatic change. Climatic Change 19:271–282.

Cook, E. R., and K. Peters. 1981. The smoothing spline: a new approach to standardizing forest interior tree ring width series for dendroclimatic studies. Tree-Ring Bulletin 41:45–53.

Cook, E., S. Shiyatov, and V. Mazepa. 1990b. Estimation of the mean chronology. Pages 123– 132 in E. R. Cook and L. A. Kairiukstis, editors. Methods of Dendrochronology: Applications in the Environmental Sciences. Kluwer Academic Publishers, Dordrecht, The Netherlands.

Crins, W. J., P. A. Gray, P. W. C. Uhlig, and M. C. Wester. 2009. The ecosystems of Ontario, Part 1: Ecozones and Ecoregions. Inventory, Monitoring and Assessment, SIB TER IMA TR-01.‎Queen’s‎Printer‎for‎Ontario,‎Peterborough,‎ON.

Crins,‎W.‎J.,‎and‎P.‎W.‎C.‎Uhlig.‎2000.‎Ecoregions‎of‎Ontario:‎modifications‎to‎Angus‎Hills’‎site regions‎and‎site‎districts:‎revisions‎and‎rationale.‎Queen’s‎Printer‎for‎Ontario,‎Peterborough,‎ ON.

Cunningham, S. C., and J. Read. 2002. Comparison of temperate and tropical rainforest tree species: photosynthetic responses to growth temperature. Oecologia 133:112–119.

D’Arrigo,‎R.‎D.,‎R.‎K.‎Kaufmann,‎N.‎Davi,‎G.‎C.‎Jacoby,‎C.‎Laskowski,‎R.‎B.‎Myneni,‎and‎P.‎ Cherubini. 2004. Thresholds for warming - induced growth decline at elevational tree line in the Yukon Territory, Canada. Global Biogeochemical Cycles 18:GB3021.

112

D’Arrigo,‎R.,‎R.‎Wilson,‎B.‎Liepert,‎and‎P.‎Cherubini.‎2008.‎On‎the‎“Divergence‎Problem”‎in‎ Northern Forests: a review of the tree-ring evidence and possible causes. Global and Planetary Change 60:289–305.

Davis, A. J., L. S. Jenkinson, J. H. Lawton, B. Shorrocks, and S. Wood. 1998. Making mistakes when predicting shifts in species range in response to global warming. Nature 391:783–786.

Day, M. E., M. S. Greenwood, and A. S. White. 2001. Age-related changes in foliar morphology and physiology in red spruce and their influence on declining photosynthetic rates and productivity with tree age. Tree Physiology 21:1195–1204.

Decker, J. P. 1944. Effect of temperature on photosynthesis and respiration in red and loblolly pines. Plant Physiology 19:679–688.

DeSoto, L., J. J. Camarero, J. M. Olano, and V. Rozas. 2012. Geographically structured and temporally unstable growth responses of Juniperus thurifera to recent climate variability in the Iberian Peninsula. European Journal of Forest Research 131:905–917.

Devall, M. S., J. M. Grender, and J. Koretz. 1991. Dendroecological analysis of a longleaf pine Pinus palustris forest in Mississippi. Vegetatio 93:1–8.

DeWitt, E., and M. Ames, editors. 1978. Tree-ring chronologies of eastern North America. Chronology Series IV, Vol. 1. Tucson, AZ.

Dickmann, D. I., and T. T. Kozlowski. 1968. Mobilization by Pinus resinosa cones and shoots of C14-photosynthate from needles of different ages. American Journal of Botany 55:900–906.

Dickmann, D. I., and T. T. Kozlowski. 1970. Mobilization and incorporation of photoassimilated 14C by growing vegetative and reproductive tissues of adult Pinus resinosa Ait . trees. Plant Physiology 45:284–288.

Doughty, C. E., and M. L. Goulden. 2008. Are tropical forests near a high temperature threshold? Journal of Geophysical Research 113:G00B07.

Drever, C. R., C. Messier, Y. Bergeron, and F. Doyon. 2006. Fire and canopy species composition in the Great Lakes-St. Lawrence forest of Témiscamingue, Québec. Forest Ecology and Management 231:27–37.

Driscoll,‎W.‎W.,‎G.‎C.‎Wiles,‎R.‎D.‎D’Arrigo,‎and‎M.‎Wilmking.‎2005.‎Divergent‎tree‎growth‎ response to recent climatic warming, Lake Clark National Park and Preserve, Alaska. Geophysical Research Letters 32:L20703.

Dubes, R., and A. K. Jain. 1979. Validity studies in clustering methodologies. Pattern Recognition 11:235–254. 113

Dudoit, S., and J. Fridlyand. 2002. A prediction-based resampling method for estimating the number of clusters in a dataset. Genome Biologyiology 3:research0036.1–0036.21.

Eilmann, B., R. Zweifel, N. Buchmann, E. G. Pannatier, and A. Rigling. 2011. Drought alters timing, quantity, and quality of wood formation in scots pine. Journal of Experimental Botany 62:2763–2771.

Ettinger,‎A.‎K.,‎and‎J.‎HilleRisLambers.‎2013.‎Climate‎isn’t‎everything: competitive interactions and variation by life stage will also affect range shifts in a warming world. American Journal of Botany 100:1344–55.

Evans, R. D., and R. W. Fonda. 1990. The innuence of snow on subalpine meadow community pattern, North Cascades, Washington. Canadian Journal of Botany 68:212–220.

Fang, J., and M. J. Lechowicz. 2006. Climatic limits for the present distribution of beech (Fagus L.) species in the world. Journal of Biogeography 33:1804–1819.

Ferguson, C. W. 1970. Concepts and techniques of dendrochronology. Pages 183–200 in R. Berger, editor. Scientific Methods in Medieval Archaeology. University of Calafornia Press, Berkeley, CA.

Flannigan, M. D., and Y. Bergeron. 1998. Possible role of disturbance in shaping the northern distribution of Pinus resinosa. Journal of Vegetation Science 9:477–482.

Flannigan, M. D., and J. B. Harrington. 1988. A study of the relation of meteorological variables to monthly provincial area burned by wildfire in Canada (1953-80). Journal of Applied Meteorology 27:441–452.

Flannigan, M. D., and F. I. Woodward. 1993. A laboratory study of the effect of temperature on red pine seed germination. Forest Ecology and Management 62:145–156.

Flannigan, M. D., and F. I. Woodward. 1994. Red pine abundance: current climatic control and responses to future warming. Canadian Journal of Forest Research 24:1166–1175.

Foster, T. E., and J. R. Brooks. 2001. Long-term trends in growth of Pinus palustris and Pinus elliottii along a hydrological gradient in central Florida. Canadian Journal of Forest Research 31:1661–1670.

Fränti, P., M. Rezaei, and Q. Zhao. 2014. Centroid index: cluster level similarity measure. Pattern Recognition 47:3034–3045.

FRI. 2009. Forest resource inventories – planning composite inventory or base model inventory. Land Information Ontario.

114

Fritts, H. C. 1966. Growth-rings of trees: their correlation with climate. Science 154:973–979.

Fritts, H. C. 1974. Relationships of ring widths in arid-site conifers to variations in monthly temperature and precipitation. Ecological Monographs 44:411–440.

Fritts, H. C. 2001. Tree rings and climate. Academic Press Inc, New York, NY.

Fritts, H. C., T. J. Blasings, B. P. Hayden, and J. E. Kutzbach. 1971. Multivariate techniques for specifying tree-growth and climate relationships and for reconstructing anomalies in paleoclimate. Journal of Applied Meteorology 10:845–864.

Fritts, H. C., and D. J. Shatz. 1975. Selecting and characterizing tree ring chronologies for dendroclimatic analysis. Tree-Ring Bulletin 35:31–40.

Fritts, H. C., D. G. Smith, J. W. Cardis, and C. A. Budelsky. 1965. Tree-ring characteristics along a vegetation gradient in Northern Arizona. Ecology 46:394–401.

Gabriel, A. O., and R. D. Kreutzwiser. 1993. Drought Hazard in Ontario: a review of impacts, 1960–1989, and management implications. Canadian Water Resources Journal 18:117–132.

Gao, Q., P. Zhao, X. Zeng, X. Cai, and W. Shen. 2002. A model of stomatal conductance to quantify the relationship between leaf transpiration , microclimate and soil water stress. Plant, Cell and Environment 25:1373–1381.

García Gonzálea, I., E. A. Díaz Vizcaíno, and A. Martínez Cortizas. 1999. Dendrochronological analysis of oak (Quercus robur L., Fagaceae) in the Serra da Carba (Galicia, NW Spain): an application of cluster analysis. Nova Acta Cientifica Compostelana (Bioloxia) 9:171–177.

García-González, I. 2008. Comparison of different distance measures for cluster analysis of tree- ring series. Tree-Ring Research 64:27–37.

Gavin, D. G., and F. S. Hu. 2006. Spatial variation of climatic and non-climatic controls on species distribution: the range limit of Tsuga heterophylla. Journal of Biogeography 33:1384–1396.

Gedalof, Z., and D. J. Smith. 2001. Dendroclimatic response of mountain hemlock (Tsuga mertensiana) in Pacific North America. Canadian Journal of Forest Research 31:322–332.

St. George, S., D. M. Meko, and M. N. Evans. 2008. Regional tree growth and inferred summer climate in the Winnipeg River basin, Canada, since AD 1783. Quaternary Research 70:158– 172.

115

St. George, S., D. M. Meko, M.-P. Girardin, G. M. MacDonald, E. Nielsen, G. T. Pederson, D. J. Sauchyn, J. C. Tardif, and E. Watson. 2009. The tree-ring record of drought on the Canadian Prairies. Journal of Climate 22:689–710.

Gilmore, D. W., and B. J. Palik, editors. 2006. A revised managers handbook for red pine in the North Central Region. General Technical Report NC-264. Saint Paul, MN.

Girardin, M. P., E. Berglund, J. C. Tardif, and K. Monson. 2005. Radial growth of tamarack (Larix laricina) in the Churchill Area, Manitoba, Canada, in relation to climate and larch sawfly (Pristiphora erichsonii) herbivory. Arctic, Antarctic, and Alpine Research 37:206– 217.

Girardin, M. P., F. Raulier, P. Y. Bernier, and J. C. Tardif. 2008. Response of tree growth to a changing climate in boreal central Canada: a comparison of empirical, process-based, and hybrid modelling approaches. Ecological Modelling 213:209–228.

Girardin, M. P., and B. M. Wotton. 2009. Summer moisture and wildfire risks across Canada. Journal of Applied Meteorology and Climatology 48:517–533.

Girardin, M., J. Tardif, and Y. Bergeron. 2001. Radial growth analysis of Larix laricina from the Lake Duparquet area, Québec, in relation to climate and larch sawfly outbreaks. Écoscience 8:127–138.

Girardin, M.-P., and J. Tardif. 2005. Sensitivity of tree growth to the atmospheric vertical profile in the Boreal Plains of Manitoba, Canada. Canadian Journal of Forest Research 35:48–64.

Girardin, M.-P., J. C. Tardif, M. D. Flannigan, and Y. Bergeron. 2006. Synoptic-scale atmospheric circulation and boreal Canada summer drought variability of the past three centuries. Journal of Climate 19:1922–1947.

Girvetz, E. H., C. Zganjar, G. T. Raber, E. P. Maurer, P. Kareiva, and J. J. Lawler. 2009. Applied climate-change analysis: the climate Wizard tool. PLoS ONE 4:e8320.

Godman, R. M., and G. A. Mattson. 1976. Seed crops and regeneration problems of 19 species in northeastern Wisconsin. Research Paper NC 123. Saint Paul, MN.

Goldblum, D., and L. S. Rigg. 2005. Tree growth response to climate change at the deciduous- boreal forest ecotone , Ontario , Canada. Canadian Journal of Forest Research 35:2709– 2718.

Goldblum, D., and L. S. Rigg. 2010. The deciduous forest - boreal forest ecotone. Geography Compass 4:701–717.

116

Government of Ontario. 2012. Amendments to remove and re-designate parts of the Wolf Lake Forest Reserve (F175) Crown land use designation and replace those parts with additions to Chiniguchi Waterway provincial park, general use area G2048 and enhanced management area E183r. Queen’s‎Printer‎for‎Ontario.

Graumlich, L. J. 1989. The utility of long-term records of tree growth for improving forest stand simulation models. Pages 39–49 in G. P. Malanson, editor. Natural Areas Facing Climate Change. SPB Academic Publishing, The Hague, The Netherlands.

Graumlich, L. J. 1993. Response of tree growth to climatic variation in the mixed conifer and deciduous forests of the upper Great Lakes region. Canadian Journal of Forest Research 23:133–143.

Greenwood, M. S. 1995. Juvenility and maturation in conifers: current concepts. Tree Physiology 15:433–438.

Griesbauer, H. P., and D. S. Green. 2010a. Assessing the climatic sensitivity of Douglas-fir at its northern range margins in British Columbia, Canada. Trees 24:375–389.

Griesbauer, H. P., and D. S. Green. 2010b. Regional and ecological patterns in interior Douglas- fir climate - growth relationships in British Columbia, Canada. Canadian Journal of Forest Research 40:308–321.

Griesbauer, H. P., and D. S. Green. 2012. Geographic and temporal patterns in white spruce climate-growth relationships in Yukon, Canada. Forest Ecology and Management 267:215– 227.

Grissino-Mayer, H. D. 2001. Evaluating crossdating accuracy: a manual and tutorial for the computer program COFECHA. Tree-Ring Research 57:205–221.

Grissino-Mayer, H. D., and D. R. Butler. 1993. Effects of climate on growth of shortleaf pine (Pinus echinata Mill.) in northern Georgia: a dendroclimatic study. Southeastern Geographer 33:65–81.

Guiot, J. 1990. Methods of calibration. Pages 165–177 in E. R. Cook and L. A. Kairiukstis, editors. Methods of Dendrochronology: Applications in the Environmental Sciences. Kluwer Academic Publishers, Dordrecht, The Netherlands.

Halkidi, M., Y. Batistakis, and M. Vazirgiannis. 2001. On clustering validation techniques. Journal of Intelligent Information Systems 17:107–145.

Hall, M., M. Räntfors, M. Slaney, S. Linder, and G. Wallin. 2009. Carbon dioxide exchange of buds and developing shoots of boreal Norway spruce exposed to elevated or ambient CO2 concentration and temperature in whole-tree chambers. Tree Physiology 29:467–481. 117

Hanson, J. S., G. P. Malanson, and M. P. Armstrong. 1989. Spatial constraints on the response of forest communities to climate change. Pages 1–23 in G. P. Malanson, editor. Natural Areas Facing Climate Change. SPB Academic Publishing, The Hague, The Netherlands.

Hartigan, J. A. 1975. Clustering Algorithms. John Wiley and Sons Inc, New York, NY.

Henderson, J. P. 2006. Dendroclimatological analysis and fire history of longleaf pine (Pinus palustris Mill.) in the Atlantic and Gulf costal plain. The University of Tennessee, Knoxville.

Henderson, J. P., and H. D. Grissino-Mayer. 2009. Climate-tree growth relationships of longleaf pine (Pinus palustris Mill.) in the Southeastern Coastal Plain, USA. Dendrochronologia 27:31–43.

Henry,‎M.,‎and‎P.‎Quinby.‎2009.‎Ontario’s‎old‎growth‎forests.‎Fitzhenry‎and‎Whiteside‎Limited,‎ Markham, ON.

Hepsen, A., and M. Vatansever. 2012. Using hierarchical clustering algorithms for Turkish residential market. International Journal of Economics and Finance 4:138–150.

Hill, R. S., J. Read, and J. R. Busby. 1988. The temperature-dependence of photosynthesis of some Australian temperate rainforest trees and its biogeographical significance. Journal of Biogeography 15:431–449.

Hills, G. A. 1959. A ready reference to the description of the land of Ontario and its productivity. , ON.

Hills, G. A. 1969. The physiographic approach to the classification of terrestrial ecosystems with respect to representative biological communities. Toronto, ON.

Hocker, H. W. J. 1956. Certain aspects of climate as related to the distribution of loblolly pine. Ecology 37:824–834.

Hofgaard, A., J. Tardif, and Y. Bergeron. 1999. Dendroclimatic response of Picea mariana and Pinus banksiana along a latitudinal gradient in the eastern Canadian boreal forest. Canadian Journal of Forest Research 29:1333–1346.

Holmes, R. L., R. K. Adams, and H. C. Fritts. 1986. Tree-ring chronologies of western North America. Califirnia, Eastern Oregon and Northern Great Basin with procedures used in the chronology development work including users manuals for computer programs COFECHA and ARSTAN. Tucson, AZ.

Horton, K. W., and W. G. E. Brown. 1960. Ecology of white and red Pine in the Great Lakes-St. Lawrence Forest Region. Tecnical Note No. 88. Ottawa, ON. 118

Huang, J., J. C. Tardif, Y. Bergeron, B. Denneler, F. Berninger, and M. P. Girardin. 2010. Radial growth response of four dominant boreal tree species to climate along a latitudinal gradient in the eastern Canadian boreal forest. Global Change Biology 16:711–731.

Hughes, M. K. 2002. Dendrochronology in climatology - the state of the art. Dendrochronologia 20:95–116.

Hughes, M. K., P. M. Kelly, J. R. Pilcher, and V. C. J. LaMarche. 1982. Introduction. Pages 1–2 in M. K. Hughes, P. M. Kelly, J. R. Pilcher, and V. C. J. LaMarche, editors. Climate from Tree Ring. Cambridge University Press, Cambridge, UK.

Hutchinson, M. F. 1995. Interpolating mean rainfall using thin plate smoothing splines. International Journal of Geographical Information Systems 9:385–403.

Iles, N. 1990. Reconnaissance inventory to locate old white and/or red pine stands in site region 4E of the Ontario. Sudbury, ON.

IPCC. 2014. Summary for Policymakers. Pages 1–35 in The Core Writing Team, R. K. Pachauri, and L. Meyer, editors. Climate Change 2014: Synthesis Report. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

Jacobi, J. C., and F. H. Tainter. 1988. Dendroclimatic examination of white oak along an environmental gradient in the piedmont of South Carolina. Castanea 53:252–262.

Jump, A. S., J. M. Hunt, and J. Peñuelas. 2006. Rapid climate change-related growth decline at the southern range edge of Fagus sylvatica. Global Change Biology 12:2163–2174.

Jump, A. S., J. M. Hunt, and J. Peñuelas. 2007. Climate relationships of growth and establishment across the altitudinal range of Fagus sylvatica in the Montseny Mountains, northeast Spain. Écoscience 14:507–518.

Kaufman, L., and P. J. Rousseeuw. 1990. Finding groups in data: an introduction to cluster analysis. John Wiley and Sons, Inc., Hoboken, NJ, USA.

Kilgore, J. S., and F. W. Telewski. 2004. Climate-growth relationships for native and nonnative Pinaceae‎in‎northern‎Michigan’s‎pine‎barrens.‎Tree-Ring Research 60:3–13.

Kipfmueller, K. F., G. P. Elliott, E. R. Larson, and M. W. Salzer. 2010. An assessment of the dendroclimatic potential of three conifer species in northern Minnesota. Tree-Ring Research 66:113–126.

Koprowski, M., and A. Zielski. 2006. Dendrochronology of Norway spruce (Picea abies (L.) Karst.) from two range centres in lowland Poland. Trees 20:383–390.

119

Kozlowski, T. T., and G. A. Borger. 1971. Effect of temperature and light intensity early in ontogeny on growth of Pinus resinosa seedlings. Canadian Journal of Forest Research 1:57–65.

Kremenetski, C. V., L. D. Sulerzhitsky, and R. Hantemirov. 1998. Holocene history of the northern range limits of some trees and shrubs in Russia. Arctic and Alpine Research 30:317–333.

Lafleur, B., D. Paré, A. D. Munson, and Y. Bergeron. 2010. Response of northeastern North American forests to climate change: will soil conditions constrain tree species migration? Environmental Reviews 18:279–289.

Larson, P. R. 1960. A physiological consideration of the springwood summerwood transition in red pine. Forest Science 6:110–122.

LeBlanc, D. C. 1993. Temporal and spatial variation of oak growth - climate relationships along a pollen gradient in the midwestern United States. Canadian Journal of Forest Research 23:772–782.

Lebourgeois, F. 2000. Climatic signals in earlywood, latewood and total ring width of corsican pine from western France. Annals of Forest Science 57:155–164.

Legendre, P., and H. J. B. Birks. 2012. From classical to canonical ordination. Pages 201–248 in H. J. B. Birks, A. F. Lotter, S. Juggins, and J. P. Smol, editors. Tracking Environmental Change using Lake Sediments. Springer Netherlands, Dordrecht, The Netherlands.

Legendre, P., and L. Legendre. 2012. Numerical Ecology. 3rd edition. Elsevier, Amsterdam, The Netherlands.

Leithead, M. D., M. Anand, and L. C. R. Silva. 2010. Northward migrating trees establish in treefall gaps at the northern limit of the temperate-boreal ecotone, Ontario, Canada. Oecologia 164:1095–106.

Leithead, M., L. C. R. Silva, and M. Anand. 2012. Recruitment patterns and northward tree migration through gap dynamics in an old-growth white pine forest in northern Ontario. Plant Ecology 213:1699–1714.

Lemmen, D. S., F. J. Warren, J. Lacroix, and E. Bush, editors. 2008. From impacts to adaptation :‎Canada‎in‎a changing climate 2007. Government of Canada, Ottawa, ON.

Lenoir, J., J. C. Gégout, P. A. Marquet, P. de Ruffray, and H. Brisse. 2008. A significant upward shift in plant species optimum elevation during the 20th century. Science 320:1768–1771.

120

Liang, E., Y. Wang, D. Eckstein, and T. Luo. 2011. Little change in the fir tree-line position on the southeastern Tibetan Plateau after 200 years of warming. New Phytologist 190:760– 769.

Linderholm, H. W. 2006. Growing season changes in the last century. Agricultural and Forest Meteorology 137:1–14.

Linderholm, H. W., and K. Linderholm. 2004. Age dependent climate sensitivity of Pinus sylvestris L. in the central Scandinavian Mountains. Boreal Environment Research 9:307– 317.

Little, E. L. J. 1971. Atlas of United States trees, volume 1, conifers and important hardwoods: Miscellaneous Publication 1146. 200 maps. U.S. Department of Agriculture.

Liu, K.-B. 1990. Holocene paleoecology of the boreal forest and Great Lakes - St . Lawrence forest in northern Ontario. Ecological Monographs 60:179–212.

Lloyd, A. H., A. G. Bunn, and L. Berner. 2011. A latitudinal gradient in tree growth response to climate warming in the Siberian taiga. Global Change Biology 17:1935–1945.

Lloyd, A. H., and C. L. Fastie. 2002. Spatial and temporal variability in the growth and climate response of treeline trees in alaska. Climatic Change 52:481–509.

Loehle, C., and D. Leblanc. 1996. Model-based assessments of climate change effects on forests :‎a‎critical‎review.‎Ecological‎Modelling‎90:1–31.

Lukac, M., C. Calfapietra, A. Lagomarsino, and F. Loreto. 2010. Global climate change and tree nutrition: effects of elevated CO2 and temperature. Tree Physiology 30:1209–1220.

Lyakh, Y., V. Gurianov, O. Gorshkov, and Y. Vihovanets. 2012. Estimating the number of data clusters via the contrast statistic. Journal of Biomedical Science and Engineering 5:95–99.

Mackey, B. G., D. W. McKenney, Y.-Q. Yang, J. P. McMahon, and M. F. Hutchinson. 1996a. Erratum:‎site‎regions‎revisited:‎a‎climatic‎analysis‎of‎Hills’‎site‎regions‎for‎the‎province‎of‎ Ontario using a parametric method. Canadian Journal of Forest Research 26:1112.

Mackey, B. G., D. W. McKenney, Y.-Q. Yang, J. P. McMahon, and M. F. Hutchinson. 1996b. Site‎regions‎revisited:‎a‎climatic‎analysis‎of‎Hills’‎site‎regions‎for‎the‎province‎of‎Ontario‎ using a parametric method. Canadian Journal of Forest Research 26:333–354.

Maeglin, R. R. 1979. Increment cores: how to collect, handle, and use them.

Mäkinen, H., P. Nöjd, H.-P. Kahle, U. Neumann, B. Tveite, K. Mielikäinen, H. Röhle, and H. Spiecker. 2002. Radial growth variation of norway spruce (Picea abies (L .) Karst .) across 121

latitudinal and altitudinal gradients in central and northern Europe. Forest Ecology and Management 171:243–259.

Manabe, S., R. T. Wetherald, P. C. D. Milly, T. L. Delworth, and R. J. Stouffer. 2004. Century- scale change in water availability: CO2-quadrupling experiment. Climatic Change 64:59– 76.

Mather, J. R., and G. A. Yoshioka. 1968. The role of climate in the distribution of vegetation. Annals of the Association of American Geographers 58:29–41.

Matthews, S. N., L. R. Iverson, A. M. Prasad, M. P. Peters, and P. G. Rodewald. 2011. Modifying climate change habitat models using tree species-specific assessments of model uncertainty and life history-factors. Forest Ecology and Management 262:1460–1472.

McDowell, N., W. T. Pockman, C. D. Allen, D. D. Breshears, N. Cobb, T. Kolb, J. Plaut, J. Sperry, A. West, D. G. Williams, and E. a. Yepez. 2008. Mechanisms of plant survival and mortality during drought: Why do some plants survive while others succumb to drought? New Phytologist 178:719–739.

McKenney, D., J. Pedlar, M. Hutchinson, P. Papadopol, K. Lawrence, K. Campbell, E. Milewska,‎R.‎F.‎Hopkinson,‎and‎D.‎Price.‎2013.‎Spatial‎climate‎models‎for‎Canada’s‎ forestry community. The Forestry Chronicle 89:659–663.

McKenney, D. W., M. F. Hutchinson, J. L. Kesteven,‎and‎L.‎A.‎Venier.‎2001.‎Canada’s‎plant‎ hardiness zones revisited using modern climate interpolation techniques. Canadian Journal of Plant Science 81:129–143.

McKenney, D. W., J. H. Pedlar, K. Lawrence, K. Campbell, and M. F. Hutchinson. 2007. Potential impacts of climate change on the distribution of North American trees. BioScience 57:939–948.

McLane, S. C., V. M. LeMay, and S. N. Aitken. 2011. Modeling lodgepole pine radial growth relative to climate and genetics using universal growth-trend response functions. Ecological Applications 21:776–788.

Meilleur, A., J. Brisson, and A. Bouchard. 1997. Ecological analyses of the northernmost population of pitch pine (Pinus rigida). Canadian Journal of Forest Research 27:1342–1350.

Millar, C. I., N. L. Stephenson, and S. L. Stephens. 2007. Climate change and forests of the future: managing in the face of uncertainty. Ecological Applications 17:2145–2151.

Milligan, G. W., and M. C. Cooper. 1985. An examination of procedures for determining the number of clusters in a data set. Psychometrika 50:159–179.

122

Mitchell, T. D., and P. D. Jones. 2005. An improved method of constructing a database of monthly climate observations and associated high-resolution grids. International Journal of Climatology 25:693–712.

Mittler, R. 2006. Abiotic stress, the field environment and stress combination. Trends in Plant Science 11:15–9.

Miyamoto, Y., H. P. Griesbauer, and D. S. Green. 2010. Growth responses of three coexisting conifer species to climate across wide geographic and climate ranges in Yukon and British Columbia. Forest Ecology and Management 259:514–523.

Mooi, E., and M. Sarstedt. 2011. Cluster analysis. Pages 237–284 A Concise Guide to Market Research. Springer-Verlag Berlin Heidelberg, Heidelberg, Germany.

Moorcroft, P. R., S. W. Pacala, and M. A. Lewis. 2006. Potential role of natural enemies during tree range expansions following climate change. Journal of Theoretical Biology 241:601– 616.

Moore, P. D. 2003. Back to the future: biogeographical responses to climate change. Progress in Physical Geography 27:122–129.

Nakićenović,‎N.,‎O.‎Davidson,‎G.‎Davis,‎A.‎Grübler,‎T.‎Kram,‎E.‎Lebre‎La‎Rovere,‎B.‎Metz,‎T.‎ Morita, W. Pepper, H. Pitcher, A. Sankovski, P. Shukla, R. Swart, R. Watson, and Z. Dadi, editors. 2000. Emission scenarios: summary for policymakers: a special report of IPCC Working Group III. Geneva, Switzerland.

Newman, J. A., M. Anand, H. A. L. Henry, S. Hunt, and Z. Gedalof. 2011. Climate change biology. CAB International, Cambridge, MA, USA.

O’Neill,‎G.‎A.,‎A.‎Hamann,‎and‎T.‎Wang.‎2008.‎Accounting‎for‎population variation improves estimates‎of‎the‎impact‎of‎climate‎change‎on‎species‎’‎growth‎and‎distribution.‎Journal‎of‎ Applied Ecology 45:1040–1049.

Oberhuber, W., M. Stumböck, and W. Kofler. 1998. Climate-tree-growth relationships of scots pine stands ( Pinus sylvestris L .) exposed to soil dryness. Trees 13:19–27.

Ohse, B., F. Jansen, and M. Wilmking. 2012. Do limiting factors at Alaskan treelines shift with climatic regimes? Environmental Research Letters 7:015505 (12pp).

OMNR. 1998. A silvicultural guide for the Great Lakes - St . Lawrence conifer forest in Ontario. Version‎1.1.‎Queen’s‎Printer‎for‎Ontario,‎Toronto,‎ON.

123

OMNR. 2003. Chiniguchi Waterway provincial park (P174), Wolf Lake old growth forest forest reserve (F175) and Kukagami Lake Forest Reserve (F181)‎fact‎sheet.‎Queen’s‎Printer‎for‎ Ontario, Sault Ste Marie, ON.

Orwig, D. A., and M. D. Abrams. 1997. Variation in radial growth responses to drought among species, site, and canopy strata. Trees 11:474–484.

Overpeck, J. T., P. J. Bartlein, and T. Webb III. 1991. Potential magnitude of future vegetation changes in Eastern North America: comparisons with the past. Science 254:692–695.

Pallardy, S. G. 2008. Photosynthesis. Pages 107–167 Physiology of Woody Plants. 3rd edition. Elsevier, San Diego, California, USA.

Palmer, M. W. 1993. Putting things in even better order: the advantages of canonical correspondence analysis. Ecology 74:2215–2230.

Pan, C., S. J. Tajchman, and J. N. Kochenderfer. 1997. Dendroclimatological analysis of major forest species of the central Appalachians. Forest Ecology and Management 98:77–87.

Parker, J. 1963. Cold resistence in woody plants. The Botanical Review 29:123–201.

Parmesan, C. 2006. Ecological and evolutionary responses to recent climate change. Annual Review of Ecology, Evolution, and Systematics 37:637–669.

Paul, V., Y. Bergeron, and F. Tremblay. 2014. Does climate control the northern range limit of eastern white cedar (Thuja occidentalis L.)? Plant Ecology 215:181–194.

Peltola, H., A. Kilpeläinen, and S. Kellomäki. 2002. Diameter growth of scots pine (Pinus sylvestris) trees grown at elevated temperature and carbon dioxide concentration under boreal conditions. Tree Physiology 22:963–972.

Pereira, J. . 1995. Gas exchange and growth. Pages 147–181 in E.-D. Schulze and M. M. Caldwell, editors. Ecophysiology of Photosynthesis. Illustrate. Springer Berlin Heidelberg.

Peterson, D. W., D. L. Peterson, and G. J. Ettl. 2002. Growth responses of subalpine fir to climatic variability in the Pacific Northwest. Canadian Journal of Forest Research 32:1503– 1517.

Piao, S., P. Ciais, M. Lomas, C. Beer, H. Liu, J. Fang, P. Friedlingstein, Y. Huang, H. Muraoka, Y. Son, and I. Woodward. 2011. Contribution of climate change and rising CO2 to terrestrial carbon balance in East Asia: a multi-model analysis. Global and Planetary Change 75:133–142.

124

Piao, S., P. Friedlingstein, P. Ciais, L. Zhou, and A. Chen. 2006. Effect of climate and CO2 changes on the greening of the Northern Hemisphere over the past two decades. Geophysical Research Letters 33:L23402 (1–6).

Pielou, E. C. 1984. The interpretation of ecological data: a premier on classification and ordination. John Wiley and Sons, Inc., Canada.

Piovesan, G., F. Biondi, M. Bernabei, A. Di Filippo, and B. Schirone. 2005. Spatial and altitudinal bioclimatic zones of the Italian peninsula identified from a beech (Fagus sylvatica L.) tree-ring network. Acta Oecologica 27:197–210.

Pretzsch, H. 2009. Forest dynamics, growth and yield: from measurement to model. Springer- Verlag Berlin Heidelberg.

Price, D. T., D. W. McKenney, L. A. Joyce, R. M. Siltanen, P. Papadopol, and K. Lawrence. 2011. High-resolution interpolation of climate scenarios for Canada derived from General Circulation Model simulations. Information Report NOR-X-421. Natural Resources Canada, Canadian Forest Service, Northern Forestry Centre, Edmonton, AB.

Price, D. T., D. W. McKenney, I. A. Nalder, M. F. Hutchinson, and J. L. Kesteven. 2000. A comparison of two statistical methods for spatial interpolation of Canadian mean climate data. Agricultural and Forest Meteorology 101:81–94.

Puritch, G. S. 1977. Cone production in conifers. Forestry Service, Environment Canada, Victoria, BC.

Quinby, P. A. 1996. Status of old growth red pine forests in eastern North America: a preliminary assessment. Forest Landscape Baseline No. 14. Toronto, ON.

Quinby, P. A., K. Cowcill, T. Martens, R. Hart, J. Paterson, P. Smylie, C. Blomme, D. Sone, and M. Anand. 2013. Wolf lake reserve species inventory 2012.

Reed, D. D., and P. V. Desanker. 1992. Ecological implications of projected climate change scenarios in forest ecosystems in northern Michigan, USA. International Journal of Biometeorology 36:99–107.

Reich, P. B., and J. Oleksyn. 2008. Climate warming will reduce growth and survival of scots pine except in the far North. Ecology Letters 11:588–597.

Revelle, W. 1979. Hierarchical cluster analysis and the internal structure of tests. Multivariate Behavioral Research 14:57–74.

Robeson, S. M. 2004. Trends in time-varying percentiles of daily minimum and maximum temperature over North America. Geophysical Research Letters 31:L04203. 125

Rolland, C., V. Petitcolas, and R. Michalet. 1998. Changes in radial tree growth for Picea abies, Larix decidua, Pinus cembra and Pinus uncinata near the alpine timberline since 1750. Trees 13:40–53.

Romero-Lankao, P., J. B. Smith, D. J. Davidson, N. S. Diffenbaugh, P. L. Kinney, P. Kirshen, P. Kovacs, and L. Villers Ruiz. 2014. North America. Pages 1439–1498 in V. R. Barros, C. B. Field, D. J. Dokken, M. D. Mastrandrea, K. J. Mach, T. E. Bilir, M. Chatterjee, K. L. Ebi, Y. O. Estrada, R. C. Genova, B. Girma, E. S. Kissel, A. N. Levy, S. MacCracken, P. R. Mastrandrea, and L. L. White, editors. Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part B: Regional Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

Rowe, J. S. 1972. Forest regions of Canada. Ottawa, ON.

Rozas, V. 2005. Dendrochronology of pedunculate oak (Quercus robur L.) in an old-growth pollarded woodland in northern Spain: tree ring growth responses to climate. Annals of Forest Science 62:209–218.

Rozas, V., L. DeSoto, and J. M. Olano. 2009. Sex-specific, age-dependent sensitivity of tree-ring growth to climate in the dioecious tree Juniperus thurifera. New Phytologist 182:687–697.

Rudolf, P. O. 1990. Red pine. Pages 442–455 in R. M. Burns and B. H. Honkala, editors. Silvics of North America. Volume 1: Conifers. Agriculture Handbook 654. Forest Service United States Department of Agriculture, Washington, DC.

Rudolph, T. D., and P. R. Laidly. 1990. Jack Pine. Pages 280–293 in R. M. Burns and B. H. Honkala, editors. Silvics of North America. Volume 1: Conifers. Agriculture Handbook 654. Forest Service United States Department of Agriculture, Washington, DC.

Rustad, L. E., J. L. Campbell, G. M. Marion, R. J. Norby, M. J. Mitchell, A. E. Hartley, J. H. C. Cornelissen, J. Gurevitch, and GCTE-NEWS. 2001. A meta-analysis of the response of soil respiration, net nitrogen mineralization, and aboveground plant growth to experimental ecosystem warming. Oecologia 126:543–562.

Sage, R. F., and D. S. Kubien. 2007. The temperature response of C3 and C4 photosynthesis. Plant, Cell and Environment 30:1086–1106.

Sakai, A., and C. J. Weiser. 1973. Freezing resistance of trees in North America with reference to tree regions. Ecology 54:118–126.

Saxe, H., M. G. R. Cannell, Ø. Johnsen, M. G. Ryan, and G. Vourlitis. 2001. Tree and forest functioning in response to global warming. New Phytologist 149:369–400.

126

Schenk, H. J. 1996. Modeling the effects of temperature on growth and persistence of tree species :‎a‎critical‎review‎of‎tree‎population‎models.‎Ecological‎Modelling‎92:1–32.

Schimel, D., J. Melillo, H. Tian, A. D. McGuire, D. Kicklighter, T. Kittel, N. Rosenbloom, S. Running, P. Thornton, D. Ojima, W. Parton, R. Kelly, M. Sykes, R. Neilson, and B. Rizzo. 2000. Contribution of increasing CO2 and climate to carbon storage by ecosystems in the United States. Science 287:2004–2006.

Schumacher, F. X., and B. B. Day. 1939. The influence of precipitation upon the width of annual rings of certain timber trees. Ecological Monographs 9:387–429.

Schuster, R., and W. Oberhuber. 2013. Age-dependent climate-growth relationships and regeneration of Picea abies in a drought-prone mixed coniferous forest in the Alps. Canadian Journal of Forest Research 43:609–618.

Schwartz, M. D., R. Ahas, and A. Aasa. 2006. Onset of spring starting earlier across the Northern Hemisphere. Global Change Biology 12:343–351.

Shao, G., and P. N. Halpin. 1995. Climatic controls of eastern North American coastal tree and shrub distributions. Journal of Biogeography 22:1083–1089.

Sigurdsson, B. D., J. L. Medhurst, G. Wallin, O. Eggertsson, and S. Linder. 2013. Growth of mature boreal Norway spruce was not affected by elevated CO2 and/or air temperature unless nutrient availability was improved. Tree Physiology 33:1192–1205.

Silva, L. C. R., M. Anand, and M. D. Leithead. 2010. Recent widespread tree growth decline despite increasing atmospheric CO2. PLoS ONE 5:e11543.

Sims, R. A., H. M. Kershaw, and G. M. Wickware. 1990. The autecology of major tree species in the north central region of Ontario. COFRDA Report 3302, NWOFTDU Technical Report 48. Northwestern Ontario Forest Technology Development Unit, Ontario Ministry of Natural Resources, , ON.

Slatyer, R. O. 1977. Altitudinal variation in the photosynthetic characteristics of snow gum, Eucalyptus pauciflora Sieb. ex Spreng. IV temperature response of four populations growth at different temperatures. Australian Journal of Plant Physiology 4:583–94.

Smith, S. P., and R. Dubes. 1980. Stability of a hierarchical clustering. Pattern Recognition 12:177–187.

Speer, J. H. 2011. Fundamentals of tree-ring research. The University of Arizona Press, Tucson, AZ.

127

Spicer, R., and B. L. Gartner. 2001. The effects of cambial age and position within the stem on specific conductivity in Douglas-fir (Pseudotsuga menziesii) sapwood. Trees 15:222–229.

Spittlehouse, D. L. 2005. Integrating climate change adaptation into forest management. The Forestry Chronicle 81:691–695.

Stephenson, N. L. 1990. Climatic control of vegetation distribution: the role of the water balance. American Naturalist 135:649–670.

Strauss, J. S., J. J. Bartko, and W. T. J. Carpenter. 1973. The use of clustering techniques for the classification of psychiatric patients. The British Journal of Psychiatry 122:531–540.

Strömgren, M., and S. Linder. 2002. Effects of nutrition and soil warming on stemwood production in a boreal Norway spruce stand. Global Change Biology 8:1195–1204.

Sugar, C. A., and G. M. James. 2003. Finding the number of clusters in a dataset. Journal of the American Statistical Association 98:750–763.

Sutton, A., R. J. Staniforth, and J. Tardif. 2002. Reproductive ecology and allometry of red pine (Pinus resinosa) at the northwestern limit of its distribution range in Manitoba, Canada. Canadian Journal of Botany 80:482–493.

Swetnam, T. W., C. D. Allen, and J. L. Betancourt. 1999. Applied historical ecology: using the past to manage for the future. Ecological Applications 9:1189–1206.

Szeicz, J. M., and G. M. MacDonald. 1994. Age dependent tree ring growth response of subarctic white spruce to climate. Canadian Journal of Forest Research 24:120–132.

Tanja, S., F. Berninger, T. Vesala, T. Markkanen, P. Hari, A. Mäkelä, H. Ilvesniemi, H. Hänninen, E. Nikinmaa, T. Huttula, T. Laurila, M. Aurela, A. Grelle, A. Lindroth, A. Arneth, O. Shibistova, and J. Lloyd. 2003. Air temperature triggers the recovery of evergreen boreal forest photosynthesis in spring. Global Change Biology 9:1410–1426.

Tardif, J., J. Brisson, and Y. Bergeron. 2001a. Dendroclimatic analysis of Acer saccharum, Fagus grandifolia, and Tsuga canadensis from an old-growth forest, southwestern Quebec. Canadian Journal of Forest Research 31:1491–1501.

Tardif, J. C., F. Conciatori, P. Nantel, and D. Gagnon. 2006. Radial growth and climate responses of white oak (Quercus alba) and northern red oak (Quercus rubra) at the northern distribution limit of white oak in Quebec, Canada. Journal of Biogeography 33:1657–1669.

Tardif, J., J. J. Camarero, M. Ribas, and E. Gutiérrez. 2003. Spatiotemporal variability in tree growth in the Central Pyrenees: climatic and site influences. Ecological Monographs 73:241–257. 128

Tardif, J., F. Conciatori, and Y. Bergeron. 2001b. Comparrative analysis of the climatic response of seven boreal tree species from northwestern Quebec, Canada. Tree-Ring Research 57:169–181.

Tardif, J., and D. Stevenson. 2001. Radial growth-climate association of Thuja occidentalis L . at the northwestern limit of its distribution, Manitoba, Canada. Dendrochronologia 19:179– 187.

Taylor, K. E., R. J. Stouffer, and G. a. Meehl. 2012. An overview of CMIP5 and the experiment design. Bulletin of the American Meteorological Society 93:485–498.

Tessier, L. 1989. Spatio-temporal analysis of climate-tree ring relationships. New Phytologist 111:517–529.

Tibshirani, R., and G. Walther. 2005. Cluster validation by prediction strength. Journal of Computational and Graphical Statistics 14:511–528.

Tibshirani, R., G. Walther, and T. Hastie. 2001. Estimating the number of clusters in a data set via the gap statistic. Journal of the Royal Statistical Society: Series B 63:411–423.

Toledo, M., M. Peña-Claros, F. Bongers, A. Alarcón, J. Balcázar, J. Chuviña, C. Leaño, J. C. Licona, and L. Poorter. 2012. Distribution patterns of tropical woody species in response to climatic and edaphic gradients. Journal of Ecology 100:253–263.

Tryon, E. H., J. O. Cantrell, and K. L. Carvell. 1957. Effect of precipitation and temperature on increment of yellow-poplar. Forest Science 3:32–44.

USGS.‎1999.‎Digital‎representation‎of‎“Atlas‎of‎United‎States‎trees”‎by‎Elbert‎L.‎Little,‎Jr.

Vaganov, E. A., M. K. Hughes, A. V. Kirdyanov, F. H. Schweingruber, and P. P. Silkin. 1999. Influence of snowfall and melting on tree growth in subartic Eurasia. Nature 400:149–151.

Viattchenin, D. A. 2013. A heuristic approach to possibilistic clustering: algorithms and applications. Springer-Verlag Berlin Heidelberg, Heidelberg, Germany.

Vila, B., M. Vennetier, C. Ripert, O. Chandioux, E. Liang, F. Guibal, and F. Torre. 2008. Has global change induced divergent trends in radial growth of Pinus sylvestris and Pinus halepensis at their bioclimatic limit? The example of the Sainte-Baume forest (south-east France). Annals of Forest Science 65:709p1–709p9.

Vincent, L. A., and É. Mekis. 2006. Changes in daily and extreme temperature and precipitation indices for Canada over the twentieth century. Atmosphere-Ocean 44:177–193.

Van Wagner, C. E. 1970. Fire and red pine. Annual Tall Timbers Fire Ecology Conference. 129

Wang, X., G. Huang, Q. Lin, and J. Liu. 2014a. High-resolution probabilistic projections of temperature changes over Ontario, Canada. Journal of Climate 27:5259–5284.

Wang, X., G. Huang, Q. Lin, X. Nie, and J. Liu. 2014b. High-resolution temperature and precipitation projections over Ontario, Canada: a coupled dynamical-statistical approach. Quarterly Journal of the Royal Meteorological Society:1–10.

Way, D. A., and R. Oren. 2010. Differential responses to changes in growth temperature between trees from different functional groups and biomes: a review and synthesis of data. Tree Physiology 30:669–88.

Wettstein, J. J., J. S. Littell, J. M. Wallace, and Z. Gedalof. 2011. Coherent region-, species-, and frequency-dependent local climate signals in northern Hemisphere tree-ring widths. Journal of Climate 24:5998–6012.

Wetzel, S., and D. Burgess. 1994. Current understanding of white and red pine physiology. The Forestry Chronicle 70:420–426.

Wigley, T. M. L., K. R. Briffa, and P. D. Jones. 1984. On the average value of correlated time series, with applications in dendroclimatology and hydrometeorology. Journal of Climate and Applied Meteorology 23:201–213.

Williamson, T., S. Colombo, P. Duinker, P. Gray, R. Hennessey, D. Houle, M. Johnston, A. Ogden,‎and‎D.‎Spittlehouse.‎2009.‎Climate‎change‎and‎Canada’s‎forests:‎from‎impacts‎to‎ adaptation. Sustainable Forest Management Network, and Natural Resource Canada, Canadian Forest Service, Northern Forestry Centre, Edmonton, AB.

Wilmking, M., G. P. Juday, V. A. Barber, and H. J. Zald. 2004. Recent climate warming forces contrasting growth responses of white spruce at treeline in Alaska through temperature thresholds. Global Change Biology 10:1724–1736.

Wilson, R., and W. Elling. 2004. Temporal instability in tree-growth/climate response in the Lower Bavarian Forest region: implications for dendroclimatic reconstruction. Trees 18:19– 28.

Wilson, R. J. S., and M. Hopfmueller. 2001. Dendrochronological investigations of Norway spruce along an elevational transect in the Bavarian Forest, Germany. Dendrochronologia 19:67–79.

Wimmer, R., and M. Grabner. 2000. A comparison of tree-ring features in Picea abies as correlated with climate. IAWA Journal 21:403–416.

WinDENDRO. 2012. WinDENDRO for tree ring analysis. Regent Instruments Canada Inc, Québec, QC. 130

Woodward, A., D. G. Silsbee, E. G. Schreiner, and J. E. Means. 1994. Influence of climate on radial growth and cone production in subalpine fir (Abies lasiocarpa) and mountain hemlock (Tsuga mertensiana). Canadian Journal of Forest Research 24:1133–1143.

Woodward, F. I. 1987. Climate and Plant Distribution. Cambridge University Press, Cambridge.

Woodward, F. I., G. E. Fogg, and U. Heber. 1990. The impact of low temperature in controlling the geographical distribution of plants [and discission]. Philosophical Transcations of the Royal Society of London. Series B, Biological Sciences 326:585–593.

Woolsey, T. S., and H. H. Chapman. 1914. Norway pine in the Lake States. Bulletin of the U. S. Department of Agriculture No. 139.

Xu, R., and D. I. Wunsch. 2005. Survey of clustering algorithms. IEEE Transactions on Neural Networks 16:645–678.

Yamori, W., K. Hikosaka, and D. A. Way. 2014. Temperature response of photosynthesis in C3, C4, and CAM plants: temperature acclimation and temperature adaptation. Photosynthesis Research 119:101–117.

Yeh, H.-Y., and L. C. Wensel. 2000. The relationship between tree diameter growth and climate for coniferous species in northern California. Canadian Journal of Forest Research 30:1463– 1471.

Yin, Z., X. Zhou, C. Bakal, F. Li, Y. Sun, N. Perrimon, and S. T. C. Wong. 2008. Using iterative cluster merging with improved gap statistics to perform online phenotype discovery in the context of high-throughput RNAi screens. BMC Bioinformatics 9:264.

Yokohata, T., M. J. Webb, M. Collins, K. D. Williams, M. Yoshimori, J. C. Hargreaves, and J. D. Annan. 2010. Structural similarities and differences in climate responses to CO2 increase between two perturbed physics ensembles. Journal of Climate 23:1392–1410.

Yu, G., Y. Liu, X. Wang, and K. Ma. 2008. Age-dependent tree-ring growth responses to climate in Qilian juniper (Sabina przewalskii Kom.). Trees 22:197–204.

Zahner, R., J. E. Lotan, and W. D. Baughman. 1964. Earlywood - latewood features of red pine grown under simulated drought and irrigation. Forest Science 10:361–370.

Zhang, X., L. A. Vincent, W. D. Hogg, and A. Niitsoo. 2000. Temperature and precipitation trends in Canada during the 20th century. Atmosphere-Ocean 38:395–429.

Zhou, L., C. J. Tucker, R. K. Kaufmann, D. Slayback, N. V Shabanov, and R. B. Myneni. 2001. Variations in northern vegetation activity inferred from satellite data of vegetation index during 1981 to 1999. Journal of Geophysical Research 106:20069–20083. 131

Zhu, K., C. W. Woodall, and J. S. Clark. 2012. Failure to migrate: lack of tree range expansion in response to climate change. Global Change Biology 18:1042–1052.

132 Appendix A Red pine chronologies with corresponding sample size. Bold red line indicates the running average of chronology index values

Redpr94 PAKPRC98

NWO25 CANA259

133

SEUPR95 SIOPRA88

NEO07 NEO09

134

NWO12 NWO11

NEO05 NEO06

135

NEO08 BLUPRC55

NEO29 IVAPRC52

136

Appendix B Dissimilarity Matrix

Squared Euclidean Distance 1 2 3 4 5 6 7 8 9 10 1:NEO02 0.000 16.814 191.921 1.148 14.933 4.984 112.459 74.110 66.516 37.760 2:NEO03 16.814 0.000 106.268 20.562 1.240 22.409 46.202 23.409 22.123 46.352 3:NEO05 191.921 106.268 0.000 201.947 115.942 203.003 18.177 51.295 63.715 161.869 4:NEO01 1.148 20.562 201.947 0.000 16.603 1.665 118.775 77.514 67.898 31.838 5:NEO04 14.933 1.240 115.942 16.603 0.000 16.228 50.788 23.806 20.107 37.993 6:NEO10 4.984 22.409 203.003 1.665 16.228 0.000 116.517 72.780 61.458 26.346 7:NEO06 112.459 46.202 18.177 118.775 50.788 116.517 0.000 8.970 16.807 102.923 8:BLUPRC55 74.110 23.409 51.295 77.514 23.806 72.780 8.970 0.000 2.397 73.873 9:IVAPRC52 66.516 22.123 63.715 67.898 20.107 61.458 16.807 2.397 0.000 57.533 10:556PRB 37.760 46.352 161.869 31.838 37.993 26.346 102.923 73.873 57.533 0.000 11:NEO07 171.220 83.884 23.604 177.600 88.950 172.567 10.517 25.107 35.381 156.752 12:NEO08 139.902 66.818 25.011 143.440 67.895 136.315 7.990 15.131 17.456 106.681 13:NEO09 195.285 103.192 17.980 200.871 107.003 193.935 14.725 33.706 41.588 160.048 14:MONPRE 43.067 76.263 255.289 35.510 66.471 32.451 176.893 133.884 111.987 17.195 15:NWO29 122.474 71.385 41.365 123.412 71.076 118.931 32.762 41.425 36.271 64.338 16:NWO12 173.203 96.392 18.604 178.081 100.814 175.047 24.619 47.086 50.482 125.092 17:NWO11 196.403 117.818 20.723 201.114 122.205 198.012 36.336 64.110 66.719 134.391 18:SGWPR 42.088 29.667 139.805 41.336 26.029 39.183 78.691 51.479 39.310 35.831 19:NWO13 40.357 30.130 147.706 39.469 26.176 37.311 83.666 54.343 41.665 36.384 20:NWO25 223.044 126.624 29.575 234.505 138.413 237.203 40.306 72.903 89.984 220.193 21:SOWPRA 114.644 51.021 54.422 123.994 58.520 126.858 28.739 32.952 40.345 136.164 22:CANA264 78.807 37.423 116.896 86.031 40.581 87.285 61.658 41.883 39.837 112.929 23:NWO28 83.634 38.944 111.915 91.382 42.672 92.808 58.478 40.476 39.759 119.528 24:SABPRA 118.382 54.715 55.652 128.311 62.716 132.027 32.144 37.662 45.386 140.947 25:NW027 91.987 42.204 100.648 100.937 47.464 103.538 53.131 40.263 42.448 130.737 137

26:NWO14 91.624 42.280 101.152 100.707 47.725 103.645 53.871 41.290 43.746 131.426 27:SIOPRA88 123.072 60.496 63.772 132.906 69.931 138.164 40.280 48.759 60.277 154.090 28:STOPRB 88.007 42.317 104.713 97.537 49.518 103.202 59.068 50.015 55.910 137.122 29:SEUPR95 140.940 73.201 62.897 151.352 83.621 157.041 44.511 57.320 70.771 170.687 30:NWO15 95.538 50.447 107.847 105.611 59.250 113.318 66.753 62.374 70.955 148.218 31:NWO26 76.085 58.559 221.246 83.969 62.419 90.090 136.738 100.284 99.972 158.974 32:CANA259 131.501 70.620 89.167 141.707 80.696 148.196 59.743 65.620 78.541 174.340 33:BIBPR 95.301 54.655 130.885 105.002 63.234 113.126 82.258 73.664 82.403 157.009 34:NWO16 99.505 56.267 124.299 109.310 65.106 117.342 78.656 72.288 81.647 158.613 35:PAKPRC98 155.847 85.387 87.389 166.105 95.687 171.313 60.901 69.900 84.757 195.907 36:CANA246 105.041 59.881 126.963 114.684 68.753 122.381 80.515 74.303 84.429 165.236 37:NWO17 109.385 60.491 109.689 119.458 69.786 127.258 71.073 69.622 80.029 161.772 38:NWO19 106.341 60.491 125.960 115.988 69.397 123.616 79.964 74.137 84.419 166.102 39:CANA263 95.652 55.563 133.104 105.301 63.864 113.375 84.127 74.960 83.188 156.776 40:CANA252 117.674 64.809 98.150 128.028 74.364 135.758 66.023 68.661 79.453 163.372 41:Redpr94 197.566 111.993 79.068 208.233 122.843 211.859 62.222 78.764 96.338 228.808 42:CANA255 97.388 57.571 138.579 106.936 65.706 114.932 87.811 77.629 85.873 160.357 43:CALPR 90.871 72.739 242.618 99.028 77.333 106.028 154.700 117.818 119.681 183.839 44:CANA245 92.250 72.084 236.659 100.530 76.788 107.393 150.500 114.833 116.973 183.626 45:LOOPR2 91.384 65.944 206.978 99.914 72.033 107.595 131.139 103.811 108.830 176.818 46:NWO24 95.422 61.417 168.154 104.445 68.786 112.398 106.459 89.295 96.719 169.084 47:JOHPRB 105.663 63.732 145.377 114.912 71.895 122.563 92.709 82.451 91.949 171.561 48:CANA265 97.289 77.547 247.346 105.434 82.246 112.342 158.605 121.786 124.372 192.358 49:LOMPR3 99.173 65.533 177.154 107.973 72.745 115.823 112.849 94.445 102.179 175.808 50:NWO20 100.223 70.057 200.056 108.685 76.473 116.129 127.344 103.217 109.924 183.995 51:NWO21 104.174 70.545 190.498 112.620 77.200 119.832 121.103 99.703 107.373 184.982 52:NWO23 105.307 72.363 198.264 113.654 78.646 120.562 125.790 102.213 109.425 188.397 53:NWO22 117.619 73.263 168.056 126.009 80.182 131.796 105.378 89.452 99.108 190.782 54:CANA253 114.455 77.938 205.195 122.639 83.345 128.100 128.974 102.783 109.530 198.522

138

Dissimilarity‎Matrix‎count…….. Squared Euclidean Distance 12 13 14 15 16 17 18 19 20 21 1:NEO02 139.902 195.285 43.067 122.474 173.203 196.403 42.088 40.357 223.044 114.644 2:NEO03 66.818 103.192 76.263 71.385 96.392 117.818 29.667 30.130 126.624 51.021 3:NEO05 25.011 17.980 255.289 41.365 18.604 20.723 139.805 147.706 29.575 54.422 4:NEO01 143.440 200.871 35.510 123.412 178.081 201.114 41.336 39.469 234.505 123.994 5:NEO04 67.895 107.003 66.471 71.076 100.814 122.205 26.029 26.176 138.413 58.520 6:NEO10 136.315 193.935 32.451 118.931 175.047 198.012 39.183 37.311 237.203 126.858 7:NEO06 7.990 14.725 176.893 32.762 24.619 36.336 78.691 83.666 40.306 28.739 8:BLUPRC55 15.131 33.706 133.884 41.425 47.086 64.110 51.479 54.343 72.903 32.952 9:IVAPRC52 17.456 41.588 111.987 36.271 50.482 66.719 39.310 41.665 89.984 40.345 10:556PRB 106.681 160.048 17.195 64.338 125.092 134.391 35.831 36.384 220.193 136.164 11:NEO07 9.549 4.474 238.263 48.701 24.512 37.733 108.287 114.302 29.746 35.322 12:NEO08 0.000 6.330 182.788 22.677 17.166 26.365 78.807 84.116 49.697 39.390 13:NEO09 6.330 0.000 250.145 40.054 17.103 25.067 120.139 126.959 33.546 46.665 14:MONPRE 182.788 250.145 0.000 119.783 199.178 212.669 46.232 44.773 307.809 194.599 15:NWO29 22.677 40.054 119.783 0.000 15.339 16.502 55.295 60.777 77.433 61.700 16:NWO12 17.166 17.103 199.178 15.339 0.000 2.171 88.347 95.361 29.862 39.638 17:NWO11 26.365 25.067 212.669 16.502 2.171 0.000 105.086 112.837 38.884 56.749 18:SGWPR 78.807 120.139 46.232 55.295 88.347 105.086 0.000 .150 147.892 63.735 19:NWO13 84.116 126.959 44.773 60.777 95.361 112.837 .150 0.000 155.490 67.409 20:NWO25 49.697 33.546 307.809 77.433 29.862 38.884 147.892 155.490 0.000 28.697 21:SOWPRA 39.390 46.665 194.599 61.700 39.638 56.749 63.735 67.409 28.697 0.000 22:CANA264 67.850 93.252 148.769 87.904 86.150 109.437 35.765 36.251 95.100 20.383 23:NWO28 65.264 88.475 158.455 88.736 83.734 107.187 40.666 41.382 88.334 17.363 24:SABPRA 43.692 50.610 200.064 64.517 41.282 57.750 67.335 71.085 28.152 .318 25:NW027 62.604 81.508 175.031 88.751 77.382 100.215 49.863 51.172 72.784 10.874 26:NWO14 64.148 83.017 175.446 89.754 78.085 100.907 50.413 51.728 72.354 10.636 139

27:SIOPRA88 59.295 63.431 209.609 78.175 50.781 68.781 79.133 83.226 25.322 5.459 28:STOPRB 77.818 95.132 176.452 97.701 83.175 106.637 55.524 57.113 65.928 10.216 29:SEUPR95 64.221 64.363 229.078 83.689 51.282 68.473 92.289 96.934 20.021 9.194 30:NWO15 91.098 106.010 185.886 105.487 87.676 110.477 66.025 68.083 62.043 14.350 31:NWO26 160.409 196.773 168.755 181.965 185.280 219.455 66.775 64.508 173.221 65.786 32:CANA259 82.503 86.509 222.128 102.087 72.188 93.093 87.565 91.194 36.697 14.060 33:BIBPR 109.337 126.287 187.614 124.035 107.321 132.835 69.783 71.239 77.912 23.805 34:NWO16 105.352 120.519 191.516 120.164 101.622 126.367 71.549 73.322 71.248 21.675 35:PAKPRC98 81.290 79.687 249.120 108.107 71.317 92.480 102.989 107.194 30.197 17.711 36:CANA246 107.314 121.087 198.024 124.275 103.887 129.378 75.534 77.380 70.726 23.425 37:NWO17 96.399 107.865 200.206 111.056 88.781 111.466 75.363 77.824 57.339 17.285 38:NWO19 106.627 119.958 199.353 123.882 103.003 128.418 76.169 78.076 69.490 23.161 39:CANA263 110.624 128.117 187.067 124.552 108.217 133.479 68.743 70.099 79.880 24.221 40:CANA252 89.827 98.815 206.649 103.041 78.391 98.797 78.264 81.264 47.632 14.440 41:Redpr94 78.416 67.934 294.192 113.655 66.602 86.072 131.296 136.705 19.666 25.302 42:CANA255 114.425 132.081 189.494 129.210 112.682 138.595 70.221 71.438 83.359 26.392 43:CALPR 182.099 216.637 189.850 206.169 205.162 241.726 82.325 79.910 183.046 76.566 44:CANA245 177.262 210.628 191.405 201.995 199.564 235.666 81.387 79.143 176.899 72.826 45:LOOPR2 159.963 187.941 189.135 180.644 173.243 206.850 77.656 76.542 145.189 57.556 46:NWO24 134.743 156.162 190.380 152.097 138.644 168.169 73.949 74.222 107.887 39.291 47:JOHPRB 120.039 135.861 199.744 137.809 118.819 146.121 77.362 78.639 84.731 30.655 48:CANA265 186.313 219.631 198.094 212.344 209.312 246.631 87.320 84.930 184.014 78.952 49:LOMPR3 141.741 163.166 195.276 160.392 146.578 177.202 78.166 78.290 113.745 44.156 50:NWO20 156.235 180.553 198.789 178.368 166.807 200.084 82.202 81.568 134.142 54.777 51:NWO21 149.593 171.420 202.462 172.241 158.179 190.713 83.087 82.866 123.682 50.575 52:NWO23 153.775 176.513 204.731 178.439 164.951 198.439 84.514 84.011 130.731 53.940 53:NWO22 130.860 146.250 216.462 158.999 138.085 169.413 86.937 87.642 99.916 40.864 54:CANA253 153.751 175.803 216.738 185.772 170.522 205.340 88.921 88.158 136.165 56.552

140

Dissimilarity‎Matrix‎count‎……. Squared Euclidean Distance 23 24 25 26 27 28 29 30 31 32 1:NEO02 83.634 118.382 91.987 91.624 123.072 88.007 140.940 95.538 76.085 131.501 2:NEO03 38.944 54.715 42.204 42.280 60.496 42.317 73.201 50.447 58.559 70.620 3:NEO05 111.915 55.652 100.648 101.152 63.772 104.713 62.897 107.847 221.246 89.167 4:NEO01 91.382 128.311 100.937 100.707 132.906 97.537 151.352 105.611 83.969 141.707 5:NEO04 42.672 62.716 47.464 47.725 69.931 49.518 83.621 59.250 62.419 80.696 6:NEO10 92.808 132.027 103.538 103.645 138.164 103.202 157.041 113.318 90.090 148.196 7:NEO06 58.478 32.144 53.131 53.871 40.280 59.068 44.511 66.753 136.738 59.743 8:BLUPRC55 40.476 37.662 40.263 41.290 48.759 50.015 57.320 62.374 100.284 65.620 9:IVAPRC52 39.759 45.386 42.448 43.746 60.277 55.910 70.771 70.955 99.972 78.541 10:556PRB 119.528 140.947 130.737 131.426 154.090 137.122 170.687 148.218 158.974 174.340 11:NEO07 72.753 39.883 65.863 67.118 47.013 74.427 48.073 83.048 163.768 64.750 12:NEO08 65.264 43.692 62.604 64.148 59.295 77.818 64.221 91.098 160.409 82.503 13:NEO09 88.475 50.610 81.508 83.017 63.431 95.132 64.363 106.010 196.773 86.509 14:MONPRE 158.455 200.064 175.031 175.446 209.609 176.452 229.078 185.886 168.755 222.128 15:NWO29 88.736 64.517 88.751 89.754 78.175 97.701 83.689 105.487 181.965 102.087 16:NWO12 83.734 41.282 77.382 78.085 50.781 83.175 51.282 87.676 185.280 72.188 17:NWO11 107.187 57.750 100.215 100.907 68.781 106.637 68.473 110.477 219.455 93.093 18:SGWPR 40.666 67.335 49.863 50.413 79.133 55.524 92.289 66.025 66.775 87.565 19:NWO13 41.382 71.085 51.172 51.728 83.226 57.113 96.934 68.083 64.508 91.194 20:NWO25 88.334 28.152 72.784 72.354 25.322 65.928 20.021 62.043 173.221 36.697 21:SOWPRA 17.363 .318 10.874 10.636 5.459 10.216 9.194 14.350 65.786 14.060 22:CANA264 .282 21.951 2.612 2.963 36.978 11.779 47.946 25.128 26.281 43.711 23:NWO28 0.000 18.830 1.306 1.623 33.254 10.042 43.606 23.029 27.436 39.953 24:SABPRA 18.830 0.000 11.654 11.252 5.097 10.092 8.520 13.463 66.683 13.477 25:NW027 1.306 11.654 0.000 .041 23.564 5.745 32.247 16.349 30.467 29.813 26:NWO14 1.623 11.252 .041 0.000 22.578 4.946 31.127 14.991 29.609 28.479 141

27:SIOPRA88 33.254 5.097 23.564 22.578 0.000 11.993 .973 8.597 70.813 3.373 28:STOPRB 10.042 10.092 5.745 4.946 11.993 0.000 18.384 2.865 26.194 13.216 29:SEUPR95 43.606 8.520 32.247 31.127 .973 18.384 0.000 12.772 84.108 2.996 30:NWO15 23.029 13.463 16.349 14.991 8.597 2.865 12.772 0.000 33.853 6.645 31:NWO26 27.436 66.683 30.467 29.609 70.813 26.194 84.108 33.853 0.000 64.267 32:CANA259 39.953 13.477 29.813 28.479 3.373 13.216 2.996 6.645 64.267 0.000 33:BIBPR 28.368 23.137 22.227 20.689 15.181 5.767 19.522 1.516 27.651 9.364 34:NWO16 29.510 20.934 22.717 21.175 12.521 5.930 16.157 1.124 32.034 6.995 35:PAKPRC98 49.120 17.590 37.739 36.579 6.293 21.634 4.019 14.860 79.942 1.954 36:CANA246 32.592 22.953 25.481 23.945 12.945 7.762 15.912 2.297 34.418 6.092 37:NWO17 31.866 16.257 23.734 22.196 7.708 6.873 9.961 1.242 42.351 3.295 38:NWO19 32.959 22.703 25.720 24.190 12.559 7.955 15.356 2.397 35.405 5.676 39:CANA263 27.504 23.334 21.572 20.015 16.283 5.629 20.838 1.707 26.267 10.467 40:CANA252 33.946 12.929 24.830 23.281 5.327 8.491 6.721 2.523 51.562 2.465 41:Redpr94 67.838 25.367 54.009 53.083 14.019 39.510 9.083 32.719 113.827 11.032 42:CANA255 28.567 25.556 22.762 21.190 18.148 6.542 22.712 2.558 25.139 11.441 43:CALPR 39.792 77.364 41.695 40.512 76.335 32.380 88.302 36.817 2.339 64.959 44:CANA245 37.528 73.540 39.090 37.926 72.774 30.093 84.382 34.535 2.347 61.670 45:LOOPR2 36.240 57.862 34.913 33.459 51.787 20.537 60.502 20.312 7.228 40.241 46:NWO24 32.792 38.983 28.551 26.962 30.588 11.700 36.442 8.342 17.018 20.788 47:JOHPRB 34.784 30.217 28.449 26.843 19.830 10.152 23.406 4.827 28.796 10.763 48:CANA265 43.040 79.840 44.563 43.343 77.453 34.236 88.859 38.000 3.980 64.686 49:LOMPR3 36.919 44.035 32.719 31.096 34.298 14.831 39.987 10.972 17.758 22.987 50:NWO20 39.592 55.092 36.872 35.340 46.305 20.287 53.337 18.128 12.884 33.487 51:NWO21 39.969 50.883 36.332 34.788 40.910 18.881 46.895 15.749 16.868 28.079 52:NWO23 40.069 54.415 37.031 35.557 45.355 20.667 51.903 18.563 14.893 32.194 53:NWO22 39.106 41.528 33.762 32.404 30.665 16.945 34.515 13.417 27.211 18.528 54:CANA253 38.390 57.425 36.076 34.872 51.267 23.462 58.363 24.158 14.422 38.186

142

Dissimilarity‎Matrix‎count…….. Squared Euclidean Distance 34 35 36 37 38 39 40 41 42 43 1:NEO02 99.505 155.847 105.041 109.385 106.341 95.652 117.674 197.566 97.388 90.871 2:NEO03 56.267 85.387 59.881 60.491 60.491 55.563 64.809 111.993 57.571 72.739 3:NEO05 124.299 87.389 126.963 109.689 125.960 133.104 98.150 79.068 138.579 242.618 4:NEO01 109.310 166.105 114.684 119.458 115.988 105.301 128.028 208.233 106.936 99.028 5:NEO04 65.106 95.687 68.753 69.786 69.397 63.864 74.364 122.843 65.706 77.333 6:NEO10 117.342 171.313 122.381 127.258 123.616 113.375 135.758 211.859 114.932 106.028 7:NEO06 78.656 60.901 80.515 71.073 79.964 84.127 66.023 62.222 87.811 154.700 8:BLUPRC55 72.288 69.900 74.303 69.622 74.137 74.960 68.661 78.764 77.629 117.818 9:IVAPRC52 81.647 84.757 84.429 80.029 84.419 83.188 79.453 96.338 85.873 119.681 10:556PRB 158.613 195.907 165.236 161.772 166.102 156.776 163.372 228.808 160.357 183.839 11:NEO07 94.000 58.063 93.363 83.884 92.320 101.394 77.892 49.445 104.543 179.021 12:NEO08 105.352 81.290 107.314 96.399 106.627 110.624 89.827 78.416 114.425 182.099 13:NEO09 120.519 79.687 121.087 107.865 119.958 128.117 98.815 67.934 132.081 216.637 14:MONPRE 191.516 249.120 198.024 200.206 199.353 187.067 206.649 294.192 189.494 189.850 15:NWO29 120.164 108.107 124.275 111.056 123.882 124.552 103.041 113.655 129.210 206.169 16:NWO12 101.622 71.317 103.887 88.781 103.003 108.217 78.391 66.602 112.682 205.162 17:NWO11 126.367 92.480 129.378 111.466 128.418 133.479 98.797 86.072 138.595 241.726 18:SGWPR 71.549 102.989 75.534 75.363 76.169 68.743 78.264 131.296 70.221 82.325 19:NWO13 73.322 107.194 77.380 77.824 78.076 70.099 81.264 136.705 71.438 79.910 20:NWO25 71.248 30.197 70.726 57.339 69.490 79.880 47.632 19.666 83.359 183.046 21:SOWPRA 21.675 17.711 23.425 17.285 23.161 24.221 14.440 25.302 26.392 76.566 22:CANA264 31.582 53.844 34.912 34.457 35.348 29.165 36.830 74.099 30.208 39.023 23:NWO28 29.510 49.120 32.592 31.866 32.959 27.504 33.946 67.838 28.567 39.792 24:SABPRA 20.934 17.590 22.953 16.257 22.703 23.334 12.929 25.367 25.556 77.364 25:NW027 22.717 37.739 25.481 23.734 25.720 21.572 24.830 54.009 22.762 41.695 26:NWO14 21.175 36.579 23.945 22.196 24.190 20.015 23.281 53.083 21.190 40.512 143

27:SIOPRA88 12.521 6.293 12.945 7.708 12.559 16.283 5.327 14.019 18.148 76.335 28:STOPRB 5.930 21.634 7.762 6.873 7.955 5.629 8.491 39.510 6.542 32.380 29:SEUPR95 16.157 4.019 15.912 9.961 15.356 20.838 6.721 9.083 22.712 88.302 30:NWO15 1.124 14.860 2.297 1.242 2.397 1.707 2.523 32.719 2.558 36.817 31:NWO26 32.034 79.942 34.418 42.351 35.405 26.267 51.562 113.827 25.139 2.339 32:CANA259 6.995 1.954 6.092 3.295 5.676 10.467 2.465 11.032 11.441 64.959 33:BIBPR .184 17.904 .668 1.892 .834 .182 4.875 38.145 .380 27.606 34:NWO16 0.000 14.737 .327 .914 .416 .460 3.297 33.516 .770 32.014 35:PAKPRC98 14.737 0.000 12.578 9.514 11.898 19.357 8.020 4.056 20.155 79.365 36:CANA246 .327 12.578 0.000 1.110 .010 1.158 3.750 30.286 1.271 33.177 37:NWO17 .914 9.514 1.110 0.000 1.042 2.207 .814 25.107 2.795 42.792 38:NWO19 .416 11.898 .010 1.042 0.000 1.352 3.588 29.227 1.469 34.095 39:CANA263 .460 19.357 1.158 2.207 1.352 0.000 4.960 39.901 .105 26.225 40:CANA252 3.297 8.020 3.750 .814 3.588 4.960 0.000 21.349 5.831 52.904 41:Redpr94 33.516 4.056 30.286 25.107 29.227 39.901 21.349 0.000 40.868 113.159 42:CANA255 .770 20.155 1.271 2.795 1.469 .105 5.831 40.868 0.000 24.329 43:CALPR 32.014 79.365 33.177 42.792 34.095 26.225 52.904 113.159 24.329 0.000 44:CANA245 29.966 75.534 31.082 40.244 31.955 24.329 49.904 108.295 22.467 .085 45:LOOPR2 15.344 52.121 15.698 23.156 16.322 11.660 31.223 81.290 10.186 3.711 46:NWO24 4.519 30.372 4.593 8.944 4.934 2.775 14.264 54.390 1.962 14.056 47:JOHPRB 1.482 17.724 .939 3.472 1.028 1.338 7.068 37.144 .873 25.759 48:CANA265 32.347 78.124 32.985 43.067 33.846 26.693 53.363 111.120 24.548 .261 49:LOMPR3 6.159 32.115 5.779 10.974 6.107 4.366 16.854 56.268 3.268 13.475 50:NWO20 12.341 43.235 11.839 18.954 12.287 9.516 26.392 69.574 7.910 7.693 51:NWO21 9.933 36.522 9.041 15.538 9.372 7.797 22.320 60.789 6.266 11.313 52:NWO23 12.552 40.744 11.651 18.763 12.023 9.964 26.021 65.560 8.235 9.166 53:NWO22 8.035 23.067 6.140 11.306 6.171 7.497 16.644 41.406 6.129 21.528 54:CANA253 18.287 45.416 17.154 24.754 17.500 15.105 32.159 68.754 12.993 8.635

144

Dissimilarity‎Matrix‎count…….. Squared Euclidean Distance Case 45 46 47 48 49 50 51 52 53 54 1:NEO02 91.384 95.422 105.663 97.289 99.173 100.223 104.174 105.307 117.619 114.455 2:NEO03 65.944 61.417 63.732 77.547 65.533 70.057 70.545 72.363 73.263 77.938 3:NEO05 206.978 168.154 145.377 247.346 177.154 200.056 190.498 198.264 168.056 205.195 4:NEO01 99.914 104.445 114.912 105.434 107.973 108.685 112.620 113.654 126.009 122.639 5:NEO04 72.033 68.786 71.895 82.246 72.745 76.473 77.200 78.646 80.182 83.345 6:NEO10 107.595 112.398 122.563 112.342 115.823 116.129 119.832 120.562 131.796 128.100 7:NEO06 131.139 106.459 92.709 158.605 112.849 127.344 121.103 125.790 105.378 128.974 8:BLUPRC55 103.811 89.295 82.451 121.786 94.445 103.217 99.703 102.213 89.452 102.783 9:IVAPRC52 108.830 96.719 91.949 124.372 102.179 109.924 107.373 109.425 99.108 109.530 10:556PRB 176.818 169.084 171.561 192.358 175.808 183.995 184.982 188.397 190.782 198.522 11:NEO07 151.920 124.080 105.956 180.846 129.436 144.259 135.764 140.226 112.586 139.810 12:NEO08 159.963 134.743 120.039 186.313 141.741 156.235 149.593 153.775 130.860 153.751 13:NEO09 187.941 156.162 135.861 219.631 163.166 180.553 171.420 176.513 146.250 175.803 14:MONPRE 189.135 190.380 199.744 198.094 195.276 198.789 202.462 204.731 216.462 216.738 15:NWO29 180.644 152.097 137.809 212.344 160.392 178.368 172.241 178.439 158.999 185.772 16:NWO12 173.243 138.644 118.819 209.312 146.578 166.807 158.179 164.951 138.085 170.522 17:NWO11 206.850 168.169 146.121 246.631 177.202 200.084 190.713 198.439 169.413 205.340 18:SGWPR 77.656 73.949 77.362 87.320 78.166 82.202 83.087 84.514 86.937 88.921 19:NWO13 76.542 74.222 78.639 84.930 78.290 81.568 82.866 84.011 87.642 88.158 20:NWO25 145.189 107.887 84.731 184.014 113.745 134.142 123.682 130.731 99.916 136.165 21:SOWPRA 57.556 39.291 30.655 78.952 44.156 54.777 50.575 53.940 40.864 56.552 22:CANA264 36.467 33.942 36.792 42.504 38.107 40.415 41.185 41.179 41.399 39.709 23:NWO28 36.240 32.792 34.784 43.040 36.919 39.592 39.969 40.069 39.106 38.390 24:SABPRA 57.862 38.983 30.217 79.840 44.035 55.092 50.883 54.415 41.528 57.425 25:NW027 34.913 28.551 28.449 44.563 32.719 36.872 36.332 37.031 33.762 36.076 26:NWO14 33.459 26.962 26.843 43.343 31.096 35.340 34.788 35.557 32.404 34.872 27:SIOPRA88 51.787 30.588 19.830 77.453 34.298 46.305 40.910 45.355 30.665 51.267 145

28:STOPRB 20.537 11.700 10.152 34.236 14.831 20.287 18.881 20.667 16.945 23.462 29:SEUPR95 60.502 36.442 23.406 88.859 39.987 53.337 46.895 51.903 34.515 58.363 30:NWO15 20.312 8.342 4.827 38.000 10.972 18.128 15.749 18.563 13.417 24.158 31:NWO26 7.228 17.018 28.796 3.980 17.758 12.884 16.868 14.893 27.211 14.422 32:CANA259 40.241 20.788 10.763 64.686 22.987 33.487 28.079 32.194 18.528 38.186 33:BIBPR 12.417 3.142 1.272 28.028 4.677 10.055 8.147 10.471 7.383 16.008 34:NWO16 15.344 4.519 1.482 32.347 6.159 12.341 9.933 12.552 8.035 18.287 35:PAKPRC98 52.121 30.372 17.724 78.124 32.115 43.235 36.522 40.744 23.067 45.416 36:CANA246 15.698 4.593 .939 32.985 5.779 11.839 9.041 11.651 6.140 17.154 37:NWO17 23.156 8.944 3.472 43.067 10.974 18.954 15.538 18.763 11.306 24.754 38:NWO19 16.322 4.934 1.028 33.846 6.107 12.287 9.372 12.023 6.171 17.500 39:CANA263 11.660 2.775 1.338 26.693 4.366 9.516 7.797 9.964 7.497 15.105 40:CANA252 31.223 14.264 7.068 53.363 16.854 26.392 22.320 26.021 16.644 32.159 41:Redpr94 81.290 54.390 37.144 111.120 56.268 69.574 60.789 65.560 41.406 68.754 42:CANA255 10.186 1.962 .873 24.548 3.268 7.910 6.266 8.235 6.129 12.993 43:CALPR 3.711 14.056 25.759 .261 13.475 7.693 11.313 9.166 21.528 8.635 44:CANA245 3.205 12.781 23.892 .289 12.291 6.871 10.242 8.178 19.767 7.443 45:LOOPR2 0.000 3.550 10.360 3.446 3.100 .946 2.410 1.841 8.714 3.698 46:NWO24 3.550 0.000 1.889 13.744 .253 2.055 1.484 2.461 3.736 6.208 47:JOHPRB 10.360 1.889 0.000 25.106 2.379 6.715 4.504 6.451 3.071 10.990 48:CANA265 3.446 13.744 25.106 0.000 12.724 6.691 10.042 7.814 19.618 6.931 49:LOMPR3 3.100 .253 2.379 12.724 0.000 1.192 .594 1.388 2.840 4.874 50:NWO20 .946 2.055 6.715 6.691 1.192 0.000 .376 .213 4.382 2.213 51:NWO21 2.410 1.484 4.504 10.042 .594 .376 0.000 .216 2.326 2.457 52:NWO23 1.841 2.461 6.451 7.814 1.388 .213 .216 0.000 3.079 1.373 53:NWO22 8.714 3.736 3.071 19.618 2.840 4.382 2.326 3.079 0.000 4.627 54:CANA253 3.698 6.208 10.990 6.931 4.874 2.213 2.457 1.373 4.627 0.000

146

Appendix C SPSS output for HCA after changing the order of sites entering the analysis

PROXIMITIES tmn01 tmn02 tmn03 tmn04 tmn05 tmn06 tmn07 tmn08 tmn09 tmn10 tmn11 tmn12 tmx01 tmx02 tmx03 tmx04 tmx05 tmx06 tmx07 tmx08 tmx09 tmx10 tmx11 tmx12 cmi01 cmi02 cmi03 cmi04 cmi05 cmi06 cmi07 cmi08 cmi09 cmi10 cmi11 cmi12 /VIEW=CASE /MEASURE=SEUCLID /PRINT NONE /ID=Chron_1 /STANDARDIZE=VARIABLE Z.

Case Processing Summarya Cases Valid Missing Total N Percent N Percent N Percent 54 100.0% 0 0.0% 54 100.0% a. Squared Euclidean Distance used

CLUSTER /METHOD COMPLETE /ID=Chron_1 /PRINT SCHEDULE /PRINT DISTANCE /PLOT DENDROGRAM. Complete Linkage Agglomeration Schedule Stage Cluster Combined Coefficients Stage Cluster First Appears Next Stage Cluster 1 Cluster 2 Cluster 1 Cluster 2 1 23 24 .010 0 0 16 2 52 54 .041 0 0 25 3 21 27 .085 0 0 10 4 40 42 .105 0 0 13 5 3 4 .150 0 0 46 6 32 37 .184 0 0 13 7 17 18 .213 0 0 12 8 25 30 .253 0 0 26 9 48 49 .282 0 0 25 10 16 21 .289 0 3 27 11 47 53 .318 0 0 40 12 17 20 .376 7 0 23 13 32 40 .770 6 4 19 14 41 50 .814 0 0 24 15 38 39 .973 0 0 31 16 23 35 1.028 1 0 19 147

17 6 10 1.148 0 0 30 18 5 11 1.240 0 0 41 19 23 32 1.482 16 13 32 20 26 34 1.954 0 0 31 21 28 45 2.171 0 0 38 22 8 33 2.397 0 0 48 23 17 19 2.410 12 0 26 24 41 43 2.523 14 0 32 25 48 52 2.963 9 2 40 26 17 25 3.550 23 8 34 27 16 29 3.980 10 0 43 28 1 12 4.474 0 0 37 29 14 22 4.627 0 0 34 30 6 31 4.984 17 0 41 31 26 38 6.293 20 15 36 32 23 41 7.068 19 24 35 33 7 15 7.990 0 0 37 34 14 17 8.714 29 26 43 35 23 46 10.152 32 0 45 36 26 36 14.019 31 0 44 37 1 7 14.725 28 33 42 38 13 28 16.502 0 21 47 39 2 51 17.195 0 0 49 40 47 48 21.951 11 25 45 41 5 6 22.409 18 30 46 42 1 9 25.011 37 0 47 43 14 16 27.211 34 27 50 44 26 44 36.697 36 0 51 45 23 47 36.830 35 40 50 46 3 5 42.088 5 41 49 47 1 13 48.701 42 38 48 48 1 8 66.719 47 22 51 49 2 3 76.263 39 46 52 50 14 23 79.840 43 45 52 51 1 26 113.655 48 44 53 52 2 14 216.738 49 50 53 53 1 2 307.809 51 52 0

148

149

Appendix D SPSS output for HCA using a matrix of 54 x 28. Analysis was run by removing randomly selected 8 (~20%) climatic variables

PROXIMITIES tmn01 tmn02 tmn04 tmn06 tmn07 tmn08 tmn09 tmn10 tmn11 tmn12 tmx03 tmx05 tmx06 tmx07 tmx09 tmx10 tmx11 tmx12 cmi01 cmi02 cmi03 cmi04 cmi05 cmi06 cmi08 cmi09 cmi10 cmi12 /VIEW=CASE /MEASURE=SEUCLID /PRINT NONE /ID=Chron_1 /STANDARDIZE=VARIABLE Z Case Processing Summarya Cases Valid Missing Total N Percent N Percent N Percent 54 100.0% 0 0.0% 54 100.0% a. Squared Euclidean Distance used

CLUSTER /METHOD COMPLETE /ID=Chron_1 /PRINT SCHEDULE /PRINT DISTANCE /PLOT DENDROGRAM VICICLE. Complete Linkage Agglomeration Schedule Stage Cluster Combined Coefficients Stage Cluster First Appears Next Stage Cluster 1 Cluster 2 Cluster 1 Cluster 2 1 36 38 .007 0 0 14 2 25 26 .028 0 0 25 3 43 44 .069 0 0 11 4 39 42 .088 0 0 18 5 18 19 .110 0 0 47 6 33 34 .127 0 0 14 7 51 52 .157 0 0 12 8 21 24 .191 0 0 35 9 46 49 .208 0 0 27 10 22 23 .211 0 0 25 11 43 48 .234 3 0 28 12 50 51 .257 0 7 22 13 1 4 .562 0 0 26 14 33 36 .591 6 1 19 15 37 40 .637 0 0 24 16 27 29 .654 0 0 31

150

17 2 5 .802 0 0 42 18 39 47 1.068 4 0 19 19 33 39 1.201 14 18 32 20 8 9 1.379 0 0 46 21 32 35 1.402 0 0 31 22 45 50 1.669 0 12 27 23 16 17 1.714 0 0 38 24 30 37 1.865 0 15 32 25 22 25 1.936 10 2 39 26 1 6 2.123 13 0 42 27 45 46 2.410 22 9 34 28 31 43 3.201 0 11 43 29 53 54 3.383 0 0 34 30 11 13 3.993 0 0 37 31 27 32 5.107 16 21 36 32 30 33 5.327 24 19 44 33 7 12 5.818 0 0 37 34 45 53 5.995 27 29 43 35 21 28 7.677 8 0 39 36 27 41 10.547 31 0 45 37 7 11 12.133 33 30 41 38 15 16 12.727 0 23 48 39 21 22 13.867 35 25 44 40 10 14 14.728 0 0 50 41 3 7 14.968 0 37 46 42 1 2 15.974 26 17 47 43 31 45 18.658 28 34 49 44 21 30 27.377 39 32 49 45 20 27 27.673 0 36 51 46 3 8 36.928 41 20 48 47 1 18 37.164 42 5 50 48 3 15 54.002 46 38 51 49 21 31 60.068 44 43 52 50 1 10 61.247 47 40 52 51 3 20 90.945 48 45 53 52 1 21 179.820 50 49 53 53 1 3 232.938 52 51 0

151

152

Appendix E SPSS output for HCA validation using split half method a) Subset – 1

PROXIMITIES tmn01 tmn02 tmn03 tmn04 tmn05 tmn06 tmn07 tmn08 tmn09 tmn10 tmn11 tmn12 tmx01 tmx02 tmx03 tmx04 tmx05 tmx06 tmx07 tmx08 tmx09 tmx10 tmx11 tmx12 cmi01 cmi02 cmi03 cmi04 cmi05 cmi06 cmi07 cmi08 cmi09 cmi10 cmi11 cmi12 /VIEW=CASE /MEASURE=CORRELATION /PRINT NONE /ID=Chron_1 /STANDARDIZE=VARIABLE Z. Case Processing Summarya Cases Valid Missing Total N Percent N Percent N Percent 27 100.0% 0 0.0% 27 100.0% a. Correlation between Vectors of Values used

CLUSTER /METHOD COMPLETE /ID=Chron_1 /PRINT SCHEDULE /PRINT DISTANCE /PLOT DENDROGRAM.

Agglomeration Schedule Stage Cluster Combined Coefficients Stage Cluster First Appears Next Stage Cluster 1 Cluster 2 Cluster 1 Cluster 2

1 22 23 1.000 0 0 2 2 22 25 .998 1 0 11 3 14 15 .998 0 0 23 4 11 12 .998 0 0 14 5 19 20 .995 0 0 10 6 26 27 .994 0 0 19 7 21 24 .980 0 0 10 8 5 6 .977 0 0 13 9 7 8 .971 0 0 13 10 19 21 .970 5 7 15 11 18 22 .969 0 2 19 12 1 3 .965 0 0 18 13 5 7 .914 8 9 17 14 9 11 .905 0 4 22 153

15 17 19 .890 0 10 21 16 13 16 .847 0 0 24 17 2 5 .827 0 13 20 18 1 4 .814 12 0 22 19 18 26 .763 11 6 21 20 2 10 .688 17 0 24 21 17 18 .404 15 19 23 22 1 9 .293 18 14 25 23 14 17 .047 3 21 26 24 2 13 -.099 20 16 25 25 1 2 -.818 22 24 26 26 1 14 -.963 25 23 0

154 b) Subset – 2

PROXIMITIES tmn01 tmn02 tmn03 tmn04 tmn05 tmn06 tmn07 tmn08 tmn09 tmn10 tmn11 tmn12 tmx01 tmx02 tmx03 tmx04 tmx05 tmx06 tmx07 tmx08 tmx09 tmx10 tmx11 tmx12 cmi01 cmi02 cmi03 cmi04 cmi05 cmi06 cmi07 cmi08 cmi09 cmi10 cmi11 cmi12 /VIEW=CASE /MEASURE=CORRELATION /PRINT NONE /ID=Chron_1 /STANDARDIZE=VARIABLE Z.

Case Processing Summarya Cases Valid Missing Total N Percent N Percent N Percent 27 100.0% 0 0.0% 27 100.0% a. Correlation between Vectors of Values used

CLUSTER /METHOD COMPLETE /ID=Chron_1 /PRINT SCHEDULE /PRINT DISTANCE /PLOT DENDROGRAM. Agglomeration Schedule Stage Cluster Combined Coefficients Stage Cluster First Appears Next Stage Cluster 1 Cluster 2 Cluster 1 Cluster 2

1 24 25 .999 0 0 6 2 6 7 .996 0 0 19 3 9 10 .993 0 0 21 4 1 2 .984 0 0 17 5 8 11 .983 0 0 19 6 22 24 .981 0 1 9 7 14 16 .980 0 0 16 8 15 21 .980 0 0 13 9 22 27 .977 6 0 15 10 3 4 .969 0 0 20 11 23 26 .959 0 0 15 12 17 19 .952 0 0 18 13 15 18 .936 8 0 18 14 13 20 .924 0 0 16 15 22 23 .914 9 11 22 16 13 14 .868 14 7 23 17 1 5 .740 4 0 20 155

18 15 17 .698 13 12 22 19 6 8 .628 2 5 23 20 1 3 .559 17 10 25 21 9 12 .508 3 0 24 22 15 22 .113 18 15 24 23 6 13 -.042 19 16 25 24 9 15 -.570 21 22 26 25 1 6 -.770 20 23 26 26 1 9 -.966 25 24 0

156

Appendix F Characteristics of 37 red pine site chronologies

Site ID Lon Lat Eco Ele Dated Period Data Source (Region) (°W) (°N) Region (meters) Series (years) NEO02 79.4795 47.1430 4E 265 19 1871-2011 This Study (NEO) (141) NEO03 79.6741 47.8596 4E 251 25 1832-2011 This Study (NEO) (180) NEO01 79.9769 46.9703 4E 305 14 1745-2011 This Study (NEO) (267) NEO04 80.1959 47.6161 4E 306 25 1838-2011 This Study (NEO) (174) 556PRB 83.4500 46.8700 4E 444 33 1837-2002 Girardin et al., (NEO) (166) 2006 MONPRE 84.6500 47.2200 4E 220 37 1773-2003 Girardin et al., (NEO) (231) 2006 SGWPR 88.7300 48.4500 4W 223 41 1805-2001 Girardin et al., (NWO) (197) 2006 NWO13 88.7862 48.4012 4W 222 30 1808-2011 This Study (NWO) (204) SOWPRA 91.1700 49.5300 3W 413 39 1640-2001 Girardin et al., (NWO) (362) 2006 CANA264 91.2310 48.6064 4W 418 75 1768-2003 St. George et (NWO) (236) al., 2008 NWO28 91.3192 48.7094 4W 425 35 1903-2011 This Study (NWO) (109) SABPRA 91.5500 49.4500 3W 452 35 1828-2000 Girardin et al., (NWO) (173) 2006 NWO27 91.7038 48.8978 4W 450 23 1877-2011 This Study (NWO) (135) NWO14 91.8133 48.9090 4W 453 33 1868-2011 This Study (NWO) (144) STOPRB 92.2300 49.3500 4S 405 37 1791-2001 Girardin et al., (NWO) (211) 2006 NWO15 92.8079 49.6525 4S 392 40 1900-2011 This Study (NWO) (112) NWO26 93.0746 48.7980 4S 356 39 1891-2011 This Study (NWO) (121) BIBPR 93.3300 49.7800 4S 362 39 1808-2001 Girardin et al., (NWO) (194) 2006 NWO16 93.3581 49.8515 4S 366 40 1890-2011 This Study (NWO) (122) CANA246 93.4870 50.0347 4S 348 46 1776-2004 St. George et (NWO) (229) al., 2008 NWO17 93.5000 49.9253 4S 390 41 1933-2011 This Study (NWO) (79) NWO19 93.5059 50.0596 4S 348 37 1778-2011 This Study (NWO) (234) CANA263 93.6700 49.6444 4S 389 27 1750-2003 St. George et (NWO) (254) al., 2008

157

CANA252 93.7390 49.8446 4S 433 34 1759-2004 St. George et (NWO) (246) al., 2008 CANA255 93.8800 49.6444 4S 391 19 1875-2003 St. George et (NWO) (129) al., 2008 CALPR 93.9200 49.0500 5S 338 17 1851-2001 Girardin et al., (NWO) (151) 2006 CANA245 94.0099 49.0484 5S 349 32 1836-2004 St. George et (NWO) (169) al., 2008 LOOPR2 94.0500 49.4200 5S 334 44 1772-2003 Girardin et al., (NWO) (230) 2006 NWO24 94.0883 49.6410 5S 359 30 1876-2011 This Study (NWO) (136) JOHPRB 94.1200 49.9200 4S 363 41 1792-2001 Girardin et al., (NWO) (210) 2006 CANA265 94.1610 49.1441 5S 331 34 1854-2004 St. George et (NWO) (151) al., 2008 LONPR3 94.2800 49.7200 5S 348 40 1830-2001 Girardin et al., (NWO) (172) 2006 NWO20 94.4069 49.6394 5S 333 40 1899-2011 This Study (NWO) (113) NWO21 94.4999 49.7795 5S 334 40 1851-2011 This Study (NWO) (161) NWO23 94.5731 49.7247 5S 333 36 1896-2011 This Study (NWO) (116) NWO22 94.5919 50.1393 4S 330 36 1827-2011 This Study (NWO) (185) CANA253 94.8700 49.6444 5S 346 38 1775-2004 St. George et (NWO) (230) al., 2008 Notes: Site‎IDs‎for‎“This‎Study”‎represent‎a‎combination‎of‎Ontario‎Region‎(NEO/NWO)‎and‎numerical‎order‎of‎ sampling.‎For‎nomenclature‎of‎“archived‎data”‎site‎IDs,‎refer‎to‎the‎respective‎data‎source Abbreviations used are Lon (longitude), Lat (latitude), and Ele (elevation).

158

Appendix G Red pine chronologies and the corresponding sampling size. Bold red line indicates the running average

NEO02 NEO03

NEO01 NEO04

159

NWO22 556PRB

MONPRE SGWPR

160

NWO13 SOWPRA

CANA264 NWO28

161

SABPRA NWO27

NWO14 STOPRB

162

NWO15 NWO26

BIBPR NWO16

163

CANA246 NWO17

NWO19 CANA263

164

CANA252 CANA255

CALPR CANA245

165

LOOPR2 NWO24

JOHPRB CANA265

166

LONPR3 NWO20

NWO21 NWO23

167

CANA253

168

Appendix H Response coefficients of red pine growth – seasonal precipitation relationships. Significant coefficients, at 95% confidence interval, are shown in bold Italic.

Site Lon (°W) Lat (°N) Eco Region Prior Summer Winter Spring Summer NEO02 79.4795 47.1430 4E 0.1205 0.0540 0.1308 0.2159 NEO03 79.6741 47.8596 4E 0.0800 -0.0543 0.0233 0.0930 NEO01 79.9769 46.9703 4E 0.2902 0.2133 -0.0354 0.2621 NEO04 80.1959 47.6161 4E 0.2793 -0.0299 0.0575 0.2918 556PRB 83.4500 46.8700 4E 0.0962 0.0149 -0.0585 -0.0334 MONPRE 84.6500 47.2200 4E 0.2060 -0.0282 0.2639 0.0724 SGWPR 88.7300 48.4500 4W 0.1960 -0.0902 0.0800 0.1111 NWO13 88.7862 48.4012 4W 0.2678 -0.0098 0.2033 0.1951 SOWPRA 91.1700 49.5300 3W 0.1739 0.1544 0.0533 -0.0082 CANA264 91.2310 48.6064 4W 0.0581 -0.0828 0.2110 -0.0251 NWO28 91.3192 48.7094 4W 0.1732 -0.1006 0.1032 0.0222 SABPRA 91.5500 49.4500 3W 0.1518 -0.1224 0.1885 0.0399 NWO27 91.7038 48.8978 4W 0.2623 -0.1553 0.1023 0.1081 NWO14 91.8133 48.9090 4W 0.1564 -0.0201 0.0873 0.0134 STOPRB 92.2300 49.3500 4S 0.1262 0.0246 0.1294 0.0083 NWO15 92.8079 49.6525 4S 0.0453 -0.2600 0.1965 0.1007 NWO26 93.0746 48.7980 4S 0.1182 -0.1925 0.1937 0.1861 BIBPR 93.3300 49.7800 4S 0.0790 -0.0610 0.2569 0.2247 NWO16 93.3581 49.8515 4S 0.0156 -0.2052 0.1335 0.3479

169

CANA246 93.4870 50.0347 4S 0.0874 0.0670 0.2951 0.1434 NWO17 93.5000 49.9253 4S 0.0975 -0.2898 0.0866 0.3297 NWO19 93.5059 50.0596 4S 0.1179 0.0558 0.2922 0.1692 CANA263 93.6700 49.6444 4S 0.0824 -0.1291 0.1678 0.2296 CANA252 93.7390 49.8446 4S 0.1227 -0.0621 -0.0622 -0.0219 CANA255 93.8800 49.6444 4S 0.0883 0.0059 0.2748 0.0292 CALPR 93.9200 49.0500 5S 0.0924 0.0147 0.3287 0.0033 CANA245 94.0099 49.0484 5S 0.1582 0.0381 0.2703 0.0124 LOOPR2 94.0500 49.4200 5S 0.0454 0.0774 0.2309 -0.0270 NWO24 94.0883 49.6410 5S 0.1714 -0.1379 0.1904 0.1074 JOHPRB 94.1200 49.9200 4S 0.0997 0.0448 0.2434 0.1021 CANA265 94.1610 49.1441 5S -0.0153 -0.0509 0.2581 -0.0144 LONPR3 94.2800 49.7200 5S -0.0371 -0.2006 0.1882 0.1364 NWO20 94.4069 49.6394 5S 0.2405 -0.0522 0.2547 0.1204 NWO21 94.4999 49.7795 5S 0.1281 -0.0319 0.1893 0.1078 NWO23 94.5731 49.7247 5S 0.1156 -0.0487 0.1101 0.2401 NWO22 94.5919 50.1393 4S 0.0196 0.0000 0.1748 0.1252 CANA253 94.8700 49.6444 5S 0.0363 -0.0580 0.2598 0.0924

170

Appendix I Response coefficients of red pine growth – seasonal Tmin relationships. Significant coefficients, at 95% confidence interval, are shown in bold Italic.

Site Lon (°W) Lat (°N) Eco Region Prior Summer Winter Spring Summer NEO02 79.4795 47.1430 4E -0.0520 0.0394 -0.0801 -0.0639 NEO03 79.6741 47.8596 4E -0.0603 0.1643 -0.0331 -0.0804 NEO01 79.9769 46.9703 4E -0.1471 -0.0993 -0.0545 -0.0910 NEO04 80.1959 47.6161 4E -0.1596 0.0034 -0.0623 -0.1125 556PRB 83.4500 46.8700 4E -0.0522 0.1441 0.0065 -0.0004 MONPRE 84.6500 47.2200 4E -0.2130 0.0197 0.0460 -0.2401 SGWPR 88.7300 48.4500 4W -0.0305 0.0110 0.0574 -0.0440 NWO13 88.7862 48.4012 4W -0.0828 0.0282 0.0273 -0.0200 SOWPRA 91.1700 49.5300 3W -0.0560 0.0553 0.0153 -0.0450 CANA264 91.2310 48.6064 4W 0.0571 -0.0302 0.0408 0.0808 NWO28 91.3192 48.7094 4W 0.0511 -0.0158 0.0886 -0.0163 SABPRA 91.5500 49.4500 3W -0.0403 -0.0408 0.0878 -0.0220 NWO27 91.7038 48.8978 4W -0.0835 -0.0717 0.0516 0.0803 NWO14 91.8133 48.9090 4W -0.0718 -0.0028 0.0732 0.0170 STOPRB 92.2300 49.3500 4S -0.0804 0.0333 0.1075 0.0574 NWO15 92.8079 49.6525 4S -0.0712 0.0203 0.0293 0.0533 NWO26 93.0746 48.7980 4S -0.0658 0.0360 0.0373 0.0409 BIBPR 93.3300 49.7800 4S -0.1152 -0.0185 0.0213 0.0522 NWO16 93.3581 49.8515 4S -0.1379 0.0190 0.0478 0.0005

171

CANA246 93.4870 50.0347 4S -0.1562 -0.0024 0.0423 -0.0452 NWO17 93.5000 49.9253 4S -0.1391 0.0414 -0.0264 0.0424 NWO19 93.5059 50.0596 4S -0.1630 0.0045 0.0328 -0.0424 CANA263 93.6700 49.6444 4S -0.1667 0.0269 0.0045 0.0279 CANA252 93.7390 49.8446 4S -0.0027 -0.0219 0.0365 0.0707 CANA255 93.8800 49.6444 4S -0.2039 0.0481 0.0933 -0.0231 CALPR 93.9200 49.0500 5S -0.1472 0.0418 0.1224 -0.0052 CANA245 94.0099 49.0484 5S -0.1431 0.0108 0.0819 0.0085 LOOPR2 94.0500 49.4200 5S -0.0307 0.0354 0.0628 0.0613 NWO24 94.0883 49.6410 5S -0.0985 0.0655 0.0557 0.0266 JOHPRB 94.1200 49.9200 4S -0.1105 0.0096 0.0231 0.0777 CANA265 94.1610 49.1441 5S -0.0121 -0.0101 0.1249 0.1173 LONPR3 94.2800 49.7200 5S -0.0419 -0.0552 -0.0188 0.1310 NWO20 94.4069 49.6394 5S -0.1352 0.0038 -0.0126 0.0187 NWO21 94.4999 49.7795 5S -0.1059 0.0043 -0.0157 0.0424 NWO23 94.5731 49.7247 5S -0.0651 0.0322 -0.0157 0.0440 NWO22 94.5919 50.1393 4S -0.0663 -0.0120 0.0340 0.1061 CANA253 94.8700 49.6444 5S -0.1050 0.0239 0.0366 0.0704

172

Appendix J Response coefficients of red pine growth – seasonal Tmax relationships. Significant coefficients, at 95% confidence interval, are shown in bold Italic.

Site Lon (°W) Lat (°N) Eco Region Prior Summer Winter Spring Summer NEO02 79.4795 47.1430 4E -0.0149 0.0566 -0.0987 -0.0973 NEO03 79.6741 47.8596 4E -0.1751 0.1170 -0.0656 -0.2187 NEO01 79.9769 46.9703 4E 0.0108 0.1025 0.0943 0.1133 NEO04 80.1959 47.6161 4E -0.0615 0.0988 -0.0036 0.0004 556PRB 83.4500 46.8700 4E -0.1070 0.1263 -0.0007 -0.0176 MONPRE 84.6500 47.2200 4E 0.0025 0.1479 0.1143 -0.0371 SGWPR 88.7300 48.4500 4W -0.0373 0.0387 0.0617 -0.0488 NWO13 88.7862 48.4012 4W -0.0626 0.0803 0.0411 -0.0425 SOWPRA 91.1700 49.5300 3W -0.2165 0.0203 -0.0353 -0.1515 CANA264 91.2310 48.6064 4W -0.0856 -0.0785 -0.0708 -0.0433 NWO28 91.3192 48.7094 4W -0.0202 -0.0190 0.0600 -0.0974 SABPRA 91.5500 49.4500 3W -0.0479 0.0104 0.0732 -0.0439 NWO27 91.7038 48.8978 4W -0.1911 -0.0452 0.0613 0.0133 NWO14 91.8133 48.9090 4W -0.1943 -0.0178 0.0303 -0.0661 STOPRB 92.2300 49.3500 4S -0.2831 -0.0360 0.0030 -0.1054 NWO15 92.8079 49.6525 4S -0.1499 0.0225 -0.0315 -0.0206 NWO26 93.0746 48.7980 4S -0.0977 0.0694 0.0030 -0.0613 BIBPR 93.3300 49.7800 4S -0.2057 -0.0097 -0.0651 -0.0265 NWO16 93.3581 49.8515 4S -0.1275 0.0728 0.0555 -0.0392

173

CANA246 93.4870 50.0347 4S -0.2317 0.0237 -0.0403 -0.1137 NWO17 93.5000 49.9253 4S -0.1684 0.0561 -0.0374 -0.0837 NWO19 93.5059 50.0596 4S -0.2916 0.0114 -0.0689 -0.1606 CANA263 93.6700 49.6444 4S -0.1847 0.0807 -0.0164 0.0419 CANA252 93.7390 49.8446 4S -0.1005 -0.0559 0.0491 0.0668 CANA255 93.8800 49.6444 4S -0.2887 0.0875 0.0352 -0.0245 CALPR 93.9200 49.0500 5S -0.2168 0.0708 0.0154 -0.0435 CANA245 94.0099 49.0484 5S -0.2684 0.0081 -0.0297 -0.0653 LOOPR2 94.0500 49.4200 5S -0.2261 -0.0358 -0.1042 -0.1440 NWO24 94.0883 49.6410 5S -0.2563 0.0522 -0.0252 -0.0695 JOHPRB 94.1200 49.9200 4S -0.2618 -0.0085 -0.0944 -0.0064 CANA265 94.1610 49.1441 5S -0.1306 -0.0440 -0.0131 -0.0198 LONPR3 94.2800 49.7200 5S -0.1711 -0.1082 -0.1380 -0.0153 NWO20 94.4069 49.6394 5S -0.2243 0.0369 -0.0823 0.0284 NWO21 94.4999 49.7795 5S -0.2020 0.0033 -0.0834 0.0041 NWO23 94.5731 49.7247 5S -0.0898 0.0496 -0.0455 0.0189 NWO22 94.5919 50.1393 4S -0.1892 -0.0599 -0.0730 -0.0077 CANA253 94.8700 49.6444 5S -0.2193 0.0109 -0.0841 -0.0552

174