Variation and Integration of Ecophysiological Traits across Scales in Tropical and Temperate : Patterns, Drivers and Consequences

Item Type text; Electronic Dissertation

Authors Messier, Julie

Publisher The University of Arizona.

Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.

Download date 10/10/2021 23:02:47

Link to Item http://hdl.handle.net/10150/594556

VARIATION AND INTEGRATION OF ECOPHYSIOLOGICAL TRAITS ACROSS SCALES IN TROPICAL AND TEMPERATE TREES: PATTERNS, DRIVERS AND CONSEQUENCES

by

Julie Messier

______Copyright © Julie Messier 2015

A Dissertation Submitted to the Faculty of the

DEPARTMENT OF ECOLOGY AND EVOLUTIONARY BIOLOGY

In Partial Fulfillment of the Requirements

For the Degree of

DOCTOR OF PHILOSOPHY

In the Graduate College

THE UNIVERSITY OF ARIZONA

2015 THE UNIVERSITY OF ARIZONA GRADUATE COLLEGE

As members of the Dissertation Committee, we certify that we have read the dissertation prepared by Julie Messier, titled “Variation and integration of ecophysiological traits across scales in tropical and temperate trees: patterns, drivers and consequences”, and recommend that it be accepted as fulfilling the dissertation requirement for the Degree of Doctor of Philosophy.

______Date: November 9th 2015 Brian J. Enquist

______Date: November 9th 2015 Brian J. McGill

______Date: November 9th 2015 Martin J. Lechowicz

______Date: November 9th 2015 Judith L. Bronstein

______Date: November 9th 2015 Katrina M. Dlugosch

Final approval and acceptance of this dissertation is contingent upon the candidate’s submission of the final copies of the dissertation to the Graduate College.

I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement.

______Date: November 9th 2015 Dissertation Director: Brian J. Enquist

______Date: November 9th 2015 Dissertation Director: Brian J. McGill

2 STATEMENT BY AUTHOR

This dissertation has been submitted in partial fulfillment of the requirements for an advanced degree at the University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.

Brief quotations from this dissertation are allowable without special permission, provided that an accurate acknowledgement of the source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the copyright holder.

SIGNED: Julie Messier

3 ACKNOWLEDGEMENTS

If I have seen further, it is by standing on the shoulders of giants. ~ Bernard de Chartres

I owe the accomplishment of this work to many wonderful people. Above all I am indebted to my three advisors, Brian McGill, Brian Enquist and Martin Lechowicz, exceptional scientists who have given me the freedom to pursue my ideas, taught me so much, and managed to gracefully navigate a complicated 3-way advising situation. Brian McGill is a truly creative thinker who showed me how to think outside the box, turn science into good stories and find the essence of complex results. He has also showed me how to keep faith in the importance and relevance of my work despite tides or rejections. Marty, an in- depth thinker, careful scientist and exceptional writer with an encyclopedic knowledge of physiology and of the ecological literature, showed me how to design thoughtful experiments, think thoroughly through problems and be prudent and skeptical with the meaning of data and results. His wisdom in understanding how one’s work fits within the greater context of science and how science itself is a lifestyle where one must balance research in the greater context of their life, continues to uplift and impress me. Brian Enquist, a scientist who dreams large and makes those dreams happen, has always encouraged me to try new things and throw myself at new challenges with confidence. His passion, energy and unwavering positive outlook on challenges really does lift mountains. I have also been blessed with the support of two scientists of excellence, Judie Bronstein and Katrina Dlugosch. Judie, with a deep mastery of fundamental ecological and evolutionary concepts is also a big picture thinker. Her sharp, critical mind can cut through the first layers of one’s research to see the big picture. She has been invaluable in reminding me to find the broader questions and messages in my research that scientists in different fields can relate to. Katrina with her unique expertise at the intersection of plant ecology and evolutionary biology has been instrumental in shaping the direction of my dissertation. She introduced me to some aspects of evolutionary biology that are now a major focus of my research interests. Katrina, your comprehensive exam question on quantitative genetics was by far the hardest, but it has been a pivotal, eye-opening milestone. She has also helped me navigate the professional and ‘business’ side of academia.

Cyrille Violle a mentor, colleague and friend, has played a crucial role in my dissertation work in many ways. Our numerous conversations have not only helped shape

4 my thoughts on ecology and evolution, but also acted as a catalyst in the writing of my dissertation chapters. He has pushed me through the finish line by offering me a writing haven in Montpellier for seven months in the last year of my doctoral research. Cyrille is selflessly generous with his time, sharing his ideas and financial support.

I thank the funding sources – Canada’s NSERC and the Organization for American States - that have allowed me to come to the rich and stimulating research environment of EEB at the University of Arizona. The Chateaubriand fellowship from the French Embassy in the United States has also allowed me the time to finish writing my dissertation, to form new work relationships in Europe and make wonderful friends.

I am indebted to experts in their field who have generously and patiently shared their knowledge. They have taught me so much and allowed all of this work to happen: Dr. Erica Bigio made all the ring work possible; Dr. Louise Comas taught me how to collect fine roots of trees in the field; Dr. David Killick showed me how to operate high-precision microscopes to document anatomy and trusted me with the expensive equipment; and John Sperry and Mel Tyree guided my efforts to measure sap transport. None of this work would have been possible without the steady and dedicated efforts of a number of field and lab assistants: Anke Roth, Émilie Lavoie, Natasha Salter, Andréanne Ferland, Carol Mordy, Ricardo Cossio, Sierra Kaszubinski, Sanga Shir, Surbhi Patel, Margretta Murphy, John Lacson, Kevin Wong, Sarah Schwenck, Anjeanette McKay, Casey Knoks, Meghan Iacueuilli, Shahrzad Badie, Irene Liang, Jordyn Celaya and Mélisanne Gagnon. David Maneli and the McGill Gault crew have made possible all the field logistics, lending me their hands, chainsaws, vehicles and ArcGIS skills when in need. My present and past lab, Vanessa Buzzard, Lindsey Sloat, Brian Maitner, Alex Brummer, Kyle Martins , Amanda Henderson, Ben Blonder, Sean Michaletz, Daniel Guaderrama, Kathy Hulshof and Christine Lamanna for the shared excitement over nature, math and ecology, the intellectual stimulation, the innumerable laughs and for standing by my side through trials and tribulations. You have made the daily grind enjoyable.

I also want to acknowledge the profound influence of my early mentor, Daniel Kneeshaw, on shaping my career path. He took me under his wing and his generosity, patience and faith launched me forward in the scientific world when I had everything to learn. I need to thank early mentors and role models: thanks to Alain Bombardier, my high School math teacher whose recognition of my stubborn perseverance bolstered me up; to

5 Farley Morris, my high school English teacher who taught me to think critically and question the world around me; to Will Boshuck my calculus Math teacher who infected me with a love of math; and to Sharon Rutherford, my college Biology teacher who held her students to high standards with confidence, whose enthusiastic teaching of biology steered my studies and whom I continue to look up to.

Last, I need to thank the people close to me who supported me through the ups and downs of the long doctoral journey. Loren Albert and Pacifica Sommers, my accountabili- buddies whose daily support through the last year of my PhD made everything seem manageable. To my mother, father and brother who never doubted for an instant my ability to succeed, to my best friends Maryse and Gerardo who reminded me to stay silly with my serious, to my boyfriend Cres who cooked for me more meals than I can count and whose supportive love gave me the freedom to work as hard and travel as much as I needed and wanted. Thanks to good friends along the journey who made Tucson and Montpellier home: Erica, Maite, Kit, Carrie, Doug, Leo, Dan H., Nancy, EFR, Jim, Briana, Jesse, Luis, Megan, Mike, Aleix, Dave K., Scott, Mark, Stacey, Jolina, Jill, Thanh, Alex, Crista, Chen, Catherine, Dave S, Marielle, Max, Ty, Simon, Grégoire, Helena, Pierre, Cassie and Alex S.L.

All of you have inspired and uplifted me. Thank you.

6 DEDICATION

This work is dedicated to the past, present and future women in science.

7 TABLE OF CONTENTS

ABSTRACT …………………….…………...…………………………...………….…………… 9

INTRODUCTION …...……………...……………………………………...……….….….....… 11

PRESENT WORK ………………………………………………….…………………………... 18

REFERENCES ………………………………………………………………………………..… 24

APPENDIX A: Trait variation and integration across scales: Is the economic spectrum present at local scales? ……………………...……………………...…………………………… 31

APPENDIX B: Interspecific integration of trait dimensions at local scales: plant architecture is at the center of the phenotypic network ……………………………………………………… 86

APPENDIX C: Variable intrapopulation phenotypic integration structure in temperate trees …………………...………………...……………………….…..…………….. 135

APPENDIX D: Testing central assumptions of trait-based ecology: Traits are good predictors of plant performance when phenotypic complexity and individual variation are taken into account ………………………………………...……………………………………..……….. 196

APPENDIX E: Detailed Protocols ……………………………..………..……………………..241

8 ABSTRACT

The overarching goal of my dissertation is to explore the potential and limits of a trait-based approach to plant ecology. Together, the different studies presented here address two explicit and implicit foundational assumptions underpinning the trait-based approach: (1) that the correlation patterns and biological significance of traits transfer across scales and (2) that the phenotypic complexity of can accurately be synthesized into a few meaningful traits to study their ecology. Moreover, the last chapter focuses on a third key assumption: (3) that traits are strong predictors of plant performance (Shipley et al. n.d.). I examine these assumptions by exploring multivariate patterns of phenotypic variation and integration across different ecological scales (e.g., individuals, populations, species) while explicitly considering the phenotypic complexity of trees, both in terms of their multidimensional and integrated nature. Two themes thus permeate this body of work: scales and phenotypic complexity. Much of what we know about the relationships among key traits comes from species-scale studies. Trait variation at smaller scales are often interpreted in the context of these interspecific relationships, but it is not clear that interspecific patterns observed at global scales apply to smaller scales. Moreover, although plants are complex, integrated organisms with intricate relationships among their traits, single traits are often studied and interpreted without considering the rest of the phenotype. Yet, examining individual traits outside of their phenotypic context might provide limited insight or be misleading. To address these shortcomings, this body of work examines multidimensional patterns of trait variation and correlation across ecological scales. It uses (1) a set of six ecophysiological leaf traits from mature trees in a lowland tropical rainforest, and (2) a set of twenty leaf, root, stem, branch and whole-plant ecophysiological traits from deciduous saplings in a temperate forest. The combination of our findings point to three main conclusions: (i) local interspecific and intra-population trait integration structures differ from each other and from

9 the global interspecific patterns reported in the literature, such that global-scale interspecific patterns cannot readily be transferred to more local scales; (ii) considering the complexity of the plant phenotype provides better insights into ecological patterns and processes than what we can learn from considering individual or a handful of traits; and (iii) traits strongly affect individual plant performance, although there is no relationship between a species’ trait correlation structure and its environmental niche, which suggests that there are multiple alternative optimal phenotypes in a given environment.

10 INTRODUCTION

The appeal of traits in ecology The overarching goal of my dissertation is to explore the potential and limits of a trait-based approach to plant ecology, a rapidly expanding approach that has taken an increasingly central role in plant ecology in the last decade (Westoby and Wright 2006, Weiher et al. 2011, Kattge et al. 2011). Functional traits are broadly defined as those traits affecting the fitness of organisms (Violle et al. 2007). The appeal of this approach also resides in its promise of generality - providing a common basis for the comparison of individuals, populations and species from different phylogenetic histories and environments (Shipley 2007) - and of predictive ability – being directly linked to environmental gradients and organismal properties at different ecological scales (e.g. population demographics, species’ life-history strategies, community assembly, ecosystem processes) (Arnold 1983, McGill et al. 2006). As the field is rapidly expanding, it is becoming increasingly important to verify the fundamental assumptions on which the trait-based approach rests (Shipley et al. n.d.). Together, the work presented here explores two of those explicit and implicit foundational assumptions: (1) the correlation patterns, biological significance and predictive ability of traits transfer across scales, (2) the phenotypic complexity of plants can be synthesized into a few meaningful traits. The themes of scales and phenotypic complexity thus permeate all the elements of this body of work. The last chapter also tests a third key assumption: that (3) traits are strong predictors of plant performance (Shipley et al. n.d.). I examine these assumptions by exploring patterns of phenotypic variation and integration across different ecological scales and explicitly considering the phenotypic complexity of trees in terms of its integrated and multidimensional nature.

Assumption 1 - Trait relationships and their biological significance bridges scales Since traits describe general properties of all plants that can be assessed at any ecological and organizational level, from a leaf metamer to a whole ecosystem, traits have

11 been heralded as a way to move across scales (Shipley 2007), a central challenge of ecology (Levin 1992). Indeed, traits are being used to study a variety of ecological patterns and processes occurring at different scales, such as community assembly (Cornwell et al. 2006, Ackerly and Cornwell 2007, Kraft et al. 2008, Cornwell and Ackerly 2009, Jung et al. 2010, Violle et al. 2012), ecosystem services (Diaz and Cabido 2001, Lavorel and Garnier 2002, Garnier et al. 2004, Diaz et al. 2007), coexistence (Kraft et al. 2015) and evolutionary processes (Silvertown et al. 2001, 2006), species abundances (Mouillot et al. 2007) and biogeography (Violle et al. 2014). However, the assumption that trait-trait and trait-process relationships hold across all scales has poorly been tested. The trait-based approach spawned from comparative ecology (Bradshaw 1987, Calow 1987, Grime et al. 1988), a discipline focused of understanding species-level patterns of similarities and differences (Grime 1965, 1979, Grime et al. 1988, Hendry 1993, Cornelissen et al. 2003). Thus, most of our theoretical and empirical understanding of trait-trait and trait- environment relationships comes from broad species-level studies. For example, three traits (specific leaf area, maximum tree height and seed size), have been highlighted as good proxies of species strategies (Westoby 1998) and this trilogy is typically used as a set of key indicators traits. However, while these traits describe global differences in species strategies, the processes of interest at other scales might affect other traits more strongly. Furthermore, it is unclear whether global-scale interspecific trait correlation structures hold at smaller organizational and spatial scales dominated by local-scale processes. The forces shaping phenotype among species globally differ from those acting at other scales, so that we should not necessarily expect them to concur (trait dimensions, groups of globally correlated traits reflecting different aspects of species strategies have been shaped by global scale processes over evolutionary timescales). There are surprisingly few studies comparing plant phenotypic covariance structure across scales and most of these focus on size-related floral traits (Armbruster 1991, Armbruster et al. 2004). In fact, the literature holds some evidence that intraspecific and local interspecific trait correlation

12 patterns may differ from the global interspecific patterns (Wright and Sutton-Grier 2012, Funk and Cornwell 2013, Grady et al. 2013, Richardson et al. 2013, Edwards et al. 2014, Kraft et al. 2015, Niinemets 2015). To properly interpret results from studies not carried global interspecific scales, it is thus crucial to clarify whether ecophysiological trait relationships and their ecological significance hold across scales.

Assumption 2 - A few traits adequately summarize phenotypic complexity A second assumption that is implicit to the trait-based approach is that a few traits can be assessed to summarize the response of the processes of interest. A few key traits were identified early in the history of this approach as depicting species strategies (Westoby et al. 2002) and most studies tend to consider only these few individual traits without considering the phenotypic context in which they occur. Yet, the whole-organism phenotype is an integrated expression of multiple traits, with coordinated development, adaptation and response to stimuli (Murren 2002, Pigliucci 2003, Merila and Bjorklund 2004, Blows 2007). Any given trait is thus not free to vary independently but is part of a web of interlinked properties (Raup and Michelson 1965, Lewontin 1977, Gould and Lewontin 1979, Cheverud 1982). Thus, because a few traits can synthesize global differences in plant strategies does not necessarily mean that a few traits, let alone the same few traits adequately summarize ecological response to distribution, assembly, coexistence, or any other processes acting at other scales. In fact, recent studies have shown that taking into account multiple integrated phenotypic traits greatly improves our predictions of performance and fitness (Laughlin 2014, Kleyer and Minden 2015, Kraft et al. 2015). While this has long been acknowledged by ecologists (Levins and Lewontin 1985, Bonser 2006, Craine 2009, Milla and Reich 2011), considering the phenotypic integration structure of organisms has not become an inherent part of the ecological approach in the same way that it has pervaded the toolbox of the

13 evolutionary biologist (Phillips and Arnold 1989, Lynch and Walsh 1998, Arnold et al. 2008, Walsh and Blows 2009). The body of work presented here explicitly takes into account the complexity of the vegetative plant body by examining the relationships among a large number of ecophysiological traits representing different basic physiological functions.

Assumption 3 - Traits predict performance The interest of studying phenotypic traits critically relies on their close association with organismal performance (Lavorel & Garnier 2002, Violle et al. 2007). A review paper by Shipley and colleagues (n.d.) highlights that while this is an implicit assumption underlying all trait-based studies, surprisingly few studies have empirically established these relationships (but see Poorter et al. 2008, Paine et al. 2015). In fact, in an echo to the appeal by Laughlin and Messier (2015), they stress that we should really be designing studies examining how the strength of trait-fitness correlations changes across environments. There are two general ways to link traits to fitness (Laughlin and Messier 2015). Most studies use a ‘likelihood’ approach to indirectly link traits to fitness. One establishes the probability densities of traits in communities and assumes that the most abundant trait value (increased likelihood) are those that are most adaptive. This is a reasonable assumption under equilibrium conditions, but communities are temporally and spatially dynamic such that currently observed traits might not be in equilibrium with the current environmental conditions. A second, more direct way to link plant phenotype to fitness is to measure fitness components, such as growth and mortality rates, in their environment. In plant ecology, this approach has recently been implemented in a handful of studies (Rüger et al. 2012, Adler et al. 2014, Lasky et al. 2014) but needs further robust testing to assess which traits and combinations of traits affect different aspects of plant fitness in which environmental conditions.

14 It is important to note that the expectation of a clear relationship between traits to performance is stems from the single-optimum paradigm. In other words a single phenotype has the highest performance in a given environment and the adaptive landscape forms a single peak. If this is true and strong natural selection is at play, we should observe strong environment-trait relationships, whether it be across species with different environmental niches, or across plots under different environmental conditions. However, an alternative paradigm has been put forward by a different authors from different fields using different analytical approaches (Niklas 1994a, 1999, Marks and Lechowicz 2006, Shoval et al. 2012). This paradigm suggests that many alternative phenotypes might perform well in a given environment. The phenotypic adaptive landscape would then show multiple rolling hills. This would result in many different phenotypes observed in a given environment and this pattern would be indistinguishable from random noise using the conventional statistical tools such as regressions, which operate under the single optimum paradigm. Surprisingly, to our knowledge no empirical study has tested these two alternative paradigms. While I do not directly address the environmental gradient-phenotype-fitness relationship in this work, I do examine their components separately in the same dataset: I examine first the relationship between a species’ environmental niche and its phenotypic structure and later assess the relationships between the phenotype and tree performance of individual trees. Together, these results allow me to draw some tentative inferences on the shape of the phenotype-performance relationship across environments.

The experimental approach of this work This body of work examines the assumptions described above by exploring multivariate patterns of phenotypic variation and integration across different ecological scales while explicitly considering the phenotypic complexity of trees in terms of its integrated and multidimensional nature. I chose to focus on plant properties affecting the ecological filtering process of community assembly, i.e., the natural selection process at

15 contemporary timescales by which a given environment retains those individuals best adapted to it. Thus, the sampling design for Appendices B, C and D is based at the individual level. The individual level is a key level for community ecology because the filtering process occurs on individuals. In fact, evidence is accruing that intraspecific trait variation is important for understanding community-level processes (Clark et al. 2004; Stubbs and Wilson 2004; Clark et al. 2007; Clark 2010; Clark et al. 2010; Lichti and Murphy 2010). Further, a tree-level sampling allows me to assess patterns of trait variation at three scales: the individual, intra-population and interspecific scales. Further, in contrast to the broad ‘functional’ traits definition that encompasses any trait affecting the fitness of organisms (Violle et al. 2007), here I focus on those traits affecting the growth and survival stage of individual plants in their environment. I therefore focus on non-reproductive morphological, physiological or chemical traits reflecting the basic and ubiquitous physiological function that all plans must fulfill in order to grow and survive, namely resource acquisition, resource transport, resource conservation and storage, defense and mechanical support (Niklas 1992, 1994b, Garnier and Navas 2013). To address the three assumptions discuss above, Appendix A examines patterns of variation across ecological scales, from to communities, Appendix B examines species- level trait integration structure in locally co-occurring species, Appendix C examines intra- population level similarities and differences in trait integration structure and Appendix D examines how individual-level traits predict tree performance. Specifically, some of the main questions I ask are: How much does trait correlation strength constrain the ability of individual traits to respond independently to drivers of variation? What are the main differences among traits in how they vary across scales? In locally co-occurring species, are trait dimensions independent from each other, or do they covary? Are there some fundamental relationships shaping intrapopulation trait correlations? Conversely, how much are differences in intrapopulation trait correlation structures related

16 to environmental niche differences and phylogeny? How well do phenotypic traits predict performance and do we find evidence of environmental filtering acting on traits?

17 PRESENT WORK

To explore the limits of trait-based approach to plant ecology, the present work examines similarities and differences of trait relationships across scales and addresses the multidimensional and integrated nature of the phenotype by simultaneously considering the correlation structure of a wide range of tree ecophysiological traits. Appendix A examines patterns of trait variation across six ecological scales, from leaves to communities and asks: “Are correlated traits constrained to respond similarly to drivers of variation?” Appendix B examines the relationships among traits and trait dimensions in locally co-occurring species and asks: “Are trait dimensions independent?” Appendix C examines the similarities and differences in intrapopulation trait correlation structures and asks: “What shapes intrapopulation trait correlations: constraints, phylogeny or the environment?” Finally, Appendix D examines some fundamental assumption of trait-based ecology and asks: “How well do traits predict plant performance?” Three main messages emerge from this body of work: (i) relationships among functional trait differ within species, among species locally and among species globally such that global-scale interspecific patterns cannot readily be transferred to more local scales; (ii) embracing the complexity of the plant phenotype provides different insights into plant ecology than considering a handful of traits; and (iii) individual-level traits affect plant performance, but there is no simple relationships between traits, performance, and environment.

First, the ensemble of my results suggest that global patterns of trait correlation, such as the leaf economic spectrum, are not necessarily present among species at local scales, nor among individuals within a population. A literature review shows that the majority of studies conducted at small scales fail to find the leaf economic spectrum among species (Appendix A). Moreover, we find that the moderate strength of correlation among the leaf economic spectrum traits allows them to retain to a large degree the ability to vary independently of

18 each other at different ecological scales (Appendix A). Further, the worldwide leaf economic spectrum (Reich et al. 1997, Wright et al. 2004) and the wood spectrum (Baas et al. 2004, Chave et al. 2009) were not found among locally-occurring species (Appendix B), nor within the populations (Appendix C). Second, taking into consideration phenotypic complexity offers unique insights into plant ecology. We find that at local scales, interspecific traits relationships form an integrated network (Appendix B) instead of orthogonal, independent dimensions. Further, we find that trait-trait interactions and inter-individual variation must be taken into account to accurately predict plant performance (Appendix D). Last, while we did not directly test this assumption, the results of this work are consistent with the hypothesis that there may be multiple alternative phenotypic solutions to optimize performance within an environment (Niklas 1994a, Marks and Lechowicz 2006, Shoval et al. 2012): we find that traits and trait-trait interactions plays a major role in predicting plant performance (Appendix D), yet species-level phenotypic integration structure is independent of environmental niche (Appendix C). This indicates that while traits do affect performance, the relationships between the phenotype and the environment does not appear to be linear. The methods, results, and conclusions of this research program are presented in manuscripts joined to this dissertation in Appendices A-D. Appendix E presents the detailed protocols used to collect the data. All work presented herein is a result of multi-authored manuscripts on which I was the primary author and contributor. The following is a summary of the main findings of each manuscript.

First, in Appendix A “Traits variation and integration across scales: Is the leaf economic spectrum present at local scales?”, I examine how much the strength of trait correlations constrains patterns of trait variation across scales and discuss the implication of these results for our ability to generalize global-scale patterns to local scales . Although the

19 concept of trait-based strategy dimensions, have been developed at global scales, trait variation at the local community scale is often interpreted in the context of global-scale patterns. Here I argue that a research priority should be to determine whether global-scale trait relationships hold at more local scales. I review recent literature suggesting that the leaf economic spectrum may be scale-dependent, and present a case study exploring the interplay between trait correlation and variation in leaf traits across ecological scales. In contrast to expectations, I find that correlation strength between two traits does not predict how similarly they respond to different drivers of variation. Instead, leaf economic spectrum traits largely retain the ability to respond independently to different drivers of phenotypic variation. Recent literature and the results of this study suggest that it is unclear if the leaf economic spectrum concept readily applies at local-scales. This is consistent with suggestions that drivers underpinning their correlation mainly act at broad spatial and taxonomic scales. I argue that it is pressing to determine at which scales the leaf economic spectrum is present. Clarifying how the interplay between trait variation and correlation changes through ecological scales is necessary for us to make the most of the emerging trait toolbox and will provide insight into the forces structuring functional diversity.

Second, in Appendix B “Interspecific integration of trait dimensions at local scales: plant architecture is at the center of the phenotypic network”, I examine how three independent global-scale patterns of trait correlations, called trait dimensions, are related at local scales. The plant phenotype is structured by the interplay between evolution and ecophysiological and biophysical tradeoffs and constraints. Patterns of trait covariation have identified closely integrated groups of traits forming trait dimensions. However, trait dimensions have typically been examined separately so that their degree of interdependence remains unclear. In this study I examine the relationships among these trait dimensions by considering simultaneously multiple traits from these dimensions. Using empirical data

20 gathered on individuals of 24 locally coexisting tree species in a temperate forest, I examine the correlation structure of 20 leaf, branch, stem and root traits. These traits are involved in three well-established trait dimensions that characterize different ecophysiological functions: resource acquisition, hydraulic, mechanical support and architecture. Using both strict and broad definitions of these trait dimensions, I test whether the phenotype is organized along distinct trait dimensions. The results indicate that patterns of trait covariation do not segregate into clear trait dimensions. Expected trait relationships that define the three trait dimensions (the leaf economic spectrum, the wood spectrum and Corner’s rules) are present but do not dominate the phenotypic structure. Instead, many traits from distinct trait dimensions are correlated together into a trait network. These results suggest that trait dimensions described across broad taxonomic and environmental scales are not independent among locally coexisting species. Importantly, two architectural traits appear as most central to the phenotypic network. This indicates that architecture may be central to resource, mechanical and hydraulic integration. These results suggest that local and global drivers and patterns of phenotypic integration may be distinct. This questions the use of the concept of trait dimensions at local scales. As an alternate approach, I suggest that a network view might better portray the phenotype locally. Such an approach offers the promise of identifying core traits that reflect the multiple trade-off and constraints shaping the phenotype locally.

Third, in Appendix C “Variable intrapopulation phenotypic integration structure in temperate deciduous trees”, I focus on similarities and differences in multivariate trait correlation structures at the intrapopulation level. Generalities and differences in intrapopulation phenotypic integration provide a lens into the fundamental constraints and trade-offs shaping trait relationship. Yet, it remains largely unknown whether there are general patterns of functional trait correlation within populations. Here I explore similarities and differences in intrapopulation phenotypic integration structures in locally coexisting tree

21 species. Using Mantel tests, I compare the correlation structure of several key functional traits within and among temperate tree species. I ask: (i) Are some intrapopulation pairwise trait correlations shared across all species to form a common ‘tree template’? and ; (ii) are species differences in their phenotypic integration structure due to differences in environmental niche or phylogenetic distance? I find that intrapopulation integration structures are weak and distinct from each other. Three trait pairs, LMA- δ13C, KS-Lumen Fraction and KBRANCH-Lumen Fraction do show consistent intrapopulation correlations across species. Interestingly, I find that the variation in intrapopulation phenotypic integration among species is not related to environmental or phylogenetic differences. The species-specific nature of intrapopulation trait integration structure questions the existence of fundamental constraints and trade-offs among the ecophysiological functions measured here and instead suggests a tremendous flexibility in plant design. Further, the lack of relationship between phenotypic integration structure and environmental niche is consistent with two contrasting interpretations: (i) phenotypic integration affects plant performance but alternative designs can produce equivalently effective solutions to optimizing performance; or (ii) local patterns of phenotypic integration within populations are decoupled from local plant performance. Distinguishing between these two alternate possibilities should be an important goal on the research agenda of ecologists since it would clarify the relationships between phenotypic integration structure and plant performance and shed light on unimodal or multimodal the shape of the phenotypic adaptive landscape.

Fourth, in Appendix D, “Testing central assumptions of trait-based ecology: Traits are good predictors of plant performance when phenotypic complexity and individual variation are taken into account”, I test the central assumption of trait-based ecology that traits are good predictors of plant performance. While trait-based plant ecology has taken an increasingly important role in plant ecology, some of its fundamental assumptions remain

22 poorly tested. Here I examine three key assumptions: (1) that traits are good predictors of plant performance, (2) that traits are better indicators of plant performance than species identity, and (3) that environmental filtering operates on traits. I measure relative growth rate (RGR) of individual tree saplings from 24 co-occurring temperate tree species and use multivariate linear regressions and variance partitioning to predict RGR based on plant traits, environment, size, age and species identity. I find that (1) traits explains 53% of variance in RGR, (2) that adding species does not improve the models but instead that species affect RGR through traits, and (3) that the environmental affects RGR largely through their interaction with traits. Further, trait-trait, environment-environment and trait-environment interactions are important in the models, suggesting that phenotypic and environmental context-dependence are key to the accurate predictive ability of a trait-based approach. Together, traits, age and environmental variables explain 75% of RGR variance, albeit with a large number of variables. These findings largely support the assumptions of trait-based ecology, but suggest that taking phenotypic complexity into account might be necessary to produce accurate predictions.

Last, given the number of the large number of ecophysiological traits and environmental variables sampled and the lack of space to describe their collection methods in each manuscript, Appendix D described in detail the protocols used to collect the data used in Appendices B, C and D.

23 REFERENCES

Ackerly, D. D., and W. K. Cornwell. 2007. A trait-based approach to community assembly: partitioning of species trait values into within- and among-community components. Ecology letters 10:135–45. Adler, P. B., R. Salguero-gómez, A. Compagnoni, J. S. Hsu, J. Ray-mukherjee, P. B. Adler, R. Salguero-gómez, A. Compagnoni, J. S. Hsu, and J. Ray-mukherjee. 2014. Functional traits explain variation in plant life history strategies. Proceedings of the National Academy of Sciences 111:10019–10019. Armbruster, S. W. 1991. Multilevel analysis of morphometric data from natural plant populations: insights into ontogenetic, genetic, and selective correlations in Dalechampia scandens. Evolution 45:1229–1244. Armbruster, S. W., C. Pélabon, T. F. Hanson, and C. P. H. Mulder. 2004. Floral Integration, Modularity, and Accuracy. Pages 23–49 in M. Pigliucci and K. Preston, editors. Phenotypic Integration: Studying the Ecology and Evolution of Comlpex Phenotypes. Oxford University Press, Oxford. Arnold, S. J. 1983. Morphology, performance and fitness. American zoologist 23:347–361. Arnold, S. J., R. Bürger, P. a. Hohenlohe, B. C. Ajie, and A. G. Jones. 2008. Understanding the evolution and stability of the G-matrix. Evolution 62:2451–2461. Baas, P., F. W. Ewers, S. D. Davis, and E. A. Wheeler. 2004. Evolution of xylem physiology. Pages 273–295 in A. R. Hemsley and I. Poole, editors. The evolution of plant physiology. Elsevier Academic Press, London. Blows, M. W. 2007. A tale of two matrices: Multivariate approaches in evolutionary biology. Journal of Evolutionary Biology 20:1–8. Bonser, S. P. 2006. Form defining function: interpreting leaf functional variability in integrated plant phenotypes. Oikos 114:187–190. Bradshaw. 1987. Functional ecology = comparative ecology? Functional Ecology 1:71–72. Calow, P. 1987. Towards a definition of functional ecology. Functional Ecology 1:57–61. Chave, J., D. Coomes, S. Jansen, S. L. Lewis, N. G. Swenson, and A. E. Zanne. 2009. Towards a worldwide wood economics spectrum. Ecology letters 12:351–66. Cheverud, J. M. 1982. Phenotypic, genetic and environmental morphological integration in the cranium. Evolution 36:499–516. Cornelissen, J. H. C., S. Lavorel, E. Garnier, S. Díaz, N. Buchmann, D. E. Gurvich, P. B. Reich, H. Ter Steege, H. D. Morgan, M. G. a. Van Der Heijden, J. G. Pausas, and H. Poorter. 2003. A handbook of protocols for standardised and easy measurement of plant

24 functional traits worldwide. Australian Journal of Botany 51:335. Cornwell, W. K., and D. D. Ackerly. 2009. Community assembly and shifts in plant trait distributions across an environmental gradient in coastal California. Ecological Monographs 79:109–126. Cornwell, W. K., L. D. W. Schwilk, and D. D. Ackerly. 2006. A trait-based test for habitat filtering: convex hull volume. Ecology 87:1465–71. Craine, J. M. 2009. Resource Strategies of Wild Plants. Princeton University Press., Woodstock. Diaz, S., and M. Cabido. 2001. Vive la difference: plant functional diversity matters to ecosystem processes. Trends in Ecology & Evolution 16:646–655. Diaz, S., S. Lavorel, F. De Bello, F. Quétier, K. Grigulis, and M. Robson. 2007. Incorporating plant functional diversity effects in ecosystem service assessments. Proceedings of the National Academy of Sciences 104:20684–20689. Edwards, E. J., D. S. Chatelet, L. Sack, and M. J. Donoghue. 2014. Leaf life span and the leaf economic spectrum in the context of whole plant architecture. Journal of Ecology 102:328–336. Funk, J. L., and W. K. Cornwell. 2013. Leaf traits within communities: Context may affect the mapping of traits to function. Ecology 94:1893–1897. Garnier, E., J. Cortez, B. Georges, M.-L. Navas, C. Roumet, M. Debussche, G. Laurent, A. Blanchard, D. Aubry, A. Bellmann, C. Neill, and J.-P. Toussaint. 2004. Plant functional markers capture ecosystem properties. Ecology 85:2630–2637. Garnier, E., and M.-L. Navas. 2013. Diversité fonctionnelle des plantes: traits des organismes, structure des communautés, propriétés des écosystèmes. De Boeck Supérieur. Gould, S. J., and R. C. Lewontin. 1979. The spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist programme. Proceedings of the Royal Society of London 205:581–98. Grady, K. C., D. C. Laughlin, S. M. Ferrier, T. E. Kolb, S. C. Hart, G. J. Allan, and T. G. Whitham. 2013. Conservative leaf economic traits correlate with fast growth of genotypes of a foundation riparian species near the thermal maximum extent of its geographic range. Functional Ecology 27:428–438. Grime, J. 1965. Comparative experiments as a key to the ecology of flowering plants. Ecology 46:513–525. Grime, J. 1979. Plant strategies and vegetation processes. John Wiley & Sons, New York. Grime, J., J. Hodgson, and R. Hunt. 1988. Comparative plant ecology: a functional approach

25 to common British species. Unwin Hyman, London. Hendry, G. 1993. Methods in comparative plant ecology: a laboratory manual. Chapman and Hall, London. Jung, V., C. Violle, C. Mondy, L. Hoffmann, and S. Muller. 2010. Intraspecific variability and trait-based community assembly. Journal of Ecology 98:1134–1140. Kattge, J., S. Díaz, S. Lavorel, I. C. Prentice, P. Leadley, G. Bönisch, E. Garnier, M. Westoby, P. B. Reich, I. J. Wright, J. H. C. Cornelissen, C. Violle, S. P. Harrison, P. M. Van BODEGOM, M. Reichstein, B. J. Enquist, N. a. Soudzilovskaia, D. D. Ackerly, M. Anand, O. Atkin, M. Bahn, T. R. Baker, D. Baldocchi, R. Bekker, C. C. Blanco, B. Blonder, W. J. Bond, R. Bradstock, D. E. Bunker, F. Casanoves, J. Cavender-Bares, J. Q. Chambers, F. S. Chapin III, J. Chave, D. Coomes, W. K. Cornwell, J. M. Craine, B. H. Dobrin, L. Duarte, W. Durka, J. Elser, G. Esser, M. Estiarte, W. F. Fagan, J. Fang, F. Fernández-Méndez, A. Fidelis, B. Finegan, O. Flores, H. Ford, D. Frank, G. T. Freschet, N. M. Fyllas, R. V. Gallagher, W. a. Green, a. G. Gutierrez, T. Hickler, S. I. Higgins, J. G. Hodgson, A. Jalili, S. Jansen, C. a. Joly, A. J. Kerkhoff, D. Kirkup, K. Kitajima, M. Kleyer, S. Klotz, J. M. H. Knops, K. Kramer, I. Kühn, H. Kurokawa, D. C. Laughlin, T. D. Lee, M. Leishman, F. Lens, T. Lenz, S. L. Lewis, J. Lloyd, J. Llusià, F. Louault, S. Ma, M. D. Mahecha, P. Manning, T. Massad, B. E. Medlyn, J. Messier, A. T. Moles, S. C. Müller, K. Nadrowski, S. Naeem, Ü. Niinemets, S. Nöllert, A. Nüske, R. Ogaya, J. Oleksyn, V. G. Onipchenko, Y. Onoda, J. Ordoñez, G. Overbeck, W. a. Ozinga, S. Patiño, S. Paula, J. G. Pausas, J. Peñuelas, O. L. Phillips, V. Pillar, H. Poorter, L. Poorter, P. Poschlod, A. Prinzing, R. Proulx, A. Rammig, S. Reinsch, B. Reu, L. Sack, B. Salgado-Negret, J. Sardans, S. Shiodera, B. Shipley, A. Siefert, E. Sosinski, J.-F. Soussana, E. Swaine, N. G. Swenson, K. Thompson, P. Thornton, M. Waldram, E. Weiher, M. White, S. White, S. J. Wright, B. Yguel, S. Zaehle, a. E. Zanne, and C. Wirth. 2011. TRY - a global database of plant traits. Global Change Biology 17:2905–2935. Kleyer, M., and V. Minden. 2015. Why functional ecology should consider all plant organs: An allocation-based perspective. Basic and Applied Ecology 16:1–9. Kraft, N. J. B., O. Godoy, and J. M. Levine. 2015. Plant functional traits and the multidimensional nature of species coexistence. Proceedings of the National Academy of Sciences 112:797–802. Kraft, N. J. B., R. Valencia, and D. D. Ackerly. 2008. Functional traits and niche-based tree community assembly in an amazonian forest. Science 322:580–582. Lasky, J. R., M. Uriarte, V. K. Boukili, and R. L. Chazdon. 2014. Trait-mediated assembly processes predict successional changes in community diversity of tropical forests. Proceedings of the National Academy of Sciences 111:5616–5621. Laughlin, D. C. 2014. The intrinsic dimensionality of plant traits and its relevance to

26 community assembly. Journal of Ecology. Laughlin, D. C., and J. Messier. 2015. Fitness of multidimensional phenotypes in dynamic adaptive landscapes. Trends in Ecology & Evolution:1–10. Lavorel, S., and E. Garnier. 2002. Predicting changes in community composition and ecosystem functioning from plant traits :545–556. Levin, S. A. 1992. The problem of pattern and scale in ecology. Ecology 73:1943–1967. Levins, R., and R. Lewontin. 1985. The dialect biologist. Harvard University Press. Lewontin, R. 1977. Adaptation. Pages 65–84 in R. Levins and R. Lewontin, editors. The Dialectical Biologist. Harvard University Press, Cambridge. Lynch, M., and B. Walsh. 1998. Genetics and Analysis of Quantitative Traits. Sinauer Associates. Marks, C. O., and M. J. Lechowicz. 2006. Alternative designs and the evolution of functional diversity. The American naturalist 167:55–66. McGill, B. J., B. J. Enquist, E. Weiher, and M. Westoby. 2006. Rebuilding community ecology from functional traits. Trends in ecology & evolution 21:178–85. Merila, J., and M. Bjorklund. 2004. Phenotypic Integration as a Constraint and Adaptation. Pages 107–129 in M. Pigliucci and K. Preston, editors. Phenotypic Integration: Studyinf the Ecology and Evolution of Comlpex Phenotypes. Oxford University Press. Milla, R., and P. B. Reich. 2011. Multi-trait interactions, not phylogeny, fine-tune leaf size reduction with increasing altitude. Annals of botany 107:455–65. Mouillot, D., N. W. H. Mason, and J. B. Wilson. 2007. Is the abundance of species determined by their functional traits? A new method with a test using plant communities. Oecologia 152:729–37. Murren, C. J. 2002. Phenotypic integration in plants. Plant Species Biology 17:89–99. Niinemets, U. 2015. Is there a species spectrum within the world-wide leaf economics spectrum ? Major variations in leaf functional traits in the Mediterranean sclerophyll Quercus ilex. New Phytologist:79–96. Niklas, K. J. 1992. Plant Biomechanics: an engineering approach to plant form and function. University of Chicago Press. Niklas, K. J. 1994a. Morphological evolution through complex domains of fitness. Proceedings of the National Academy of Sciences of the United States of America 91:6772–6779. Niklas, K. J. 1994b. Plant Allometry: The scaling of form and process. The University of

27 Chicago Press, Chicago. Niklas, K. J. 1999. Evolutionary walks through a land plant morphospace. Journal of Experimental Botany 50:39–52. Paine, C. E. T., L. Amissah, H. Auge, C. Baraloto, M. Baruffol, N. Bourland, H. Bruelheide, K. Daïnou, R. C. de Gouvenain, J.-L. Doucet, S. Doust, P. V. a. Fine, C. Fortunel, J. Haase, K. D. Holl, H. Jactel, X. Li, K. Kitajima, J. Koricheva, C. Martínez-Garza, C. Messier, A. Paquette, C. Philipson, D. Piotto, L. Poorter, J. M. Posada, C. Potvin, K. Rainio, S. E. Russo, M. Ruiz-Jaen, M. Scherer-Lorenzen, C. O. Webb, S. J. Wright, R. a. Zahawi, and A. Hector. 2015. Globally, functional traits are weak predictors of juvenile tree growth, and we do not know why. Journal of Ecology 103:978–989. Phillips, P. C., and S. J. Arnold. 1989. Visualizing Multivariate Selection. Evolution 43:1209– 1222. Pigliucci, M. 2003. Phenotypic integration: studying the ecology and evolution of complex phenotypes. Ecology Letters 6:265–272. Poorter, L., S. J. Wright, H. Paz, D. D. Ackerly, R. G. Condit, G. Ibarra-Manríquez, K. E. Harms, J. C. Licona, M. Martínez-Ramos, S. J. Mazer, H. C. Muller-Landau, M. Peña- Claros, C. O. Webb, and I. J. Wright. 2008. Are functional traits good predictors of demographic rates? Evidence from five neotropical forests. Ecology 89:1908–20. Raup, D. M., and a Michelson. 1965. Theoretical Morphology of the Coiled Shell. Science (New York, N.Y.) 147:1294–5. Reich, P. B., M. B. Walters, and D. S. Ellsworth. 1997. From tropics to tundra: global convergence in plant functioning. Proceedings of the National Academy of Sciences of the United States of America 94:13730–4. Richardson, S. J., R. B. Allen, R. P. Buxton, T. a. Easdale, J. M. Hurst, C. W. Morse, R. D. Smissen, and D. a. Peltzer. 2013. Intraspecific Relationships among Wood Density, Leaf Structural Traits and Environment in Four Co-Occurring Species of Nothofagus in New Zealand. PLoS ONE 8:24–26. Rüger, N., C. Wirth, S. J. Wright, and R. Condit. 2012. Functional traits explain plasticity of growth rates in tropical tree species. Ecology 93:2626–2636. Shipley, B. 2007. Comparative plant ecology as a tool for integrating across scales. Annals of Botany 99:965–966. Shipley, B., F. De Bello, J. H. C. Cornelissen, E. Laliberté, D. C. Laughlin, and P. B. Reich. (n.d.). Reinforcing loose foundation stones in trait-based plant ecology. Oecologia (In Press). Shoval, O., H. Sheftel, G. Shinar, Y. Hart, O. Ramote, a. Mayo, E. Dekel, K. Kavanagh, and

28 U. Alon. 2012. Evolutionary Trade-Offs, Pareto Optimality, and the Geometry of Phenotype Space. Science 336:1157–1160. Silvertown, J., M. Dodd, and D. Gowing. 2001. Phylogeny and the niche structure of meadow plant communities. Journal of Ecology 89:428–435. Silvertown, J., K. McConway, D. Gowing, M. Dodd, M. F. Fay, J. a Joseph, and K. Dolphin. 2006. Absence of phylogenetic signal in the niche structure of meadow plant communities. Proceedings. Biological sciences / The Royal Society 273:39–44. Violle, C., B. J. Enquist, B. J. McGill, L. Jiang, C. H. Albert, C. Hulshof, V. Jung, and J. Messier. 2012. The return of the variance: intraspecific variability in community ecology. Trends in Ecology & Evolution 27:1–9. Violle, C., M.-L. Navas, D. Vile, E. Kazakou, C. Fortunel, I. Hummel, and E. Garnier. 2007. Let the concept of trait be functional! Oikos 116:882–892. Violle, C., P. B. Reich, S. W. Pacala, B. J. Enquist, and J. Kattge. 2014. The emergence and promise of functional biogeography. Proceedings of the National Academy of Sciences 11:13690–13696. Walsh, B., and M. W. Blows. 2009. Abundant Genetic Variation plus Strong Selection = Multivariate Genetic Constraints: A Geometric View of Adaptation. Annual Review of Ecology Evolution and Systematics 40:41–59. Weiher, E., D. Freund, T. Bunton, A. Stefanski, T. Lee, and S. Bentivenga. 2011. Advances, challenges and a developing synthesis of ecological community assembly theory. Philosophical transactions of the Royal Society of London. Series B, Biological sciences 366:2403–13. Westoby, M. 1998. A leaf-height-seed (LHS) plant ecology strategy scheme. Plant and Soil 199:213–227. Westoby, M., D. S. Falster, A. T. Moles, P. A. Vesk, and I. J. Wright. 2002. Plant ecological strategies: Some leading dimensions of variation between species. Annual Review of Ecology and Systematics 33:125–159. Westoby, M., and I. J. Wright. 2006. Land-plant ecology on the basis of functional traits. Trends in ecology & evolution 21:261–8. Wright, I. J., P. B. Reich, M. Westoby, D. D. Ackerly, Z. Baruch, F. Bongers, J. Cavender- Bares, T. Chapin, J. H. C. Cornelissen, M. Diemer, J. Flexas, E. Garnier, P. K. Groom, J. Gulias, K. Hikosaka, B. B. Lamont, T. Lee, W. Lee, C. Lusk, J. J. Midgley, M.-L. Navas, U. Niinemets, J. Oleksyn, N. Osada, H. Poorter, P. Poot, L. Prior, V. I. Pyankov, C. Roumet, S. C. Thomas, M. G. Tjoelker, E. J. Veneklaas, and R. Villar. 2004. The worldwide leaf economics spectrum. Nature 428:821–827.

29 Wright, J. P., and A. Sutton-Grier. 2012. Does the leaf economic spectrum hold within local species pools across varying environmental conditions? Functional Ecology 26:1390– 1398.

30 APPENDIX A:

Trait variation and integration across scales: Is the leaf economic spectrum present at local scales? Submitted as: Messier, Julie, McGill, Brian J., Enquist, Brian J., Lechowicz, Martin J. In Review. Trait variation and integration across scales: Is the leaf economic spectrum present at local scales? Ecography

Authors: 1Julie Messier, 2Brian J. McGill, 1Brian J. Enquist & 3Martin J. Lechowicz

Affiliations : 1 Ecology and Evolutionary Biology, University of Arizona, Tucson, AZ 85721, USA

2 School of Biology and Ecology, University of Maine, Orono, ME 04469, USA. 3 Biology Department, McGill University, Montréal, QC, H3A1B1, Canada.

Corresponding author: Julie Messier University of Arizona 1041 E. Lowell Street, Tucson, Arizona, 85711, USA 520-626-3336 [email protected]

Keywords: Trait Variation, Phenotypic Integration, Scales, Leaf Economic Spectrum

Short title: Traits variation and integration across scales

31 ABSTRACT

Trait-based approaches have been taking an increasingly dominant role in community ecology. Although the concept of trait-based strategy dimensions, such as the leaf economic spectrum, have been developed with global-scale data, trait variation at the community scale is often interpreted in this context. Here we argue that a research priority should be to determine whether global-scale trait relationships hold at more local scales. We review recent literature suggesting that the leaf economic spectrum may be scale dependent, and present a case study exploring the interplay between trait correlation and variation in leaf traits across ecological scales. In contrast to expectations, we find that the strength of correlations between pairs of traits does not predict whether these pairs respond similarly to different drivers of variation. Instead, leaf economic spectrum traits largely retain the ability to respond independently to different drivers of phenotypic variation. Recent literature and our results suggest that the leaf economic spectrum concept might not apply at local-scales. It is pressing to determine at which scales the leaf economic spectrum is present. Clarifying how the interplay between trait variation and correlation changes through ecological scales is necessary for us to make the most of the emerging trait toolbox and will provide insight into the forces structuring functional diversity.

32 INTRODUCTION

A central goal of community ecology is to disentangle the different processes creating patterns of functional diversity within and among communities and across spatial and temporal scales (Weiher and Keddy 1999). In the ecological literature, our understanding of functional diversity has been linked to advances in the concept of strategy dimensions (Grime 1974), and more recently to that of trait-based strategy dimensions (Westoby 1998). The trait-based approach has quickly been adopted in comparative ecology since it offers a general ways to describe ecological strategies (Lavorel and Garnier 2002, McGill et al. 2006). A core assumption of trait-based ecology is that trade-offs and constraints have shaped trait variation into different trait dimensions. Trait dimensions are defined by sets of correlated traits and reflect distinct aspects of a plant’s ecological strategy (Westoby et al. 2002). Trait dimensions have been characterized across broad spatial and taxonomic scales to summarize global patterns of phenotypic diversity (Reich et al. 1999, Wright et al. 2004, Reich 2014), but the processes shaping communities are expected to primarily act at more local spatial and taxonomic scales. Thus, it is unclear whether the trait relationships forming trait dimensions at large spatial and taxonomic scales are also present at the scales of community assembly processes. This gap in the literature is especially important since individual traits are commonly used to study community assembly processes (Swenson and Enquist 2007, Siefert et al. 2014, Enquist et al. 2015, Kraft et al. 2015) and the variance of these individual traits often interpreted in the context of trait dimensions (e.g. Falster et al. 2011, Long et al. 2011, Hulshof et al. 2013). The leaf economic spectrum, LES, reflects a trade-off between the leaf’s lifespan and its maximum photosynthetic rate, thus reflecting an investment in a fast or slow rate of carbon return. Some of the traits associated with the LES, such as Leaf mass per area (LMA), Leaf Nitrogen Content (LNC) and leaf dry matter content (LDMC), are widely used in ecological studies as indicators of this strategy dimension. It is therefore important to assess whether variation in the LES traits at local scales can be interpreted in the context of the trait dimension. In this paper, we present two types of evidence suggesting that the LES might not occur at local scales. First we review recent literature examining patterns of trait correlation and variation across scales. Next, we provide new analyses exploring the interplay between trait correlation

33 and variation across scales. Based on the evidence from the literature and the results of the case study, we argue that we need to clarify the scales at which the LES is present in order to appropriately interpret variation in the LES trait in community ecology.

Phenotypic integration is central to this issue because asking “At what scale does the LES appear?” is tantamount to asking “At what scales are the drivers shaping the LES dominant?” Different drivers of trait correlation (also known as phenotypic integration) constrain the ability of any one trait to vary in response to drivers of variation (Box 1). The scales at which trait dimensions appear depend on the drivers shaping those trait correlations. Patterns of trait correlation can result from different drivers of phenotypic integration that act at different ecological and spatial scales and impose more or less flexible and variable constraints (Box 1). Since different sets of constraints exist at each scale, the relationship among traits can also change across scales (Figure B1). Thus, the scales at which trait dimensions are present hinge upon which drivers shape those trait correlations and their relative importance at different scales. Ultimately, variation in any given trait reflects the balance between drivers of phenotypic variation (Box 2) and drivers of phenotypic integration (Box 1). Considering trait correlation is necessary to properly understand the observed variation in a trait at a given scale (e.g. Laughlin and Messier 2015). Two types of evidence appearing in the recent literature question whether the LES is present within communities: (1) the LES might only appear at large scales; and (2) within studies, the LES traits show contrasting patterns of variance across scales, which suggests the LES traits respond differently to drivers of phenotypic variation. Below, we review these literature-based evidences, then present a case study further examining the second argument.

LITTERATURE REVIEW: RECENT STUDIES QUESTION WHETHER THE LES IS PRESENT WITHIN

COMMUNITIES.

Two lines of evidence question whether the LES should be present within communities. First, recent studies fail to find the LES correlation patterns at smaller ecological and spatial scales. A meta-analysis by Funk & Cornwell (2013) argues that the strength of the LES

34 relationships depends on the variation in leaf life span (LLS) in the study. This variability in LLS can be genetic and is often an adaptation to distinct climatic regimes or life-histories. If this argument is true, we would only expect to find the LES relationships when there is a large variability in LLS. Specifically, we can make three predictions. We expect that the LES relationships should be: (i) absent within an individual within an environment; (ii) present among ecotypes within species, when these ecotypes span bioclimatic zones, (iii) present among species within localities when sufficient differences in LLS occur among the coexisting populations. Unfortunately, it is not possible to make more specific prediction for (iii) since the minimum range of LLS at which the LES appears was not specified in Funk & Cornwell (2013). The findings of the studies examining LES trait correlations at each of these three scales (within individuals, among ecotypes within species and among species within localities) are summarized in Table 1. The study examining LES trait correlations within individuals supported prediction (i) in 4/6 clones. Six out of eight trait correlations among ecotypes within species from four different studies supported prediction (ii). The only one of three studies examining trait correlations within communities in which we could evaluate prediction (iii) supported the prediction. The patterns observed in the literature largely agree with the predictions based on the argument that the LES only arises with sufficient variation in the LLS. Clearly, more studies are needed examining the LES relationships within locality both among and within species to understand at when the LES appears. Funk & Cornwell’s argument consistent with the suggestion that adaptive evolution plays a major role in creating the observed trait associations (Reich and Wright 2003, Donovan et al. 2011, Vasseur et al. 2012). If adaptive evolution is largely responsible for synchronizing LLS and maximum photosynthetic rate (AMAX), then the LES relationships clearly expressed over a wide range of species, bioclimatic zones or trait values may be obscured in less extensive data sets (Funk and Cornwell 2013). As the signal-to-noise ratio of the global-scale LES decreases at smaller ecological and spatial scales, other drivers of phenotypic variation and integration might become dominant (e.g. see Figure 3 in Wright et al. 2004). Smaller ecological scales usually imply a smaller range of trait values expressed, smaller spatial scales and less environmental

35 heterogeneity (Figure B3) such that it might not be possible to distinguish the effects of ecological scale, spatial scales and trait coverage on trait correlation patterns (Box 3).

Second, studies examining the fraction of total variance occurring at different ecological scales (hereafter, called variance structure across scales) point to a lack of coordination among LES traits. Interestingly, each of the six studies examining variance structure across scales in the LES traits found different patterns of variation for these traits (Roche et al. 2004, Hulshof and Swenson 2010, Messier et al. 2010, Albert et al. 2010, Auger and Shipley 2012, Kang et al. 2014). To illustrate this, Table 2 shows the fraction of the total variance occurring at the species scale for two traits - Leaf Mass per Area (LMA) and Leaf Dry Matter Content (LDMC). The table shows that within each study, contrasting variance structures were found for these two LES traits. These differences suggests that the different LES traits are sensitive to different drivers of variation, and invites us to further explore how much the strength of trait correlation constrain traits to respond similarly to drivers of variation.

CASE STUDY: DOES CORRELATION STRENGTH CONSTRAIN THE RESPONSE OF INDIVIDUAL TRAITS TO

DRIVERS OF VARIATION?

To further explore how much the strength of trait correlation constrains the response of individual traits to drivers of variation, here we examine the relationship between the Pearson’s correlation between pairs of traits and the similarity of their variance structure across scales. The main focus of this case study is to parse out the sensitivity of related traits to different drivers of variation. These results have bearing on whether the LES, a global-scale correlation pattern, might also be present at local scales. If the trait correlation imposes in the LES traits a coordinated response to drivers of phenotypic variation at all scales, then the LES is likely to be present within communities. Conversely, if traits correlated at global scale retain the ability to respond differently to local-scale drivers of variation, then the LES might not appear locally because the locally dominant drivers of variation may have distinct effects on the traits. We expect that as the strength of correlation between two traits increases, so should the similarity of their response to drivers of variation. As a result, increasingly correlated traits

36 should show increasingly similar variance structure across scales. Specifically, we expect that the variance structure across scales of the LES traits (LMA, LDMC and LNC) should be more similar to each other than to uncorrelated traits (LCC and Leaf Area). Further, given the mixed evidence within and among studies regarding the relative importance of intra- and inter-specific variance (Hulshof and Swenson 2010, Messier et al. 2010, Auger and Shipley 2012, Kang et al. 2014), we expect that one of the main differences among traits will be the relative importance of the species-scale variance (i.e. in the amount of variation attributable to differences among species).

Methods To examine whether related foliar traits respond similarly to drivers of variation, we compare and contrast the variance structure across scales in five foliar traits. The sampling design allows us to largely isolate the effect of different drivers to individual scales (Box 2; Figure A1). We pay special attention to untangling the different scales of intraspecific variation because the different intraspecific scales are subject to distinct drivers of variation, thereby allowing us to examine whether LES traits respond similarly to drivers of trait phenotypic variation (Figure B2). We compare five leaf functional traits measured on the same leaf samples: Leaf Mass per Area (LMA), Leaf Nitrogen Content (LNC), Leaf Dry Matter Content (LDMC), Leaf Area and Leaf Carbon Content (LCC). LMA, LNC and LDMC reflect the leaf economic spectrum, LCC reflect mechanical support and Leaf Area reflect architecture. (See Appendix 1 for further discussion of the study traits). We measured variation in these traits across six ecological scales: (1) among leaves within a stratum; (2) between strata within a tree; (3) among trees within a species; (4) among species; (5) among plots within a site and (6) among sites spanning two biomes (Figure A1). The three sites span a strong precipitation gradient, whereas the plots within each site are located within the same habitat (Table A1). Leaves of the sun and shade strata were sampled for each tree within the plots. Sampling design and ecological scales. Although it was not possible to fully cover all the scales detailed in Figure B2, this nested sampling design covers most of them. Importantly, the hierarchical design allows us to measure and remove the effects of lower-scale drivers from

37 higher scales. This comprehensive approach allows us to link each scale to a few drivers of phenotypic variation much better than would otherwise be possible. The following summarizes the main drivers of phenotypic variation associated with each of the scales in this study given its sampling design: The leaf scale mainly reflects the ontogenetic effects of metamer and module position. The strata scale mostly reflects plastic effects due to the vertical light gradient in the canopy. The tree scale mainly reflects genetic differences within a species and micro-environmental gradients. The species scale mainly reflects genetic differences resulting from adaptive evolution and drift. The plot scale reflects the natural variability in trait composition within habitats. The site scale reflects plastic and/or filtering response to the strong precipitation and seasonality gradient. See Appendix 1 for details on the sampling design and the scales studied. Statistical analyses. For each trait, we partitioned the total variance among ecological scales to determine what fraction was explained by each of them, independently of the other scales. To do so, we carried out a variance partitioning analysis on a general linear model with nested and crossed factors using the vegan & lme4 packages in R (R Development Core Team 2011) . Specifically, the three lower scales were nested within the species scale (leaf within strata, within tree, within species) and the plot level was nested within sites. The species/tree/strata/leaf hierarchy was then crossed with the site/plot hierarchy. This model is realistic because it allows species to occur in multiple plots and sites. The confidence intervals around the variance component values were calculated by bootstrapping and presented in Table A5. We then ran a principal component analysis on the variance component table (i.e. on Table A5) to identify the main differences among the variance structures of the five traits. See Appendix 1 for details of the statistical analyses. See Appendix 1 for details on the statistical analyses. Data description. In our dataset, LMA and LNC respectively span 14% and 87% of the range covered in GLOPNET (Wright et al. 2004). Total variation differs among traits: LCC is the least variable trait, with a ca. 2-fold range of variation and a coefficient of variation of 0.08; Leaf Area is the most variable trait, with a ca. 2000-fold range of variation and a coefficient of variation of 2.19 (Table A2; Figure 1). Bivariate trait correlations at the leaf level agree with the

38 patterns broadly reported in the literature: LNC, LMA and LDMC are moderately correlated with each other but weakly with LCC and not at all with Leaf Area (Table A4).

Case Study Results and Discussion Foliar traits differ in their species-specificity and in the scale of response to environmental gradients The five foliar traits, LMA, LNC, LDMC, LCC and Leaf Area, showed differences in their relative sensitivity to environmental, genetic and ontogenic drivers of variation (Figure 1). The principal difference among the traits lies in the fraction of variance occurring at the species scale (Figure 2). With 79% of the total variance occurring among species, the variance structure of Leaf Area is dominated by interspecific differences (Figure 1). This is consistent with this trait being predominantly determined by genetic drivers, whether in the form of adaptive evolution or drift (Figure B2). In contrast, LMA is much more labile, with only 30% of the total variance occurring at the species scale. This composite trait is affected by the different processes influencing leaf mass and leaf area. LMA’s sensitivity to all environmental gradients, whether at the site, tree, stratum or leaf scale, is consistent with its integrative nature. Variation at the site scale mainly reflects the strong precipitation gradient and variation at the leaf and strata scales mainly reflect micro-environmental gradients, such as irradiance, within the tree and forest canopies. The second important difference among the five foliar traits is in the spatial scale at which they respond to environmental gradients. Some traits, such as LNC and LCC, are more sensitive to large scale environmental gradients such as the climate or soil type. This suggests that these traits are best to detect large-scale ecological filtering. Other traits, such as LDMC, are more sensitive to small scale (leaf and strata) environmental gradients such as insolation within the canopy. This suggests that it is the best suited trait to detect plastic and ontogenetic effects. Note that the three morphological traits - LMA, LDMC and Leaf Area – tend to be more sensitive to small-scale environmental gradients whereas the two stoichiometric traits (LNC, LCC) respond to large-scale environmental gradients (Figure 2).

39 The upshot of the contrasting sensitivity of LES traits to drivers of variation is that in future studies, we can take advantage of these differences to fine-tune the trait toolbox and select the traits that best respond to the processes of interest at the scale of interest; of the three leaf economic spectrum traits examined in this study, LNC is best suited to detect community-level responses to climate, LDMC is best suited to detect the plastic response within individuals to micro-environmental gradients, and LMA being the least species-specific trait is broadly sensitive to environmental and ontogenetic drivers of variation.

LES traits largely retain their ability to respond independently to drivers of phenotypic variation We find that, up to the correlation values expressed here in the LES, traits that are more strongly correlated do not necessarily respond more similarly to drivers of phenotypic variation: (1) the LES traits do not cluster together in principal component space (Figure 2); and (2) the three pairs of LES are as or more dissimilar to each other as three LES-non LES trait pairs (Figure 3, Table A4). For example, the variance structure across scales of uncorrelated LCC and LNC group together in principal component space and are more similar to each other than any other pair of traits (Figures 1 and 2). Figure 3 shows that the strength of trait correlation does not predict whether two traits will respond similarly to drivers of variation (SMA regression, p-value = 0.29). Instead, the data show a triangular relationship (0.90 quantile: p-value = 0.028). This indicates that the strength of trait correlation found for the LES imposes a minimum coordination in the traits’ response to drivers of phenotypic variation. It makes sense that uncorrelated or weakly correlated traits can have similar or dissimilar variance structures across scales (such as LNC-LCC and LMA-Leaf Area, respectively) but that strongly correlated traits cannot have entirely dissimilar variance structure across scales. In other words, while the LES traits are restricted from having very different variance structures, the moderate level of trait correlation observed here and typical of this trait dimension (i.e. r= ca.0.5-0.6) largely allow traits to respond distinctly to local drivers of trait variation. The breadth of correlation strengths in the dataset, which range from 0.03-0.59, is a typical range for mathematically independent traits in regional datasets (e.g., Grime et al. 1997,

40 Jakobsson and Eriksson 2000, Laughlin et al. 2010)(but see White 1983a). Our ability to detect the triangular relationship is limited by the small number of traits, so that it would be worthwhile to verify this pattern in other traits beyond the five studied here as well as in other ecosystems. Nonetheless, the data presented here suggest that traits typically considered to be part of the LES largely retain their ability to respond differently to drivers of phenotypic variation. If this conclusion holds, this would imply that while LES traits may be correlated at large scales, they may respond differently enough to local drivers of variation to be less or not correlated at local scales.

CONCLUSIONS

In this paper we have presented evidence from the literature and from a case study suggesting that we need to verify whether the LES is present at local scales. We have argued that the trait dimensions that are present at broad scales may represent an overall tendency that can be locally variable at smaller scales. The global interspecific patterns of trait relationships are associated with a large ecological scale, large spatial scale and large trait coverage. The relationship between any two LES traits can change or even reverse in a local dataset due to the action of distinct drivers of phenotypic integration and variation acting at those ecological scales and/or to the increased noise around the relationship. In such case, variation in LES traits at local scales might not reflect the LES strategy dimension, but the effect of other drivers of variation. Recent literature has highlighted that without spanning a large range of trait values, the leaf economic spectrum trait relationships might be weak or absent. This is also consistent with indications that adaptive constraints shape the LES. The analyses presented here find that even when sampling leaves across two biomes, the LES traits largely retain their ability to respond differently to drivers of phenotypic variation. If this case study holds generally, we should be careful to infer the behavior of a suite of traits from studying any one of them at local scales. It is not news that traits respond to numerous drivers of variation, nor that these distinct drivers act at different ecological scales. Accordingly, trait sampling protocols attempt to minimize variation from sources that are not of interest (Lavorel and Garnier 2002, Cornelissen et al. 2003, Pérez-Harguindeguy et al. 2013). Nonetheless, here we emphasize the importance of these well-known principles to question whether we can interpret local scale trait variation in

41 the context of globally-described patterns. We highlight that if the LES mainly appears across broad scales with modest correlation strength, local sources of variation may blur the broad relationships we are attempting to detect if we measure them locally. Functional traits are a powerful tool in the community ecologist’s toolbox but to use this tool most efficiently, it should be a research priority to verify at what scales trait dimensions appear. Meanwhile, we should interpret with caution leaf trait variation at smaller ecological scales.

42 ACKNOWLEDGEMENTS

Many thanks to Ricardo Cossio, Omar Lopez, Joe Wright, Mirna Samaniego, Oris Acevedo and José Baharona for invaluable help with field work in Panama. Thanks to Mélisanne Gagnon for assistance with leaf grinding and to Kyle Martin for help with the elemental analyses. Our gratitude also goes to Katrina Dlugosch, Judith Bronstein and Cyrille Violle for insightful discussions of the manuscript and to Joshua Scholl, Kelsey Yule, Simon Stump, Nicolas Kortessis and as well as to anonymous reviewers for detailed comments on previous versions of this manuscript which helped improve it.

43 REFERENCES

Ackerly, D. D. and Cornwell, W. K. 2007. A trait-based approach to community assembly: partitioning of species trait values into within- and among-community components. - Ecol. Lett. 10: 135–45. Albert, C. H. et al. 2010. A multi-trait approach reveals the structure and the relative importance of intra- vs. interspecific variability in plant traits. - Funct. Ecol. 24: 1192–1201. Antonovics, J. and van Tienderen, P. H. 1991. Ontoecogenophyloconstraints? The chaos of constraint terminology. - Trends Ecol. Evol. 6: 166–168. Armbruster, S. W. 1990. Estimating and testing the shapes of adaptive surfaces - the morphology and pollination of Dalechampia blossoms. - Am. Nat. 135: 14–31. Armbruster, S. W. 1991. Multilevel analysis of morphometric data from natural plant populations: insights into ontogenetic, genetic, and selective correlations in Dalechampia scandens. - Evolution (N. Y). 45: 1229–1244. Armbruster, S. W. et al. 2004. Floral Integration, Modularity, and Accuracy. - In: Pigliucci, M. and Preston, K. (eds), Phenotypic Integration: Studying the Ecology and Evolution of Comlpex Phenotypes. Oxford University Press, pp. 23–49. Auger, S. and Shipley, B. 2012. Inter-specific and intra-specific trait variation along short environmental gradients in an old-growth temperate forest (F de Bello, Ed.). - J. Veg. Sci. 24: 419–428. Barthélémy, D. and Caraglio, Y. 2007. Plant architecture: A dynamic, multilevel and comprehensive approach to plant form, structure and ontogeny. - Ann. Bot. 99: 375–407. Berg, R. L. 1960. The ecological significance of correlation pleiades. - Evolution (N. Y). 14: 171– 180. Blonder, B. et al. 2013. Assessing the causes and scales of the leaf economics spectrum using venation networks in tremuloides. - J. Ecol. 101: 981–989. Brouat, C. et al. 1998. Corner’s rules revisited: ontogenetic and interspecific patterns in leaf-stem allometry. - New Phytol. 139: 459–470. Chao, A. et al. 2005. A new statistical approach for assessing similarity of species compostition with incidence and abundance data. - Ecol. Lett. 8: 148–159. Cheverud, J. M. 1982. Phenotypic, genetic and environmental morphological integration in the cranium. - Evolution (N. Y). 36: 499–516. Cheverud, J. M. 1984. Quantitative genetics and developmental constraints on evolution by selection. - J. Theor. Biol. 110: 155–171. Coley, P. D. and Barone, J. a. 1996. Herbivory and Plant Defenses in Tropical Forests. - Annu. Rev. Ecol. Syst. 27: 305–335. Condit, R. et al. 2004. Tropical forest dynamics across a rainfall gradient and the impact of an El Ni ˜ no dry season.: 51–72. Cornelissen, J. H. C. et al. 2003. A handbook of protocols for standardised and easy measurement of plant functional traits worldwide. - Aust. J. Bot. 51: 335. Corner, E. J. H. 1949. The durian theory or the origin of the modern tree. - Ann. Bot. 13: 367– 414. de Kroon, H. et al. 2005. A modular concept of phenotypic plasticity in plants. - New Phytol. 166: 73–82.

44 Donovan, L. a et al. 2011. The evolution of the worldwide leaf economics spectrum. - Trends Ecol. Evol. 26: 88–95. Edwards, E. J. et al. 2014. Leaf life span and the leaf economic spectrum in the context of whole plant architecture. - J. Ecol. 102: 328–336. Enquist, B. J. et al. 1999. Allometric scaling of production and life-history variation in vascular plants. - Nature 401: 907–911. Enquist, B. J. et al. 2015. Scaling from traits to ecosystems: Developing a general Trait Driver Theory via integrating trait-based and metabolic scaling theories. - Adv. Ecol. Res. in press. Erickson, R. O. and Michelini, F. J. 1957. The Plastochron Index. - Am. J. Bot. 44: 297–305. Evans, J. R. 1989. Photosynthesis and nitrogen relationships in leaves of C3 plants. - Oecologia 78: 9–19. Falster, D. S. et al. 2011. Influence of four major plant traits on average height, leaf-area cover, net primary productivity, and biomass density in single-species forests: A theoretical investigation. - J. Ecol. 99: 148–164. Funk, J. L. and Cornwell, W. K. 2013. Leaf traits within communities: Context may affect the mapping of traits to function. - Ecology 94: 1893–1897. Gardner, K. M. and Latta, R. G. 2007. Shared quantitative trait loci underlying the genetic correlation between continuous traits. - Mol. Ecol. 16: 4195–209. Gould, S. J. and Lewontin, R. C. 1979. The spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist programme. - Proc. R. Soc. London 205: 581–98. Grime, J. 1974. Vegetation classification by reference to strategies. - Nature 250: 26–30. Grime, J. et al. 1997. Integrated screening validates primary axes of specialisation in plants. - Oikos 79: 259–281. Huber, H. and Stuefer, J. F. 1997. Shade-induced changes in the branching pattern of a stoloniferous herb: Functional response or allometric effect? - Oecologia 110: 478–486. Hulshof, C. M. and Swenson, N. G. 2010. Variation in leaf functional trait values within and across individuals and species: an example from a Costa Rican dry forest. - Funct. Ecol. 24: 217–223. Hulshof, C. M. et al. 2013. Intra-specific and inter-specific variation in specific leaf area reveal the importance of abiotic and biotic drivers of species diversity across elevation and latitude. - J. Veg. Sci. 24: 921–931. Huxley, J. S. 1935. The Problems of Relative Growth. - Methuen & Co.ltd. Jakobsson, A. and Eriksson, O. 2000. A comparative study of seed number, seed size, seedling size and recruitment in grassland plants. - OIKOS 88: 494–502. Kang, M. et al. 2014. Trait variability differs between leaf and wood tissues across ecological scales in subtropical forests (S Bartha, Ed.). - J. Veg. Sci. 25: 703–714. Kaplan, D. R. 2001. The science of plant morphology: History, definition, and role in modern biology. - Am. J. Bot. 88: 1711–1741. Keddy, P. 1992. Assembly and response rules: two goals for predictive community ecology. - J. Veg. Sci. 3: 157–164. Kimball, S. et al. 2013. Phenotypic selection favors missing trait combinations in coexisting annual plants. - Am. Nat. 182: 191–207. Klingenberg, C. P. 2004. Integration, Modules, and Development. - In: Pigliucci, M. and Preston, K. (eds), Phenotypic Integration: Studyinf the Ecology and Evolution of Comlpex

45 Phenotypes. Oxford University Press, pp. 213–227. Kollmann, J. 2000. Dispersal of fleshy-fruited species: a matter of spatial scale? - Perspect. Plant Ecol. Evol. Syst. 3: 29–51. Kraft, N. J. B. et al. 2015. Plant functional traits and the multidimensional nature of species coexistence. - Proc. Natl. Acad. Sci. 112: 797–802. Lambers, H. et al. 2008. Plant physiological ecology. - Springer. Lande, R. and Arnold, S. J. J. 1983. The measurement of selection on correlated characters. - Evolution (N. Y). 37: 1210–1226. Laughlin, D. C. and Messier, J. 2015. Fitness of multidimensional phenotypes in dynamic adaptive landscapes. - Trends Ecol. Evol.: 1–10. Laughlin, D. C. et al. 2010. A multi-trait test of the leaf-height-seed plant strategy scheme with 133 species from a pine forest flora. - Funct. Ecol. 24: 493–501. Lavorel, S. and Garnier, E. 2002. Predicting changes in community composition and ecosystem functioning from plant traits: revisiting the Holy Grail. - Funct. Ecol. 16: 545–556. Leigh, E. G. et al. 2004. Barro Colorado Island Forest Dynamics Plot, Panama. - In: Losos, E. C. and Leigh Jr., E. G. (eds), Tropical forest diversity and dynamism: Findings from a large- scale plot network. University of Chicago Press, pp. 451–463. Lepš, J. et al. 2011. Community trait response to environment: disentangling species turnover vs intraspecific trait variability effects. - Ecography (Cop.). 34: 856–863. Lewontin, R. 1977. Adaptation. - In: Levins, R. and Lewontin, R. (eds), The Dialectical Biologist. Harvard University Press, pp. 65–84. Long, W. et al. 2011. Within- and among-species variation in specific leaf area drive community assembly in a tropical cloud forest. - Oecologia 167: 1103–1113. Lucas, P. W. et al. 2000. Mechanical Defences to Herbivory. - Ann. Bot. 86: 913–920. Markow, T. A. 1995. Evolutionary Ecology and Developmental Instability. - Annu. Rev. Entomol. 40: 105–120. Martin, R. E. et al. 2007. Genetic variation in leaf pigment, optical and photosynthetic function among diverse phenotypes of Metrosideros polymorpha grown in a common garden. - Oecologia 151: 387–400. McGill, B. J. 2008. Exploring predictions of abundance from body mass using hierarchical comparative approaches. - Am. Nat. 172: 88–101. McGill, B. J. et al. 2006. Rebuilding community ecology from functional traits. - Trends Ecol. Evol. 21: 178–85. McGuigan, K. 2006. Studying phenotypic evolution using multivariate quantitative genetics. - Mol. Ecol. 15: 883–96. Merila, J. and Bjorklund, M. 1999. Population divergence and morphometric integration in the green-finch (Carduelis chloris) - evolution against the trajectory of least resistance ? - J. Evol. Biol. 12: 103–112. Merila, J. and Bjorklund, M. 2004. Phenotypic Integration as a Constraint and Adaptation. - In: Pigliucci, M. and Preston, K. (eds), Phenotypic Integration: Studyinf the Ecology and Evolution of Comlpex Phenotypes. Oxford University Press, pp. 107–129. Messier, J. et al. 2010. How do traits vary across ecological scales? A case for trait-based ecology. - Ecol. Lett. 13: 838–48. Murren, C. J. 2002. Phenotypic integration in plants. - Plant Species Biol. 17: 89–99.

46 Niinemets, U. 2015. Is there a species spectrum within the world-wide leaf economics spectrum ? Major variations in leaf functional traits in the Mediterranean sclerophyll Quercus ilex. - New Phytol.: 79–96. Niinemets, U. et al. 2007. Do we underestimate the importance of leaf size in plant economics? Disproportional scaling of support costs within the spectrum of leaf physiognomy. - Ann. Bot. 100: 283–303. Niklas, K. J. 1992. Plant Biomechanics: an engineering approach to plant form and function. - University of Chicago Press. Niklas, K. J. 1994. Morphological evolution through complex domains of fitness. - Proc. Natl. Acad. Sci. U. S. A. 91: 6772–6779. Niklas, K. J. and Kerchner, V. 1984. Mechanical and photosynthetic constraits on the evolution of plant shape. - Paleobiology 10: 79–101. Olson, E. C. and Miller, R. L. 1958. Morphological Integration. - University of Chicago Press. Olson, M. E. et al. 2009. Universal foliage-stem scaling across environments and species in dicot trees: plasticity, biomechanics and Corner’s Rules. - Ecol. Lett. 12: 210–9. Pérez-Harguindeguy, N. et al. 2013. New handbook for standardised measurement of plant functional traits worldwide. - Aust. J. Bot. 61: 167–234. Pigliucci, M. 2001. Phenotypic Plasticity: Beyond Nature and Nurture. - Johns Hopkins University Press. Pigliucci, M. 2003. Phenotypic integration: studying the ecology and evolution of complex phenotypes. - Ecol. Lett. 6: 265–272. Pigliucci, M. and Preston, K. A. 2004. Phenotypic Integration: Studying the Ecology and Evolution of Complex Phenotypes. - Oxford University Press. Poorter, L. and Rozendaal, D. M. a 2008. Leaf size and leaf display of thirty-eight tropical tree species. - Oecologia 158: 35–46. R Development Core Team, R. 2011. R: A Language and Environment for Statistical Computing (RDC Team, Ed.). - R Found. Stat. Comput. 1: 409. Raup, D. M. and Michelson, a 1965. Theoretical Morphology of the Coiled Shell. - Science 147: 1294–5. Reich, P. B. 2014. The world-wide “fast-slow” plant economics spectrum: a traits manifesto (H Cornelissen, Ed.). - J. Ecol. 102: 275–301. Reich, P. B. and Wright, I. J. 2003. The evolution of plant functional variation: traits, spectra, and strategies. - Int. J. Plant Sci. 164: S143–S164. Reich, P. B. et al. 1999. Generality of leaf trait relationships: A test across six biomes. - Ecology 80: 1955–1969. Ridley, M. 2003. Evolution. - Wiley-Blackwell. Roche, P. et al. 2004. Congruency analysis of species ranking based on leaf traits: which traits are the more reliable? - Plant Ecol. Former. `Vegetatio’ 174: 37–48. Roff, D. 2000a. Short Review The evolution of the G matrix : selection or drift ? in press. Roff, D. 2000b. The evolution of the G matrix: selection or drift? - Heredity (Edinb). 84: 135– 142. Santiago, L. S. and Mulkey, S. S. 2005. Leaf productivity along a precipitation gradient in lowland Panama: patterns from leaf to ecosystem. - Trees 19: 349–356. Santiago, L. S. et al. 2004. Coordinated changes in photosynthesis, water relations and leaf

47 nutritional traits of canopy trees along a precipitation gradient in lowland tropical forest. - Oecologia 139: 495–502. Shipley, B. et al. 2006. Fundamental trade-offs generating the worldwide leaf economics spectrum. - Ecology 87: 535–41. Siefert, A. et al. 2014. Community Functional Responses to Soil and Climate at Multiple Spatial Scales : When Does Intraspecific Variation Matter ? - PLoS One 9: 1–9. Smith, J. M. et al. 1985. Developmental constraints and evolution. - Q. Rev. Biol. 60: 265–287. Smithsonian Tropical Research Institute 2007. Physical Monitoring Program. Sutera, S. P. and Skalak, R. 1993. The History of Poiseuille Law. - Annu. Rev. Fluid Mech. 25: 1. Swenson, N. and Enquist, B. J. 2007. Ecological and evolutionary determinants of a key plant functional trait: wood density and its communitywide variation across latitutde and elevation. - Am. J. Bot. 94: 451–459. Vasseur, F. et al. 2012. A common genetic basis to the origin of the leaf economics spectrum and metabolic scaling allometry. - Ecol. Lett. 15: 1149–57. Vertregt, N. and Devries, F. 1987. A rapid method for determining the efficiency of biosynthesis of plant biomass. - J. Theor. Biol. 128: 109–119. Vile, D. et al. 2005. Specific leaf area and dry matter content estimate thickness in laminar leaves. - Ann. Bot. 96: 1129–36. Violle, C. et al. 2012. The return of the variance: intraspecific variability in community ecology. - Trends Ecol. Evol. 27: 1–9. Walsh, B. and Blows, M. W. 2009. Abundant Genetic Variation plus Strong Selection = Multivariate Genetic Constraints: A Geometric View of Adaptation. - Annu. Rev. Ecol. Evol. Syst. 40: 41–59. Weiher, E. and Keddy, P. A. 1999. Ecological Assembly rules: Perspectives, Advances, Retreats. - Cambridge University Press. West, G. B. et al. 1999. A general model for the structure and allometry of plant vascular systems. - Nature 400: 664–667. Westoby, M. 1998. A leaf-height-seed (LHS) plant ecology strategy scheme. - Plant Soil 199: 213–227. Westoby, M. et al. 2002. Plant ecological strategies: Some leading dimensions of variation between species. - Annu. Rev. Ecol. Syst. 33: 125–159. White, P. S. 1983a. Evidence that temperate east North-American evergreen woody-plants follow Corner’s rules. - New Phytol. 95: 139–145. White, P. S. 1983b. Corner’s rules in eastern deciduous trees : Allometry and its implications for the adaptive architecture of trees. - Bull. Torrey Bot. Club 110: 203–212. Wiens, J. A. 1989. Spatial scaling in ecology. - Funct. Ecol. 3: 385–397. Wright, J. P. and Sutton-Grier, A. 2012. Does the leaf economic spectrum hold within local species pools across varying environmental conditions? - Funct. Ecol. 26: 1390–1398. Wright, I. J. et al. 2004. The worldwide leaf economics spectrum. - Nature 428: 821–827.

48 FIGURES

100% 11 11 10 7 90% 17 6 6 8 7 80% 17 11 16 12 70% Leaf & Error 13 60% 21 Strata 50% Tree 45 44 Species 40% 79 44 by eachscaleby 30 Plot 30% 3 Site 20% 0 3 0 10% 21 22 15 20 Fraction of total variance explainedofvarianceFractiontotal 0 0% 1 LMA LDMC LNC LCC Area Study Trait Figure 1. Variance structure of five leaf functional traits across six ecological scales. LMA – Leaf Mass per Area; LDMC – Leaf Dry Matter Content; LNC – Leaf Nitrogen Concentration; LCC – Leaf Carbon Concentration; Area – Leaf Area. The black bar represents the coefficients of variation of each trait shown as the fraction of the highest coefficient of variation.

49

Figure 2. Principal component analysis of the variance component matrix of five leaf functional traits across six ecological scales. The length of the x and y axes are proportional to the amount of variance explained by each. Principal component 1, which explains 58% of the total variance in the variance component matrix (Table A6), reflects the relative importance of interspecific variance for each trait. Principal component 2, which explains 27% of the total variance among traits reflects whether the traits tends to show higher site- and plot- level variance or higher leaf- and strata-level variance.

50

Figure 3. Relationship between the strength of trait correlations and the dissimilarity of their variance structure across scales. Trait correlation was measured as the absolute value of the Pearson’s correlation coefficient. Dissimilarity structure across scales was measured as the Euclidian distance between trait pairs in principal component space. A quantile regression is shown by the dashed grey line (0.90 quantile p=0.028). The trait pairs that are known to be part of the LES are marked with a filled circle. The trait pairs that are not considered to be part of the LES are marked with a hollow circle. A triangular relationship, delineated by the quantile regression, would indicate that the strength of trait correlations establishes an upper limit to how different traits variance structures across scales can be, but not a lower limit.

51 TABLES

Table 1. Summary of LES trait correlation patterns in recent studies examining the LES at local ecological scales: within individuals, among ecotypes within species and among species within communities. Ecotypes are populations found in different environments. In Vasseur (2012), different genotypes were created (recombinant lines) by mixing genes from 2 different bioclimatic zones. In Wright & Suton-Grier (2012) three trait correlations were examined in six different environments for a total of 18 trait pair-by-environments combinations. In this study, three trait pairs were examined in three environments for a total of nine trait pair-by-environments combinations. Grey shading highlights the trait pattern corresponding to the predictions made by the hypothesis that the LES should be observed when a sufficient LLS range is spanned. No predictions are made for Wright & Suton-Grier and for this study because the extent of variation in LLS in these studies is unknown. Trait patterns Study design Study Organisms Environment Agree Disagree with LES with LES

Intra-individual Blonder et al. , 6 clones along elevation gradient, LMA-AMAX (among ramets (2013) tree Rocky mountains 4/6 clones n.s. within individual) Inter-population Martin et al. Metrosideros polymorpha, elevation gradient, LMA -LDMC (among ecotypes (2007) tree Hawai’i island within species) Albert et al. 12 species, elevation gradient, LMA-LDMC LMA-LNC (2010) many life-forms French Alps Vasseur et al. Arabidopsis thaliana, 2 bioclimatic zones, LMA-AMAX; (2012) forb Europe & South Africa LMA-LNC Niinemets. Quercus ilex, Across Q. ilex bioclimatic range, LMA-AMAX; LMA-LNC (2015) tree Europe LMA-LDMC Within community Wright & 22 species, Six wetland communities along **15/18 n.s. or (among species Suton-Grier. forbs & grasses flooding and N gradients, sign. opposite within (2012) greenhouse environment) Edwards et 32 deciduous Viburnum Living collection within Arboretum AMAX-NMASS AMAX-LMA; al. (2014) species. LLS range: 19-26 wks AMAX-LLS This study 10, 35 & 55 species, Three tropical communities along **5/9 n.s. trees precipitation & seasonality (Table A7) gradient, Panama

52

Table 2. Table 2. Summary of relative importance of interspecific variance in LMA and LDMC in six different studies partitioning variance in foliar traits across scales. We provided the number of ecological scales studied as a reference because studying a larger number of scales will affect the values reported for the interspecific scale: by definition, partitioning the total variance among a larger number of ecological scales will result in less variance at each scale.

# of ecological Species scale variance Study scales in study Biome & growth form LMA LDMC Difference Roche et al. 2004 3 Shrubland, various forms 65% 92% 27% Hulshof & Swenson 2010 4 Tropical Dry Forests 63% 20% 43% Messier et al. 2010 6 Tropical Dry & Rain Forests 21% 35% 14% Albert et al. 2010 3 Various biomes and forms ~65% ~79% 14% Auger & Shipley 2012 5 Temperate Forests 49% 70% 21% Kang et al. 2014 6 Subtropical Forests 51% 39% 12%

53

BOXES ======Beginning Box 1 ======Box 1. Drivers of phenotypic integration across scales The scales at which trait dimensions appear depend on the drivers shaping those trait correlations. In thinking about the drivers underpinning trait correlations, it is helpful to cast trait dimensions in the context of phenotypic integration. This field, which has a deep and broad history anchored in an evolutionary framework (Olson and Miller 1958, Pigliucci 2003) focuses on genetic and phenotypic patterns of covariance and the drivers underpinning them. Phenotypic integration posits that the functioning of complex phenotypes requires numerous traits to simultaneously ‘work’ together (Gould and Lewontin 1979, Cheverud 1982) so that they change in a coordinated fashion (Lewontin 1977, Murren 2002, Pigliucci and Preston 2004, Walsh and Blows 2009). Being part of a cohesive phenotype, individual traits cannot vary freely in response to different drivers of variation affecting them. Instead the realized phenotypic space to certain axes of variation (Raup and Michelson 1965). See Reich et al. (2003) and Donovan et al. (2011) for a discussion the drivers of phenotypic integration in the leaf economic spectrum. Clearly, traits can be correlated in response to the action of one or many drivers of phenotypic integration. These drivers create constraints which act at different ecological scales and are more or less flexible (Figure B2). There are many alternative ways to classify the variety of constraints (Antonovics and van Tienderen 1991). The list presented below provides a broad classification that stresses the variability of these types of constraint within and across scales. ● Biophysical constraints lead to obligatory relationships among traits that are impossible to avoid (Huxley 1935, Niklas and Kerchner 1984, Niklas 1992). For example, the photosynthetic biomachinery links maximum photosynthetic rate to leaf nitrogen concentration (Lambers et al. 2008) and Hagen-Poiseuille’s Law links the rate of sap flow to the average vessel diameter (Sutera and Skalak 1993). These “hard” constraints are consistent within and across all ecological scales.

54

● Genetic constraints (e.g. pleiotropy, epistasis or linkage) produce obligatory covariance relationships over ecological timescales but can be modified by natural selection or drift over evolutionary timescales (Smith et al. 1985, Roff 2000a, McGuigan 2006, Gardner and Latta 2007). They should lead to trait correlations from the metamer to population and species scale. The different genetic makeup of populations and species can create distinct trait relationships (in terms of slope, intercept and dispersion) within and among populations or taxa (Armbruster 1991, Merila and Bjorklund 2004). Here genetic constraints include developmental controls which are often considered to be epigenetic effects (Klingenberg 2004) and physiological constraints which are ultimately determined by genetics. ● Ecological filtering , i.e. natural selection for beneficial trait combinations in a given environment on ecological timescales, leads to trait associations within an environments (Olson and Miller 1958, Berg 1960). This creates “soft” constraints at the individual to community scales that can change over short temporal and spatial scales as the environment changes. Over time, consistent ecological filtering can lead to genetic correlations, which create more permanent trait associations (Cheverud 1984). Populations of a given species that exist in different environment (i.e. ecotypes) are subject to distinct ecological filtering. Different drivers of phenotypic integration can act at different ecological scales and lead to distinct trait correlation patterns (e.g. Figure B3 Panel B) (Armbruster 1991, Armbruster et al. 2004). Figure B1 Panel B illustrates the shape of hypothetical constraint for the three broad categories of drivers of phenotypic integration at three ecological scales. Studies like those of Kimball et al. (2013) and Armbruster (1991) provide elegant examples of how comparing patterns of trait correlation in nested ecological scales can sheds light on the drivers shaping phenotypic integration patterns at different scales.

55

Figure B1. Drivers of phenotypic integration across scales. A) Broad types of drivers of phenotypic integration and the scales at which they act. Neighborhood refers to adjacent individuals that directly influence each other. Community to refer to individuals interacting

56

more indirectly within a larger area. The species scale is portrayed outside of the ecological scale nesting because it does not fit clearly within the spatial hierarchy. The corresponding scales sampled in the case study are shown at the topic. These drivers produce trait correlations that are easy (soft constraints) to impossible (hard constraints) to escape. Over time, trait correlations due to ecological filtering can become genetic correlations, which are harder to escape. B) Illustration of the individual and combined effects of the broad types of drivers of phenotypic integration on trait correlations at different ecological scales (within populations, among populations in a species and among species). The triangles show the unavailable trait space due to each constraint. Biophysical constraints are consistent within and among scales. Genetic constraints can vary among individuals, populations and species in terms of the height of the regression (shown at the intra-population scale), the slope of the regression (shown at the inter-population scale), and the dispersion along the regression (shown at the interspecific scale). Ecological filtering constraints should be consistent across scales within an environment, but vary among environments. No ecological filtering constraint is shown for the species level since a species is made of different ecotypes and different species are found in different environments. The diagrams illustrates that different drivers of integration can be constant or variable within and across scales and combine to produce different net constraints at different scales (e.g. Fig. B3 panel B); the specific location, size and shapes of the constraints in the panels do not reflect specific hypotheses. ======End Box 1 ======

57

======Beginning Box 2 ======Box 2. Drivers of phenotypic variation across scales Functional traits vary across all ecological scales in response to a wide range of ecological and evolutionary drivers of phenotypic variation. Figure B2 shows the scales at which the broad categories of drivers of variation act. First, environmental gradients can cause trait variation via: ● Plasticity, the variable expression of a given genotype in response to the environment, acts at the metamer (leaves, flowers and ), module (a repeating unit composed of integrated parts, such as a branch and its leaves), and individual scales (Pigliucci 2001, de Kroon et al. 2005) ; ● Ecological filtering, the natural selection on ecological timescales of the individuals best suited to a given environment (Keddy 1992), acts at the individual, population, and community scales. Second, different genetic mechanisms can also cause trait variation among individuals, populations and species. ● Sexual reproduction, the different mechanisms reshuffling the parents’ alleles in the offspring (e.g. meiotic shuffling of alleles and recombination), create differences among individuals (Ridley 2003); ● Adaptive evolution, the selection for directional shifts in trait means over time, leads to differences among populations and species through time (Lande and Arnold 1983, Armbruster 1990, Merila and Bjorklund 1999, Walsh and Blows 2009); ● Genetic drift, the change in frequency of alleles due to random sampling, can lead to differences among populations and species (Roff 2000b). Third, ontogenetic mechanisms cause phenotypic differences at smaller scales, from metamers to individuals; ● The position of organs along the architectural plan of the plant leads to differences among metamers and modules along the ontogenetic progression (Barthélémy and Caraglio 2007).

58

This results from distinct internal environments (e.g. physiological and hormonal) of the metamers and modules along the plant axis. For example, leaf size varies with node position along the axis (Kaplan 2001); ● Age or stage of individuals and their organs, (Erickson and Michelini 1957, Huber and Stuefer 1997) create differences among mematers, modules and individuals; ● Developmental instability, the loss of ability by an organism to regulate its development during ontogeny, leads to differences among metamers, modules and individuals (Markow 1995). It is not feasible to directly measure all the drivers of phenotypic variation. Nonetheless, since different drivers of phenotypic variation act at different ecological scales, inferences on their relative effect sizes can be drawn from the variance structure across scales (i.e. the fraction of total variance occurring at different ecological scales) (Ackerly and Cornwell 2007, Lepš et al. 2011, Violle et al. 2012). With an appropriate sampling design, the variance structure across scales reflects the relative contribution of different drivers of phenotypic variation (McGill 2008). The rationale behind this approach is that if a given trait is largely affected by a given driver, then variation in this trait will also be large at the ecological scale(s) at which the driver acts. For example, if variation in Leaf Area is primarily driven by the precipitation regime, then we would expect that most of the variation in Leaf Area will occur among sites differing in their climate. However, if Leaf Area is mainly influenced by irradiance, then we would expect a large amount of variation between the sun and shade strata within a plant canopy. Note that confounding factors may act across scales, but with the appropriate hierarchical sampling design these effects can be accounted for in the variance partitioning analysis. For example, if variation in the mean trait value of different sites is due to differences in both environment and species composition, then when species composition is accounted for, the remaining variance at the site scale reflects environmental effects. This hierarchical approach can thus reasonably disentangle the effects of multiple drivers of

59

phenotypic variation and provide an alternative approach when the effects of the drivers of variation cannot all directly be measured.

Figure B2. Drivers of phenotypic variation across scales. Ecological scales are shown at the bottom and the corresponding sample scales sampled for the case study are shown at the top. Neighborhood refers to adjacent individuals that they directly influence each other. This corresponds to the plot sampling scale. In contrast, here we use community to refer to individuals interacting more indirectly within a larger area. This corresponds to the site sampling scale (See Appendix 1). “Environmental drivers” refers to both the biotic and abiotic components of the environment. The species scale is portrayed outside the hierarchy because this taxonomic scale it does not fit clearly within the spatial hierarchy. Specifically, populations can be grouped either (1) within communities by combining local populations of different species, or (2) within species by combining the different populations of a given species across its environmental range (see fig. B3A). ======End Box 2 ======

60

======Beginning Box 3 ======Box 3. Scaling frameworks: ecological scales, spatial scales and trait coverage Working across scales may not be straightforward because it involves overlapping scaling frameworks: biological and ecological hierarchy, spatial extent and trait coverage. Moving from smaller to larger ecological scales typically implies covering a larger spatial extent as well as a larger range of trait values (see Wiens 1989). Increasing spatial extent (Figure B3 Panel C) is typically associated with increasing environmental heterogeneity and increasing trait range because environmental turnover typically leads to species and trait turnover. Consequently, the effects of changing ecological scales can be difficult, if not impossible, to separate from the effects of increasing trait coverage (Figure B3, Panel B) and spatial extent. Figure B3 illustrates the overlap among these three scaling frameworks. In addition, the clear nested biological hierarchy organizing metamers into modules into individuals into populations becomes less clear above the population level: Populations of different species coexisting within an environment can be grouped into communities and biomes if the study focuses on habitats, localities, biodiversity or species interactions; Alternatively, the populations of a given species from distinct environments can be grouped into a species if the study focuses on species ranges, phylogeny, etc (Figure B3, Panel A). Here, we call these ‘ecological’ scales as they include both a biological hierarchy and units defined by interaction among individuals or within an environmental gradient (e.g. neighborhood, community).

61

Figure B3. Illustration of three different scaling frameworks: nested ecological scales, trait coverage and spatial scales. A) Ecological scales are mainly nested but populations can be grouped within communities or within species. The dashed lines indicate these alternate classifications. B) Trait coverage (ranges of trait values covered) typically increases with

62

larger ecological scale. C) Larger ecological scale typically involve larger spatial scale and therefore increased environmental heterogeneity. Spatial scales follow Kollmann (2000). ======End Box 3 ======

63

SUPPLEMENTARY MATERIAL - APPENDIX 1. MATERIALS AND METHODS

Leaf traits For each leaf samples, we measured five leaf functional traits reflecting different aspects of plant strategies and leaf function. LMA, the foliar dry mass per unit fresh area (g/m2), reflects the amount of biomass investment per unit of light capture area. LDMC, the ratio of a leaf’s dry mass to its water-saturated mass (g/g), reflects the tradeoff in investing resources in structural tissues versus liquid-phase processes. LDMC has been argued to be the central variable underpinning correlations among the traits in the leaf economic spectrum (Shipley et al. 2006) and has been shown to be a good proxy for leaf tissue density (Vile et al. 2005). Leaf nitrogen content (LNC), the fraction of a leaf’s total dry weight accounted by nitrogen (g/g), reflects photosynthetic capacity because foliar nitrogen is mostly present in RUBISCO and chlorophyll (Evans 1989). In leaves, nitrogen is also found in inducible anti- herbivory compounds such as alkaloids but they usually constitute a small amount of total nitrogen in healthy leaves. LMA, LDMC and LNC are correlated because they are part of the same ecological strategy dimension, known as the leaf economic spectrum. On one end of the spectrum, leaves have a high photosynthetic rate (high LNC), which usually entails thinner and/or less dense leaves (low LMA, low LDMC), as well as more vulnerable and shorter lived leaves. On the other end of the spectrum, leaves have a low photosynthetic rate (low LNC), which involves thicker (high LMA, high LDMC), more durable and longer lived leaves (Reich et al. 1999, Wright et al. 2004). Leaf carbon content (LCC), the fraction of a leaf’s total dry weight accounted by carbon (g/g), mostly reflects the leaves’ investment in structural support (Niinemets et al. 2007) and mechanical defense from herbivory (Coley and Barone 1996, Lucas et al. 2000). In the leaf, carbon is primarily found in the cell wall in the forms of lignin, cellulose, hemicellulose, and in carbohydrate compounds such as sugars and starch. Lignin, cellulose and hemicellulose are cell wall components that provide structural support and mechanical

64

protection to the leaf. Part of the leaf carbon content is also found in proteins, where it constitutes ca. 53% of protein’s weight (Vertregt and Devries 1987). Leaf Area, the projected area of one side of the leaf blade (cm2), is an architectural trait that is part of a strategy dimension known as Corner’s Rules (White 1983b, Brouat et al. 1998, Olson et al. 2009). Corner (Corner 1949) described two architectural rules: (1) the larger the plant appendage (, flower, leaf), the larger the twig or branch to which it is attached and (2) the more highly branched the twigs, the smaller their sizes. These allometric relationships have been suggested by Olson et al. (2009) to result from the combined effects of metabolic scaling (West et al. 1999) and constant carbon assimilation rate per unit crown area (Enquist et al. 1999), together leading to a trade-off between the mechanical support and transport functions of the stems and leaves. Plant architecture is ecologically important because it affects three fundamental plant functions: mechanical support, light capture through leaf spatial arrangement, and reproduction through flower display and seed dispersal (Niklas 1994). Finally, while Leaf Area is a component of LMA, it is well established as an independent axis of variation among species (e.g. Westoby et al. 2002, Poorter and Rozendaal 2008).

Sampling Design The sampling design spanned six important ecological scales: (1) among leaves within a stratum; (2) between strata within a tree; (3) among trees within a species; (4) among species; (5) among plots within a site and (6) among sites within a biome (Figure A1). Three sites were sampled: Parque Natural Metropolitano (PNM) located close to the Pacific coast, Barro Colorado Island (BCI) located on the Panama Canal and Parque Natural San Lorenzo located on the Atlantic coast (PNSL). These sites are located in lowland tropical rainforests along the Panama Canal and follow a strong precipitation and seasonality gradient. Further details on the study sites are provided in Table A1. Four (in PNM) or eight (in BCI and

PNSL) 400 m2 plots, systematically located 60-80 m apart center to center, were sampled at

65

each of the three sites. Each site covers ca. 1 (PNM) to ca. 3 ha in area (BCI and PNSL). Within each plot, all trees with a diameter at breast height (dbh) greater or equal to 10 cm were sampled. For each tree, three healthy and fully mature leaves were randomly sampled from a branch collected from each of the sun and shade stratum, yielding a total of 1910 leaf samples across 124 species (See Table A3 for species list). For elemental analysis, petioles were removed and each leaf was homogenized and ground to a fine powder using a Thomas Wiley Mini-Mill. The carbon and nitrogen content of the leaf blades were determined on 1.0-2.0 mg of ground leaf material using a Fison EA 1108 CHNS-O Elemental Analyzer. Elemental analyses were standardized using acetanilide, atropine and BBOT. All trait values were transformed using the natural logarithm to improve normality. The following describes which drivers of phenotypic variation are associated with each of the scales in this study. This results from the specifics of the sampling design. Differences among leaves within a stratum reflect developmental instability, the ontogenetic effects of metamer and module position along the plant and potentially the plastic response to micro-environmental gradients. To minimize trait variation due to leaf age we sampled leaves that are all fully mature but not senescing. Differences among strata within a tree mostly reflect the plastic response to light because strata were defined based on their exposure to sunlight. Developmental instability and other micro-environmental gradients can also cause variation among strata. Differences among trees within a species reflect developmental instability, tree age, sexual genetic mixing, and plastic and filtering responses to micro-environmental gradients. To minimize the effect of tree age on trait measurements, we only sampled mature trees (dbh >10cm). Unfortunately, the high alpha and beta diversities of the study system do not allow us to measure variation among populations of a species. Differences among species mainly reflect genetic differences resulting from adaptive evolution and drift. In this study, due to the high species richness and turnover, the species scale is mostly nested within sites and partly nested within plots. This means that variance at

66

the species scale also somewhat reflects the environmental effects due to differences among sites. However, Table A8 shows that the environment effect on species level variance is minor: When removing the species level from the analyses, the species level variance gets reassigned to the tree, plot and species levels and the majority of the variance originally attributed to the species level gets re-attributed to the tree level (66% for LDMC to 89% for Leaf Area) and the site and plot level variances increase but modestly. Among plots within a site, the average Sorensen’ similarity indices (Chao et al. 2005) are 0.33 for PNM, 0.13 for BCI and 0.23 for PNSL. Among sites, the indices are 0.02 between PNM and BCI, 0.26 between BCI and PNSL and 0.00 between PNM and PNSL. Despite this high species turnover, the species level does not fall entirely within the ecological hierarchy presented in Figure A1. Thus, in the analyses the species scale was crossed with the site scale (see the statistical analyses section for full details on the structure of the statistical model). In this study, the plots were located within a habitat with no noticeable environmental gradients among them. Differences due to species composition are accounted for at the species level and the effects of differences in species composition is removed from the plot-level variance. In the study, plot variance therefore reflects undetectable differences in the environment and the natural variability in trait composition that occurs within habitats. Since three study sites are located along a strong precipitation gradient, differences among them reflect community-level ecological filtering along a climatic gradient. Although there is a strong species turnover among sites, differences among species is accounted for at the species level and the effects of species composition is removed from the site-level variance.

Statistical analyses Note that we do not observe meaningful qualitative changes between this model crossing the species and site hierarchies and the model used in Messier et al. (2010), which

67

nests the species level within plots. Further, Table A8 shows that removing the species level from the analysis mainly leads to an increase in the variance at the tree level, with little change in the variance at the site level. This indicates that despite the species turnover across sites, the model is largely parses out species level variance from site-level variance. To calculate confidence intervals, tor each trait we created 500 simulated datasets by resampling the data with replacement. We then calculated the variance components on each simulated dataset. For each scale, the 95% confidence intervals were calculated from the results of the 500 variance components analyses. Palm fronds were too large to collect intact, and hence were excluded from analyses for Leaf Area. Some leaves were too small to provide sufficient material for the stoichiometry analyses and were also excluded from the statistical analyses. Sample size was thus n=1890 for LMA, n=1896 for LDMC, n=1860 for LNC, LCC and 1784 for Leaf Area. We used the rda() function of the vegan package in R (R Development Core Team 2011). We used the correlation matrix of the data in order to give each ecological scale an equivalent weight. Last, we calculated a traits dissimilarity index for their variance structure across scales by measuring the Euclidian distances between trait pairs along principal components 1 and 2. We weighted the distance between trait pairs along each principal component by the principal component’s eigenvalue.

68

Appendix 1 References

Ackerly, D. D. and Cornwell, W. K. 2007. A trait-based approach to community assembly: partitioning of species trait values into within- and among-community components. - Ecol. Lett. 10: 135–45. Albert, C. H. et al. 2010. A multi-trait approach reveals the structure and the relative importance of intra- vs. interspecific variability in plant traits. - Funct. Ecol. 24: 1192– 1201. Antonovics, J. and van Tienderen, P. H. 1991. Ontoecogenophyloconstraints? The chaos of constraint terminology. - Trends Ecol. Evol. 6: 166–168. Armbruster, S. W. 1990. Estimating and testing the shapes of adaptive surfaces - the morphology and pollination of Dalechampia blossoms. - Am. Nat. 135: 14–31. Armbruster, S. W. 1991. Multilevel analysis of morphometric data from natural plant populations: insights into ontogenetic, genetic, and selective correlations in Dalechampia scandens. - Evolution (N. Y). 45: 1229–1244. Armbruster, S. W. et al. 2004. Floral Integration, Modularity, and Accuracy. - In: Pigliucci, M. and Preston, K. (eds), Phenotypic Integration: Studying the Ecology and Evolution of Comlpex Phenotypes. Oxford University Press, pp. 23–49. Auger, S. and Shipley, B. 2012. Inter-specific and intra-specific trait variation along short environmental gradients in an old-growth temperate forest (F de Bello, Ed.). - J. Veg. Sci. 24: 419–428. Barthélémy, D. and Caraglio, Y. 2007. Plant architecture: A dynamic, multilevel and comprehensive approach to plant form, structure and ontogeny. - Ann. Bot. 99: 375– 407. Berg, R. L. 1960. The ecological significance of correlation pleiades. - Evolution (N. Y). 14: 171–180. Blonder, B. et al. 2013. Assessing the causes and scales of the leaf economics spectrum using venation networks in Populus tremuloides. - J. Ecol. 101: 981–989. Brouat, C. et al. 1998. Corner’s rules revisited: ontogenetic and interspecific patterns in leaf- stem allometry. - New Phytol. 139: 459–470. Chao, A. et al. 2005. A new statistical approach for assessing similarity of species compostition with incidence and abundance data. - Ecol. Lett. 8: 148–159. Cheverud, J. M. 1982. Phenotypic, genetic and environmental morphological integration in the cranium. - Evolution (N. Y). 36: 499–516. Cheverud, J. M. 1984. Quantitative genetics and developmental constraints on evolution by selection. - J. Theor. Biol. 110: 155–171.

69

Coley, P. D. and Barone, J. a. 1996. Herbivory and Plant Defenses in Tropical Forests. - Annu. Rev. Ecol. Syst. 27: 305–335. Condit, R. et al. 2004. Tropical forest dynamics across a rainfall gradient and the impact of an El Ni ˜ no dry season.: 51–72. Cornelissen, J. H. C. et al. 2003. A handbook of protocols for standardised and easy measurement of plant functional traits worldwide. - Aust. J. Bot. 51: 335. Corner, E. J. H. 1949. The durian theory or the origin of the modern tree. - Ann. Bot. 13: 367–414. de Kroon, H. et al. 2005. A modular concept of phenotypic plasticity in plants. - New Phytol. 166: 73–82. Donovan, L. a et al. 2011. The evolution of the worldwide leaf economics spectrum. - Trends Ecol. Evol. 26: 88–95. Edwards, E. J. et al. 2014. Leaf life span and the leaf economic spectrum in the context of whole plant architecture. - J. Ecol. 102: 328–336. Enquist, B. J. et al. 1999. Allometric scaling of production and life-history variation in vascular plants. - Nature 401: 907–911. Enquist, B. J. et al. 2015. Scaling from traits to ecosystems: Developing a general Trait Driver Theory via integrating trait-based and metabolic scaling theories. - Adv. Ecol. Res. in press. Erickson, R. O. and Michelini, F. J. 1957. The Plastochron Index. - Am. J. Bot. 44: 297–305. Evans, J. R. 1989. Photosynthesis and nitrogen relationships in leaves of C3 plants. - Oecologia 78: 9–19. Falster, D. S. et al. 2011. Influence of four major plant traits on average height, leaf-area cover, net primary productivity, and biomass density in single-species forests: A theoretical investigation. - J. Ecol. 99: 148–164. Funk, J. L. and Cornwell, W. K. 2013. Leaf traits within communities: Context may affect the mapping of traits to function. - Ecology 94: 1893–1897. Gardner, K. M. and Latta, R. G. 2007. Shared quantitative trait loci underlying the genetic correlation between continuous traits. - Mol. Ecol. 16: 4195–209. Gould, S. J. and Lewontin, R. C. 1979. The spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist programme. - Proc. R. Soc. London 205: 581– 98. Grime, J. 1974. Vegetation classification by reference to strategies. - Nature 250: 26–30. Grime, J. et al. 1997. Integrated screening validates primary axes of specialisation in plants. - Oikos 79: 259–281.

70

Huber, H. and Stuefer, J. F. 1997. Shade-induced changes in the branching pattern of a stoloniferous herb: Functional response or allometric effect? - Oecologia 110: 478–486. Hulshof, C. M. and Swenson, N. G. 2010. Variation in leaf functional trait values within and across individuals and species: an example from a Costa Rican dry forest. - Funct. Ecol. 24: 217–223. Hulshof, C. M. et al. 2013. Intra-specific and inter-specific variation in specific leaf area reveal the importance of abiotic and biotic drivers of species diversity across elevation and latitude. - J. Veg. Sci. 24: 921–931. Huxley, J. S. 1935. The Problems of Relative Growth. - Methuen & Co.ltd. Jakobsson, A. and Eriksson, O. 2000. A comparative study of seed number, seed size, seedling size and recruitment in grassland plants. - OIKOS 88: 494–502. Kang, M. et al. 2014. Trait variability differs between leaf and wood tissues across ecological scales in subtropical forests (S Bartha, Ed.). - J. Veg. Sci. 25: 703–714. Kaplan, D. R. 2001. The science of plant morphology: History, definition, and role in modern biology. - Am. J. Bot. 88: 1711–1741. Keddy, P. 1992. Assembly and response rules: two goals for predictive community ecology. - J. Veg. Sci. 3: 157–164. Kimball, S. et al. 2013. Phenotypic selection favors missing trait combinations in coexisting annual plants. - Am. Nat. 182: 191–207. Klingenberg, C. P. 2004. Integration, Modules, and Development. - In: Pigliucci, M. and Preston, K. (eds), Phenotypic Integration: Studyinf the Ecology and Evolution of Comlpex Phenotypes. Oxford University Press, pp. 213–227. Kollmann, J. 2000. Dispersal of fleshy-fruited species: a matter of spatial scale? - Perspect. Plant Ecol. Evol. Syst. 3: 29–51. Kraft, N. J. B. et al. 2015. Plant functional traits and the multidimensional nature of species coexistence. - Proc. Natl. Acad. Sci. 112: 797–802. Lambers, H. et al. 2008. Plant physiological ecology. - Springer. Lande, R. and Arnold, S. J. J. 1983. The measurement of selection on correlated characters. - Evolution (N. Y). 37: 1210–1226. Laughlin, D. C. and Messier, J. 2015. Fitness of multidimensional phenotypes in dynamic adaptive landscapes. - Trends Ecol. Evol.: 1–10. Laughlin, D. C. et al. 2010. A multi-trait test of the leaf-height-seed plant strategy scheme with 133 species from a pine forest flora. - Funct. Ecol. 24: 493–501. Lavorel, S. and Garnier, E. 2002. Predicting changes in community composition and ecosystem functioning from plant traits: revisiting the Holy Grail. - Funct. Ecol. 16:

71

545–556. Leigh, E. G. et al. 2004. Barro Colorado Island Forest Dynamics Plot, Panama. - In: Losos, E. C. and Leigh Jr., E. G. (eds), Tropical forest diversity and dynamism: Findings from a large-scale plot network. University of Chicago Press, pp. 451–463. Lepš, J. et al. 2011. Community trait response to environment: disentangling species turnover vs intraspecific trait variability effects. - Ecography (Cop.). 34: 856–863. Lewontin, R. 1977. Adaptation. - In: Levins, R. and Lewontin, R. (eds), The Dialectical Biologist. Harvard University Press, pp. 65–84. Long, W. et al. 2011. Within- and among-species variation in specific leaf area drive community assembly in a tropical cloud forest. - Oecologia 167: 1103–1113. Lucas, P. W. et al. 2000. Mechanical Defences to Herbivory. - Ann. Bot. 86: 913–920. Markow, T. A. 1995. Evolutionary Ecology and Developmental Instability. - Annu. Rev. Entomol. 40: 105–120. Martin, R. E. et al. 2007. Genetic variation in leaf pigment, optical and photosynthetic function among diverse phenotypes of Metrosideros polymorpha grown in a common garden. - Oecologia 151: 387–400. McGill, B. J. 2008. Exploring predictions of abundance from body mass using hierarchical comparative approaches. - Am. Nat. 172: 88–101. McGill, B. J. et al. 2006. Rebuilding community ecology from functional traits. - Trends Ecol. Evol. 21: 178–85. McGuigan, K. 2006. Studying phenotypic evolution using multivariate quantitative genetics. - Mol. Ecol. 15: 883–96. Merila, J. and Bjorklund, M. 1999. Population divergence and morphometric integration in the green-finch (Carduelis chloris) - evolution against the trajectory of least resistance ? - J. Evol. Biol. 12: 103–112. Merila, J. and Bjorklund, M. 2004. Phenotypic Integration as a Constraint and Adaptation. - In: Pigliucci, M. and Preston, K. (eds), Phenotypic Integration: Studyinf the Ecology and Evolution of Comlpex Phenotypes. Oxford University Press, pp. 107–129. Messier, J. et al. 2010. How do traits vary across ecological scales? A case for trait-based ecology. - Ecol. Lett. 13: 838–48. Murren, C. J. 2002. Phenotypic integration in plants. - Plant Species Biol. 17: 89–99. Niinemets, U. 2015. Is there a species spectrum within the world-wide leaf economics spectrum ? Major variations in leaf functional traits in the Mediterranean sclerophyll Quercus ilex. - New Phytol.: 79–96. Niinemets, U. et al. 2007. Do we underestimate the importance of leaf size in plant

72

economics? Disproportional scaling of support costs within the spectrum of leaf physiognomy. - Ann. Bot. 100: 283–303. Niklas, K. J. 1992. Plant Biomechanics: an engineering approach to plant form and function. - University of Chicago Press. Niklas, K. J. 1994. Morphological evolution through complex domains of fitness. - Proc. Natl. Acad. Sci. U. S. A. 91: 6772–6779. Niklas, K. J. and Kerchner, V. 1984. Mechanical and photosynthetic constraits on the evolution of plant shape. - Paleobiology 10: 79–101. Olson, E. C. and Miller, R. L. 1958. Morphological Integration. - University of Chicago Press. Olson, M. E. et al. 2009. Universal foliage-stem scaling across environments and species in dicot trees: plasticity, biomechanics and Corner’s Rules. - Ecol. Lett. 12: 210–9. Pérez-Harguindeguy, N. et al. 2013. New handbook for standardised measurement of plant functional traits worldwide. - Aust. J. Bot. 61: 167–234. Pigliucci, M. 2001. Phenotypic Plasticity: Beyond Nature and Nurture. - Johns Hopkins University Press. Pigliucci, M. 2003. Phenotypic integration: studying the ecology and evolution of complex phenotypes. - Ecol. Lett. 6: 265–272. Pigliucci, M. and Preston, K. A. 2004. Phenotypic Integration: Studying the Ecology and Evolution of Complex Phenotypes. - Oxford University Press. Poorter, L. and Rozendaal, D. M. a 2008. Leaf size and leaf display of thirty-eight tropical tree species. - Oecologia 158: 35–46. R Development Core Team, R. 2011. R: A Language and Environment for Statistical Computing (RDC Team, Ed.). - R Found. Stat. Comput. 1: 409. Raup, D. M. and Michelson, a 1965. Theoretical Morphology of the Coiled Shell. - Science 147: 1294–5. Reich, P. B. 2014. The world-wide “fast-slow” plant economics spectrum: a traits manifesto (H Cornelissen, Ed.). - J. Ecol. 102: 275–301. Reich, P. B. and Wright, I. J. 2003. The evolution of plant functional variation: traits, spectra, and strategies. - Int. J. Plant Sci. 164: S143–S164. Reich, P. B. et al. 1999. Generality of leaf trait relationships: A test across six biomes. - Ecology 80: 1955–1969. Ridley, M. 2003. Evolution. - Wiley-Blackwell. Roche, P. et al. 2004. Congruency analysis of species ranking based on leaf traits: which traits are the more reliable? - Plant Ecol. Former. `Vegetatio’ 174: 37–48. Roff, D. 2000a. Short Review The evolution of the G matrix : selection or drift ? in press.

73

Roff, D. 2000b. The evolution of the G matrix: selection or drift? - Heredity (Edinb). 84: 135– 142. Santiago, L. S. and Mulkey, S. S. 2005. Leaf productivity along a precipitation gradient in lowland Panama: patterns from leaf to ecosystem. - Trees 19: 349–356. Santiago, L. S. et al. 2004. Coordinated changes in photosynthesis, water relations and leaf nutritional traits of canopy trees along a precipitation gradient in lowland tropical forest. - Oecologia 139: 495–502. Shipley, B. et al. 2006. Fundamental trade-offs generating the worldwide leaf economics spectrum. - Ecology 87: 535–41. Siefert, A. et al. 2014. Community Functional Responses to Soil and Climate at Multiple Spatial Scales : When Does Intraspecific Variation Matter ? - PLoS One 9: 1–9. Smith, J. M. et al. 1985. Developmental constraints and evolution. - Q. Rev. Biol. 60: 265– 287. Smithsonian Tropical Research Institute 2007. Physical Monitoring Program. Sutera, S. P. and Skalak, R. 1993. The History of Poiseuille Law. - Annu. Rev. Fluid Mech. 25: 1. Swenson, N. and Enquist, B. J. 2007. Ecological and evolutionary determinants of a key plant functional trait: wood density and its communitywide variation across latitutde and elevation. - Am. J. Bot. 94: 451–459. Vasseur, F. et al. 2012. A common genetic basis to the origin of the leaf economics spectrum and metabolic scaling allometry. - Ecol. Lett. 15: 1149–57. Vertregt, N. and Devries, F. 1987. A rapid method for determining the efficiency of biosynthesis of plant biomass. - J. Theor. Biol. 128: 109–119. Vile, D. et al. 2005. Specific leaf area and dry matter content estimate thickness in laminar leaves. - Ann. Bot. 96: 1129–36. Violle, C. et al. 2012. The return of the variance: intraspecific variability in community ecology. - Trends Ecol. Evol. 27: 1–9. Walsh, B. and Blows, M. W. 2009. Abundant Genetic Variation plus Strong Selection = Multivariate Genetic Constraints: A Geometric View of Adaptation. - Annu. Rev. Ecol. Evol. Syst. 40: 41–59. Weiher, E. and Keddy, P. A. 1999. Ecological Assembly rules: Perspectives, Advances, Retreats. - Cambridge University Press. West, G. B. et al. 1999. A general model for the structure and allometry of plant vascular systems. - Nature 400: 664–667. Westoby, M. 1998. A leaf-height-seed (LHS) plant ecology strategy scheme. - Plant Soil 199:

74

213–227. Westoby, M. et al. 2002. Plant ecological strategies: Some leading dimensions of variation between species. - Annu. Rev. Ecol. Syst. 33: 125–159. White, P. S. 1983a. Evidence that temperate east North-American evergreen woody-plants follow Corner’s rules. - New Phytol. 95: 139–145. White, P. S. 1983b. Corner’s rules in eastern deciduous trees : Allometry and its implications for the adaptive architecture of trees. - Bull. Torrey Bot. Club 110: 203–212. Wiens, J. A. 1989. Spatial scaling in ecology. - Funct. Ecol. 3: 385–397. Wright, J. P. and Sutton-Grier, A. 2012. Does the leaf economic spectrum hold within local species pools across varying environmental conditions? - Funct. Ecol. 26: 1390–1398. Wright, I. J. et al. 2004. The worldwide leaf economics spectrum. - Nature 428: 821–827.

75

SUPPLEMENTARY MATERIAL - APPENDIX 2. SUPPLEMENTARY TABLES AND FIGURES

Table A1. Description of the three study sites located along the Panama Canal. PNM: Parque Natural San Lorenzo, BCI: Barro Colorado Island, PNSL: Parque Natural San Lorenzo, MAP: Mean Annual Precipitation, MDSL: Mean Dry Season Length. Information from: 1 – (Smithsonian Tropical Research Institute 2007), 2- (Santiago and Mulkey 2005), 3- (Santiago et al. 2004), 4- (Condit et al. 2004), 5- (Leigh et al. 2004). * : calculated as the mean interval during which potential evapotranspiration (PET) exceeds rainfall. ** : species with individuals with stems ≥10 cm in diameter.

Site Location MAP MDSL* Parent Elevation Richness Tree density** (mm) (days) material (m) (sp/ha) (#/ha) PNM 18°59'N 11850 2129 2Volcanic 360 336 3318 79°33'W BCI 19°10'N 12620 4118 5Volcanic 540 591 5429 79°51'W PNSL 19°17'N 13020 2102 2Sedimentary 3140 387 3659 79°58'W

76

Table A2. Comparison of the Minimum, Maximum, Mean, Max/Min and Coefficient of Variation of the five leaf functional traits studied with the GLOPNET dataset. The statistics presented are for not-transformed data. Of the five study traits, only LMA and LNC values are available in the GLOPNET dataset. Leaf Mass per Area, LMA (g·m-2); Leaf Dry Matter Content, LDMC (g·g-1); Leaf Nitrogen Content, LNC (g·g-1·100-1); Leaf Carbon Content, LCC

(g·g-1·100-1), Leaf Area, Area (m2).

PANAMA GLOPNET

LMA LDMC LNC LCC Leaf Area LMA LNC

Min 22.29 0.08 0.68 33.16 0.0002 14.45 0.25 Max 235.6 0.72 5.96 58.03 0.39 1514 6.35 1402 Variance 970 0.005 0.49 13.97 0.0008 8 0.96 Max/Min 10.5 8.53 9.61 1.75 1938 104 25.40 CV (sd/mean) 0.36 0.20 0.33 0.08 2.19 0.93 1.05

77

Table A3. Species List Alibertia edulis Guarea guidonia Picramnia latifolia Alseis blackiana Guatteria dumetorum Piper reticulatum Anacardium excelsum Guazuma ulmifolia Pittoniotis trichantha Andira inermis Guettarda foliacea Platypodium elegans Apeiba membranacea Hamelia axillaris Poulsenia armata Aspidosperma cruentum Hasseltia floribunda Pourouma bicolor Aspidosperma spruceanum Heisteria acuminata Pouteria reticulata Astrocaryum standleyanum Heisteria concinna Protium costaricense Astronium graveolens Herrania purpurea Protium panamense Beilschmiedia pendula Hieronyma alchorneoides Protium tenuifolium alicastrum Hirtella triandra Psychotria horizontalis Brosimum guianensis Humiriastrum diguense Pterocarpus rohrii Brosimum utile Hybanthus prunifolius Quararibea asterolepis Carapa guianensis Inga nobilis Randia armata Cassipourea elliptica Inga pezizifera Simarouba amara Castilla elastica Inga sapindoides Sloanea meianthera Cecropia insignis Jacaranda copaia Sloanea terniflora Cecropia obtusifolia Lacistema aggregatum Socratea exorrhiza Cespedesia spathulata Lacmellea panamensis Spondias mombin Chamguava schippii Lindackeria laurina Symphonia globulifera Chimarrhis parviflora Luehea seemannii Tabebuia guayacan Chrysophyllum argenteum Manilkara bidentata Tabernaemontana arborea Cinnamomum triplinerve Maquira guianensis Tachigali versicolor Cordia alliodora Maranthes panamensis Tapirira guianensis Cordia bicolor Marila laxiflora Terminalia oblonga Croton billbergianus Matayba apetala Theobroma bernoullii Cupania scrobiculata Miconia elata Tovomita longifolia Dendropanax arboreus Miconia ligulata Trattinnickia aspera Desmopsis panamensis Miconia minutiflora Trichilia poeppigii Diospyros artanthifolia Miconia sp Trichilia tuberculata Dussia sp Mortoniodendron anisophyllum Turpinia occidentalis Eugenia coloradoensis Nectandra purpurea Unonopsis panamensis Eugenia nesiotica Ochroma pyramidale Unonopsis pittieri Eugenia oerstediana Ocotea cernua Virola elongata Faramea occidentalis Ocotea dendrodaphne Virola multiflora Ficus insipida Ocotea ira Virola sebifera Ficus maxima Oenocarpus mapora Virola surinamensis Garcinia intermedia Ormosia coccinea Vochysia ferruginea Garcinia madruno Palicourea guianensisv Xylopia macrantha Guapira standleyana Perebea xanthochyma

78

Table A4. Correlations and variances among the five study traits. Trait variances are in bold along the diagonal, the Pearson correlation coefficients (r), are located below the diagonal and statistical significance of the correlations (p-values) are above the diagonal. Significant correlations are also marked with a star. Each data point is a leaf-level measurement. Data were natural log transformed.

LMA LDMC LNC LCC Leaf Area

LMA 0.131 < 2.2e-16 < 2.2e-16 < 2.2e-16 0.227 LDMC 0.585* 0.046 < 2.2e-16 < 2.2e-16 5.06e-15 LNC -0.526* -0.411* 0.092 0.017 0.829 LCC 0.292* 0.262* -0.055* 0.006 < 2.2e-16 Leaf Area 0.029 -0.185* 0.005 -0.252* 1.062

79

Table A5. Variance partitioning analyses of the five leaf traits giving the percentage of total variance explained by each scale and their confidence intervals (0.025 and 0.975 quantiles) calculated by bootstrapping. Data was normalized using natural log transformation.

Scale LMA LDMC LNC LCC Leaf Area

Leaf & Error 11 (7-10) 17 (10-17) 11 (7-9) 10 (6-10) 7 (5-6) Strata 17(15-21) 11 (10-17) 6 (6-11) 8(8-14) 6 (6-9) Tree 21 (18-25) 13 (10-20) 16 (13-19) 12 (5-14) 7 (5-9) Species 30 (26-34) 44 (39-49) 44 (40-48) 45 (42-49) 79 (77-80) Plot 0 (0-1) 0 (0-1) 3 (2-5) 3 (2-4) 0 (0-1) Site 21 (19-25) 15 (11-19) 20 (18-24) 22 (20-24) 1 (1-2)

80

Table A6. Loadings of Principal Components Analysis on variance component table

PC1 PC2 Proportion of total variance 0.58 0.27 Loadings Leaf & Error 0.29 0.25 Strata 042 0.42 Tree 0.49 -0.02 Species - 0.53 0.10 Plot 0.01 - 0.78 Site 0.46 - 0.37

81

Table A7. Trait standardized major axis regressions (sma function, smatr package, R software) for three leaf economic spectrum traits among species in three communities spanning a precipitation and seasonality gradient along the panama canal (Table A1). PNM n=10, BCI n=58, SANLO n=35. r-values, p-values and slopes are given only for the significant correlations.

Traits Community

PNM BCI PNSL

LMA- LNC n.s. r= 0.49 r= 0.72 slope= -1.06 slope = -1.04

LMA-LDMC n.s. r= 0.55 n.s. slope = 1.58

LNC-LDMC n.s. r= 0.53 n.s. slope = - 1.49

82

Table A8. Variance partitioning analyses of the five leaf traits without the species scale.

Scale LMA LDMC LNC LCC Leaf Area

Leaf & Error 11 16 11 9 9 Strata 17 11 6 9 6 Tree 43 52 50 46 77 Plot 0 3 6 7 6 Site 29 19 26 31 1

83

Figure A1. Sampling design illustrating the six ecological scales and their nested structure.

84

Figure A2 – Principal component analysis of the five leaf traits at the leaf level. Data natural log transformed and standardized

85 APPENDIX B:

Interspecific integration of trait dimensions at local scales: plant architecture is at the center of the phenotypic network Submitted as: Messier, Julie, Lechowicz, Martin J., McGill, Brian J., Violle, Cyrille, Enquist, Brian, J. In Review. Interspecific integration of trait dimensions at local scales: plant architecture is at the center of the phenotypic network. Axios Review.

Authors: 1Julie Messier, 2Martin J. Lechowicz, 3Brian J. McGill, 4Cyrille Violle & 1Brian J. Enquist

Affiliations : 1Ecology and Evolutionary Biology, University of Arizona, Tucson, AZ 85721, USA

2Biology Department, McGill University, Montréal, QC, H3A1B1, Canada. 3School of Biology and Ecology, University of Maine, Orono, ME 04469, USA.

4Centre d’Écologie Fonctionelle et Évolutive, CNRS, Montpellier, 34293, France

Corresponding author: Julie Messier University of Arizona 1041 E. Lowell Street, Tucson, Arizona, 85711, USA 520-626-3336 [email protected]

Keywords: architecture; Corner’s Rules; deciduous trees; Leaf Economic Spectrum; local scales; trait dimensions; phenotypic integration; phenotypic network; Wood Spectrum

Short title: integration of trait dimensions

86 ABSTRACT

1. The plant phenotype is structured by the interplay between evolution and ecophysiological and biophysical tradeoffs and constraints. Patterns of trait covariation have identified closely integrated groups of traits forming trait dimensions. However, trait dimensions have typically been examined separately so that their degree of interdependence remains unclear. In this study we examine the relationships among these trait dimensions by considering them simultaneously.

2. Using empirical data gathered on individuals of 24 locally coexisting tree species in a temperate forest, we examine the correlation structure of 20 leaf, branch, stem and root traits. These traits make up three well established trait dimensions that characterize different ecophysiological functions: resource acquisition, hydraulic, mechanical support and architecture. Using both narrow and broad definitions of trait dimensions, we test whether the phenotype is organized along these distinct trait dimensions.

3. Patterns of trait covariation do not segregate into clear trait dimensions. Instead, many traits from distinct trait dimensions are correlated together into a trait network. The expected trait relationships defining the trait dimensions are absent in the wood spectrum, weak in the leaf economic spectrum and present for Corner’s rules.

4. Our results indicate that trait dimensions described across broad taxonomic and environmental scales are not independent among locally coexisting species. Importantly, two architectural traits appear as most central to the phenotypic network. This suggests that architecture may be central to resource, mechanical and hydraulic integration.

5. These results suggest that local and global drivers and patterns of phenotypic integration may be distinct. This questions the use of the concept of trait dimensions at local scales. Instead, a network view might better portray the phenotype locally. Such an approach offers the promise of identifying core traits that reflect the multiple trade-off and constraints shaping the phenotype locally.

87 INTRODUCTION

The plant phenotype is structured by ecophysiological and biophysical constraints and tradeoffs that can be captured by trait dimensions (Chapin et al. 1993). Describing different aspects of plant ecological strategies has been an important research goal in plant comparative ecology (Grime 1977, Southwood 1988, Westoby et al. 2002, Craine 2009). While Grime (1977) offered a description of plant strategy dimensions based on the environmental challenges they face (disturbances, physiological stress or competition), Westoby (1998) proposed trait-based strategy dimensions as a generalizable way to compare strategies across species. Despite the importance of trait dimensions in comparative plant ecology, the position of trait dimensions within the broader phenotype is not clear: the independence of these dimensions as well as their individual limits are unknown. To explore these issues, we examine three important trait dimensions that together organize the non-reproductive tree phenotype: the leaf economic spectrum (LES; Reich et al. 1999, Wright et al. 2004), the wood spectrum (WS; Chave et al. 2009) and Corner’s rules (CR; Corner 1949, Hallé et al. 1978, White 1983a). These axes represent dimensions along which plant species differ from each other (Westoby et al. 2002, Westoby and Wright 2006) (cf. Fig 1). Of these axes, the LES is the most studied and best understood. The LES reflects a trade-off between two vital physiological functions, resource acquisition and conservation. In essence, it captures a trade-off between maximum photosynthetic rate and leaf life-span. Other traits, such as leaf mass per area and leaf nitrogen concentration, are also associated with these traits. The traits that define the Wood Spectrum are measured in the xylem of branches and stems. Centered on wood density, the Wood Spectrum is thought to reflect trade-offs among three important aspects of physiological functions served by the xylem: transport safety, transport efficiency, and mechanical support (cf. Fig 1). Based on first principles, a set of trade-offs is expected among traits reflecting these physiological functions (Baas et

88 al. 2004, Chave et al. 2009). However, strong empirical evidence for the presence of one or more trait dimensions in the xylem is lacking (Chave et al. 2009). Last, Corner’s Rules is an architectural dimension describing relationships among leaf size, branch diameter and branching distance. These relationships are broadly observed across environments and phylogeny (White 1983a, b, Midgley and Bond 1989, Brouat et al. 1998, Ackerly and Donoghue 1998, Cornelissen 1999). Architectural traits are thought to not be directly related to physiological functions but instead to describe alternate designs to display leaves and reproductive organs. Architecture in general arises as natural selection shapes plant form to optimize conflicting functions, namely sap transport, mechanical support, light harvesting and seed dispersal (Niklas 1992, 1994b, Enquist and Bentley 2012). A mechanism shaping the relationships among these traits that includes branch wood density has been proposed by Olson and colleagues (Olson et al. 2009). It essentially involves metabolic proportionalities described by the Metabolic Scaling Theory (West et al. 1997, 1999, Enquist and Niklas 2002). White (1983b) proposed that the different architectural designs across environments was adaptive: plants with distinct statures and shade tolerance levels express differential allocation of biomass for structural support between leaves and stems.

There are two limitations to the current focus on trait-based strategy dimensions. First, despite extensive work on plant strategies in comparative ecology, trait dimensions have typically been studied separately. As a result, it is not clear how independent trait dimensions are. Ecological strategy theory and evolutionary biology frameworks lead to contrasting expectations regarding the relationships among trait dimensions. On one hand, ecological strategy theories suggest some independence among major trait dimensions (Berg 1960, Westoby 1998). Empirical evidence of independence among plant size and leaf economics has indeed been shown (e.g. Wright et al. 2007, Price et al. 2014). Importantly, a common methodology for identifying trait dimensions is to use ordination. By definition, the family of ordination techniques identifies orthogonal axes of variation and thus explicitly assumes orthogonality among trait dimensions. On the other hand, we

89 know from evolutionary biology that performance and fitness depends on the constellation of traits that co-vary to differing degrees and together characterize the whole phenotype. This framework thus predicts a minimum integration among trait dimensions (Cheverud 1982, Murren 2002). For example, Vasseur et al. (2012) found the same genes to be closely related with variation in leaf economic traits and plant size in Arabidopsis thaliana, two dimensions a priori assumed to be independent (Westoby 1998). In this study we address the relationships among trait dimensions by simultaneously examining different trait dimensions.

Second, the limits delineating any single trait dimension within the broader context of an integrated phenotype are also unclear. In fact, the boundaries of trait membership within any single trait dimensions are unknown. Ultimately, interpreting trait variability will best be advanced by examining trait covariation within the context of the integrated plant phenotype, an approach taken by studies examining intraspecific (P matrices) and interspecific (Q matrices) trait covariation matrices (Merila and Bjorklund 2004). Any examination of the LES, WS, and CR must take into account how any given trait operates within the broader biological context of the individual plant (Bonser 2006). For example, although two major trait dimensions, the LES and the WS, are well studied and understood in terms of fundamental trade-offs among vital physiological functions (cf. Fig 1)(Reich et al. 1999, Baas et al. 2004), most studies use a narrow set of traits to quantify these trait dimensions. Within the context of an integrated phenotype, the question thus remains open as to which traits can and cannot be included within these trait dimensions. Do the LES, WES, and CR trait dimensions include all traits closely reflecting the underlying trade-off (hereafter, ‘broadly defined’ trait dimension), or do they only include the traditionally studied set of traits (hereafter, ‘narrowly defined’ trait dimension)? Trade-offs in physiological functions underpin the LES and WS dimensions, while CR dimension involves architectural designs providing alternate ways to effectively display leaves and reproductive structures (Niklas and Kerchner 1984, Niklas 1992). We

90 might reasonably expect the LES dimension to extend beyond leaves to also include traits from other organs reflecting resource acquisition and conservation (Freschet et al. 2010). For example, is specific root length (SRL) part of the LES dimension? This might be expected since it is the absorptive root analogue to LMA. Contrasting results have been found for SRL membership within the LES dimension (Craine et al. 2001, Laughlin et al. 2010, Kembel and Cahill 2011, Fortunel et al. 2012), possibly because fine roots acquire a variety of resources that possess different dynamics, because the soil environment is much more complex and heterogeneous than the above ground environment, or because root growth is continuous and the roles they play change across ontogeny (see Poorter and Ryser 2015 for a complete discussion). The boundaries delineating trait membership in the WS are also ambiguous (Chave et al. 2009). Much work has examined xylem trait relationships to clarify the interplay among sap transport efficiency, safety, and mechanical support (e.g. Davis et al. 1999, Hacke and Sperry 2001, Hacke et al. 2001, Pratt et al. 2007), but the set of traits involved in the wood spectrum and the contingencies defining their relationships remains uncertain. To address this, here we assess the limits delineating each of the three trait dimensions under study by comparing the empirical support for ‘narrowly defined’ trait dimensions versus ‘broadly defined’ trait dimensions. In this study, we explore the independence of trait dimensions and the limits of trait membership, by examining trait covariation in locally co-occurring angiosperm tree species, and ask: (1) are the LES, WS and CR trait dimensions independent from each other? ; and (2) do the LES, WS and CR trait dimensions include only the traditionally studied traits or do they encompass all traits reflecting to the imputed underlying trade- offs? To answer these questions, we examine the interspecific correlation structure among 20 traits in saplings from 24 locally coexisting temperate tree species. These traits, measured on different organs (leaves, branches, stems, roots), were sampled on 15 to 20 individuals per species to characterize plant architecture and four vital physiological functions: resource acquisition, conservation, sap transport and mechanical support. We use three statistical approaches to describe and test trait correlation structure, namely

91 principal component analyses to look for groups of traits with orthogonal variation, Mantel tests to compare the support for broad and narrow trait dimensions and network analyses to describe the trait correlation network, examine trait centrality and test for modularity of the trait dimensions.

92 METHODS

Study Site. We measured traits for each of 380 tree saplings located on or around Mont Saint-Hilaire, Canada (45°33′8″N 73°9′3″W). This hill, located in Southwestern Québec in the transition zone between the boreal forests to the north and the eastern deciduous forests to the south, is of plutonic origin. It includes a series of seven hilltops that rise up to 414m above the sedimentary floor of the surrounding Saint-Lawrence river valley (Feininger et al. 1995). The mixed wood forest on the mountain is the largest remnant of primeval forest in the region. It is surrounded by an agricultural and suburban mosaic containing multiple forest fragments of varying successional status. The mountain’s unusual geological history created a broad diversity of habitats on and around the mountain (Maycock 1961, Arii and Lechowicz 2002, Arii et al. 2005). A local diversity hotspot, the 10 km2 nature reserve contains 650-900 of the 1,600 regional species of terrestrial vascular forest plants (Maycock 1961, Marie-Victorin 1995, Elliott and Davies 2014) Sampling Design. We conducted an individual-based sampling design. Being at the tail-end of the establishment phase, a critical growth phase with high mortality (Grubb 1977), saplings express traits that are probably linked to successful way of life in the environment. Fifteen to 20 healthy tree saplings from the 24 most abundant deciduous species were sampled, for a total of 380 saplings. Sapling size was restricted to 1-5cm diameter at breast height (DBH) and a height of less than 66% of the canopy height in order to sample individuals whose crowns were in the sub-canopy layer and standardize ontogenetic stage. To maximize the range of trait values, the individuals were randomly selected from a diversity of habitats and microhabitats on and near the base of the mountain. All sampling dates were recorded and we removed the phenological effects on trait values when applicable. Tree age was determined by counting tree rings of cross- sections of the base of the tree. Traits. The twenty traits measured on each sapling are listed in Table 2. The specific methods for the measurement of these traits are provided in the supplementary material. The traits were chosen to include (i) the traits traditionally measured in the LES,

93 WS and CR dimensions, and (ii) additional traits reflecting the vital biological functions underpinning the LES and WS dimensions (resource acquisition, resource conservation, sap transport and mechanical support) and process underpinning the CR dimension. The traits were measured from all plant organs, namely the leaves, roots, stems and branches. For resource acquisition and conservation, leaf mass per area (LMA), leaf thickness, leaf nitrogen content (LNC), leaf phosphorus content (LPC) are the traditional traits typically included in the LES. We also measured other resource acquisition and conservation traits: carbon isotope ratio (δ13C) in leaves, which is as an integrated index of water use efficiency; specific root length (SRL) of first and second order absorptive roots; leaf thickness of the leaf blade between primary and secondary veins and leaf carbon concentration (LCC). For mechanical support and transport, stem wood density (Stem

WD), stem Modulus of elasticity (MOE) and conductivity per sapwood area (KS) are the set of traits typically included in the WS. In addition to this set of traits, we also measured other mechanical support and sap transport safety and efficiency traits: conductivity per branch (KBRANCH), vessel diameter (VD), branch wood density (Branch WD), Lumen

Fraction, vessel span-to-thickness ratio ((t/d)2) and coarse root wood density (Root WD). For architectural traits, individual leaf area, branching distance and branch diameter are the traditional traits measured in CR. We also measured branching angle as an architectural trait. We compiled shade, drought and waterlogging tolerance indices for our study species from Niinemets & Valladares (2006) in order to relate these phenotypic traits to these aspects of their natural history. Data Transforms. Following Kerkhoff and Enquist (2009), size and growth related traits were natural-log transformed (LMA, Leaf Area, Branch diameter, KS and KBRANCH ) and all other traits were transformed following individual boxcox transformations to optimize normality when necessary. Tree age had a marginally statistically significant yet minimal effects on traits correlations (r2 = 0.006 p = 0.07) and tree height had a significant but minimal effect on trait correlations (r2 = 0.0016 p=0.007), so the trait values were not corrected for these factors. The effect of sampling date was removed by taking the residuals of a linear regression of individual traits against the date for those traits for

94 which the regression was significant. The height aboveground at which each branch was sampled did not have an effect on δ13C values as can potentially occur close to the ground (Farquhar et al. 1989). Ordinations. All statistical analyses were conducted in R version 3.1.2 (R Development Core Team 2011). Principal Component Analyses (PCA) and redundancy analyses (RDA) on all traits and on trait subsets were conducted using the rda() function of the vegan package. Significance test of RDA’s constrained axes were conducted using the anova() function on the rda output of the same package. We used PCAsignificance() function of the BiodiversityR package to evaluate the number of significant PCA axes.

This function compares the eigenvalue of the ith axis with the corresponding eigenvalue expected from a broken-stick model and only axes with larger eigenvalues then the null model are considered significant. This criterion has been shown to be one of the most robust (Jackson 1993, Legendre and Legendre 1998). The significance of the contribution of each variable to each principal component in a two dimensional plane was determined using the ordiequilibriumcircle() function of the BiodiversityR package. It uses as the null hypothesis that each variable contributes equally to all the components in the plane and retains only variables with value larger than the expected contribution (Legendre and Legendre 1998). The integration index (I) was calculated as the variance of the eigenvalues of the principal components (Wagner 1984, Cheverud et al. 1989). It gives an index of the overall level of correlation of a set of variables (Cheverud et al. 1989). The rcorr() function of the Hmisc package was used to calculate the statistical significance of trait correlations. Mantel Tests. We tested two alternate hypotheses regarding the structure of phenotypic integration (i.e. narrowly vs broadly defined trait dimensions, i.e. HNARROW and

HBROAD) with standardized Mantel’s statistics. These calculate the Pearson’s correlation coefficient between two correlation matrices (Zuur et al. 2007). Here we compared a hypothesized correlation matrix (H) describing the theoretical network of relationships derived from each of our hypotheses and the matrix (D) calculated from our data

퐿퐸푆 퐿퐸푆 (Cheverud et al. 1989). For the LES, we built the 퐻푁퐴푅푅푂푊 and 퐻퐵푅푂퐴퐷 matrices based on

95 the hypothesis that traits favoring resource acquisition trade-off with traits favoring

푊푆 푊푆 resource conservation (Fig S2). For the WS, we built the 퐻푁퐴푅푅푂푊 and 퐻퐵푅푂퐴퐷 matrices based on the hypothesis that traits favoring mechanical support trade off with traits favoring conductive efficiency and are positively correlated to traits favoring conductive

퐶푅 퐶푅 safety (Fig S3). For CR, we built the 퐻푁퐴푅푅푂푊 and 퐻퐵푅푂퐴퐷 trait matrices based on the mechanism proposed in Olson (2009) and detailed in Fig S4. We do not have specific hypotheses regarding the relative strengths of trait correlations so we included values of -

1, 0 and 1 in the hypotheses. See Tables S7-S9 for the HNARROW and HBROAD matrices for each trait dimension. Network Analyses. Network graphs and analyses were performed with the package igraph on the statistically significant trait correlations. The Kamada-Kawai layout was used to calculate coordinates for the nodes of the network. The algorithm finds an optimal placement in a two dimensional plane that reflects the relative correlation strengths of the set of nodes in the network. We used the modularity() function to test whether trait dimensions formed distinct modules. It calculates the fraction of edges within the defined modules minus the expected fractions if the edges were random. Values range from -1 to 0.5 and are positive when the observed fraction exceeds the expected fraction based on chance. We calculated two indicators of network centrality for the traits using the degree() and betweenness() functions. The degree is the number of edges of a node and betweenness gives the number of shortest paths from all nodes to all others passing through the focal node.

96 RESULTS

Testing the orthogonality of trait dimensions: principal component analyses results Our analyses show a relatively homogenous phenotypic covariation structure not dominated by distinct trait dimensions. The general properties of the principal component analysis suggest that the phenotype is ‘round’ in multivariate trait space (Fig 1), with numerous correlations of equivalent magnitudes linking most traits. This is reflected by the low integration index of I = 2.09 (relative to the expected value for a random dataset of I = 1.15 (Cheverud et al. 1989)), which indicates many significant axes of similar importance and many traits of similar loading along each axis. Four principal components explain significantly more variance than expected from a broken-stick null model. Six components have eigenvalues greater than 1 and six are necessary to account for 80% of total variance or more (Table S3). Known trait dimensions, namely the LES, the WS and CS, do not segregate from each other along the principal components (Fig 1, table S3). Instead, the LES and CR relationships are present, but found embedded within broader trait associations (Fig 2, Table S3), and the WS traits do not appear within a given principal component. The first axis shows a negative relationship between resource acquisition traits (Leaf Thickness and SRL) and support traits (Stem WD and MOE). Five other traits contribute marginally to this axis: two hydraulic traits (KBRANCH and Lumen Fraction) two resource acquisition traits, (LMA, WUE) and one mechanical support trait (Coarse Root WD). Species sorting along this axis suggests an early-late successional gradient. The species scores along this axes are correlated with the shade-tolerance index of species compiled in Niinements & Valladares (2006) (Fig 2) (SMA regression: r = 0.67, p=0.0005), but not with other stress-tolerance indices. The second principal component describe positive relationships among the three traditional CR traits (Leaf Area, Branch Diameter and Branching Distance) and two sap transport traits (KS and VD). Species sorting along this axis follow a distinction in wood anatomy construction between ring- and diffuse-porous species.

97 The third principal component includes LES traits (namely LMA, LPC and LNC) correlating with each other the way we would expect. However, it also includes Branch WD. Species scores along PC3 are also significantly correlated with their shade-tolerance index (SMA regression: r = 0.42, p=0.05). The fourth and last significant principal component reflects relationships among Branching Angle, LCC and Lumen fraction. Species sorting along the 4th axes does not highlight any notable differences in the natural histories of these species that we can discern. To determine whether the known functional dimensions would appear if smaller trait subsets were considered, we ran principal component analyses separately on each of the broadly defined functional dimensions (including traits known to be part of a functional dimension in addition to other traits that also reflect the associated functions). Within the architectural traits, CR appeared along the first principal component, but the LES and the WS did not appear within the acquisition/conservation and support/transport set of traits (Table S4).

Testing the hypotheses of narrowly and broadly trait defined trait dimensions: Mantel test results Mantel tests comparing the observed trait correlation matrices with those given the narrow and broad definitions of trait dimensions provide moderate support for the presence of LES and WS correlations and strong but marginally significant support for the presence of the CR correlations (Table 1, rM = 0.54, p=0.03; rM =0.63 p=0.02 and rM =0.85 p=0.10, respectively). The trait correlation structure agrees with the narrowly defined LES but not significantly with the broadly defined LES. This appears to be because correlations between SRL and leaf traits are opposite to the hypotheses’ predictions (Tables S7B-C). The data are consistent with the broad, but not narrow definition of the WS. This might be partly driven by the strong correlations among the wood densities of the stem, root and branch (Table S8C). In terms of the relationships among architectural traits, the data matrices are strongly correlated with both hypothesis matrices, but surprisingly, these strong correlations are not statistically significant. This is likely due to

98 the low number of traits (five) in the trait correlation matrix resulting in few permutations for the null distribution (Table 1). The relationships between Branch WD and other CR traits clearly follow the expectations based on Olson’s (2009) mechanism. Branching angle is not significantly correlated with any other architectural trait. This is contrary to our expectation that increased branching angle should also be linked with Leaf Area to minimize leaf overlap and optimize light capture (Fig S9). After removing the effect of Leaf Area on both branching angle and branching distance, the two traits are marginally correlated (SMA r = 0.34, p=0.11), as suggested by PC3 (Fig 2). This indicates that Branching Angle and Branching Distance are marginally correlated when holding leaf area constant.

Describing the phenotype as a trait network: network analyses results Network representations of the trait relationships show that trait dimensions do not form distinct modules (Figs 2 and 3). In fact, modularity analyses on the fraction of significant trait correlations within and among trait dimensions (giving the fraction of the edges that fall within the given groups minus the expected such fraction if edges were distributed at random) indicate that the fraction of correlations within trait dimensions are not different from random (modularity value, Q = 0.0058). This means that the traits are as likely to be connected within trait dimensions as among them. Table 2 gives the centrality indices of each trait in the network shown in Figure 4. Three traits appear as more central to the phenotype than others, with large betweenness values: Modulus of elasticity (MOE), Leaf Area and Branch Diameter. Other traits with high centrality indices include LMA, Leaf Thickness and KS.

99 DISCUSSION

Lack of support for distinct and independent trait dimensions Collectively, the analyses indicate that the ensemble of the phenotype is well integrated, with traits reflecting different physiological functions and from different organs correlated with each other. We do find some evidence for the presence of the trait relationships forming trait dimensions: (1) the trait dimensions are moderately (LES, WS) and strongly (CR) supported by Mantel tests (Table 1); and (2) the traditional CR and LES traits group along PC2 and PC3 in the principal component analysis conducted on all twenty traits (Fig 2). However, the LES, CR and WS dimensions remain embedded within a broader phenotype with many strong correlations to traits outside of these trait dimensions: (1) the modularity test shows that connections between traits are as common within as among trait dimensions (Q= 0.0058); (2) other traits are correlated with the LES and CR on PC2 and PC3, which furthermore indicates that the effects of PC1 must be removed for these trait relationships to appear (Fig 2); so that accordingly, (3) the LES and WES traits do not cluster along a single PCA axis when analyses are conducted on trait subsets (Table S4). In our dataset, trait dimensions do not segregate into distinct groups of traits, but instead appear to be woven into a broader network of correlations. For example, as expected from Corner’s rules, Branch Diameter is indeed correlated with Leaf

Area and Branching Distance, but it is also correlated with KS and LMA, two relationships not predicted from the literature. Importantly, note that many of the relationships expected in the LES and WS are indeed present, but our results emphasize that in this study they do not form independent, isolated modules dominating phenotypic structure. A closer examination of the relationships shows that most correlations between the mechanical support and resource acquisition traits are negative, suggesting a trade-off among these two functions (Figs 3 and 4). This is notable since we are aware of only one other study finding a relationship between these two physiological functions (Santiago et al. 2004). Further, most correlations between acquisition and transport efficiency traits are positive. This supports Reich’s (2014) hypothesis that fast vs. slow pace of life is a major aspect characterizing plant strategies and that it includes sap transport properties.

100 Indeed, we find that the first principal component axis is well correlated with the shade tolerance index of the study species. Shade-tolerance is related to successional status, a central concept in forest ecology literature and to the ruderal vs. competitive side of the CSR triangle, a central framework in comparative ecology (Grime 1977). Although the individual dimensions do not segregate, it is reassuring that when we examine the ensemble of phenotype and include a number of vital physiological functions, the main axis of trait variation is consistent with shade-tolerance index, a major descriptor of tree natural history. It is also noteworthy that there are no significant direct correlations between conductive efficiency and support traits. Instead, these traits are linked indirectly through resource acquisition and architectural traits (Fig 3 and Fig 4). This is surprising because a trade-off between the support and transport functions of the xylem has been suggested by Baas (2004) and is a central element underpinning the wood spectrum. Some empirical studies have questioned this relationship: Preston et al (2006) found that wood density was not related to vessel properties and Pratt et al. (2007) found that stem mechanical support was only associated with conductive safety, not efficiency. A caveat in our ability to assess transport-support relationships is that three of the mechanical support traits reflect wood density in different organs. It would be interesting to confirm these patterns in other traits reflecting mechanical support. Further, the independence of the two conductivity traits stands out. KBRANCH is associated with resource acquisition traits in leaves whereas KS correlates with architectural traits. The independence of KS and KBRANCH

(Bhaskar et al. 2007) and the links between leaf traits and KBRANCH (or similarly, with Huber Value) have been noted before (Grady et al. 2013). The contrasting trait associations between the two measures of conductivity highlights the complexity of hydraulics and suggest that its different nuances reflect different aspects of the plant’s ecophysiology. The absence of a clear WS signal in our dataset is in line with inconsistent evidence for these relationships at broader scales (Chave et al. 2009) and highlights the necessity to elucidate the relationships in wood traits among mechanical support, transport safety and

101 efficiency. While wood density, like LMA, is a high-level “summary” trait affected by a large number of underlying traits, in this dataset it is only correlated with the mechanical support trait and not with any of the hydraulic safety or efficiency traits. This lack of correlation questions the role of wood density as an indicator of the relationships among transport safety, efficiency and mechanical support at local scales. In summary, our results suggest that at the scale of our study, the non-reproductive phenotype is a web of traits, some of which are more closely linked to each other without forming clear, distinct trait dimensions. This questions (1) whether the trait dimension paradigm is applicable to local-scale studies such as most of those of community assembly, and (2) whether using analytical tools forcing orthogonality is the most effective way to summarize ecological data.

Differences in trait patterns at local scales This study differs from the majority of trait studies in two main respects: (i) we simultaneously consider many traits from different plant organs reflecting a variety of plant ecophysiological functions; and (ii) we work at a more local scale and thereby with a narrower range of trait values. The fact that the expected LES trait relationships does not appear when analyzing a trait subset eliminates the possibility that our results are driven by the larger number and variety of traits included. This suggests that differences in scale underlie the difference between our results and those found at global-scales (Reich et al. 1999, Wright et al. 2004). Trait dimensions are typically studied across bioclimatic zones (Reich et al. 1999, Wright et al. 2004, Violle et al. 2014), whereas the saplings sampled here span a range of microenvironments within a homogeneous bioclimatic zone. Nonetheless, although the spatial scale and bioclimatic variability is greatly reduced, the dataset covers a significant fraction of global trait variation (Table S6). The range of leaf trait values covered in our dataset relative to the global dataset GLOPNET (Wright et al. 2004), is 20% for LMA, 49% for LNC and 80% for LPC. The range of wood trait values covered in our dataset relative to Chave’s global compilation (Chave et al. 2009) is 16% for VD, 29% for Stem

102 WD and 42% for MOE. Although ecological scale and trait ranges are necessarily coupled, the substantial trait range of this local study suggests that differences between the phenotypic patterns observed here and those observed at global scales are likely due to distinct scale-dependent processes acting in our study system more than to the relatively small range in trait values.

While this is the first study to examine the interactions among three trait dimensions, our findings are consistent with earlier work. Other studies focusing on leaf traits have failed to find the LES among species at local scales. Wright and Sutton-Grier (2012) examined the interspecific relationships among leaf traits in different soil environments and failed to observe the LES fifteen out of eighteen times. In a meta-analysis, Funk and Cornwell (2013) found that the strength of the LES decreased with the range of trait values. Conversely, Santiago and Mulkey (2005) found interspecific relationships among leaf traits typically considered to be part of distinct dimensions. Similarly Kraft et al. (2015), who looked at interspecific relationships among whole-plant and organ level traits, did not observe the LES relationships, but reported instead a relatively even correlation among the traits (Kraft et al. 2015 Fig S3). Edwards et al. (2014) also failed to find a number of trait relationships expected from the LES dimension in Viburnum species. So why are these trait relationships so different at broad and local scales? Trait correlation patterns are the outcome of many drivers of phenotypic integration. Hard, absolute constraints such as biophysical laws are ubiquitous across all taxon and scales. Softer, malleable constraints can vary across scales so that trait relationships arising from them are expected to be less consistent across scales or taxon. For example, the force of natural selection changes across environments and genetic constraints can vary across taxon and over time. We thus surmise that moving from global to local scale entails a shift in the dominating drivers of phenotypic integration and that the locally dominant drivers of phenotypic integration favor distinct relationships from those observed globally.

103 For the LES, ongoing natural selection has been suggested to play an important role in shaping these trait relationships (Reich and Wright 2003, Donovan et al. 2011, Vasseur et al. 2012). More specifically, Funk and Cornwell (2013) suggested that natural selection optimizing photosynthetic rate for a given leaf lifespan drives the LES so that a broad leaf lifespan range is required to observe a strong LES. Our data are consistent with this explanation. Our study system does encompass a range of leaf lifespan (Lechowicz 1984), but by studying deciduous broadleaf trees located in a strongly seasonal bioclimatic zone, this range is still limited relative to the worldwide variation. Thus, natural selection shaping the LES may not be the dominant driver of leaf trait relationships at the scale of our study. The scale-dependence of correlation strength is probably not a phenomenon unique to trait dimensions. Tilman and colleagues (2004) pointed out that as the range of body size decreases, the predictive power of the metabolic theory of ecology declines and other drivers become more important in explaining variation at smaller scales. The results for CR contrast sharply with those for the LES. CR was strongly supported, suggesting that the forces shaping these relationships are also dominant at local scales. Plant architecture in general is typically understood in the context of physical laws (e.g. Fick’s first law, Euler column equation, etc.), constraining the optimization of the various physiological functions of the plant (Niklas 1992, 1999). In fact, Olson et al. (2009) suggests that metabolic scaling, which explains how the optimization of metabolic rates is constrained by the properties of fractal transport networks, is a driving force behind CR (Fig S4). The limits imposed by biophysical constraints cannot be escaped and should be ubiquitous across scales. Hence, the strong correlations among architectural traits within this local study are consistent with the central role of biophysical constraints in shaping CR and architecture.

Plant architecture at the crossroads of the trait network Given that at the local scale of our study traits do not clearly segregate into distinct groups, our results argue for a shift in perspective in local trait-based comparative ecology from axes towards an integrated network view. In fact, a number of well-established and

104 robust ecological frameworks argue for the integration of many plant functions. First , as noted above, plant architecture and biomechanics describes the phenotype as the compromise between intimately linked acquisition, support, transport and reproductive functions (Niklas 1992, 1994a, 1999). Second, metabolic scaling theory explains a broad spectrum of scaling relationships from the optimization of two vital physiological functions (resource acquisition and transport efficiency) within a set of biophysical constraints, notably a fractal transport network (West et al. 1997). In fact, Olson et al. (2009) argue that architectural relationships observed in CR arise from the optimization of carbon acquisition, mechanical support and sap transport within the constraints imposed by metabolic scaling. Third, plant physiology tightly couples hydraulic properties with resource acquisition because stomatal conductance controls both the rate of evapotranspiration and CO2 exchange for photosynthesis (Brodribb and Feild 2000, Sack and Holbrook 2006, Lambers et al. 2008, Maherali et al. 2008). Yet, based on trait dimensions, these two functions are considered along two distinct spectra, the LES and the WS. Here we indeed find KBRANCH, a hydraulic trait, to be closely related to a number of resource acquisition-conservation traits (see also Wright et al. 2006). In summary, biomechanical, metabolic and physiological constraints argue for the integration of multiple physiological functions. Our results are consistent with the idea that these hard and moderate constraints dominate the phenotypic integration structure locally by showing that physiological functions are linked into a network. Further, our results suggest that architectural traits are phenotypic keystones: we find that two architectural traits, Leaf Area and Branch Diameter, are among the most central to the phenotypic network, linking mechanical support, resource acquisition and transport traits. Since architecture is the outcome of the close phenotypic integration of multiple vital physiological functions within sets of biophysical constraints, it is not surprising that architectural traits should be at the crossroads of different physiological functions. Indeed, Edwards et al. (2014) found that leaf life-span was better correlated with plant architectural traits than with any of the LES traits in locally coexisting Viburnum species.

105

Based on the network nature of the phenotype, we suggest that to adequately characterize trait hypervolume (cf. Blonder et al. 2014, Laughlin 2014) or functional niche (cf. Violle and Jiang 2009, Winemiller et al. 2015) we should include traits central to the phenotypic network in addition to those describing ‘classical’ trait dimensions. Architecture, at the crossroads of plant functions, offers an integrated measure of the compromises shaping the plant’s phenotype (Fig 5). If our results hold across deciduous trees in general, the fact that architectural traits are absent from most large and small trait measurement campaigns (Kattge et al. 2011, Pérez-Harguindeguy et al. 2013) might be an important loss for local-scale comparative ecology.

106 ACKNOWLEDGEMENTS

We would like to thank Andréanne Ferland, Émilie Lavoie, Natasha Salter, Carol Mordy, Anke Roth, Sierra Kaszubinski, Sanga Shir, Surbhi Patel, Margretta Murphy, John Lacson, Kevin Wong, Sarah Schwenck, Anjeanette McKay, Casey Knoks, Meghan Iacueuilli, Shahrzad Badie, Irene Liang, Jordyn Celaya and David Maneli for their instrumental role in data collection, Erica Bigio for invaluable assistance with the tree ring work, John Sperry, Melvin Tyree and David Killick for guidance and assistance with the tedious conductance measurements and Louise Comas for guidance on root trait field sampling methods.

107 REFERENCES Ackerly, D. D. and Donoghue, M. J. 1998. Leaf size, sapling allometry, and Corner’s rules: phylogeny and correlated evolution in (Acer). - Am. Nat. 152: 767–91. American Public Health Association 1999. Standard Methods for the Examination of Water and Wastewater. - Port City Press. Arii, K. and Lechowicz, M. J. 2002. The influence of overstory trees and abiotic factors on the sapling community in an old-growth Fagus-Acer forest. - Ecoscience 9: 386–396. Arii, K. et al. 2005. Environmental correlates of canopy composition at Mont St. Hilaire, Quebec, Canada. - J. Torrey Bot. Soc. 132: 90–102. Baas, P. et al. 2004. Evolution of xylem physiology. - In: Hemsley, A. R. and Poole, I. (eds), The evolution of plant physiology. Elsevier Academic Press, pp. 273–295. Berg, R. L. 1960. The ecological significance of correlation pleiades. - Evolution (N. Y). 14: 171–180. Bhaskar, R. et al. 2007. Evolution of hydraulic traits in closely related species pairs from Mediterranean and nonMediterranean environments of North America. - New Phytol. 176: 718–26. Blonder, B. et al. 2014. The n-dimensional hypervolume. - Glob. Ecol. Biogeogr. 23: 595– 609. Bonser, S. P. 2006. Form defining function: interpreting leaf functional variability in integrated plant phenotypes. - Oikos 114: 187–190. Brodribb, T. J. and Feild, T. S. 2000. Stem hydraulic supply is linked to leaf photosynthetic capacity: Evidence from New Caledonian and Tasmanian rainforests. - Plant, Cell Environ. 23: 1381–1388. Brouat, C. et al. 1998. Corner’s rules revisited: ontogenetic and interspecific patterns in leaf-stem allometry. - New Phytol. 139: 459–470. Chapin, F. S. et al. 1993. Evolution of suites of traits in response to environmental stress. - Am. Nat. 142: S78–S92. Chave, J. et al. 2009. Towards a worldwide wood economic spectrum. - Ecol. Lett. 12: 351–366. Cheverud, J. M. 1982. Phenotypic, genetic and environmental morphological integration in the cranium. - Evolution (N. Y). 36: 499–516. Cheverud, J. M. et al. 1989. Methods for the comparative analysis of variation patterns. - Syst. Zool. 38: 201–213. Comas, L. H. and Eissenstat, D. M. 2009. Patterns in root trait variation among 25 co- existing North American forest species. - New Phytol. 182: 919–28. Cornelissen, J. H. C. 1999. A triangular relationship between leaf size and seed size among

108 woody species: allometry, ontogeny, ecology and . - Oecologia 118: 248– 255. Corner, E. J. H. 1949. The durian theory or the origin of the modern tree. - Ann. Bot. 13: 367–414. Craine, J. M. 2009. Resource Strategies of Wild Plants. - Princeton University Press. Craine, J. M. et al. 2001. The relationships among root and leaf traits of 76 grassland species and relative abundance along fertility and disturbance gradients. - Oikos 93: 274–285. Davis, S. D. et al. 1999. The relationship between xylem conduit diameter and cavitation caused by freezing. - Am. J. Bot. 86: 1367–1372. Donovan, L. a et al. 2011. The evolution of the worldwide leaf economics spectrum. - Trends Ecol. Evol. 26: 88–95. Edwards, E. J. et al. 2014. Leaf life span and the leaf economic spectrum in the context of whole plant architecture. - J. Ecol. 102: 328–336. Eissenstat, D. and Yanai, R. 1997. The ecology of root lifespan. - Adv. Ecol. Res. 27: 1–59. Elliott, T. L. and Davies, J. T. 2014. Challenges to barcoding an entire flora. - Mol. Ecol. Resour. 14: 883–891. Enquist, B. J. and Niklas, K. J. 2002. Global allocation rules for patterns of biomass partitioning in seed plants. - Science (80-. ). 295: 1517–1520. Enquist, B. J. and Bentley, L. P. 2012. Land Plants: New Theoretical Directions and Empirical Prospects. - In: Brown, J. et al. (eds), Metabolic Theory of Ecology. Oxford University Press, pp. 164–187. Enquist, B. J. et al. 2007. A general integrative model for scaling plant growth, carbon flux, and functional trait spectra. - Nature 449: 218–222. Farquhar, G. D. et al. 1989. Carbon discrimination and photosynthesis. - Annu. Rev. Plant Physiol. Plant Mol. Biol. 40: 503–537. Feininger, T. et al. 1995. The eight classical Monteregian hills at depth and the mechanism of their intrusion1 Distance from. 1364: 1350–1364. Fortunel, C. et al. 2012. Leaf, stem and root tissue strategies across 758 Neotropical tree species. - Funct. Ecol. 26: 1153–1161. Freschet, G. T. et al. 2010. Evidence of the “plant economics spectrum” in a subarctic flora. - J. Ecol. 98: 362–373. Funk, J. L. and Cornwell, W. K. 2013. Leaf traits within communities: Context may affect the mapping of traits to function. - Ecology 94: 1893–1897. Grady, K. C. et al. 2013. Conservative leaf economic traits correlate with fast growth of genotypes of a foundation riparian species near the thermal maximum extent of its

109 geographic range. - Funct. Ecol. 27: 428–438. Grime, J. P. 1977. Evidence for existence of 3 primary strategies in plant and its relevance to ecological and evolutionary theory. - Am. Nat. 111: 1169–1194. Grubb, P. J. 1977. The maintenance of species-richness in plant communities: the importance of the regeneration niche. - Biol. Rev. Camb. Philos. Soc. in press. Hacke, U. G. and Sperry, J. S. 2001. Functional and ecological xylem anatomy. - Perspect. Plant Ecol. Evol. Syst. 4: 97–115. Hacke, U. G. et al. 2001. Trends in wood density and structure are linked to prevention of xylem implosion by negative pressure. - Oecologia 126: 457–461. Hallé, F. et al. 1978. Tropical trees and forests: An architectural analysis. - Springer- Verlag. Jackson, D. A. 1993. Stopping rules in principal components analysis: A comparison of heuristical and statistical approaches. - Ecology 74: 2204–2214. Kattge, J. et al. 2011. TRY - a global database of plant traits. - Glob. Chang. Biol. 17: 2905– 2935. Kembel, S. W. and Cahill, J. F. 2011. Independent evolution of leaf and root traits within and among temperate grassland plant communities. - PLoS One 6: e19992. Kerkhoff, A. J. and Enquist, B. J. 2009. Multiplicative by nature: Why logarithmic transformation is necessary in allometry. - J. Theor. Biol. 257: 519–521. Kraft, N. J. B. et al. 2015. Plant functional traits and the multidimensional nature of species coexistence. - Proc. Natl. Acad. Sci. 112: 797–802. Lambers, H. et al. 2008. Plant physiological ecology. - Springer. Laughlin, D. C. 2014. The intrinsic dimensionality of plant traits and its relevance to community assembly. - J. Ecol. in press. Laughlin, D. C. et al. 2010. A multi-trait test of the leaf-height-seed plant strategy scheme with 133 species from a pine forest flora. - Funct. Ecol. 24: 493–501. Lechowicz, M. J. 1984. Why Do Temperate Deciduous Trees Leaf Out at Different Times? Adaptation and Ecology of Forest Communities. - Am. Nat. 124: 821. Legendre, P. and Legendre, L. 1998. Numerical Ecology. - Elsevier. Lucas, P. W. et al. 2000. Mechanical Defences to Herbivory. - Ann. Bot. 86: 913–920. Maherali, H. et al. 2008. Leaf hydraulic conductivity and photosynthesis are genetically correlated in an annual grass. - New Phytol. 180: 240–247. Marie-Victorin, F. 1995. Flore Laurentienne. - Gaetan Morin. Maycock, P. F. 1961. Botanical studies on Mont St.Hilaire, Rouville county, Quebec. - Can. J. Bot. 39: 1293–1325.

110 Merila, J. and Bjorklund, M. 2004. Phenotypic Integration as a Constraint and Adaptation. - In: Pigliucci, M. and Preston, K. (eds), Phenotypic Integration: Studyinf the Ecology and Evolution of Comlpex Phenotypes. Oxford University Press, pp. 107– 129. Midgley, J. and Bond, W. 1989. Leaf size and inflorescence size may be allometrically related traits. - Oecologia 78: 427–429. Murren, C. J. 2002. Phenotypic integration in plants. - Plant Species Biol. 17: 89–99. Niinemets, Ü. and Valladares, F. 2006. Tolerance to shade, drought, and waterlogging of temperate northern hemisphere trees and . - Ecol. Monogr. 76: 521–547. Niinemets, U. et al. 2007. Do we underestimate the importance of leaf size in plant economics? Disproportional scaling of support costs within the spectrum of leaf physiognomy. - Ann. Bot. 100: 283–303. Niklas, K. J. 1992. Plant Biomechanics: an engineering approach to plant form and function. - University of Chicago Press. Niklas, K. J. 1994b. Plant Allometry: The scaling of form and process. - The University of Chicago Press. Niklas, K. J. 1994a. Morphological evolution through complex domains of fitness. - Proc. Natl. Acad. Sci. U. S. A. 91: 6772–6779. Niklas, K. J. 1999. Evolutionary walks through a land plant morphospace. - J. Exp. Bot. 50: 39–52. Niklas, K. J. and Kerchner, V. 1984. Mechanical and photosynthetic constraits on the evolution of plant shape. - Paleobiology 10: 79–101. Olson, M. E. et al. 2009. Universal foliage-stem scaling across environments and species in dicot trees: plasticity, biomechanics and Corner’s Rules. - Ecol. Lett. 12: 210–9. Pearcy, R. W. and Yang, W. 1996. A three-dimensional crown architecture model for assessment of light capture and carbon gain by understory plants. - Oecologia 108: 1– 12. Pérez-Harguindeguy, N. et al. 2013. New handbook for standardised measurement of plant functional traits worldwide. - Aust. J. Bot. 61: 167–234. Poorter, H. and Ryser, P. 2015. The limits to leaf and root plasticity : what is so special about specific root length ? - New Phytol. 206: 1188–1190. Pratt, R. B. et al. 2007. Relationships among xylem transport, biomechanics and storage in stems and roots of nine Rhamnaceae species of the California chaparral. - New Phytol. 174: 787–98. Preston, K. a et al. 2006. Wood density and vessel traits as distinct correlates of ecological strategy in 51 California coast range angiosperms. - New Phytol. 170: 807–18. Price, C. A. et al. 2014. Are leaf functional traits “invariant” with plant size and what is

111 “invariance” anyway? - Funct. Ecol. 28: 1330–1343. R Development Core Team, R. 2011. R: A Language and Environment for Statistical Computing (RDC Team, Ed.). - R Found. Stat. Comput. 1: 409. Reich, P. B. 2014. The world-wide “fast-slow” plant economics spectrum: a traits manifesto (H Cornelissen, Ed.). - J. Ecol. 102: 275–301. Reich, P. B. and Wright, I. J. 2003. The evolution of plant functional variation: traits, spectra, and strategies. - Int. J. Plant Sci. 164: S143–S164. Reich, P. B. et al. 1999. Generality of leaf trait relationships: A test across six biomes. - Ecology 80: 1955–1969. Sack, L. and Holbrook, N. M. 2006. Leaf hydraulics. - Annu. Rev. Plant Biol. 57: 361–381. Santiago, L. S. and Mulkey, S. S. 2005. Leaf productivity along a precipitation gradient in lowland Panama: patterns from leaf to ecosystem. - Trees 19: 349–356. Santiago, L. S. et al. 2004. Leaf photosynthetic traits scale with hydraulic conductivity and wood density in Panamanian forest canopy trees. - Oecologia 140: 543–50. Southwood, T. 1988. Tactics, strategies and templets. - Oikos 52: 3–18. Sperry, J. S. et al. 2006. Size and function in conifer tracheids and angiosperm vessels. - Am. J. Bot. 93: 1490–1500. Tilman, D. et al. 2004. Does metabolic theory apply to community ecology? It’s a matter of scale. - Ecology 85: 1797–1799. Tyree, M. T. and Zimmermann, M. H. 2002. Xylem Structure and the Ascent of Sap. - Springer-Verlag. Valladares, F. and Pearcy, R. W. 1999. The role of crown architecture for light harvesting and carbon gain in extreme light environments assessed with a realistic 3-D model. - An. del Jardín Botánico Madrid 58: 3–16. Valladares, F. et al. 2002. Convergence in light capture efficiencies among tropical forest understory plants with contrasting crown architectures: A case of morphological compensation. - Am. J. Bot. 89: 1275–1284. van Gelder, H. a et al. 2006. Wood mechanics, allometry, and life-history variation in a tropical rain forest tree community. - New Phytol. 171: 367–78. Vasseur, F. et al. 2012. A common genetic basis to the origin of the leaf economics spectrum and metabolic scaling allometry. - Ecol. Lett. 15: 1149–57. Vile, D. et al. 2005. Specific leaf area and dry matter content estimate thickness in laminar leaves. - Ann. Bot. 96: 1129–36. Violle, C. and Jiang, L. 2009. Towards a trait-based quantification of species niche. - J. Plant Ecol. 2: 87–93. Violle, C. et al. 2014. The emergence and promise of functional biogeography. - Proc.

112 Natl. Acad. Sci. 11: 13690–13696. Wagner, G. P. 1984. On the eigenvalue distribution of genetic and phenotypic dispersion matrices: Evidence for a nonrandom organization of quantitative character variation. - J. Math. Biol. 172: 77–95. 2002. Plant roots: the hidden half (Y Waisel, A Eshel, and U Kafkafi, Eds.). - Marcel Dekker. West, G. B. et al. 1997. A General Model for the Origin of Allometric Scaling Laws in Biology. - Science (80-. ). 276: 122–126. West, G. B. et al. 1999. A general model for the structure and allometry of plant vascular systems. - Nature 400: 664–667. Westoby, M. 1998. A leaf-height-seed (LHS) plant ecology strategy scheme. - Plant Soil 199: 213–227. Westoby, M. and Wright, I. J. 2006. Land-plant ecology on the basis of functional traits. - Trends Ecol. Evol. 21: 261–8. Westoby, M. et al. 2002. Plant ecological strategies: Some leading dimensions of variation between species. - Annu. Rev. Ecol. Syst. 33: 125–159. White, P. S. 1983a. Evidence that temperate east North-American evergreen woody- plants follow Corner’s rules. - New Phytol. 95: 139–145. White, P. S. 1983b. Corner’s rules in eastern deciduous trees : Allometry and its implications for the adaptive architecture of trees. - Bull. Torrey Bot. Club 110: 203– 212. Winemiller, K. O. et al. 2015. Functional traits, convergent evolution, and periodic tables of niches. - Ecol. Lett. 18: 737–751. Wright, J. P. and Sutton-Grier, A. 2012. Does the leaf economic spectrum hold within local species pools across varying environmental conditions? - Funct. Ecol. 26: 1390– 1398. Wright, I. J. et al. 2004. The worldwide leaf economics spectrum. - Nature 428: 821–827. Wright, I. J. et al. 2006. Cross-species patterns in the coordination between leaf and stem traits, and their implications for plant hydraulics. - Physiol. Plant. 127: 445–456. Wright, I. J. et al. 2007. Relationships among ecologically important dimensions of plant trait variation in seven neotropical forests. - Ann. Bot. 99: 1003–15. Zuur, A. F. et al. 2007. Analysing Ecological Data. - Springer.

113 TABLES

Table 1. Standardised Mantel’s statistic (rM) and associated p-value comparing two hypotheses regarding the structure of phenotypic integration. HNARROW - Narrowly defined trait dimensions: only the traits traditionally included in the literature as part of the trait dimension are correlated. HBROAD – Broadly defined trait dimension: all traits reflecting the underpinning trade-off are correlated. For each trait dimension, the best supported hypothesis is highlighted in grey. Neither are supported for the LES. Orange: mechanical support/transport dimension; Grey: architecture dimension; Green: resource acquisition/resource conservation dimension

rM p-value

RESOURCE ACQUISITION VS CONSERVATION (Leaf Economic Spectrum)

HNARROW Correlated Traits: LMA, LNC, LPC 0.54 0.03

HBROAD Correlated Traits: LMA, LNC, LPC, Leaf Thickness, LCC, WUE, SRL 0.40 0.07

SAP TRANSPORT VS MECHANICAL SUPPORT (Wood Spectrum):

HNARROW Correlated Traits: Stem WD, VD, Ks, MOE 0.16 0.19

HBROAD Correlated Traits: Stem WD, VD, Ks, MOE, Branch WD, (t/d)2, 0.63 0.02

KBRANCH, Lumen Fraction, Coarse Root WD

ARCHITECTURE (Corner’s Rules)

HNARROW Correlated traits: Leaf Area, Branch Diameter, Branching Distance 0.85 0.10

HBROAD Correlated Traits: Leaf Area, Branch Diameter, Branching Distance, 0.72 0.20 Branching Angle, Branch WD

114 Table 2. Node centrality indicators for the trait network. ‘Degree’ gives the number of edges of each node in the network. ‘Betweenness’ takes into account path length: the more edges and the shorter they are, the largest the ‘Betweenness’ values. The traits are sorted in decreasing order of Betweenness, then Degree, then alphabetically. Numbers in parentheses give the number of significant correlations that are mathematically independent. Orange: mechanical support/transport dimension; Grey: architecture dimension; Green: resource acquisition/resource conservation dimension Trait Degree Betweenness MOE 7 (6) 40 Branch Diameter 6 23 Leaf Area 6 21 LMA 6 16 Leaf Thickness 6 16 KS 6 (5) 14 Branch WD 6 11 SRL 6 11 Lumen Fraction 5 9 Coarse Root WD 5 1 Stem WD 5 (4) 0 Branching Distance 4 0 KBRANCH 4 0 LNC 4 0 LPC 4 0 VD 4 (3) 0 Branching Angle 2 0 δ13C 2 0 LCC 1 0 (t/d)2 1 0

115 FIGURES

Figure 1 – The three trait dimensions. The leaf economic spectrum and the wood spectrum are understood to represent trade-offs among physiological functions (shown in the boxes). Architectural traits do not directly serve specific physiological functions per se, but show “natural selection operating on the relation between form and function” (Niklas 1994b) within constraints established by biomechanics and metabolic scaling (Olson et al. 2009). Distinct architectural designs provide alternate ways to effectively display leaves and reproductive structures (White 1983b, Niklas 1992, Valladares et al. 2002). The direction of the correlations are represented with + and – in the arrows for positive and negative associations respectively. Traits associated with each of these strategy dimensions are detailed in Table S2.

116

Figure 2 – Eigenvalues and variance explained by six first principal components of principal component analysis. Bivariate plots show interspecific trait correlation structure for first 4 axes. Traits shown in blue were not included in the original PCA due to their smaller sample size, but correlated post-hoc with the principal components. Orange: mechanical support/transport dimension; Grey: architecture dimension; Green: resource acquisition/resource conservation dimension. The insets show species position along the PC axes.

117

Figure 3 – Trait correlations. Only significant correlations are shown. Full blue lines show positive correlations. Dashed red lines show negative correlations. Line thickness gives the correlation strength. Orange: mechanical support/transport dimension; Grey: architecture dimension; Green: resource acquisition/resource conservation dimension.

118

Figure 4 – Trait correlation network. The location of the nodes with respect to each other was optimized for a 2D plane with network analysis tools (igraph package, R). Red and blue edges show negative and positive correlations, respectively. Correlation strength is represented by edge thickness and distance among traits. Only significant correlations at the alpha = 0.05 level are shown, which corresponds for this dataset to r values greater than 0.41. Traits typically considered to be part of each trait dimension are outlined in black. Orange: mechanical support/transport dimension; Grey: architecture dimension; Green: resource acquisition/resource conservation dimension.

119

Figure 5 – Proposed trait integration structure. Since architecture arise as the result of trade-offs and constraints among mechanical support, sap transport and resource acquisition, architectural traits reflect the integration of these physiological functions.

120 SUPPLEMENTARY MATERIAL – APPENDIX I. TABLES AND FIGURES Table S1. List of 24 study species.

Xylem Acronym Scientific name Family Common name Leaf type Porosity ACP Acer pensylvanicum Aceraceae Stripped Simple Diffuse ACR Aceraceae Red Maple Simple Diffuse ACSA Acer saccharum Aceraceae Sugar Maple Simple Diffuse ACSP Acer spicatum Aceraceae Mountain Maple Simple Diffuse AMSP Amelanchier laevis Rosaceae Serviceberry Simple Diffuse BEA Betula alleghaniensis Betulaceae Yellow birch Simple Diffuse BEPA Betulaceae White birch Simple Diffuse BEPO Betula populifolia Betulaceae Grey birch Simple Diffuse CAC Carya cordiformis Juglandaceae Bitternut hickory Compound Ring COA Cornus alternifolia Cornaceae Alternate-leaf dogwood Simple Diffuse FAG Fagus grandifolia Fagaceae American beech Simple Diffuse FRA Fraxinus americana Oleaceae American ash Compound Ring OSV Ostrya virginiana Betulaceae Ironwood Simple Diffuse POB Populus balsamifera Balsam Poplar Simple Diffuse POD Populus deltoides Salicaceae Cottonwood Simple Diffuse POG Populus grandidentata Salicaceae Large-tooth Aspen Simple Diffuse POT Populus tremuloides Salicaceae Trembling Aspen Simple Diffuse PRP Prunus pensylvanica Rosaceae Pin cherry Simple Diffuse PRS Prunus serotina Rosaceae Black cherry Simple Diffuse PRV Prunus virginiana Rosaceae Choke cherry Simple Diffuse QUR Quercus rubrum Fagaceae Red Oak Simple Ring SOA Sorbus americana Rosaceae Mountain-Ash Compound Diffuse TIA Tilia americana Tiliaceae Linden, Basswood Simple Diffuse ULA Ulmus americana Ulmaceae American Elm Simple Ring

121 Table S2.The twenty study traits and their corresponding organs, physiological functions and sample sizes. The signs indicate whether we expect each trait to be positively or negatively correlated with the physiological function. These relationships were used to build the hypothesis matrices for Mantel’s tests (Appendix III). The transport function has both efficiency and a safety aspects. The sign of correlation with the transport efficiency aspect is shown. Orange: mechanical support/transport dimension; Grey: architecture dimension; Green: resource acquisition/resource conservation dimension. Physiological Functions

Organ Trait Architecture n refs

Resource Resource Acquisition Mechanical Support Transport Resource Conservation Stem Wood Density (Stem WD) + - 380 1-4 Stem Modulus of elasticity (MOE) + 380 5,27 Branching Distance + 366 6-8 Branch Diameter + 380 6-8 Branching Angle + 269 9-12 Branch Wood Density (Branch WD) + + - 380 1-4, 13-14 Branch Vessel Diameter (VD) + 372 1,9,14 Lumen Fraction - + 372 1,14-15 Conductivity per Sapwood area (KS) + 372 1,15 Conductivity per Branch (KBRANCH) + 372 15 Vessel Thickness-to-Span Ratio (t/d)2 - 373 1,2,14 Leaf Area + 380 6-8 Leaf Mass per Area (LMA) - + 380 16-18 Leaf Thickness + 380 19 Leaf Leaf Carbon Concentration (LCC) + 380 20-21 Leaf Nitrogen Concentration (LNC) + 380 16-18 Leaf Phosphorus Concentration (LPC) + 380 16-18 Water Use Efficiency (WUE) - 380 22 Root Wood Density (Root WD) + + 191 23-24 Root Specific Root Length (SRL) + - 146 25-26 References: 1- Sperry et al. (2006), 2- Hacke and Sperry (2001), 3- Chave et al. (2009), 4 – Niklas (1992), 5- Niklas (1994b), 6- White (1983b), 7- White(1983a), 8-Ackerly & Donoghue (1998), 9 – Niklas & Kerchner (1984), 10- Niklas (1999), 11- Pearcy & Yang (1996); 12- Valladares & Pearcy (1999); 13- Olson et al. (2009); 14 – Preston (2006); 15 – Tyree & Zimmerman (2002) ; 16 – Reich et al. (1999); 17 – Wright et al. (2004) ; 18 –Enquist et al. (2007); 19- Vile et al. (2005) ; 20- Niinemets et al. (2007); 21- Lucas et al. (2000); 22 –

122 Farquhar et al. (1989); 23- Pratt et al. (2007); 24 – Waisel et al. (2002) ; 25- Eissenstat & Yanai (1997); 26- Comas & Eissenstat (2009); 27 - van Gelder et al. (2006).

123 Table S3. Eigenvalue, variance and loadings of functional traits on principal components corresponding to Figure 1. Significant principal components (as determined by comparison with a broken stick model) are marked with a star, and significant loadings are bolded. Orange: mechanical support/transport dimension; Grey: architecture dimension; Green: resource acquisition/resource conservation dimension Principal Component *PC1 *PC2 *PC3 *PC4 *PC5 PC6 Eigenvalue 4.907 4.096 3.129 2.051 1.444 1.050 % variance explained 0.245 0.205 0.156 0.103 0.072 0.052 Cumulative variance 0.245 0.450 0.607 0.709 0.781 0.834

PC1 PC2 PC3 PC4 PC5 PC6 Organ Trait Loadings Stem MOE 0.314 -0.168 0.186 -0.116 0.102 -0.123 Stem Stem WD 0.338 -0.126 0.172 0.192 -0.290 -0.129 Branch Branch WD -0.215 0.141 0.369 0.258 -0.150 -0.078 Branch Branch Diameter -0.169 -0.323 0.031 -0.258 -0.141 0.240 Branch Branching Distance -0.020 -0.372 0.059 -0.274 0.117 -0.092 Branch Branching Angle -0.111 -0.130 0.018 0.535 0.205 0.231 Branch VD 0.038 -0.421 0.096 -0.001 0.078 -0.283 Branch Lumen Fraction -0.284 -0.157 0.101 0.306 -0.142 -0.122

Branch KS -0.102 -0.414 0.103 0.191 0.033 -0.273

Branch KBRANCH 0.278 0.004 0.259 0.233 0.282 -0.205 Branch (t/d)2 0.004 -0.276 -0.094 0.076 -0.367 0.476 Leaf Leaf Area 0.102 -0.398 -0.133 -0.183 0.235 0.124 Leaf WUE -0.284 0.002 0.129 -0.210 -0.404 -0.304 Leaf LMA -0.272 -0.030 0.391 -0.157 -0.183 0.053 Leaf Leaf Thickness 0.323 0.033 0.281 -0.064 -0.126 0.283 Leaf LNC -0.177 -0.047 -0.403 0.206 -0.114 -0.283 Leaf LPC -0.168 0.069 -0.399 -0.101 -0.281 -0.301 Leaf LCC 0.000 0.195 0.254 0.321 0.236 -0.182 Root SRL -0.322 0.143 0.077 0.009 0.142 0.091 Root Coarse Root WD 0.312 0.035 0.171 -0.049 -0.360 -0.037

124 Table S4. Principal component analysis on each of the support-transport, resource acquisition-conservation, and architectural traits. Variables with significant contribution to an axis (as determined by the ordiequilibriumcircle() function of the Biodiversity R package) are shown in black and bold whereas variables without significant contributions are shown in grey. The significant principal components are black and bolded.

S4.1 - All resource acquisition-conservation traits (Broadly defined Leaf Economic Spectrum) Eigenvalue 2.860 2.009 0.918 0.712 0.287 0.149 Variance explained 0.409 0.287 0.131 0.102 0.041 0.021 Cumulative variance 0.409 0.696 0.827 0.928 0.969 0.991 Organ Trait PC1 PC2 PC3 PC4 PC5 PC6 Leaf LMA 0.553 -0.083 0.150 -0.139 0.382 -0.173 Leaf LNC -0.125 0.632 -0.011 0.151 0.638 -0.322 Leaf LPC -0.021 0.649 -0.250 -0.204 -0.197 0.554 Leaf Leaf Thickness 0.520 0.001 0.317 0.293 0.199 0.593 Leaf LCC 0.260 -0.234 -0.833 -0.216 0.276 0.078 Leaf WUE 0.418 0.263 0.191 -0.622 -0.315 -0.304 Root SRL 0.406 0.219 -0.290 0.630 -0.440 -0.331 S4.2 - All support-transport traits (Broadly defined Wood Spectrum) Eigenvalue 3.073 2.611 1.597 1.125 0.593 0.437 variance explained 0.307 0.261 0.160 0.113 0.059 0.044 Cumulative variance 0.307 0.568 0.728 0.841 0.900 0.944 Organ Trait PC1 PC2 PC3 PC4 PC5 PC6 Stem Stem WD 0.519 -0.132 0.114 -0.167 0.018 -0.018 Branch VD 0.133 -0.545 -0.141 0.385 -0.119 0.292 Branch KS -0.004 -0.629 0.002 0.091 0.018 0.121 Stem MOE 0.448 -0.120 -0.056 0.363 0.237 -0.668 Branch Branch WD 0.372 0.062 0.535 -0.199 -0.423 -0.151 Branch (t/d)2 0.108 -0.269 -0.491 -0.617 -0.407 -0.133 Branch KBRANCH -0.292 -0.242 0.565 0.115 0.367 -0.031 Branch Lumen Fraction -0.229 -0.363 0.305 -0.483 0.608 -0.181 Root Coarse Root WD 0.473 0.093 0.155 -0.134 0.282 -0.613 S4.3 - All architectural traits (Broadly defined Corner’s Rules) Eigenvalue 2.464 1.070 0.799 0.439 0.228 Variance explained 0.493 0.214 0.160 0.088 0.046 Cumulative variance 0.493 0.707 0.867 0.954 1.000 Organ Trait PC1 PC2 PC3 PC4 PC5 Leaf Leaf Area -0.552 0.074 -0.339 0.310 -0.692 Branch Branch Diameter -0.510 0.162 0.231 -0.811 -0.053 Branch Branching Distance -0.503 -0.280 -0.493 0.078 0.648 Branch Branch WD 0.427 -0.042 -0.743 -0.476 -0.194 Branch Branching Angle 0.001 0.942 -0.193 0.117 0.248

125 Table S5. Species scale trait correlation structure. Orange: mechanical support/transport dimension; Grey: architecture dimension; Green: resource acquisition/resource conservation dimension. n=24.

2

BRANCH

VD (t/d) Ks K Lumen Fraction Branch WD Stem WD Coarse Root WD MOE LCC LNC LPC WUE LMA Leaf Thickness SRL Branch Diameter Leaf Area Branching Angle Branching Distance VD 1.00 ns 0.87 ns ns ns ns ns ns ns ns ns ns ns ns ns 0.44 0.61 ns 0.61 (t/d)2 ns 1.00 ns ns ns ns ns ns ns -0.42 ns ns ns ns ns ns ns ns ns ns Ks 0.87 ns 1.00 ns 0.52 ns ns ns ns ns ns ns ns ns ns ns 0.43 0.46 0.41 0.55

KBRANCH ns ns ns 1.00 0.51 ns ns ns ns ns ns ns ns 0.50 0.53 0.46 ns ns ns ns Lumen Fraction ns ns 0.52 0.51 1.00 ns ns ns ns ns ns ns ns 0.47 0.42 ns ns ns 0.56 ns Branch WD ns ns ns ns ns 1.00 0.68 0.58 ns ns -0.50 -0.55 ns ns ns ns -0.51 -0.42 ns ns Stem WD ns ns ns ns ns 0.68 1.00 0.70 0.65 ns ns -0.41 ns ns ns -0.60 ns ns ns ns Coarse Root WD ns ns ns ns ns 0.58 0.70 1.00 0.50 ns -0.44 ns ns ns ns -0.45 ns ns ns ns MOE ns ns ns ns ns ns 0.65 0.50 1.00 ns -0.43 -0.41 ns ns -0.42 -0.44 ns 0.43 ns ns LCC ns -0.42 ns ns ns ns ns ns ns 1.00 0.00 0.00 ns ns ns ns ns ns ns ns LNC ns ns ns ns ns -0.50 ns -0.44 -0.43 ns 1.00 0.74 ns ns ns ns ns ns ns ns LPC ns ns ns ns ns -0.55 -0.41 ns -0.41 ns 0.74 1.00 ns ns ns ns ns ns ns ns WUE ns ns ns ns ns ns ns ns ns ns ns ns 1.00 0.66 0.51 ns ns ns ns ns LMA ns ns ns 0.50 0.47 ns ns ns ns ns ns ns 0.66 1.00 0.83 0.47 0.44 ns ns ns Leaf Thickness ns ns ns 0.53 0.42 ns ns ns -0.42 ns ns ns 0.51 0.83 1.00 0.60 ns ns ns ns SRL ns ns ns 0.46 ns ns -0.60 -0.45 -0.44 ns ns ns ns 0.47 0.60 1.00 ns ns ns ns Branch Diameter 0.44 ns 0.43 ns ns -0.51 ns ns ns ns ns ns ns 0.44 ns ns 1.00 0.54 ns 0.46 Leaf Area 0.61 ns 0.46 ns ns -0.42 ns ns 0.43 ns ns ns ns ns ns ns 0.54 1.00 ns 0.70 Branching Angle ns ns 0.41 ns 0.56 ns ns ns ns ns ns ns ns ns ns ns ns ns 1.00 ns Branching Distance 0.61 ns 0.55 ns ns ns ns ns ns ns ns ns ns ns ns ns 0.46 0.70 ns 1.00

126 Table S6. Summary of trait values (calculated on untransformed data). Branch WD: Branch Wood Density; VD: Vessel Diameter; KS: Conductivity per sapwood area; KBRANCH: Conductivity per branch; (t/d)2: vessel thickness to diameter ratio; Stem WD: Stem Wood Density; MOE: Modulus of elasticity; LMA: Leaf Mass per Area; LCC: Leaf Carbon Concentration; LNC : Leaf Nitrogen Concentration; LPC : Leaf Phosphorus Concentration; WUE: Water Use Efficiency; SRL: Specific Root Length; Root WD: Root Wood Density. ‘-fold variation’ refers to the values of the 3 quantile to hat of the first quantile, which have been selected to avoid the spurious effects of outliers. Organ BRANCH STEM

Branching Branch Branching Branch Lumen 2 Stem Trait VD KS KBRANCH (t/d) MOE Distance WD Angle Diameter Fraction WD

Units cm g/cm3 degrees mm μm cm2/cm2 mm2/s*kPa mm4/s*kPa μm/μm g/cm3 N/mm2 mean 26.83 0.48 52.27 5.67 57.49 0.13 1.11 11.14 0.02 0.59 3650.29 1st Q. 7.16 0.32 0.00 3.35 30.98 0.07 0.23 3.47 0.01 0.42 700.50 3rd Q. 100.00 0.60 82.66 12.32 109.66 0.23 4.45 29.56 0.04 0.82 12963.62 -fold variation 13.96 1.84 Inf 3.68 3.54 3.14 19.49 8.53 3.83 1.95 18.51 (3rd/1st Q.) sd 23.21 0.07 17.86 2.10 55.53 0.04 1.39 7.43 0.01 0.11 2990.30 C.V. 0.87 0.14 0.34 0.37 0.97 0.32 1.25 0.67 0.34 0.19 0.82

Organ LEAF ROOT Leaf Leaf Tissue Coarse Trait LMA LCC LNC LPC WUE SRL Area Density Root WD

Units cm2 g/cm2 g/mm3 g/g g/g g/g δ13C‰ m/g g/cm3 mean 62.35 55.08 0.39 47.62 2.66 16.37 -30.11 107.57 0.58 1st Q. 7.74 25.43 0.22 42.46 1.80 8.46 -32.34 20.57 0.41 3rd Q. 294.60 111.99 0.62 51.55 3.74 32.67 -27.34 259.43 0.77 -fold variation 38.07 4.40 2.88 1.21 2.08 3.86 1.18 12.61 1.88 (3rd/1st Q.) sd 83.30 24.84 0.11 2.37 0.52 5.74 1.42 61.78 0.10 C.V. 1.34 0.45 0.28 0.05 0.20 0.35 -0.05 0.57 0.16

127

…

Figure S1 – Interspecific bivariate trait correlations for the few traditional traits that form the basis of the three trait dimensions and that are expected to be correlated. SMA regressions are given only for significant correlations. Orange: mechanical support/transport dimension; Grey: architecture dimension; Green: resource acquisition/resource conservation dimension.

128 SUPPLEMEMENTARY MATERIAL – APPENDIX II. DETAILED METHODS

Branch Collection. For each tree, the vertical height DBH (diameter at breast height) and DGH (diameter at ground height) were recorded. One healthy lateral branch of at least 50cm in length and 5mm in diameter, located in the top third of the tree crown (but not the leader) was randomly selected, pruned off and taken back to the lab for trait measurements the same day. From each branch a 50cm long segment (measured from the tip of the longest branch) was cut and all following measurements made on it. Leaf Traits. Total leaf area of the branch as well as average individual leaf area (LA) was measured using a Li-3100C leaf area meter at a resolution of 1mm2. LA measurements included the petiole on simple leaves and the rachis and petiolule on compound leaves. Leaves on each branch were divided sorted as healthy and mature or non-healthy and/or immature. A leaf with less than ca.10% of its area damaged was considered healthy. All individual leaf measurements were done on the healthy subset. The thickness of leaf blades (Leaf thickness) was measured by taking 4 micrometer measurements on each of five randomly selected healthy leaves. The measurements were taken halfway between the mid-vein and the edge of the leaf, avoiding major secondary veins. Total branch mass, total leaf mass, and average individual leaf mass of healthy leaves were measured after drying the material for 72hrs or more of at 60°C in a drying oven. Leaf mass fraction (LMF) was calculated as the ratio of the total leaf mass of a branch to its total branch mass, leaf mass per area (LMA) as the ratio of leaf mass per leaf area of individual healthy leaves. Stoichiometric composition was measured on the bulk sample of healthy leaves after removing petioles and mid-veins and grinding them using a Thomas Wiley mini-Mill. Leaf nitrogen content (LNC), carbon content (LCC) and carbon isotope ratio (δ13C) were conducted in the isotope lab of the University of Arizona. Leaf phosphorus content (LPC) was measured on triplicates by colorimetric assays after persulfate oxidation and reaction with acid molybdate (American Public Health Association 1999) using a ThermoScientific Genesys20 spectrophotometer. Branch Traits. The average branching distance of each branch was calculated by measuring the total branch length of the segments, including all lateral segments and

129 dividing it by the number of bifurcations. Twig diameter at the base of the 50cm segments was recorded as the average of two perpendicular measurements taken with electronic calipers measuring to 2 decimals after the mm. Average branching angle (Branching Angle) of each branch was measured on pictures of the defoliated branches on a white background using ImageJ. Only angles lying flat against the plane were measured and for each branch, the smallest and largest angles were excluded. Branch wood density (Branch WD) was measured by volumetric replacement methods on a 12cm long segment of branch of ca.5mm diameter. This section was taken from the 50cm branch segment when possible or from further up along the branch when the diameter at the base of the 50cm segment was less than 5mm. A 10 mm segment with an average diameter of ca.5mm was cut and maintained frozen for xylem anatomy measurements. This segment was also taken on the 50cm segment when possible and further up the branch otherwise. Xylem Anatomy Traits. Safranin-stained cross sections of ca. 100 microns in thickness cut from each branch using a microtome and permanently mounted on microscope slides. Digital pictures were taken at 20x, 50x and 200x using an Olympus BX- 50 microscope, an Optronics Microfire microscope camera and the PictureFrame software. The 20x pictures were used to calculate the % sapwood area of the branch with Image J. Bark and pith area were excluded from the sapwood area but since the branches were 1-10 yrs old, we assumed all xylem area to be conductive for diffuse-porous species and only calculated the last year of xylem area for ring-porous species. The 50x pictures were used to calculate the ratio of cell wall thickness to vessel diameter (t/d2) in twelve adjacent vessel pairs of similar size (area within 30% of each other) for each sample. The 200x pictures were used to calculate for each sample the vessel diameter (VD) of all the vessels within two to three pie-shaped xylem areas going from pith to bark. This was done using STEM_GUI, a software developed for ImageJ. Only conduits greater than 25microns in diameter were considered as vessels (Sperry et al. 2006). STEM_GUI calculates total conductivity from the individual vessel diameters using the Hagen- Poisseuille equation. We then calculate conductivity per sapwood area (Ks) and per

130 Branch (KBRANCH). Lumen Fraction was calculated as the sum of the vessel area over the pie-area. Root Traits. Before the 2012 leaf flush (from April 23rdth to May 11th), root growth bags (8-9 per study species) were installed on 5-8mm diameter roots on 200 of the study saplings. The roots were identified by following the main stem, were cut with pruners, placed in the root bag with a mixture of the original soil and ca.30% sand, watered, relocated in their original location and covered with soil and litter. Cutting the coarse roots of this size before leaf flush stimulates growth of fine absorptive roots. The 20cm x 25cm root bags were fabricated with landscape fabric and nylon thread in order to prevent the roots of the focal tree from growing roots outside the bag and neighboring roots from growing into the bag. Three months later (July 16th to 30th) the bags were harvested in the same order as they were installed. The roots were rinsed and the 1st and 2nd order roots were stained with Neutral Red and scanned using WinRhizo to calculate total fine-root length. They were then oven dried for >72hrs at 60°C and weighed. SRL was calculated by dividing the fine roots total weight by their total length. Of the 200 root bags installed, 159 produced 1st and 2nd order absorptive roots. The 5-8mm coarse roots to which the fine roots were attached were used to calculate coarse root wood density (Coarse Root WD). Stem Traits. At the end of the 2012 growing season, the trees were cut to collect a section of the stem at the base. This section was used to age the trees, calculate yearly basal area increment and calculate stem wood density (Stem WD). After removing the outer bark, Stem WD was measured using volumetric replacement methods. Modulus of elasticity of the stem (MOE) was calculated using Euler’s equation from Tree Height, DGH and Stem WD.

131 SUPPLEMENTARY MATERIAL – APPENDIX III. HYPOTHESES FOR MANTEL TESTS

Leaf Economic Spectrum

Figure S2 – Imputed trade-offs among physiological functions underpinning the LES

Table S7- Hypothesis and data matrices for LES S7a. Hypothesis matrix for narrowly S7b.-Hypothesis matrix for broadly defined LES defined LES

Thickness

C C

13 13

LMA LNC LPC Leaf LCC δ SRL LMA LNC LPC Leaf Thickness LCC δ SRL LMA 1 -1 -1 0 0 0 0 LMA 1 -1 -1 1 1 1 -1 LNC -1 1 1 0 0 0 0 LNC -1 1 1 -1 -1 -1 1 LPC -1 1 1 0 0 0 0 LPC -1 1 1 -1 1 -1 1 Leaf Thickness 0 0 0 1 0 0 0 Leaf Thickness 1 -1 -1 1 1 1 -1 LCC 0 0 0 0 1 0 0 LCC 1 -1 1 1 1 1 -1 δ13C 0 0 0 0 0 1 0 δ13C 1 -1 -1 1 1 1 -1 SRL 0 0 0 0 0 0 1 SRL -1 1 1 -1 -1 -1 1

S7c. Observed correlation matrix for LES traits

C

13

LMA LNC LPC Leaf Thickness LCC δ SRL LMA 1.00 -0.25 -0.18 0.83 0.38 -0.66 0.47 LNC -0.25 1.00 0.74 -0.14 -0.36 -0.08 0.14 LPC -0.18 0.74 1.00 -0.12 -0.11 -0.35 0.24 Leaf Thickness 0.83 -0.14 -0.12 1.00 0.13 -0.51 0.60 LCC 0.38 -0.36 -0.11 0.13 1.00 -0.11 0.28 δ13C -0.66 -0.08 -0.35 -0.51 -0.11 1.00 -0.33 SRL 0.47 0.14 0.24 0.60 0.28 -0.33 1.00

132 Wood Spectrum

Figure S3 – Imputed trade-offs among physiological functions underpinning the WS Table S8. Hypothesis and data matrices for WS S8a. Hypothesis matrix for narrowly S8b. Hypothesis matrix for broadly defined WS defined WS

2 2

S BRANCH S BRANCH

Stem WD VD K MOE Twig WD Lumen Fraction K (t/d) Root WD Stem WD VD K MOE Twig WD Lumen Fraction K (t/d) Root WD Stem WD 1 -1 -1 1 0 0 0 0 0 Stem WD 1 -1 -1 1 1 1 -1 -1 1 VD -1 1 1 -1 0 0 0 0 0 VD -1 1 1 -1 -1 -1 1 1 -1 KS -1 1 1 -1 0 0 0 0 0 KS -1 1 1 -1 -1 -1 1 1 -1 MOE 1 -1 -1 1 0 0 0 0 0 MOE 1 -1 -1 1 1 1 -1 -1 1 Twig WD 0 0 0 0 1 0 0 0 0 Twig WD 1 -1 -1 1 1 1 -1 -1 1 Lumen Lumen 0 0 0 0 0 1 0 0 0 1 -1 -1 1 1 1 -1 -1 1 Fraction Fraction KBRANCH 0 0 0 0 0 0 1 0 0 KBRANCH -1 1 1 -1 -1 -1 1 1 -1 (t/d)2 0 0 0 0 0 0 0 1 0 (t/d)2 -1 1 1 -1 -1 -1 1 1 -1 Root WD 0 0 0 0 0 0 0 0 1 Root WD 1 -1 -1 1 1 1 -1 -1 1

S8c. Observed correlation matrix for WS traits

2

S BRANCH

Stem WD VD K MOE Twig WD Lumen Fraction K (t/d) Root WD Stem WD 1.00 0.31 0.19 0.65 0.68 -0.12 -0.34 0.23 0.70 VD 0.31 1.00 0.87 0.38 -0.09 0.12 0.12 0.29 0.02 KS 0.19 0.87 1.00 0.17 -0.12 0.52 0.37 0.34 -0.14 MOE 0.65 0.38 0.17 1.00 0.36 -0.28 -0.33 0.09 0.50 Twig WD 0.68 -0.09 -0.12 0.36 1.00 -0.09 0.07 -0.12 0.58 Lumen Fraction -0.12 0.12 0.52 -0.28 -0.09 1.00 0.51 0.11 -0.26 KBRANCH -0.34 0.12 0.37 -0.33 0.07 0.51 1.00 -0.29 -0.36 (t/d)2 0.23 0.29 0.34 0.09 -0.12 0.11 -0.29 1.00 0.02 Root WD 0.70 0.02 -0.14 0.50 0.58 -0.26 -0.36 0.02 1.00

133 Corner’s Rules

Figure S4 – Imputed relationships underpinning CR traits resulting from the interplay between form, function and metabolic and biomechanical constraints. Mechanism proposed by Olson (2009). Table S9. Hypothesis and data matrices for CR S9a. Hypothesis matrix for narrowly defined CR S9b. Hypothesis matrix for broadly defined CR

Leaf Area Branch Diameter Branching Distance Branch Wood Density Branching Angle Leaf Area Branch Diameter Branching Distance Branch Wood Density Branching Angle Leaf Area 1 1 1 0 0 Leaf Area 1 1 1 -1 1 Branch Diameter 1 1 1 0 0 Branch Diameter 1 1 1 -1 1 Branching Distance 1 1 1 0 0 Branching Distance 1 1 1 -1 1 Branch Wood Density 0 0 0 1 0 Branch Wood Density -1 -1 -1 1 -1 Branching Angle 0 0 0 0 1 Branching Angle 1 1 1 -1 1

S9c. Observed correlation matrix for CR traits

Angle

Leaf Area Branch Diameter Branching Distance Branch Wood Density Branching Leaf Area 1.00 0.54 0.70 -0.42 0.10 Branch Diameter 0.54 1.00 0.46 -0.51 0.08 Branching Distance 0.70 0.46 1.00 -0.27 -0.17 Branch Wood Density -0.42 -0.51 -0.27 1.00 0.04 Branching Angle 0.10 0.08 -0.17 0.04 1.00

134

APPENDIX C:

Variable intrapopulation phenotypic integration structure in temperate deciduous trees

Prepared for submission in: American Journal of Botany

Authors : Julie Messier1, Cyrille Violle2, Brian Enquist1, Martin Lechowicz3 & Brian McGill4

Affiliations:

1 Ecology and Evolutionary Biology, University of Arizona, Tucson, AZ 85721, USA 2 Centre d’Écologie Fonctionelle et Évolutive, CNRS, Montpellier, 34293, France

3 Biology Department, McGill University, Montréal, QC, H3A1B1, Canada. 4 School of Biology and Ecology, University of Maine, Orono, ME 04469, USA.

Corresponding author: Julie Messier University of Arizona 1041 E. Lowell Street, Tucson, Arizona, 85711, USA 520-626-3336 [email protected]

Keyword: Functional traits, static integration, evolutionary integration, trade-offs, temperate trees

Short title: Intrapopulation integration structure

135

ABSTRACT

Premise of the study: Generalities across species in intrapopulation phenotypic integration highlight the fundamental constraints and trade-offs shaping phenotypic diversity. Yet, it remains largely unknown whether there are general patterns of functional trait correlation within populations. Here we explore similarities and differences in intrapopulation phenotypic integration structures in locally coexisting tree species. Methods: Using Mantel tests, we compare the correlation structure of several key functional traits within and among temperate tree species. We ask: (i) Are some intrapopulation pairwise trait correlations shared across all species to form a common ‘tree template’? and ; (ii) are species differences in phenotypic integration structure associated with differences in environmental niche or phylogenetic distance? Key Results: We find that intrapopulation integration structures of the different species are weak and distinct from each other. Three trait pairs, Leaf Mass per Area - Carbon isotope ratio, Conductivity per Sapwood Area -Lumen Fraction and Conductivity per Branch - Lumen Fraction, do show consistent intrapopulation correlations in all species. Variation in intrapopulation phenotypic integration among species is not related to environmental or phylogenetic differences. Conclusions: The species-specific nature of intrapopulation trait integration structure challenges the existence of fundamental constraints and trade-offs among the ecophysiological functions measured here and instead suggests flexibility in plant design. Further, the lack of relationship between phenotypic integration structure and environmental niche is consistent with two contrasting interpretations: (i) phenotypic integration affects plant performance, but alternative designs can produce equivalently effective solutions to optimizing performance; or (ii) local patterns of phenotypic integration within populations are decoupled from local plant performance.

136

INTRODUCTION

A goal of comparative plant ecology is to find generalities in the relationships among key ecophysiological traits, called functional traits (e.g. Díaz et al. 2004, Violle et al. 2007, Garnier and Navas 2012). While the field has searched widely for universal interspecific relationships between functional traits (e.g. Reich et al. 1999, Wright et al. 2004) and their relationships with the environment (Violle et al. 2014), we have barely explored whether universal patterns of integration for functional traits exist within populations. Further, comparing intrapopulation phenotypic correlations is a robust way to explore relationships among functional traits. Since these correlations are not shaped by evolutionary divergence among species, they reflect fundamental genetic, developmental or physiological trade-offs and constraints (Armbruster et al. 2004). In fact, phenotypic correlations at this scale (called ‘static’ correlations) often serve as a basis for comparison in integration studies (Klingenberg 2014). Understanding the generalities and differences in the static integration patterns of functional traits is a fundamental yet largely unexplored goal with important ecological and evolutionary implications. First, ubiquitous static integration patterns would suggest fundamental functional trait relationships among plant ecophysiological functions, whether they arise from fundamental constraints or consistent patterns of selection (Westoby et al. 1995). Is there a common ‘tree sapling trait template’ around which all tree species have evolved? Second, similarities and differences in static and interspecific (called ‘evolutionary’) integration patterns of functional trait integration can point to the drivers shaping interspecific phenotypic integration (Armbruster et al. 2004). For example, fundamental constraints should lead to similar static and evolutionary integrations (Shoval et al. 2012). Adaptive divergences among species should lead to differences in static integration related to environmental niches. Phylogenetic constraints (a.k.a. conservatism) also could lead to differences in static integration that are phylogenetically structured. Third, static trait correlation patterns can also have important consequences for species’ distributions and response to environmental gradients. For example, whether two traits covary in the same direction as the selection

137

gradient in an adaptive landscape affects whether the population can respond to that gradient (Davis and Shaw 2001, Laughlin and Messier 2015).

Comparing integration structures across scales is an approach mainly adopted for allometric or morphometric traits within and among structures (e.g. Cheverud 1982a, Merila and Bjorklund 1999, Klingenberg 2014). In plants, comparisons have been centered on the integration (or modularity) of morphometric floral traits (Armbruster 1991, Herrera et al. 2002, Conner et al. 2014) in species with specialized vs. generalized pollination (e.g. Berg 1960, Armbruster et al. 1999). Patterns of static phenotypic integration have hardly been studied in functional traits (but see Fajardo and Piper 2011, Richardson et al. 2013). Yet, they are central to comparative ecology because they reflect physiological function, describe adaptive strategies and affect ecosystem functions. We thus have much to gain in understanding the role of fundamental trade-offs and constraints in shaping the relationships among functional traits. We explore the generality of static integration structure of functional trait in tree saplings by measuring trait correlations in locally co-occurring angiosperm tree species. We ask: (1) Are some static trait correlation patterns consistent across species? ; (2) Are differences among static correlation structures related to environmental niche differences or to phylogenetic distance? ; and (3) Are static and evolutionary traits correlation patterns similar?

To answer these questions, we examined the ‘static’ (among individuals of the same developmental stage within a population) and ‘evolutionary’ (among species) correlation structure of 20 functional traits in saplings from 24 temperate tree species coexisting within a 20 km2 area on and immediately around a topographically diverse hill. These traits were sampled on 15 to 20 individuals per species, were measured on different organs (leaves, branches, stems, roots), and reflect plant architecture and four vital physiological functions, namely resource acquisition, conservation, sap transport and mechanical support. We also characterized the above- and belowground abiotic and competitive environment of each tree to quantify the environmental niche of the

138

populations. We tested for consistent pairwise and multivariate static correlations patterns with regressions and ordinations. We quantified the overall similarity of the correlation structures using Mantel tests and related these similarities to phylogenetic distance and environmental niche similarity. Last, we compared static and evolutionary correlations and examined relationships between intra- and interspecific trait correlation strengths.

139

METHODS

Study Site We collected traits for each of 380 tree saplings in a ca. 20 km2 area located on or around Mont Saint-Hilaire, Canada (45°33′8″N 73°9′3″W). This hill, located in Southwestern Québec in the transition zone between the boreal forests to the north and the eastern deciduous forests to the south, is of plutonic origin. It includes a series of seven hilltops that rise up to 414m above the sedimentary floor of the surrounding Saint- Lawrence river valley (Feininger and Goodacre 1995). The mixed wood forest on the mountain is the largest remnant of primeval forest in the region. It is surrounded by an agricultural and suburban mosaic containing multiple forest fragments of varying successional status. The mountain’s unusual geological history created a broad diversity of habitats on and around the mountain (Maycock 1961, Arii and Lechowicz 2002, Arii et al. 2005). A local diversity hotspot, the nature reserve contains 650-900 of the 1,600 regional species of terrestrial vascular forest plants (Maycock 1961, Elliott and Davies 2014). Sampling Design We conducted an individual-based sampling. Being at the tail-end of the establishment phase, a critical growth phase with high mortality (Grubb 1977), saplings express traits that are probably linked to successful way of life in the environment (Poorter 2007, Sterck et al. 2014). Fifteen to 20 healthy tree saplings from the 24 most abundant deciduous tree species in the area were sampled, for a total of 380 saplings (Table S1). Sapling size was restricted to 1-5cm diameter at breast height (DBH) and a height of less than two thirds of the canopy height in order to sample individuals whose crowns are in the sub-canopy layer and to standardize ontogenetic stage (Palow et al. 2012, Mason et al. 2013). To maximize the range of trait values within each species and overlap in environment among species, healthy individuals were randomly selected from a diversity of habitats and microhabitats on and near the base of the mountain (Figure S2). All sampling dates were recorded and we removed the phenological effects on trait values when applicable. Tree age was determined by counting tree rings of cross-sections from the base of the tree.

140

Traits The 20 traits measured on each sapling are listed in Table 3. The specific methods for the measurement of these traits are provided in the supplementary material. The traits include those traditionally measured to reflect architecture and four vital biological functions: resource acquisition, resource conservation, sap transport and mechanical support. The traits were measured from all vegatative plant organs, namely the leaves, roots, stems and branches. For resource acquisition and conservation, leaf mass per area (LMA), leaf thickness, leaf nitrogen content (LNC), leaf phosphorus content (LPC), carbon isotope ratio (δ13C) in leaves (an integrated index of water use efficiency), specific root length (SRL), Leaf Thickness and leaf carbon concentration (LCC) were measured. For mechanical support, stem wood density (Stem WD), branch wood density (Branch WD), coarse root wood density (Root WD) and stem Modulus of Elasticity (MOE) were measured. For sap transport safety and efficiency, vessel diameter (VD), vessel span-to- thickness ratio ((t/d)2) conductivity per sapwood area (KS), conductivity per branch

(KBRANCH) and Lumen Fraction were measured. For architectural traits, individual Leaf Area, Branching Distance, Branch Diameter and Branching Angle were measured. We compiled shade, drought and waterlogging tolerance indices for our study species from Niinemets & Valladares (2006) in order to relate the phenotype to natural history. Environmental Variables For each sapling, a set of nineteen environmental variables were measured (Table S2). These have been identified as important drivers of community composition at the study site (Enright and Lewis 1985, Arii and Lechowicz 2002, Gilbert and Lechowicz 2004, Arii et al. 2005, Karst et al. 2005, Radovski 2010). The variables reflect the aboveground (irradiance, slope, aspect and air temperature), and belowground abiotic environments (soil nutrient composition, pH, humidity and depth) as well as two competition indices measured as the basal area of symmetric and asymmetric neighbors. Detailed methods for these measurements are included in Appendix III of the Supplemental Materials. We reduced the set of environmental variables to six principal component which were then used into all further analyses.

141

Phylogeny We generated a phylogeny for our species list using Phylomatic (Webb and Donoghue 2005) based on the Bell et al. (2010) supertree (Figure S1). We scaled branch lengths using known node ages from Wickstrom et al. (Wikström et al. 2001) with the BLADJ procedure in Phylocom 3.40 (Webb et al. 2008). Therefore, the phylogenetic distance estimates are in millions of years. Statistical Analyses All statistical analyses were conducted in R version 3.1.2 (R Development Core Team 2011). Following Kerkhoff and Enquist (2009), traits reflecting size and rates were natural-log transformed (LMA, Leaf Area, Branch diameter, KS and KBRANCH ). All other traits were transformed following individual boxcox transformations to optimize normality when necessary (Box and Cox 1964) using the powerTransform() function of the car package. Tree age had a marginally statistically significant yet minimal effect on trait correlations (r2 = 0.006, p = 0.07) and tree height had a significant but minimal effect on trait correlations (r2 = 0.0016, p=0.007), so the trait values were not corrected for these factors. The effect of sampling date was removed for traits with a significant date effect by taking the residuals of a linear regression of individual traits against sampling date. The height aboveground at which each branch was sampled did not have an effect on δ13C values as can potentially occur (Farquhar et al. 1989). The rcorr() function of the Hmisc package was used to calculate the statistical significance of static pairwise trait correlations. A trait pair was considered consistently correlated across all species when the mean static correlation of the 24 species was 0.5 or larger. We chose this threshold value because the minimum statistically significance pairwise static correlation varied from 0.43 to 0.55 across trait pairs and species, with an average of 0.50. Principal component analyses and redundancy analyses were conducted using the rda() function of the vegan package. We used PCASignificance() function of the BiodiversityR package to evaluate the number of significant PCA axes. This function uses a broken-stick null model, a method that has been shown to be one of the most robust

142

(Jackson 1993, Legendre and Legendre 1998). The significance of the contribution of each variable to each principal component axis was determined using the ordiequilibriumcircle() function of the same package. We calculated the similarity of the static integration matrices of the species using standardized Mantel correlations and report these values in Table 2. Mantel correlations give the Pearson’s correlation coefficient (rM) between two correlation matrices and use permutation tests to calculate significance (Zuur et al. 2007). We also used this test to calculate the correlations between the species similarity matrix (i.e. Table 2) and both the phylogenetic distance matrix and environmental niche similarity.

We calculated the niche similarity of the different species in two ways. First we calculated the Euclidian distance between the environmental niche centroid of species pairs. We measured the Euclidian distances between species pairs along the first six principal components of the PCA on all environmental variables. We weighted the distance between two species centroids along each principal component by its eigenvalue. Six principal components were used because on average the distance between species pairs in environmental space plateaued at 6 dimensions. Second, we calculated the overlap between the environmental hypervolumes of species pairs using the hypervolume(), hypervolume_set() and hypervolume_sorensen_overlap() functions of hypervolume package (Blonder et al. 2014). We used a bandwidth of 0.06 and 4 dimensions because high-dimensional spaces rapidly become sparse as dimensionality increases, decreasing our ability to calculate overlap. Environmental Effects Trait variation in natural populations has genetic, environmental and gene-by- environment components (Schlichting 1989, Waitt and Levin 1993). The overlap in environmental hypervolume among species is large, but not complete (Figure S3), Sorensen’s similarity from 0.3-0.7). Thus, to assess the relationship between the environmental niche of a species and its trait correlation structure, for each study sapling we measured nineteen environmental variables in the immediate vicinity the tree (Table S2). We partitioned trait variance among species, environment, and their interaction

143

using the varpart() function of the vegan package for individual traits as well as for the multivariate trait space. Forward stepwise selection of traits against environmental principal components was used to retain only the significant environmental components using the step() function of the stats package for individual traits and the ordistep() function of vegan package for the multivariate trait space. To verify the extent to which environmental effects were affecting trait integration and influencing the results, all analyses were done with both uncorrected data, T, and environment-free data, TE. The results of the analyses conducted with the TE data are presented in Appendix II of the supplementary material because they are qualitatively identical to those without environmental corrections. To calculate environment-free data, TE, we first isolated the “pure” environmental effects from the environment-by-species interaction by taking the residuals of models correlating individual environmental variables against species. Due to the species turnover along environmental gradients, a fraction of environmental and species variance is shared (see Figure S5). Isolating the “pure” environmental effect avoids removing the fraction of interspecific variation associated with the species’ environmental affinity. We then conducted forward stepwise selection for each trait against the “pure” environmental variables, where only the environmental variables with significant partial correlations were retained. The residuals of these models then constitute the environment-free trait values, TE. We assessed the effect of the environment on the phenotypic integration of each species by measuring the similarity of the [T] and [TE] correlation matrices of each species with Mantel tests.

144

RESULTS

Consistent bivariate static trait correlations are rare The majority of functional trait pairs exhibit weak or no static correlations. Across species, most trait pairs have an average static correlation of 0 and a standard deviation around 0.2 (Figure 1, Table S3), indicating very variable static correlations. Notably, some trait pairs show significant positive and negative static correlations. Across species, 85% of all pairwise static correlations are non-significant. Only six trait pairs are consistently correlated within populations across all species (Figure 1): KS-KBRANCH, VD- KS, LMA- Leaf

Thickness and LMA-δ13C, KS-Lumen Fraction and KBRANCH-Lumen Fraction. The first three pairs are dimensionally related: KS-KBRANCH share a common numerator, VD goes into the calculation of KS and LMA is related to Leaf Thickness because leaf mass is the product of leaf tissue density and thickness. Principal component analyses on trait variation within individual species also show unique trait correlation structures for each species (Fig S4). Only a few groups of traits are repeatedly associated with each other across species. These groups are consistent with the bivariate correlations shown in Figure 1. In PCA space, Ks, KBRANCH and Lumen Fraction often appear together, as well as with VD in a number of species. LMA is frequently associated with either δ13C or Leaf Thickness, although Leaf thickness and δ13C are seldom associated. Stem WD and Branch WD appear together in roughly half of the species. Last, two or three traits typically associated with Corner’s rules (MOE, Leaf Area, Branching Distance and Branch Diameter) are often associated in varying combinations.

Static phenotypic integration structures are species-specific Species differ in their multivariate static integration structure. Mantel correlations, rM, vary from not significant to r=0.54 for the Prunus virginiana - Populus deltoides pair (Table 2). Overall 21% of all species pairs are not significantly correlated. For those species pairs significantly correlated, the average integration structure similarity is 0.27.

Phenotypic integration is not related to the environmental niche, natural history or phylogeny

145

The similarity of species’ static phenotypic integration is not related to their phylogenetic distance (Mantel test, p = 0.80, Figure 2A) although individual traits showed phylogenetic conservatism (Table S4). Similarity of species’ phenotypic integration is not related to the similarity of species’ environmental niches, either assessed as the distance between environmental niche centroids (Mantel test p = 0.85, Figure 2B), or as the overlap in environmental niche hypervolume (Mantel test p = 0.76, Figure 2C). Further, similarity in phenotypic integration are not correlated with similarity in stress tolerances as measured by shade, drought or waterlogging tolerance indices (Mantel tests, p = 0.166, p = 0.94, and p = 0.41, respectively). The same results are obtained when considering only the traits that are consistently correlated within all species.

Static and evolutionary integration structures are largely different For all species, the similarity between static and evolutionary integration structures is low. Measures of similarity of integration based on Mantel correlations, rM, vary from not significant to 0.44 for Acer saccharum (Table 2). The static integration structure of four of the 24 species are different (not significantly correlated) from the evolutionary integration structure. For the remaining 20 species, the average similarity between the static integration structure and the evolutionary integration structure is low at 0.26. Moreover, the relationship between the strength of static and evolutionary pairwise trait correlations is weak (Figure 3). In fact, 40 of the 45 statistically significant evolutionary trait correlations are not consistently correlated at the static level (Table S3).

Similarities and differences in phenotypic integration are not due to environmental effects Removing the environmental effects on phenotypic integration did not affect the results. The same results are found when using environment-free traits, TE, as when using uncorrected trait values, T (Figure S8). Appendix II of the Supplementary Materials shows the weak environmental effects on individual traits (average environmental effect = 5%, Figure S5), on the multivariate phenotype (environmental effect = 5%, Figure S5) and on the static integration structure of each species (Table S5) and each trait pair (Figure S7). Figure S6 shows that species similarity and static trait correlations are slightly larger on

146

average with environmental effects but trait and species responses are largely independent. Thus, environmental effects are present but weak. Further, they do not affect the consistency of pairwise trait correlation, nor the relationships between species similarity and phylogeny, environmental niche, or natural history.

147

DISCUSSION

The static integration structure of functional traits is species-specific in northeast temperate tree saplings Across the twenty traits measured, only three (dimensionally independent) pairs are consistently correlated across species. On one hand, this suggests the existence of a minimal ‘tree template’ defining fundamental relationships among functional traits. On the other hand, the limited subset of shared trait correlations points to a set of alternative hypotheses (discussed below) for functional trait variation within species.

The shared trait correlations, LMA - δ13C (푟̅ = 0.61), KBRANCH-Lumen Fraction (푟̅ =

0.66) and KS – Lumen Fraction (푟̅ = 0.81) (Figure 4), span both leaf and xylem traits. The δ13C is typically considered to reflect water use efficiency of the individual, although other factors are known to also affect this property (e.g. Hausmann et al. 2005). The consistency of the association between LMA, a trait reflecting carbon economics, and

δ13C, a trait reflecting water use, agrees with the broad coordination of resource use along a fast/slow spectrum suggested by Reich (2014). However, all resource use traits (such as LNC and SRL) are not correlated with each other as this hypothesis would predict. The close association between these two traits might reflect the physiological association between water use and carbon gain mediated by common stomatal control of these functions. However, this is the first study that we know of to show a consistent relationships between LMA and δ13C within species. The other two pairs of traits relate xylem anatomy and conductive efficiency. Conductivity can be increased by increasing xylem Lumen Fraction or Vessel Diameter. Lumen Fraction is consistently correlated with both KS (푟̅ = 0.81) and KBRANCH (푟̅ = 0.63). Vessel Diameter is also consistently correlated with KS but less strongly on average (푟̅ = 0.57). Larger conduits are more sensitive to freeze- and drought-induced cavitation (Hacke et al. 2006, Pittermann and

Sperry 2006). Thus, the fact that KS is more strongly correlated with Lumen Fraction than with VD suggests that selection favors maintaining conductive safety as conductive efficiency increases. A faster increase in Lumen Fraction than in VD also implies an increase in the number of vessels, a trait that we did not measure directly. Increased

148

vessel redundancy should offer an additional protection against loss of function due to cavitation. Other than these three consistently correlated trait pairs, the data show very variable and dissimilar patterns of static integration across species for functional traits, both in terms of bivariate and overall phenotypic integration structures. Since the traits reflect a number of important ecophysiological properties of temperate angiosperm trees, the results indicate flexibility in the relationships among these vital plant properties within species. This is the first study to show a diversity of static correlations in a broad suite of functional traits across locally coexisting tree species. While this lability is surprising given the ecophysiological importance of these functional traits, such lability in static integration structure is not uncommon in the literature. Indeed, a number of studies have found distinct static integration patterns among closely related populations and species in plants for allometric and floral traits (Primack and Antonovics 1981, Mazer and Wheelwright 1993, Herrera et al. 2002, Murren et al. 2002), as well as in different animals such as mammals (Cheverud 1982b, Martin and Harvey 1985), birds (Merila and Bjorklund 1999, Badyaev and Hill 2000, Green et al. 2001) and insects (Klingenberg and Zimmermann 1992, Simmons and Tomkins 1996, Hosken et al. 2005, Higginson et al. 2015). In addition to being variable, we found that static integration of functional traits tends to be weak. The average static correlation across all pairs of traits and all species is r = 0.02, and 85% of all static pairwise correlations were not significant. Our results are in contrast to Conner’s (2014) report of an average r= ca. 0.53 for the static correlations in morphometric vegetative traits, but this is not surprising. The results of integration studies are strongly contingent on the choice of traits, and the nature of the set of traits in this study is distinct from most phenotypic integration studies. Typically, phenotypic integration studies examine relationships among functionally or developmentally linked traits. Morphometric and allometric evolutionary studies focusing on size-based traits make up a large fraction of the integration literature. In contrast, here the objective was to characterize diverse aspects of plant ecological strategy, such that traits were selected

149

to span variety of plant physiological function and organs. Further, we avoided including redundant traits to avoid relationships that were not biologically meaningful. This selection of diverse, non-redundant traits necessarily leads to weaker average correlations, which highlight the inherently malleable character of the overall phenotype.

Static integration structure is not related to phylogeny, stress tolerance, nor local environmental niche Although twelve of the twenty traits show some phylogenetic conservatism among species (Table S4), the static integration structure do not show a phylogenetic signal (Figure 2A). This suggests that phylogenetic history imposes no or only loose restrictions on the overall static integration structure for these traits. While this appears surprising, a number of studies find that multivariate correlation matrices do not reflect phylogenetic history (Armbruster et al. 1999, Badyaev and Hill 2000, Callahan and Waller 2000, Murren 2002). These results suggest that phenotypic integration structures are weakly conserved phylogenetically. Static integration structure is not linked to stress tolerance either, at least not in terms of shade-tolerance, drought-tolerance or waterlogging stress (Mantel tests, p = 0.17, p = 0.94 and p = 0.41, respectively). Nor do other ecological features appear to be associated with integration structure. For example, five of the six most similar species pairs (Populus deltoides – Prunus virginiana, Amelanchier laevis – Populus balsamifera, Betula populifolia – Acer saccharum, Betula papyrifera – Acer pensylvanicum, Betula populifolia – Prunus serotina, and Quercus rubra - Tilia americana) do not share a priori natural history similarities that we are aware of. Notice that two species, (Carya cordiformis and Ulmus americana) have unique static integration structures (rM = n.s. with 14/24 species). They are both ring-porous and shade-tolerant, but are very weakly similar to each other (rM = 0.16). While the important natural history aspects discussed above are not associated with static integration structure, the possibility remains that

150

static integration structure is related to other aspects of the niche (such as the reproductive niche) not captured in this study. Perhaps more surprisingly, integration structure is also not related to the environmental niche (Figure 2B-C). These results contradict the common idea that the phenotype is largely adaptive and that the environment plays an important role in shaping the phenotype. However, others have also found it difficult to link static integration patterns to either phylogeny or ecology (Armbruster et al. 1999, Callahan and Waller 2000, Murren et al. 2002). This lack of association can lead to a number of interpretations. First, our study site is located within a bioclimatic envelope, so that these results can suggest that environmental differences at the scales of microhabitats (alpha- scale) and habitats (beta-scales) do not affect static integration structure. Second, we may have failed to measure the important environmental variables affecting static integration. For example, biotic agents such as mutualists and natural enemies are also central to plant ecology and can play an important role in shaping the phenotype (e.g. Hetrick et al. 1991, Poorter et al. 2004). In this study, we characterized the most important abiotic gradients that are known to affect plant growth and performance for this site and group of species (see methods), as well as competition, an important element of biotic interactions. The variation in the study system covers a large fraction of the total local range for pH (3.6-

7.2), light (0.36-36.18 Moles m-2d-1) and soil water availability (0-100%), so that it is unlikely the studied environment suffers from a lack of habitat heterogeneity. Environmental variability is also evident in the range of environments from which each species was sampled (Figure S3) and in the range of traits expressed (Figure S2). Nonetheless, despite the extensive environmental characterization, we cannot exclude the possibility that important environmental variables were not measured. This would lead to the disheartening conclusion that the most commonly assessed and accessible environmental variables are poor indicators of the environmental variables affecting the phenotypes of saplings. Third, the high variability in trait associations within species and the lack of relationships between static integration structure and environmental niche may suggest that ecophysiological functions can be combined in a variety of ways to that

151

allow trees to grow and survive in their environment. This is consistent with the “alternative optima” hypothesis proposed by Marks & Lechowicz (2006) stating that a number of alternative plant designs can be successful in a given environment. Under this interpretation, the local environment does affect phenotypic integration but since there is not a single best trait combinations for a given set of environmental variables, the alternative optima are not apparent in a traditional regression approaches. Last, on the contrary, the results could indicate the relaxation or absence of selection on phenotypic integration in saplings at among individuals within a population. In other words, these results could suggest that the relationships among functional traits at that scale have little to no bearing on sapling performance. In fact, examining a global dataset, Paine and colleagues (2015) found that functional traits only predict 3% of sapling growth rates. Further investigation of the links between functional trait integration and plant performance is required to differentiate among the latter two important and contrasting interpretations.

Static and evolutionary integration structures are independent The results suggest that static integration of functional traits has a minimal effect on evolutionary integration. The overall static and evolutionary integration structures are largely independent, with the strongest Mantel correlation between static and evolutionary integration only 0.44 (Table 2). Evolutionary theory predicts that we should expect a triangular relationship between evolutionary and static integration where trait pairs consistently correlated within species are also correlated among species, but weakly or un-correlated traits within species can be correlated or not between species (Zeng 1988, Merila and Bjorklund 2004). The data partly agree with this pattern because evolutionary pairwise correlations are high for the mean static correlation that are high (Figure 3). However, this pattern is only evident when considering mean static correlations but not individual static correlations, which vary broadly among species for any given trait pair. The fact that only a handful of static correlations are consistently high indicates that for these functional traits, interspecific relationships are not driven by

152

intraspecific correlation patterns but instead arise from mechanisms acting at the species level (Figure 1 &Figure 3).

Figure 4 illustrates the large degree of independence between static and evolutionary pairwise correlations. Panel A shows that the three traits pairs consistently correlated within species that are independent (LMA - δ13C, KS-Lumen Fraction and

KBRANCH- Lumen Fraction) are also significantly correlated among species (Figure 4 , r = 0.66, 0.52 and 0.51). However, we cannot tell whether the presence of these correlation both within and among species is due to uniform stabilizing and directional selection within and among species, or to fundamental (genetic, developmental or biophysical) constraints being expressed at both scales (Westoby et al. 1995). In contrast, only two of the three consistently correlated trait pairs that are dimensionally related (LMA-Leaf Thickness and Ks-VD) are also correlated among species (r = 0.33 and 0.87 respectively).

A third dimensionally related trait pair, Ks-KBRANCH, is tightly correlated within and among species in diffuse-porous species (evolutionary correlation r = 0.97, p = 3e-13), but has been decoupled among ring-porous species as well as within three of the four species with this wood anatomy (Fraxinus americana, Ulmus americana and Carya cordiformis). Interestingly, note that another dimensionally related trait pair, MOE-Stem WD, is significantly correlated among species, but does not show consistent static correlations. This suggests that although stem wood density is an important component of the elasticity modulus, MOE is likely adjusted within species by changing other properties such as diameter and height instead of Stem WD. The absence of evolutionary integration for KS-KBranch and of static integration for MOE-Stem WD show that two dimensionally correlated traits may nonetheless not be empirically correlated within or among species.

Panel B of Figure 4 illustrates the decoupling of static and evolutionary correlations in traits expected to be correlated among species based on ecological theory on plant strategies. LNC, LMA and SRL should be correlated based on the hypothesis that plants coordinate their traits affecting resource economics (Ackerly and Reich 1999, Wright et al. 2004, Reich 2014). LMA and LNC are significantly correlated within half of the species and not among species in this dataset. In contrast SRL and LMA, are significantly

153

correlated among species but not within any species. Leaf Area, Branching Distance and Branch Diameter should also be correlated according to Corner’s rules (White 1983a, 1983b), which describes architectural relationships arising from constraints linked to resource acquisition, mechanical support and metabolic scaling relationships (Olson et al. 2009). Leaf Area and Branching Distance are positively correlated among species and within half of the species. Leaf Area and Branch Diameter are positively correlated among species but show non-significant, significantly positive and significantly negative static correlations. Last, Stem WD, MOE and KS are expected to be related via trade-offs between the support and transport functions of the xylem (Baas et al. 2004, Chave et al.

2009). KS and Stem WD are not correlated either within or among species. MOE and Stem WD are positively correlated among species and the two significant static correlations are negative. These examples show that the expected evolutionary relationships are largely absent within species and are thus likely the result of species-level processes such as natural selection instead of the result of constraints within species carrying over to larger scales. Different static and evolutionary integration structures are often suggested to indicate that natural selection is at play (Merila and Bjorklund 2004). If this is true, here we do not find evidence that the selection gradient causing the species to differ is related to environmental niche. The independence of these two scales suggest that for these functional traits the drivers of integration acting at each scale are distinct and that static integration impose minimal constraints on evolutionary integration. Trait correlations observed among species in this set of functional traits are thus likely to be the result of processes acting at the species level. Further, given that the majority of functional traits are not correlated among individuals within populations, including those expected from ecological theory, the strategy dimensions described across broad ecological scales do not seem to apply at the scale of individuals within a population. Last, the fact that we cannot infer intraspecific trait correlations from interspecific patterns and vice versa has implications for the use of plant traits in community and ecosystem ecology. Global vegetation models rely on trait relationships often need to

154

make the assumption that intra- and inter-specific patterns of trait correlation are similar (e.g. Kempes et al. 2011). It is important to verify the validity of these assumptions and the strength of trait spectra because they can strongly affect the output of the model (Osnas et al. 2013). Moreover, some gap filling methods (e.g. Schrodt et al. 2015) predict missing values of individual populations from known trait values based on the assumption that the intra-population trait relationships follow documented interspecific patterns. This study suggests that it is unsafe to assume that species-scale trait relationships also occur among individuals within species.

CONCLUSION

In conclusion, we find that among northeastern North American deciduous temperate tree saplings, (i) bivariate and multivariate static trait correlations are weak or absent and highly variable among species; (ii) species display singular multivariate integration structures; (iii) different plant designs are not related to phylogeny, stress tolerance nor to the realized alpha and beta environmental niche and (iv) evolutionary integration structures are not predicted or limited by static integration structures. These results indicate that the drivers shaping the static and evolutionary integration of functional traits are different. Moreover, the variability of static integration structures and their lack of correspondence with the species’ ecology may point to two contrasting conclusions. On one hand, these results could imply a relaxation of selection on functional traits in static integration because integration at this scale does not affect plant performance. On the other hand, these results could also suggest that there are multiple successful ways to combine functional traits within a given environment when multiple tasks are simultaneously optimized (Niklas and Simpson 1994, Marks and Lechowicz 2006, Shoval et al. 2012). These two opposite possibilities deserve further exploration as they each have important implications: static trait relationships being mere noise could imply that functional traits only reflect plant strategies at the level of the population and higher; multiple optimal plant strategies in a given environment would make us rethink

155

much ecological theory because it is largely based on the assumption of a single optimal phenotype for a given environment (Enquist et al. 2015, Laughlin and Messier 2015).

156

REFERENCES

Ackerly, D. D., and P. B. Reich. 1999. Convergence and correlations among leaf size and function in seed plants: A comparative test using independent contrasts. American Journal of Botany 86:1272–1281.

American Public Health Association. 1999. Standard Methods for the Examination of Water and Wastewater. Port City Press.

Arii, K., B. Hamel, and M. J. Lechowicz. 2005. Environmental correlates of canopy composition at Mont St. Hilaire, Quebec, Canada. The Journal of the Torrey Botanical Society 132:90–102.

Arii, K., and M. J. Lechowicz. 2002. The influence of overstory trees and abiotic factors on the sapling community in an old-growth Fagus-Acer forest. Ecoscience 9:386– 396.

Armbruster, S. W. 1991. Multilevel analysis of morphometric data from natural plant populations: insights into ontogenetic, genetic, and selective correlations in m

Armbruster, S. W., V. S. DiStilio, J. D. Tuxill, T. C. Flores, and J. L. Velásquez Runk. 1999. Covariance and decoupling of floral and vegetative traits in nine neotropical plants: A re-evaluation of Berg’s correlation-pleiades concept. American Journal of Botany 86:39–55.

Armbruster, S. W., C. Pélabon, T. F. Hanson, and C. P. H. Mulder. 2004. Floral Integration, Modularity, and Accuracy. Pages 23–49 in M. Pigliucci and K. Preston, editors. Phenotypic Integration: Studying the Ecology and Evolution of Comlpex Phenotypes. Oxford University Press, Oxford.

Baas, P., F. W. Ewers, S. D. Davis, and E. A. Wheeler. 2004. Evolution of xylem physiology. Pages 273–295 in A. R. Hemsley and I. Poole, editors. The evolution of plant physiology. Elsevier Academic Press, London.

Badyaev, a V, and G. E. Hill. 2000. The evolution of sexual dimorphism in the house finch. I. Population divergence in morphological covariance structure. Evolution; international journal of organic evolution 54:1784–1794.

Bell, C. D., D. E. Soltis, and P. S. Soltis. 2010. The age and diversification of the angiosperms re-revisited. American Journal of Botany 97:1296–1303.

Berg, R. L. 1960. The ecological significance of correlation pleiades. Evolution 14:171– 180.

157

Blonder, B., C. Lamanna, C. Violle, and B. J. Enquist. 2014. The n-dimensional hypervolume. Global Ecology and Biogeography 23:595–609.

Box, G., and D. Cox. 1964. An analysis of transformations. Journal of the Royal Statisistical Society, Series B 26:211–46.

Bremner, J. M. 1960. Determination of nitrogen in soil by the Kjeldahl method. The Journal of Agricultural Science 55:11.

Callahan, H. S. ., and D. M. Waller. 2000. Phenotypic Integration and the Plasticity of Integration in an Amphicarpic Annual. International Journal of Plant Sciences 161:89–98.

Chave, J., D. Coomes, S. L. Lewis, N. G. Swenson, and A. E. Zanne. 2009. Towards a worldwide wood economic spectrum. Ecology Letters 12:351–366.

Cheverud, J. M. 1982a. Phenotypic, genetic and environmental morphological integration in the cranium. Evolution 36:499–516.

Cheverud, J. M. 1982b. Relationships among ontogenetic, static, and evolutionary allometry. American journal of physical anthropology 59:139–149.

Conner, J. K., I. a Cooper, R. J. La Rosa, S. G. Pérez, A. M. Royer, P. T. R. S. B, and S. G. Pe. 2014. Patterns of phenotypic correlations among morphological traits across plants and animals Patterns of phenotypic correlations among morphological traits across plants and animals.

Cottam, G., and J. T. Curtis. 1956. The Use of Distance Measures in Phytosociological Sampling. Ecology 37:451–460.

Davies, B. E. 1974. Loss-on-ignition as an estimate of soil organic matter. Soil Science Society of America Journal 38:150–151.

Davis, M. B., and R. G. Shaw. 2001. Range shifts and adaptive responses to Quaternary climate change. Science (New York, N.Y.) 292:673–679.

Díaz, S., J. G. Hodgson, K. Thompson, M. Cabido, J. H. C. Cornelissen, A. Jalili, G. Montserrat-Martí, J. P. Grime, F. Zarrinkamar, Y. Asri, S. R. Band, S. Basconcelo, P. Castro-Díez, G. Funes, B. Hamzehee, M. Khoshnevi, N. Pérez-Harguindeguy, M. C. Pérez-Rontomé, A. Shirvany, F. Vendramini, S. Yazdani, R. Abbas-Azimi, A. Bogaard, S. Boustani, M. Charles, M. Dehghan, L. de Torres-Espuny, V. Falczuk, J. Guerrero-Campo, A. Hynd, G. Jones, E. Kowsary, F. Kazemi-Saeed, M. Maestro- Martínez, A. Romo-Díez, S. Shaw, B. Siavash, P. Villar-Salvador, and M. R. Zak.

158

2004. The plant traits that drive ecosystems: Evidence from three continents. Journal of Vegetation Science 15:295.

Elliott, T. L., and J. T. Davies. 2014. Challenges to barcoding an entire flora. Molecular Ecology Resources 14:883–891.

Enquist, B. J., J. Norberg, S. P. Bonser, C. Violle, C. T. Webb, A. Henderson, L. L. Sloat, and V. M. Savage. 2015. Scaling from traits to ecosystems: Developing a general Trait Driver Theory via integrating trait-based and metabolic scaling theories. Advances in Ecological Research 52.

Enright, N. J., and J. E. Lewis. 1985. Aspect-related vegetation patterns and climate at Mont St. Hilaire, Quebec. The Canadian Geographer 29:249–256.

Fajardo, A., and F. I. Piper. 2011. Intraspecific trait variation and covariation in a widespread tree species (Nothofagus pumilio) in southern Chile. New Phytologist 189:259–271.

Farquhar, G. D., J. R. Ehleringer, and K. T. Hubick. 1989. Carbon discrimination and photosynthesis. Annual review of Plant Physiology and Plant Molecular Biology 40:503–537.

Feininger, T., and A. K. Goodacre. 1995. The eight classical Monteregian hills at depth and the mechanism of intrusion. Canadian Journal of Earth Sciences 8:440–446.

Frazer, G. W., C. D. Canham, and K. P. Lertzman. 1999. Gap Light Analyzer (GLA).

Garnier, E., and M.-L. Navas. 2012. A trait-based approach to comparative functional plant ecology: concepts, methods and applications for agroecology. A review. Agronomy for sustainable development 32:365–399.

Gilbert, B., and M. J. Lechowicz. 2004. Neutrality, niches, and dispersal in a temperate forest understory. Proceedings of the National Academy of Sciences of the United States of America 101:7651–7656.

Green, A. J., J. Figuerola, and R. King. 2001. Comparing interspecific and intraspecific allometry in the Anatidae. Journal für Ornithologie 142:321–334.

Grubb, P. J. 1977. The maintenance of species-richness in plant communities: the importance of the regeneration niche. Biological reviews of the Cambridge Philosophical Society.

Hacke, U. G., J. S. Sperry, J. K. Wheeler, and L. Castro. 2006. Scaling of angiosperm xylem structure with safety and efficiency. Tree physiology 26:689–701.

159

Hausmann, N. J., T. E. Juenger, S. Sen, K. a Stowe, T. E. Dawson, and E. L. Simms. 2005. Quantitative trait loci affecting delta13C and response to differential water availibility in Arabidopsis thaliana. Evolution; international journal of organic evolution 59:81–96.

Herrera, C. M., X. Cerda, M. B. Garcia, J. Guitian, M. Medrano, P. J. Rey, and a M. Sanchez-Lafuente. 2002. Floral integration, phenotipic covariance structure and pollination variation inbumblebee-pollinated Helleborus foetidus. Journal of Evolutionary Biology 15 :108–121.

Hetrick, B. a. D., G. W. T. Wilson, and J. F. Leslie. 1991. Root architecture of warm- and cool-season grasses: relationship to mycorrhizal dependence. Canadian Journal of Botany 69:112–118.

Higginson, D. M., A. V. Badyaev, K. a. Segraves, and S. Pitnick. 2015. Causes of Discordance between Allometries at and above Species Level: An Example with Aquatic Beetles. The American Naturalist 186:176–186.

Hosken, D. J., a. M. Minder, and P. I. Ward. 2005. Male genital allometry in Scathophagidae (Diptera). Evolutionary Ecology 19:501–515.

Jackson, D. A. 1993. Stopping rules in principal components analysis: A comparison of heuristical and statistical approaches. Ecology 74:2204–2214.

Karst, J., B. Gilbert, and M. J. Lechowicz. 2005. Fern community assembly: The roles of chance and the environment at local and intermediate scales. Ecology 86:2473–2486.

Kempes, C. P., G. B. West, K. Crowell, and M. Girvan. 2011. Predicting maximum tree heights and other traits from allometric scaling and resource limitations. PLoS ONE 6.

Kerkhoff, A. J., and B. J. Enquist. 2009. Multiplicative by nature: Why logarithmic transformation is necessary in allometry. Journal of Theoretical Biology 257:519–521.

Klingenberg, C. P. 2014. Studying morphological integration and modularity at multiple levels: concepts and analysis. Philosophical transactions of the Royal Society of London. Series B, Biological sciences 369:2010249.

Klingenberg, C. P., and M. Zimmermann. 1992. Static, ontogenetic, and evolutionary allometry: A multivariate comparison in nine species of water striders. American Naturalist 140:601–620.

Laughlin, D. C., and J. Messier. 2015. Fitness of multidimensional phenotypes in dynamic

160

adaptive landscapes. Trends in Ecology & Evolution:1–10.

Legendre, P., and L. Legendre. 1998. Numerical Ecology. 2nd Englis. Elsevier, Amsterdam, The Netherlands.

Marks, C. O., and M. J. Lechowicz. 2006. Alternative designs and the evolution of functional diversity. The American naturalist 167:55–66.

Martin, R. D., and P. H. Harvey. 1985. Brain Size Allometry: ontogeny and phylogeny. Pages 147–173 Size and scaling in primate biology. Springer, New York.

Mason, C. M., S. E. McGaughey, and L. a Donovan. 2013. Ontogeny strongly and differentially alters leaf economic and other key traits in three diverse Helianthus species. Journal of Experimental Botany 64:4089–4099.

Maycock, P. F. 1961. Botanical studies on Mont St.Hilaire, Rouville county, Quebec. Canadian Journal of Botany 39:1293–1325.

Mazer, S. J., and N. T. Wheelwright. 1993. Fruit size and shape: Allometry at different taxonomic levels in bird-dispersed plants. Evolutionary Ecology 7:556–575.

Merila, J., and M. Bjorklund. 1999. Population divergence and morphometric integration in the green-finch (Carduelis chloris) - evolution against the trajectory of least resistance ? Journal of evolutionary biology 12:103–112.

Merila, J., and M. Bjorklund. 2004. Phenotypic Integration as a Constraint and Adaptation. Pages 107–129 in M. Pigliucci and K. Preston, editors. Phenotypic Integration: Studyinf the Ecology and Evolution of Comlpex Phenotypes. Oxford University Press.

Murren, C. J. 2002. Phenotypic integration in plants. Plant Species Biology 17:89–99.

Murren, C. J., N. Pendleton, and M. Pigliucci. 2002. Evolution of phenotypic integration in Brassica (Brassicaceae). American Journal of Botany 89:655–663.

Niinemets, Ü., and F. Valladares. 2006. Tolerance to shade, drought, and waterlogging of temperate northern hemisphere trees and shrubs. Ecological Monographs 76:521– 547.

Niklas, K. J., and G. G. Simpson. 1994. Morphological evolution through complex 91:6772–6779.

Olson, M. E., R. Aguirre-Hernández, and J. a Rosell. 2009. Universal foliage-stem scaling across environments and species in dicot trees: plasticity, biomechanics and Corner’s Rules. Ecology letters 12:210–9.

161

Osnas, J. L. D., J. W. Lichstein, P. B. Reich, and S. W. Pacala. 2013. Global leaf trait relationships: mass, area, and the leaf economics spectrum. Science (New York, N.Y.) 340:741–4.

Paine, C. E. T., L. Amissah, H. Auge, C. Baraloto, M. Baruffol, N. Bourland, H. Bruelheide, K. Daïnou, R. C. de Gouvenain, J.-L. Doucet, S. Doust, P. V. a. Fine, C. Fortunel, J. Haase, K. D. Holl, H. Jactel, X. Li, K. Kitajima, J. Koricheva, C. Martínez-Garza, C. Messier, A. Paquette, C. Philipson, D. Piotto, L. Poorter, J. M. Posada, C. Potvin, K. Rainio, S. E. Russo, M. Ruiz-Jaen, M. Scherer-Lorenzen, C. O. Webb, S. J. Wright, R. a. Zahawi, and A. Hector. 2015. Globally, functional traits are weak predictors of juvenile tree growth, and we do not know why. Journal of Ecology 103:978–989.

Palow, D. T., K. Nolting, and K. Kitajima. 2012. Functional trait divergence of juveniles and adults of nine Inga species with contrasting soil preference in a tropical rain forest. Functional Ecology 26:1144–1152.

Pittermann, J., and J. S. Sperry. 2006. Analysis of freeze-thaw embolism in conifers. The interaction between cavitation pressure and tracheid size. Plant physiology 140:374– 382.

Poorter, L. 2007. Are species adapted to their regeneration niche, adult niche, or both? The American naturalist 169:433–42.

Poorter, L., M. van de Plassche, S. Willems, and R. G. a Boot. 2004. Leaf traits and herbivory rates of tropical tree species differing in successional status. Plant biology (Stuttgart, Germany) 6:746–54.

Primack, R. B., and J. Antonovics. 1981. Experimental ecological genetics in Plantago. V. Components of seed yield in the ribwort plantain Plantago lanceolata L. Evolution 35:1069–1079.

R Development Core Team, R. 2011. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing.

Radovski, M. 2010. Testing a trait-based model of fern community assembly. McGill, Montreal.

Reich, P. B. 2014. The world-wide “fast-slow” plant economics spectrum: a traits manifesto. Journal of Ecology 102:275–301.

Reich, P. B., D. S. Ellsworth, M. B. Walters, J. M. Vose, C. Gresham, J. C. Volin, and W. D. Bowman. 1999. Generality of leaf trait relationships: A test across six biomes. Ecology 80:1955–1969.

162

Richardson, S. J., R. B. Allen, R. P. Buxton, T. a. Easdale, J. M. Hurst, C. W. Morse, R. D. Smissen, and D. a. Peltzer. 2013. Intraspecific Relationships among Wood Density, Leaf Structural Traits and Environment in Four Co-Occurring Species of Nothofagus in New Zealand. PLoS ONE 8:24–26.

Schlichting, C. D. 1989. Phenotypic plasticity in Phlox - II. Plasticity of character correlations. Oecologia 78:496–501.

Schrodt, F., J. Kattge, H. Shan, F. Fazayeli, J. Joswig, A. Banerjee, M. Reichstein, G. Bönisch, S. Díaz, J. Dickie, A. Gillison, A. Karpatne, S. Lavorel, P. Leadley, C. B. Wirth, I. J. Wright, S. J. Wright, and P. B. Reich. 2015. BHPMF - a hierarchical Bayesian approach to gap-filling and trait prediction for macroecology and functional biogeography. Global Ecology and Biogeography:n/a–n/a.

Shoval, O., H. Sheftel, G. Shinar, Y. Hart, O. Ramote, a. Mayo, E. Dekel, K. Kavanagh, and U. Alon. 2012. Evolutionary Trade-Offs, Pareto Optimality, and the Geometry of Phenotype Space. Science 336:1157–1160.

Simmons, L. W., and J. L. Tomkins. 1996. Sexual selection and the allometry of earwig forceps. Evolutionary Ecology 10:97–104.

Sperry, J. S., U. G. Hacke, and J. Pittermann. 2006. Size and function in conifer tracheids and angiosperm vessels. American Journal of Botany 93:1490–1500.

Sterck, F. J., L. Markesteijn, M. Toledo, F. Schieving, and L. Poorter. 2014. Sapling performance along resource gradients drives tree species distributions within and across tropical forests. Ecology:140311112851001.

Tran, T. Sen, M. Giroux, J. Guilbeault, and P. Audesse. 1990. Evaluation of Mehlich‐III extractant to estimate the available P in Quebec soils. Communications in Soil Science and Plant Analysis 21:1–28.

Violle, C., M.-L. Navas, D. Vile, E. Kazakou, C. Fortunel, I. Hummel, and E. Garnier. 2007. Let the concept of trait be functional! Oikos 116:882–892.

Violle, C., P. B. Reich, S. W. Pacala, B. J. Enquist, and J. Kattge. 2014. The emergence and promise of functional biogeography. Proceedings of the National Academy of Sciences 11:13690–13696.

Waitt, D. E., and D. A. Levin. 1993. Phenotypic Integration and Plastic Correlations in Phlox drummondii (Polemoniaceae). American Journal of Botany 80:1224–1233.

Webb, C. O., D. D. Ackerly, and S. W. Kembel. 2008. Phylocom: Software for the analysis

163

of phylogenetic community structure and trait evolution. Bioinformatics 24:2098– 2100.

Webb, C. O., and M. J. Donoghue. 2005. Phylomatic: Tree assembly for applied phylogenetics. Molecular Ecology Notes 5:181–183.

Westoby, M., M. R. Leishman, and J. M. Lord. 1995. On Misinterpreting the “Phylogentic Correction.” Journal of Ecology 83:531–534.

White, P. S. 1983a. Evidence that temperate east North-American evergreen woody- plants follow Corner’s rules. New Phytologist 95:139–145.

White, P. S. 1983b. Corner’s rules in eastern deciduous trees : Allometry and its implications for the adaptive architecture of trees. Bulletin of the Torrey Botanical Club 110:203–212.

Wikström, N., V. Savolainen, and M. W. Chase. 2001. Evolution of the angiosperms: calibrating the family tree. Proceedings. Biological sciences / The Royal Society 268:2211–2220.

Wright, I. J., P. B. Reich, M. Westoby, D. D. Ackerly, Z. Baruch, F. Bongers, J. Cavender- Bares, T. Chapin, J. H. C. Cornelissen, M. Diemer, J. Flexas, E. Garnier, P. K. Groom, J. Gulias, K. Hikosaka, B. B. Lamont, T. Lee, W. Lee, C. Lusk, J. J. Midgley, M.-L. Navas, U. Niinemets, J. Oleksyn, N. Osada, H. Poorter, P. Poot, L. Prior, V. I. Pyankov, C. Roumet, S. C. Thomas, M. G. Tjoelker, E. J. Veneklaas, and R. Villar. 2004. The worldwide leaf economics spectrum. Nature 428:821–827.

Zeng, Z.-B. 1988. Long-Term correlated response, interpopulation covariation, and interspecific allometry. Evolution.pdf. Evolution 42:363–374.

Zuur, A. F., E. N. Ieno, and G. M. Smith. 2007. Analysing Ecological Data. Springer.

164

ACKNOWLEDGEMENTS

We would like to thank Andréanne Ferland, Émilie Lavoie, Natasha Salter, Carol Mordy, Anke Roth, Sierra Kaszubinski, Sanga Shir, Surbhi Patel, Margretta Murphy, John Lacson, Kevin Wong, Sarah Schwenck, Anjeanette McKay, Casey Knoks, Meghan Iacueuilli, Shahrzad Badie, Irene Liang, Jordyn Celaya and David Maneli for their instrumental role in data collection, Erica Bigio for invaluable assistance with the tree ring work, John Sperry, Melvin Tyree and David Killick for guidance and assistance with the tedious conductance measurements and Louise Comas for guidance on root trait field sampling methods.

165

TABLES

Table 1. Twenty study traits

Growth & Survival Function

Organ Functional Trait Units n

Resource Resource Acquisition Mechanical Support Transport Structural Defense Stem Wood Density 3,4,9 g/cm3 ✓ ✓ 380 Stem Modulus of Elasticity (MOE) N/mm2 ✓ 380 Branching Distance 7,9 cm ✓ 366 Branch Wood Density 3,4,9 g/cm3 ✓ ✓ 380 Branch Diameter 2 mm ✓ 380 Stem Vessel Diameter 3,5 μm ✓ 372 Lumen Fraction 5 cm2/cm2 ✓ ✓ 372 Conductivity per Sapwood Area (KS) 3 9 ml/s*cm2 ✓ 372 Conductivity per Branch (KBRANCH) 3,9 μm/μm ✓ 372 Vessel Thickness-to-Span Ratio (t/d2) 3,4 μm/μm ✓ 373 Leaf Area 2,9 cm2 ✓ ✓ 380 Leaf Mass per Area 1,6,9 g/cm2 ✓ ✓ 380 Leaf Thickness mm ✓ ✓ 380 Leaf Leaf Carbon Content 12 g/g ✓ ✓ 380 Leaf Nitrogen Content 1,9 g/g ✓ 380 Leaf Phosphorus Content 1,9 g/g ✓ 380 δ13C 9,13 δ13C‰ ✓ 380 Root Wood Density 9,11 g/cm3 ✓ ✓ 191 Root Specific Root Length 9,11 mg/mm ✓ ✓ 146

166

Table 2.Overall similarity, given as Mantel correlations (rM), in intra-population integration structure between species pairs and between each species and the interspecific pattern. Raw trait data are used [T]. The boxes show the correlations for groups the species of the same family. The darkness of each box is proportional to the strength of correlation. The species full name for each abbreviation are given in Table S1.

ACP ACR ACSA ACSP BEA BEPA BEPO OSV CAC COA FAG FRA POB POD POG POT AMSP PRP PRS PRV SOA QUR TIA ULA inter ACP 0.24 0.39 0.21 0.27 0.49 0.44 0.30 n.s. 0.26 0.19 n.s. 0.25 0.35 0.38 0.24 0.32 0.20 0.31 0.21 0.24 0.28 0.26 n.s. 0.35 ACR 0.33 0.23 0.13 0.31 0.28 n.s. n.s. 0.31 0.27 0.17 0.16 0.32 0.17 0.28 0.15 n.s. 0.26 0.24 n.s. 0.23 0.16 0.25 0.34 ACSA 0.35 0.23 0.40 0.50 0.38 0.15 0.38 0.15 0.25 0.32 0.27 0.34 0.33 0.36 0.24 0.24 0.20 0.17 0.40 0.30 n.s. 0.44 ACSP Aceraceae 0.14 0.37 0.22 n.s. n.s. 0.31 0.39 n.s. 0.36 0.45 0.17 0.27 0.33 0.38 0.17 0.39 0.20 0.23 0.20 0.24 n.s. BEA 0.21 n.s. 0.36 n.s. 0.27 n.s. n.s. 0.20 0.16 0.16 0.21 n.s. n.s. 0.19 n.s. n.s. 0.15 0.16 n.s. 0.31 BEPA 0.41 0.31 n.s. 0.22 0.38 0.15 0.23 0.38 0.21 0.34 0.16 0.28 0.32 0.19 0.29 0.25 0.32 n.s. 0.29 BEPO 0.37 0.15 0.27 0.31 0.36 0.32 0.39 0.39 0.35 0.29 0.31 0.49 0.32 n.s. 0.31 0.22 n.s. 0.34 OSV Betulaceae n.s. 0.24 0.21 0.20 0.36 0.26 0.29 0.19 0.27 0.25 0.32 n.s. 0.38 0.43 0.27 n.s. 0.24 CAC n.s. n.s. 0.31 0.24 n.s. n.s. 0.14 0.24 n.s. 0.20 n.s. n.s. 0.23 n.s. 0.16 0.25 COA 0.28 0.25 0.23 0.15 0.36 0.28 0.17 0.23 0.30 0.20 n.s. 0.19 0.27 n.s. 0.28 FAG 0.17 n.s. 0.27 0.18 0.26 n.s. 0.15 0.23 0.21 0.29 0.26 0.21 n.s. n.s. FRA 0.17 n.s. 0.16 0.19 n.s. n.s. 0.23 0.24 n.s. 0.29 0.24 n.s. 0.24 POB 0.39 0.41 0.43 0.51 0.45 0.43 0.30 0.21 0.39 0.22 0.14 0.23 POD 0.22 0.29 0.40 0.31 0.34 0.54 0.28 0.38 n.s. 0.34 0.12 POG 0.25 0.34 0.22 0.26 n.s. n.s. 0.21 n.s. n.s. 0.19 POT Salicaceae 0.30 0.32 0.24 0.27 n.s. 0.34 0.39 n.s. 0.26 AMSP 0.41 0.27 0.26 0.27 0.37 0.19 0.26 0.24 PRP 0.29 0.27 0.23 n.s. n.s. n.s. 0.16 PRS 0.24 0.22 0.26 0.22 0.17 0.28 PRV 0.23 0.21 0.17 n.s. n.s. SOA Rosaceae 0.32 0.37 0.16 0.15 QUR 0.48 0.22 0.30 TIA n.s. 0.26 ULA n.s. inter

167

FIGURES

Figure 1. Density distributions of static correlation coefficients for trait pairs across species. Each distribution represents the Pearson’s correlation coefficients of the 24 study species for a given pair of traits. Within species, on average pairwise trait correlations are not significant below r = 0.5. Trait pairs with mean correlation coefficient of 0.5 or larger and -0.5 or smaller are shown in red. (d) indicates dimensionally related trait pairs. Because the correlations are calculated on standardized data, the correlation coefficients r and slopes are identical. The mean values and standard deviation of the distributions of all pairwise trait correlations are given in Table S3.

168

Figure 2. Relationships between static integration structure similarity of pairs of species and (A) their phylogenetic distance, (B) the distance among the centroids of their environmental niche, and (C) the overlap of their environmental niche hypervolume, measured with Sorensen’s index. Species similarity in static integration structure is not related to their phylogenetic distance, nor to the similarity of their environmental niche. The dotted line in the cartoon of Panel A shows the distance between a pair of species. The dotted line in the cartoon of Panel B shows the distance between the centroids of the two species. The black area in the cartoon of Panel C shows the overlap among the hypervolumes of the two species.

169

Figure 3. Evolutionary correlations (Pearson’s r) between trait pairs as a function of static correlations. Each point is the Pearson correlation for a trait pair. Black points represents intraspecific correlations calculated for each species, so that each evolutionary pairwise correlation is associated with twenty four static pairwise correlations. Grey points represent the mean of the static pairwise correlations. Static correlations between -0.5 and 0.5, and evolutionary correlations between -0.40 and 0.40 are not significant. The relationship between static and evolutionary pairwise correlations is weak, whether using individual (SMA regression: r = 0.28, p = 2e-16), or mean static correlations (SMA regression, r=0.47, p = 6 e-09).

170

Figure 4. Bivariate trait correlations within species (thin lines) and among species (thick line). Solid lines show significant correlations and dashed lines show non-significant correlations. Panel A: consistent static pairwise correlations. Panel B: Trait pairs expected to be related based on ecological theory. The first two regressions are among traits that should be related based on resource acquisition strategy (Reich 2014), the next two regressions relate traits that should be related given architectural considerations (White 1983b) and the last two regressions are among traits that should be related based on trade-offs among wood functions (Baas et al. 2004).

171

SUPPLEMENTARY MATERIAL – APPENDIX I. SUPPLEMENTARY TABLES AND FIGURES

Table S1. Twenty-four study species

Xylem Code Species Family Orders Common name Leaf type Porosity ACP Acer pensylvanicum Moosewood Simple Diffuse ACR Acer rubrum Sapindaceae Sapindales Red Maple Simple Diffuse ACSA Acer saccharum Sapindaceae Sapindales Sugar Maple Simple Diffuse ACSP Acer spicatum Sapindaceae Sapindales Mountain Maple Simple Diffuse AMSP Amelanchier laevis Rosaceae Serviceberry Simple Diffuse BEA Betula alleghaniensis Betulaceae Fagales Yellow birch Simple Diffuse BEPA Betula papyrifera Betulaceae Fagales White birch Simple Diffuse BEPO Betula populifolia Betulaceae Fagales Grey birch Simple Diffuse CAC Carya cordiformis Juglandaceae Fagales Bitternut hickory Compound Semi or Ring COA Cornus alternifolia Cornaceae Cornales Alt. leaf dogwood Simple Diffuse FAG Fagus grandifolia Fagaceae Fagales American beech Simple Diffuse FRA Fraxinus americana Oleaceae Lamiales American ash Compound Ring OSV Ostrya virginiana Betulaceae Fagales Ironwood Simple Diffuse POB Populus balsamifera Salicaceae Balsam Poplar Simple Diffuse POD Populus deltoides Salicaceae Malpighiales Deltoid Poplar Simple Diffuse POG Populus grandidentata Salicaceae Malpighiales Large-tooth Aspen Simple Diffuse POT Populus tremuloides Salicaceae Malpighiales Trembling Aspen Simple Diffuse PRP Prunus pensylvanica Rosaceae Rosales Fire cherry Simple Diffuse PRS Prunus serotina Rosaceae Rosales Black cherry Simple Diffuse PRV Prunus virginiana Rosaceae Rosales Choke cherry Simple Diffuse QUR Quercus rubra Fagaceae Fagales Red Oak Simple Ring SOA Sorbus americana Rosaceae Rosales Mountain-Ash Compound Diffuse TIA Tilia americana Malvaceae Malvales Linden, Basswood Simple Diffuse ULA Ulmus americana Ulmaceae Rosales American Elm Simple Ring

172

Table S2. Nineteen environmental variables measured on each individual sapling, except for temperatures which were measured for each of the 23 general areas where the saplings were located. Environmental Variables Abiotic Aboveground Elevation North and East Aspect Slope % Canopy Opening Total transmitted light Summer Max Temperature Fall Min Temperature Average June-Dec Temp Belowground Soil Humidity pH Organic C N total P K Ca Mg Soil Depth Biotic Above & Asymmetric Belowground Competition Symmetric Competition

173

Table S3. Variability among species in static pairwise trait correlation. Mean Pearson’s correlation coefficients are shown in the upper triangle and standard deviation of Pearson’s correlation coefficients are shown in the lower diagonal. Darker hues indicate stronger mean in the upper right triangle and lower standard deviation in the lower left triangle. On average, static integration structures larger than 0.5 are significant. The pairwise trait correlations that are significant among species are enclosed in a black box.

2

C

S BRANCH

13

VD (t/d) K K Fraction Lumen Twig WD WD Stem Root WD MOE LCC LNC LPC δ LMA Leaf Thickness SRL Diameter Branch Leaf Area Angle Branching Distance Branching VD -0.17 0.57 0.47 0.23 -0.17 -0.02 0.02 0.10 0.04 0.03 0.04 -0.14 -0.04 0.00 0.01 -0.07 0.09 -0.16 0.03 (t/d)2 0.27 -0.27 -0.30 -0.19 0.14 0.05 -0.02 -0.01 -0.11 0.04 -0.04 0.02 -0.02 -0.10 -0.08 0.00 0.02 0.05 -0.03

KS 0.26 0.23 0.83 0.81 -0.19 -0.04 -0.02 0.07 0.01 -0.07 0.06 -0.09 -0.02 0.04 -0.07 0.01 0.02 -0.04 -0.03

KBRANCH 0.30 0.19 0.31 0.63 -0.09 -0.06 -0.05 0.03 0.03 -0.09 0.06 0.01 0.12 0.17 -0.07 0.02 0.00 -0.09 0.01 Lumen Fraction 0.34 0.27 0.10 0.43 -0.14 -0.06 -0.05 -0.05 -0.02 -0.08 0.03 -0.05 -0.01 0.02 -0.18 0.05 -0.06 0.03 -0.06 Branch WD 0.29 0.28 0.27 0.28 0.27 0.36 0.06 -0.06 -0.10 -0.30 -0.28 0.09 0.13 -0.04 0.00 0.10 -0.25 0.04 -0.23 Stem WD 0.25 0.28 0.27 0.31 0.25 0.29 0.19 -0.04 -0.10 -0.24 -0.09 0.09 0.07 -0.04 -0.14 0.02 -0.05 -0.01 -0.05 Root WD 0.35 0.33 0.40 0.45 0.41 0.31 0.35 0.07 0.02 -0.05 -0.06 0.03 0.09 0.06 -0.13 0.06 0.08 -0.05 0.12 MOE 0.23 0.30 0.26 0.29 0.29 0.31 0.28 0.43 -0.02 0.12 0.06 -0.13 -0.14 -0.14 -0.13 -0.15 0.11 0.09 0.12 LCC 0.28 0.29 0.28 0.26 0.31 0.23 0.23 0.32 0.26 0.25 0.07 0.12 0.11 0.00 0.17 0.09 0.01 -0.07 0.06 LNC 0.23 0.30 0.23 0.26 0.26 0.23 0.28 0.30 0.24 0.26 0.21 -0.25 -0.35 -0.23 0.00 -0.14 0.12 0.06 0.17 LPC 0.27 0.24 0.26 0.27 0.28 0.26 0.31 0.41 0.20 0.27 0.29 0.06 -0.08 0.03 -0.14 -0.05 0.11 -0.11 0.17 δ13C 0.22 0.30 0.24 0.29 0.28 0.29 0.24 0.39 0.31 0.27 0.25 0.28 0.61 0.42 0.02 0.20 -0.02 -0.05 0.02 LMA 0.23 0.30 0.24 0.26 0.28 0.29 0.30 0.26 0.32 0.31 0.28 0.29 0.20 0.70 -0.02 0.24 -0.04 -0.04 0.06 Leaf Thickness 0.27 0.28 0.25 0.26 0.26 0.22 0.30 0.38 0.28 0.28 0.27 0.24 0.28 0.21 0.00 0.14 0.06 -0.05 0.15 SRL 0.45 0.45 0.43 0.46 0.45 0.41 0.53 0.52 0.47 0.50 0.43 0.55 0.47 0.46 0.51 0.03 -0.10 0.22 -0.05 Branch Diameter 0.32 0.26 0.23 0.28 0.22 0.29 0.27 0.42 0.27 0.20 0.30 0.29 0.25 0.30 0.27 0.42 -0.03 -0.09 -0.18 Leaf Area 0.24 0.23 0.24 0.28 0.20 0.22 0.28 0.37 0.28 0.25 0.24 0.24 0.34 0.28 0.33 0.42 0.36 0.11 0.41 Branching Angle 0.22 0.36 0.25 0.24 0.30 0.36 0.37 0.45 0.42 0.35 0.36 0.34 0.35 0.37 0.31 0.58 0.32 0.32 -0.09 Branching Distance 0.30 0.25 0.24 0.26 0.31 0.28 0.28 0.39 0.26 0.22 0.23 0.26 0.37 0.33 0.25 0.47 0.28 0.28 0.32

174

Table S4. Interspecific phylogenetic conservatism of individual traits. Underlined values are significantly phylogenetically conserved.

Tailed

-

K Bloomsberg' Observed PIC varianc e of variance mean pic randomized One p.value Lumen Fraction 0.94 0.00 0.01 0.00 Leaf Thickness 0.65 0.01 0.01 0.00 LMA 0.63 0.01 0.01 0.00 Stem WD 0.59 0.01 0.02 0.00 KBRANCH 0.57 0.00 0.01 0.00 KS 0.68 0.01 0.02 0.01 Leaf Area 0.67 0.01 0.02 0.01 Branch WD 0.62 0.01 0.01 0.01 Branch Diameter 0.56 0.01 0.01 0.01 SRL 0.57 0.01 0.01 0.02 Branching Angle 0.43 0.00 0.01 0.04 VD 0.52 0.01 0.02 0.05 Root WD 0.43 0.01 0.01 0.09 MOE 0.41 0.01 0.01 0.09 Branching Distance 0.43 0.00 0.01 0.10 δ13C 0.40 0.01 0.01 0.11 (t/d)2 0.36 0.01 0.01 0.17 LNC 0.35 0.00 0.01 0.24 LPC 0.32 0.01 0.01 0.29 LCC 0.37 0.01 0.01 0.36

175

Figure S1. Phylogenetic tree of 24 study species.

176

177

Figure S2. Distribution of trait values per species. Notice that most species have similar ranges for most traits.

178

179

Figure S3. Distribution of environmental variables per species. Most species are found across large environmental ranges and overlap with each other in their distribution. This large overlap is second best to a fully crossed design where all species are located in all environments. A fully crossed design was not possible in natural populations due to different environmental affinities of the different species.

180

181

Figure S4. Principal component analysis biplots for static trait correlations for 24 study species. Seventeen of the twenty traits were used because sample size was significantly reduced by the two root traits and Branching Angle.

182

SUPPLEMENTARY MATERIAL – APPENDIX II: ANALYSES WITH TRAIT DATA CORRECTED FOR

ENVIRONMENTAL EFFECTS

Environmental effects on trait variance and integration structures and data analyses The environment explains a small amount in variance for individual traits (r2 ranging from 0 - 19%), and in the multivariate phenotype (5%, Figure S5). The effect of species turnover across environments (species-by-environment interaction) is responsible for an additional 9% of the total phenotypic variance. These values are surprisingly low given the range of environments covered (Figure S3). The environment has a minor to moderate impact on overall static integration structures for most species (Table S5, average rM between intrapopulation [T] and [TE] = 0.74), except for eastern cottonwood (Populus deltoïdes) which responds strongly to the environment. The most sensitive species to environmental effects is eastern cottonwood

(Populus deltoides, rM = 0.46 between [T] and [TE]) and the least sensitive species is red maple (Acer rubrum, rM = 0.92 between [T] and [TE]). The environment also has a minimal impact on evolutionary integration structure (rM between interspecific [T] and interspecific [TE] = 0.98, p=0.001). Comparing species similarity with and without environmental effects (Figure S6A) shows that on average species similarity increases when environmental effects are present (slope = 1.25). This suggests that part of the species’ response to the environment is similar. Nonetheless, there is much scatter about the relationship (r = 0.55), suggesting that species largely respond differently to the environment. Similarly, comparing static trait correlations with and without environmental effects (Figure S6B) shows that on average, static correlations are stronger when environmental effects are taken into account (slope = 1.35). Again, this indicates that on average, part of the traits’ response to the environment is coordinated, but given the large scatter about this relationships (r = 0.64) much of each trait’s response to the environment is different. Conducting the analyses with the trait data corrected for environmental effects,

TE, produced the same results as with the uncorrected trait data, T. First, using environment-free trait correlations [TE] to calculate similarity between species in

183

intrapopulation integration structure produce very similar patterns as with [T].

Specifically, 35% (97/276) of non-significant correlations an average rM of 0.23 (Table S6). Second, the same trait pairs show consistent static correlations with and without environmental effects removed (Figure S78). Third, removing environmental effects did not change the relationships between trait integration similarity and phylogenetic distance or environmental niche similarity (Figure S8).

184

Table S5. Environmental effects on static correlation structure for each of the 24 study species for nineteen traits (Branching angle was left out of the analysis because of missing values). For each species, using Mantel tests the trait correlation structure with effect of environmental variables removed, [TE], were compared with the trait correlation structure [T] based on the same trees. Species are sorted in increasing order of similarity.

[T] vs [TE] Species rM p-value POD 0.46 0.001 POB 0.67 0.001 BEPO 0.75 0.001 POG 0.76 0.001 QUR 0.76 0.001 POT 0.76 0.001 AMSP 0.78 0.001 ULA 0.79 0.001 ACSA 0.80 0.001 PRV 0.81 0.001 COA 0.81 0.001 BEPA 0.82 0.001 SOA 0.82 0.001 ACSP 0.82 0.001 BEA 0.83 0.001 PRP 0.84 0.001 OSV 0.87 0.001 ACP 0.88 0.001 FAG 0.88 0.001 CAC 0.88 0.001 FRA 0.91 0.001 TIA 0.91 0.001 PRS 0.91 0.001 ACR 0.92 0.001 Average 0.738

185

Table S6. Overall similarity, given as Mantel’s correlations (rM), in intra-population integration structure between species pairs and between each species and the interspecific pattern. Trait data with the effect of the environment is used [TE]. The boxes show the correlations for groups.

ACP ACR ACSA ACSP BEA BEPA BEPO OSV CAC COA FAG FRA POB POD POG POT AMSP PRP PRS PRV SOA QUR TIA ULA INTER ACP 0.22 0.21 0.25 n.s. 0.33 0.27 0.18 n.s. 0.19 0.24 n.s. 0.13 0.39 0.25 0.29 0.31 0.19 0.25 0.28 0.25 0.35 0.35 n.s. 0.25 ACR 0.25 0.13 n.s. n.s. 0.14 n.s. n.s. n.s. n.s. n.s. 0.18 n.s. n.s. 0.17 n.s. n.s. 0.18 n.s. n.s. 0.23 0.16 0.21 0.22 ACSA 0.30 0.26 0.19 0.37 0.24 n.s. 0.16 0.18 0.19 0.14 0.30 0.25 0.23 0.41 n.s. 0.24 n.s. n.s. 0.38 0.20 n.s. 0.39 ACSP Aceraceae 0.29 0.30 n.s. 0.14 n.s. 0.29 0.36 0.21 n.s. n.s. n.s. 0.26 n.s. 0.21 n.s. 0.22 n.s. n.s. 0.21 n.s. n.s. BEA n.s. 0.21 0.35 n.s. 0.13 0.17 n.s. n.s. 0.25 n.s. n.s. 0.15 n.s. 0.24 n.s. n.s. 0.17 n.s. n.s. 0.32 BEPA 0.15 0.23 n.s. n.s. 0.28 n.s. n.s. 0.24 0.16 n.s. n.s. 0.25 0.18 0.14 0.33 0.24 0.25 0.24 n.s. BEPO 0.34 n.s. n.s. 0.20 0.24 n.s. 0.42 0.29 0.26 0.24 n.s. 0.46 0.40 n.s. 0.42 0.22 n.s. 0.39 OSV Betulaceae n.s. 0.19 0.21 n.s. n.s. n.s. n.s. n.s. 0.41 n.s. 0.27 n.s. 0.24 0.26 0.31 n.s. 0.23 CAC 0.24 n.s. 0.26 n.s. 0.26 n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. 0.26 COA n.s. 0.20 0.25 0.20 0.24 0.16 0.27 n.s. 0.19 0.20 0.18 n.s. 0.26 n.s. 0.21 FAG n.s. n.s. n.s. 0.17 0.14 n.s. 0.15 0.25 n.s. 0.30 0.19 0.25 n.s. 0.23 FRA n.s. n.s. n.s. 0.17 n.s. n.s. 0.18 0.18 n.s. n.s. 0.15 n.s. 0.13 POB n.s. 0.31 0.18 0.40 0.25 0.17 n.s. n.s. 0.22 0.15 n.s. n.s. POD 0.33 0.32 0.24 0.14 0.28 0.44 n.s. 0.38 0.17 0.29 0.44 POG n.s. 0.29 0.18 0.23 0.17 n.s. 0.30 n.s. n.s. 0.24 POT Salicaceae 0.25 n.s. n.s. 0.41 n.s. 0.16 0.21 0.14 n.s. AMSP 0.15 0.27 n.s. n.s. 0.35 0.21 n.s. 0.29 PRP 0.22 0.14 0.23 n.s. 0.25 0.20 0.13 PRS 0.23 0.25 0.33 0.17 0.15 0.31 PRV n.s. n.s. 0.18 0.14 n.s. SOA Rosaceae n.s. 0.44 0.22 0.25 QUR 0.30 n.s. 0.26 TIA 0.25 0.21 ULA 0.13 INTER

186

100%

80% 0 2 Residuals 0 5 10 1 3 5 12 60% 32 Environment 5 5 7 9 5 Env * Spe 40% 41 32 5 1 2 12 5 18 48 72 76 72 8 6 19 36 66 1 0 59 Species 4 45 20% 40 1 38 39 10 28 30 31 34 22 19 17 15 12 0%

Figure S5. Variance partitioning analysis of traits among species, environment and their interaction. Environmental principal components were used. For multivariate variance partitioning, a redundancy analysis was conducted with species and environment as constraints.

187

Figure S6. Environmental effects on species integration similarity and bivariate static correlations. Panel A: Each point is a species pair. Nineteen traits were included in the trait correlation matrices (Branching angle was removed from the analysis because of some missing values). The x-axis gives the similarity of phenotypic integration matrices

[TE] with environmental effects removed (i.e. values in Table S6). The y-axis gives the corresponding species similarity calculated with the raw trait measurements [T] (i.e. values in Table 2). On average, the environment increases the similarity of species integration because the regression line is significantly above the 1:1 line (slope =1.25, CI=1.13-1.38). This suggest that species tend to respond similarly to the environment. Panel B: Each point is the static trait correlation for a given species. Seventeen traits were included in the pairwise correlations (Branching Angle, SRL and Root WD were removed) to avoid high correlations driven by low sample size. The x-axis gives trait correlations with environment-free traits, TE. The y-axis gives the corresponding trait correlations with uncorrected data, T. On average, the environment increases trait correlations because the regression line is significantly steeper than the 1:1 line (slope = 1.35, CI= 1.31-1.39). This suggests that traits tend to be more strongly correlated in response to the environment.

188

Figure S7. Density curves of intrapopulation Pearson’s correlation coefficients for trait pairs with data where the environmental effects on each trait have been removed (TE). Each line represents the correlation coefficients of the 24 study species for a given pair of traits. Within species, on average pairwise trait correlations are not significant below r = 0.5.

189

Figure S8. Relationships between static intragration structure similarity measured using [TE] (A) phylogenetic distance, (B) distance among environmental niche centroids and (C) environmental niche hypervolume overlap measured with Sorensen’s index. The dotted line in the cartoon of Panel A shows the distance between a pair of species, the dotted line in the cartoon

190

of Panel B shows the distance between the centroids of the two species and the black area in the cartoon of Panel C shows the overlap among the hypervolumes of the two species.

SUPPLEMENTARY MATERIAL – APPENDIX III. DETAILED METHODS

Trait measurements Branch Collection. For each tree, the vertical height and DBH (diameter at breast height) were recorded. One healthy lateral branch of at least 50cm in length and 5mm in diameter, located in the top third of the tree crown (but not the leader) was randomly selected, pruned off and taken back to the lab for trait measurements the same day. From each branch a 50cm long segment (measured from the tip of the longest branch) was cut and all following measurements made on it. Leaf Traits. Total leaf area of the branch as well as average individual leaf area (Leaf Area) was measured using a Li-3100C leaf area meter at a resolution of 1mm2. Leaf Area measurements included the petiole on simple leaves and the rachis and petiolule on compound leaves. Leaves on each branch were divided sorted as healthy and mature or non-healthy and/or immature. A leaf with less than ca.10% of its area damaged was considered healthy. All individual leaf measurements were done on the healthy subset. The thickness of leaf blades (Leaf thickness) was measured by taking 4 micrometer measurements on each of five randomly selected healthy leaves. The measurements were taken halfway between the mid-vein and the edge of the leaf, avoiding major secondary veins. Total branch mass, total leaf mass, and average individual leaf mass of healthy leaves were measured after drying the material for 72hrs or more of at 60°C in a drying oven. Leaf mass fraction (LMF) was calculated as the ratio of the total leaf mass of a branch to its total branch mass, leaf mass per area (LMA) as the ratio of leaf mass per leaf area of individual healthy leaves. Stoichiometric composition was measured on the bulk sample of healthy leaves after removing petioles and mid-veins and grinding them using a Thomas Wiley mini-Mill. Leaf nitrogen content (LNC), carbon content (LCC) and carbon isotope ratio (δ13C) were conducted in the isotope lab of the University of Arizona. Leaf phosphorus content (LPC) was measured on triplicates by colorimetric assays after

191

persulfate oxidation and reaction with acid molybdate (American Public Health Association 1999) using a ThermoScientific Genesys20 spectrophotometer. Branch Traits. The average branching distance of each branch was calculated by measuring the total branch length of the segments, including all lateral segments and dividing it by the number of bifurcations. Twig diameter at the base of the 50cm segments was recorded as the average of two perpendicular measurements taken with electronic calipers measuring to 2 decimals after the mm. Average branching angle (Branching Angle) of each branch was measured on pictures of the defoliated branches on a white background using ImageJ. Only angles lying flat against the plane were measured and for each branch, the smallest and largest angles were excluded. Branch wood density (Branch WD) was measured by volumetric replacement methods on a 12cm long segment of branch of ca.5mm diameter. This section was taken from the 50cm branch segment when possible or from further up along the branch when the diameter at the base of the 50cm segment was less than 5mm. A 10 mm segment with an average diameter of ca.5mm was cut and maintained frozen for xylem anatomy measurements. This segment was also taken on the 50cm segment when possible and further up the branch otherwise. Xylem Anatomy Traits. Safranin-stained cross sections of ca. 100 microns in thickness cut from each branch using a microtome and permanently mounted on microscope slides. Digital pictures were taken at 20x, 50x and 200x using an Olympus BX- 50 microscope, an Optronics Microfire microscope camera and the PictureFrame software. The 20x pictures were used to calculate the % sapwood area of the branch with Image J. Bark and pith area were excluded from the sapwood area but since the branches were 1-10 yrs old, we assumed all xylem area to be conductive for diffuse-porous species and only calculated the last year of xylem area for ring-porous species. The 50x pictures were used to calculate the ratio of cell wall thickness to vessel diameter (t/d2) in twelve adjacent vessel pairs of similar size (area within 30% of each other) for each sample. The 200x pictures were used to calculate for each sample the vessel diameter (VD) of all the vessels within two to three pie-shaped xylem areas going from pith to bark. This was done using STEM_GUI, a software developed for ImageJ. Only conduits greater than

192

25microns in diameter were considered as vessels (Sperry et al. 2006). STEM_GUI calculates total conductivity from the individual vessel diameters using the Hagen- Poisseuille equation. We then calculate conductivity per sapwood area (Ks) and per

Branch (KBRANCH). Lumen Fraction was calculated as the sum of the vessel area over the pie-area. Root Traits. Within the two weeks before 2012 leaf flush, eight root growth bags per species were installed on 5-8mm diameter roots on 200 of the study saplings. The roots were identified by following the main stem, cut with pruners, placed in the root bag with a mixture of the original soil and ca.30% sand, then watered, relocated in their original location and covered with soil and litter. Cutting the coarse roots of this size before leaf flush stimulates growth of fine absorptive roots. The 20cm x 25cm root bags were made with landscape fabric and nylon thread in order to prevent the roots of the focal tree from growing roots outside the bag and neighboring roots from growing into the bag. Three months later the bags were harvested in the same order as they were installed. The roots were rinsed and the 1st and 2nd order roots were stained with Neutral Red and scanned using WinRhizo to calculate total fine-root length. They were then oven dried for >72hrs at 60°C and weighed. Specific root length (SRL) was calculated as the ratio of total root mass per total root length. Of the 200 root bags installed, 159 produced 1st and 2nd order absorptive roots. The 5-8mm coarse roots to which the fine roots were attached were used to calculate coarse root wood density (Coarse Root WD) using the same volumetric replacement methods described for Branch WD. Stem Traits. At the end of the 2012 growing season, the trees were cut to collect a section of the stem at the base. This section was used to age the trees, calculate yearly basal area increment and calculate stem wood density (Stem WD). After removing the outer bark, Stem WD was measured using the same volumetric replacement methods described for Branch WD. Young’s modulus elasticity of the stem (ESTEM) was calculated using Euler’s equation from Tree Height, DGH and Stem WD.

Environmental measurements

193

All environmental variables described below were measured on each study sapling. Symmetric and asymmetric competition was assessed by quantifying the basal area occupied around the focal tree by competitors of similar size (1-10cm dbh) and larger size (>10cm dbh), respectively. This area was measured using the point quarter method (Cottam and Curtis 1956). Soil humidity was measured for each tree with a ML2x Theta Probe in the top soil, after brushing off the leaf litter. All measurements were taken within 36 consecutive hours following 10 days without rain. For each tree, we calculated the average of three measurements taken 1 feet from the stem of the focal tree. Elevation was recorded using a Garmin GPS. The slope of each tree was measured using a clinometer on the 20m plane around the focal tree. The aspect of the slope was then measured with a compass and converted into easting using the sine of the angle and northing as the cosine of the angle. To measure canopy opening, hemispherical photos were taken below the crown of each tree 1m aboveground with a Nikon Fisheye lense and a Nikon CoolPix 4500 Digital camera. Gap Light Analyzer v2.0 was then used to calculate % canopy opening and total transmitted light (Frazer et al. 1999). Soil depth around the tree was measured by taking the largest of three measurements taken with a 52cm soil probe ca. 30cm around the tree. Soil chemistry measurements were take on 12 individuals per species. For each sapling, four soil samples were taken in each cardinal direction from the top 15cm of soil. Each of the four samples was located at increasing distances from the tree. Those distances were adjusted to increase with tree size: from 1,2,3 and 4 ft from the tree for saplings 1cm in dbh up to 3, 6, 9 and 12 feet from the tree for saplings 5cm in dbh. The four soil samples were oven- dried, sieved through a 100 µm mesh and homogenized. Hydrogen concentration was determined in H2O, Organic carbon was determined by loss on ignition (Davies 1974), total nitrogen was determined using the Kjeldahl method (Bremner 1960) and exchangeable ion concentrations (P, K, Ca, Mg) were determined using the Melich-3 method (Tran et al. 1990). All analyses were conducted at the “laboratoire de chimie et de pédologie” of Université Laval.

194

Thermochron iButtons (model DS1921G-F50, accuracy ±1°C from -30°C to +70°C) were located in each of the 23 general areas where the study trees were located. In each area, three iButtons were installed ca.1.7m aboveground on the North-facing side of trunks of three dominant neighboring trees. The iButtons were placed in small opaque plastic containers in order to create shading and recorded air temperature at 4hr intervals from mid-June to early December 2015. For each area, we compiled monthly minimum, maximum and average temperatures across the 3 iButtons for each months. We ran a principal component analyses on these values and based on the results included August Maximum, November Minimum and the Mean Temperature across months as environmental variables.

195 APPENDIX D:

Testing central assumptions of trait-based ecology: Traits are good predictors of plant performance when phenotypic complexity and individual variation are taken into account

Prepared for submission in: Proceedings of the National Academy of Science of the United States of America

Authors: Julie Messier1, Brian J. McGill2, Martin J. Lechowicz3 and Brian J.Enquist1

Affiliations:

1 Ecology and Evolutionary Biology, University of Arizona, Tucson, AZ 85721, USA 2 School of Biology and Ecology, University of Maine, Orono, ME 04469, USA.

3 Biology Department, McGill University, Montréal, QC, H3A1B1, Canada.

Corresponding author: Julie Messier University of Arizona 1041 E. Lowell Street, Tucson, Arizona, 85711, USA 520-626-3336 [email protected]

Keyword: Trait-based ecology, functional traits, phenotypic complexity, phenotypic integration, intra-population variation, plant performance, relative growth rate.

Short title: Testing central assumptions of trait-based ecology

196 ABSTRACT:

While the trait-based approach has taken an increasingly important role in plant ecology, some of its fundamental assumptions remain poorly tested. Here we examine three key assumptions: (1) that traits are good predictors of plant performance, (2) that traits are better predictors of plant performance than species identity, and (3) that ecological filtering operates on traits. We measure relative growth rate (RGR), ecophysiological traits, environmental variables, species identity, size and age of individual tree saplings from 24 co-occurring temperate tree species. We use multivariate linear regressions and variance partitioning to predict RGR from these five categories of variables. Together, traits, age and environmental variables explain 89% of RGR variance, albeit with a large number of variables. We find that: (1) traits explains 64% of variance in RGR and that intraspecific level trait variance predicts 40% of that variance; (2) species identity and mean traits are poor predictor of RGR, and; (3) that environmental variables mainly affect RGR through their association with traits. Further, trait-trait interactions are important to accurately predict RGR, suggesting that the phenotypic context is key to accurately predict plant performance from traits. While these findings largely support the assumptions of trait-based ecology, they highlight that taking phenotypic complexity and individual-level variation into account is necessary for accurate predictions.

197 SIGNIFICANCE STATEMENT

Although the trait-based approach has been embraced to study numerous processes in plant ecology, some of its fundamental assumptions remain poorly tested. We test the assumption that traits are good predictors to plant performance in tree saplings of the temperate forest. We find this to be the case only when a large number of traits, trait- trait interactions and individual-level variation are considered. Moreover, species trait means are poor predictors of individual-level performance. Our results provide support for the trait-based approach but highlight the need to consider phenotypic complexity and inter-individual variation to predict plant performance accurately.

198 INTRODUCTION

In the last decade, the trait-based approach has been increasingly adopted in plant ecology (1–3). This approach is used to study a variety of ecological patterns and processes such as community assembly (4–9), ecosystem services (10–13), coexistence (14, 15) and evolutionary processes (16, 17), species abundances (18) and biogeographic distribution (19). The trait-based approach is appealing because it offers generality and predictability: by focusing on species’ shared properties instead of their identity, we can compare different species and link these shared properties to ecological processes. Although the usefulness of this approach hinges on the close links between traits and organismal performance (10, 20, 21, 22, Figure 1), these links remain poorly tested (23–25). This lack of scrutiny might be because a central tenet of ecology is that phenotypes are shaped by natural selection, such that we rarely question whether the phenotype affects performance. Some studies found that traits may be weak predictors of plant performance (26, 27), but these studies used species mean trait values measured on different populations from those where performance measures were taken. Intraspecific trait variability has been demonstrated to be sizeable in many traits (28–31) and to affect ecological processes (4, 32, 33). Thus, decoupling traits and performance measures probably blurs their association and traits and performance metrics. Hence, it remains unclear whether ecophysiological traits are closely linked to plant performance. Recent literature has stressed that our ability to accurately predict plant performance from traits might hinge on taking into account trait interactions (32). The effect of a given trait on performance might strongly depend on the phenotypic context, (i.e., on the set of phenotypic traits associated with a focal trait in an organism) because the functioning of organisms is intrinsically multidimensional (15, 34–39). Nonetheless, while the importance of trait-trait interactions on plant performance are clear (35, 40–42) their effects are rarely taken into consideration in ecological studies using a trait approach (43, but see 44, 45). It is unclear how ignoring trait-trait interactions affects our ability to predict plant performance (Figure 1).

199 To address these issues, we evaluate three key postulates of functional ecology: (1) that ecophysiological traits are good predictors of individual plant performance; (2) that traits predict performance better than species identity; and (3) that the environment selects individuals based on their traits (trait-based ecological filtering). Further, we examine the role of trait-trait interactions on our ability to predict individual plant performance. We use relative growth rate (RGR) as a proxy for plant performance, a metric that has been shown to be closely associated with mortality in tree saplings (46, 47) and survival in an annual plant (48). In order to evaluate these assumptions, we measure the RGR, phenotypic traits and environmental variables on 380 individual tree saplings from 24 coexisting temperate tree species. We test how well five different categories of predictor variables (tree size, tree age, species identity, ecophysiological traits and environmental variables) and their interactions predict RGR. Eighteen key ecophysiological traits from different organs (leaves, branches and stems) were sampled on each individuals across species. These traits reflect differences in plant architecture and four vital physiological functions, namely resource acquisition, resource conservation, sap transport and mechanical support (Table 1). Following the assumptions of trait-based ecology and the premise that trait interactions are important, we test the predictions that: (1) functional traits are the best predictors of RGR; (2) that trait-trait interactions are important in predicting RGR; and (3) that the effect of the environment on RGR mainly occurs via trait-environment covariance (a change in trait along environmental gradients indicates ecological filtering).

200 RESULTS

The best model explains 89% of the variance in relative growth rate (RGR) and includes traits, environmental variables and tree age (Table 3 & S3). The model includes five first order environmental variables, eighteen first order trait variables, tree age, 34 trait-trait interaction terms (T:T), three environment-environment interaction terms (E:E), and 44 trait-environment interaction terms (T:E). Functional traits are the best predictors of RGR Ecophysiological traits is the best category of variables predicting sapling RGR (Figure 2). The best traits-only model explains 64% of the variance in RGR (Table 3 & S4). In contrast, species is a poor predictor of RGR. The species-only model explains 14% of variance in RGR (Table 3) and the species term is not retained by model selection when added to the trait-only or environment-only models. Variance partitioning of RGR between traits and species shows that that 92% of the species effect on RGR occurs via traits (Figure 3B). The relationship between the variance in RGR explained and the number of traits measured starts to saturate after three traits (Figure 4A). Together, the 1st order effects of Leaf Mass per Area, Branch Wood Density (not Stem Wood Density) and Branch Diameter, three commonly and easily measured traits, account for 36% of the variance in RGR (Figure 4A). After traits, tree age and environmental variables are the best predictors of RGR: tree age alone explains 52% of variance in RGR and the best environment-only model explains 26% (Table 3).

Trait-trait interactions are important to predict RGR The best trait-only model, which explains 64% of the variance in RGR (Table 3 & S4), contains 68 T:T interaction terms. Adding the T:T interaction terms improves the trait-only model by 23 percentage points, from 41% for the 1st order trait model (Table 3). Partitioning the variance explained by T and T:T shows that they account for 40% vs. 39% of the variance in RGR, respectively (Table 3A). Their interaction accounts for 16% of the variance due to positive non-linear interactions (i.e. the effect of a T:T interaction on RGR increases as the value of one of its component traits increases) (Figure 3A).

201 Environmental effects on RGR are associated with trait turnover Alone, environmental variables are weak predictors of RGR. The best environment-only model explains 26% of the variance in sapling RGR (Table 3 and S5).

Soil fertility is the environmental variable with the largest 1st order effect on RGR (Figure 4B). Interactions among environmental variables substantially improve RGR predictions: adding the E:E interaction terms to the environment-only model improved it by 11 percentage points, from 15% for the 1st order environment model (Table 3). However, partitioning the variance explained by E and E:E shows that they account for 15% and 5%, respectively (Figure 3D). The net improvement of the environmental model when adding the interaction terms is due to the fact that the covariance between E and E:E shows negative non-linear interactions (i.e. the effect of the environment-environment interaction on RGR decrease as the value of one of the component environment- environment increases). Last, partitioning the variance explained by the traits and environmental variables retained by the best trait-environment model (without T:E interaction terms) shows that 81% of the environmental effect on RGR co-varies with the traits (Figure 2C). Thus, 21% of the variance in RGR is explained by trait turnover along environmental gradients (the coordinated change in both traits and environmental variables).

202 DISCUSSION

Relative growth rate is well predicted by traits when we consider phenotypic complexity Traits can predict a large fraction relative growth rate when we consider the high dimensionality of the phenotype and trait interactions. The trait-only model explains 62% of the variance in plant relative growth rate (Table 3, Figures 2A), but this model includes 18 traits and 68 trait-trait interactions (Table S4). Similarly, the best model predicting RGR from all categories of variables (traits, age and environment) explains 89% of the total variance in sapling RGR (Table S3), but this model contains 18 traits, 34 trait-trait interactions, five environmental variables and 44 trait-environment interactions. Indeed, variance partitioning shows that the trait-trait interactions explain 23% of the total variance in RGR and the trait-environment interactions explain 14% (Figure 2A and 3A). The large number and importance of interaction terms indicate that the effect of any given trait on plant performance depends on the phenotypic and environmental context in which the trait occurs. These results highlight that measuring phenotypic complexity, in terms of a large of number of traits and their interactions, is necessary to accurately predict plant performance (Figure 1).

The assumptions of trait based ecology are supported Despite the caveat that phenotypic complexity must be taken into account to accurately predict RGR, the results support the predictions arising from the fundamental assumptions of trait-based ecology. First, the data support the prediction that functional traits are good predictors of plant performance, with traits explaining 62% of the variance in plant relative growth rate (Table 3, Figures 2A). Second, we do find that traits are much better predictors of RGR than species identity, with species explaining only 14% of variance in RGR. In fact, 92% of the species’ explanatory power covaries with traits, indicating that the traits sampled capture most of the interspecific effects on RGR (Figure 3B). Third, ca. 3/4 of the environmental effects on RGR is expressed as trait-environment covariance. This covariance, which indicates a trait turnover along environmental gradients affecting RGR, indicates ecological filtering on traits (Figures 2C).

203

Ontogeny of traits Interestingly, tree age alone explains 52% of the variance in RGR (Figure 3). Saplings range between 5 and 66 years of age within the 1-5cm dbh size class studied. However, tree age and tree size are weakly correlated (Adjusted R2 = 0.05; p-value = 8e-6) so that the effect of age on RGR does not reflect a size effect. Moreover, variance partitioning shows that the pure effect of tree age on RGR is only 9%, while the age-trait covariance is 39% (Figure 2B). This covariance reflects a trait turnover with tree age that affects RGR and indicates a trait ontogeny affecting plant performance.

Intraspecific variability strongly influence plant performance Importantly, the covariance between traits and species indicates that interspecific trait differences accounts for only 1/5 of the effect of traits on RGR. Conversely, this shows that inter-individual differences within a population account for 4/5 of the effect of traits on RGR. Thus, species-level traits are poor predictors of plant performance at the individual level (2B). These results highlight that it is important to consider inter- individual trait differences to accurately predict individual plant performance. These findings are consistent with a growing body of literature finding that taking into account intraspecific variability is important to evaluate ecological processes such as species coexistence (24, 49) and community assembly processes (9, 30).

Our results contrast with a recent global study linking three functional traits in saplings to their height relative growth rate (26). Paine and colleagues found that functional traits explained only 3% of variation in juvenile growth rate (26). The differences in the methods used and the results obtained between this study and Paine’s support our interpretation that phenotypic complexity and inter-individual differences are important to accurately predicting RGR. First, they measured three traits, leaf mass per area, seed size and wood density, and did not include trait interactions. Here, we find that considering a large number of traits and their interactions is necessary to predict

204 plant performance well. In fact, we find that the variance in RGR explained by wood density and leaf mass per area alone is only 26% (Adjusted R2 = 0.26, p=2.2e-16, AIC = 1197). Second, Paine and colleagues conducted a global study which entailed species-level analyses, as well as trait and growth rate variables often measured on different individuals. In contrast, our study was conducted at the individual level and finds that 4/5 of traits’ predictive power of is associated with intraspecific differences. In fact, when we partition RGR variance between species and the two traits used in Paine et al. (Leaf Mass per Area and Branch Wood Density), we find that 10% of variance is explained by these species-level traits. Our results are therefore very similar to Paine’s when using similar methods. Thus, the fact that Paine’s study found that three traits measured at the species- level are poor predictor of RGR corroborates our findings that phenotypic dimensionality and complexity as well as inter-individual differences are essential to accurately predict RGR.

CONCLUSION

Our results indicate that ecophysiological traits are good predictors of plant performance, but only when a large number of phenotypic traits and their interactions with each other and with the environment are considered. Moreover, much of the trait variance affecting RGR occurs among individuals within a species. This suggests that species-level trait means are not good predictors of individual plant performance and questions the use of traits as predictors of RGR at global scales where individual plant variation is often impossible to take into account. Of the eighteen ecophysiological traits measured reflecting resource acquisition, resource conservation, sap transport and mechanical support, leaf mass per area, branch wood density and branch diameter have the largest effects on RGR. This confirms the importance of these frequently and easily measured traits in trait-based ecology. This study suggests that in order for trait-based ecology to become the powerful tool it promises to be, sampling and analytical methods will need to change to explicitly consider the phenotype’s high dimensionality, trait interactions and individual variation.

205 METHODS

Study Site We collected traits for each of 380 tree saplings in a ca. 20 km2 area located on or around Mont Saint-Hilaire, Canada (45°33′8″N 73°9′3″W). This hill, located in Southwestern Québec in the transition zone between the boreal forests to the north and the eastern deciduous forests to the south, is of plutonic origin. It includes a series of seven hilltops that rise up to 414m above the sedimentary floor of the surrounding Saint- Lawrence river valley (50). The mixed wood forest on the mountain is the largest remnant of primeval forest in the region. It is surrounded by an agricultural and suburban mosaic containing multiple forest fragments of varying successional status. The mountain’s unusual geological history created a broad diversity of habitats on and around the mountain (51–53). A local diversity hotspot, the nature reserve contains 650-900 of the 1,600 regional species of terrestrial vascular forest plants (51, 54). Sampling Design We conducted an individual-based sampling. We studied 24 temperate tree species co-occurring within a 20 km2 area on and immediately around Mont Saint-Hilaire (Table S1). The sapling life-stage is at the tail-end of the establishment phase, a critical growth phase with high mortality (55). Saplings therefore express traits that are probably linked to successful way of life in the environment (56, 57). Fifteen to 20 healthy tree saplings from the 24 most abundant deciduous species in the area were sampled, for a total of 380 saplings (Table). Sapling size was restricted to 1-5cm diameter at breast height (DBH) and a height of less than two thirds of the canopy height in order to sample individuals whose crown were in the sub-canopy layer and to standardize ontogenetic stage (58, 59). To maximize the range of trait values within each species and overlap in environment among species, healthy individuals were randomly selected from a diversity of habitats and microhabitats on and near the base of the mountain. Tree age was determined by counting tree rings of cross-sections from the base of the tree. Traits

206 The eighteen traits measured on each sapling are listed in Table. The detailed methods for the measurement of these traits are provided in the supplementary material. The traits measured reflect architecture and four vital biological functions: resource acquisition, resource conservation, sap transport and mechanical support. The traits were measured from all vegetative plant organs, namely the leaves, roots, stems and branches. For resource acquisition and conservation, leaf mass per area (LMA), leaf thickness, leaf nitrogen content (LNC), leaf phosphorus content (LPC), carbon isotope ratio (δ13C) in leaves (an integrated index of water use efficiency), leaf thickness and leaf carbon concentration (LCC) were measured. For mechanical support, stem wood density (Stem WD), branch wood density (Branch WD), coarse root wood density (Root WD) and stem Modulus of Elasticity (MOE) were measured. For sap transport safety and efficiency, vessel diameter (VD), vessel span-to-thickness ratio ((t/d)2) conductivity per sapwood area

(KS), conductivity per branch (KBRANCH) and Lumen Fraction were measured. For architectural traits, individual Leaf Area, Branching Distance and Branch Diameter were measured. Environmental Variables For each sapling, a set of nineteen environmental variables were measured (Table S2). These have been identified as important drivers of community composition at the study site (52, 53, 60–63). The individuals of each species cover a range of environments (Figure S10). The variables reflect the aboveground (irradiance, slope, aspect and air temperature), and belowground abiotic environments (soil nutrient composition, pH, humidity and depth) as well as two competition indices measured as the basal area of symmetric and asymmetric neighbors. Detailed methods for these measurements are included in Appendix III of the Supplemental Materials. Using the rda() function of the vegan package, we reduced the set of environmental variables to six principal components (Table 2). These environmental principal components were then used for all analyses. Statistical Analyses All statistical analyses were conducted in R version 3.1.2 (64). Following Kerkhoff and Enquist (65), tree size and traits reflecting size and rates were natural-log transformed

207 (LMA, Leaf Area, Branch diameter, KS and KBRANCH). All other traits and environmental variables were transformed following individual boxcox transformations to optimize normality when necessary (66) using the powerTransform() function of the car package. The effect of sampling date was removed for traits with a significant date effect by taking the residuals of a linear regression of individual traits against sampling date. The height aboveground at which each branch was sampled did not have an effect on δ13C values as can potentially occur (67). We tested how well the different categories of variables and their interactions predicted RGR using multivariate linear regressions, model selection and variance partitioning. The lm() and lme() functions of the stats and nlme packages were used to build fixed-effect and mixed-effect multiple regression models. The varpart() function of vegan, which calculates Adjusted R2 (68), was used to parse out the effect of different categories of variables. The AIC() function of the stats package was used to calculate Aikaike Information Criterion of each model. To determine the best (i.e. minimal adequate) model for each combination of explanatory variables, a stepwise model selection procedure was used. First we used the step() function of the stats package starting with the maximal model and following a two-way selection procedure. Then, we followed the manual variable elimination procedure described in Crawley (69) to further eliminate superfluous variables. Hypothesis testing To test the hypothesis that RGR is driven by traits, (1) we fit the best model predicting RGR from each of the category of variables individually (trait-only model, environment-only model, species-only model, age-only model and size-only model) and compared their Adjusted R2 and AIC values. For the trait-only and environment-only models, we tested the fit of models including only 1st order terms, 1st order and interaction terms and 1st order, interaction and quadratic terms. We retained the most parsimonious model within 2 AIC points of the model with the lowest AIC. (2) We partitioned RGR variance among the five categories of explanatory variables. For the traits and environment categories, we included only the terms retained by the best fitting

208 trait-only and environment-only models. If our hypothesis is true, traits should explain a large fraction of RGR variance and be better predictor than species identity. To test the hypothesis that the effect of a given trait value on RGR depends on the phenotypic context, we (1) compared the AIC and adjusted R2 of the best trait model fitted using only first order terms and the best trait model fitted using both first order and T-T interaction terms, and (2) we partitioned RGR variance among the first order terms and trait-trait (T-T) interaction terms. If our hypothesis is true, including the significant T-T interaction terms should lead to a large improvements in the model. Ecological filtering refers to the process of natural selection in the context of community assembly, where the environment retains the best suited individuals. To test the hypothesis that ecological filtering occurs on traits, we (1) compared the best trait- environment models (model including both traits and environmental variables) without and with T-E interaction terms , and (2) we partitioned RGR variance among the terms of the best environment-only model and the terms of the best trait-only model. If the hypothesis of strong ecological filtering is true, we predict that we will observe a large effect of trait-environment interaction (T-E) on RGR and small or absent pure environmental effect on RGR.

209 ACKNOWLEDGEMENTS

We would like to thank Andréanne Ferland, Émilie Lavoie, Natasha Salter, Carol Mordy, Anke Roth, Sierra Kaszubinski, Sanga Shir, Surbhi Patel, Margretta Murphy, John Lacson, Kevin Wong, Sarah Schwenck, Anjeanette McKay, Casey Knoks, Meghan Iacueuilli, Shahrzad Badie, Irene Liang, Jordyn Celaya and David Maneli and the Gault Nature Reserve staff for their instrumental role in data collection, Erica Bigio for invaluable assistance with the tree ring work, John Sperry, Melvin Tyree and David Killick for guidance and assistance with the tedious conductance measurements, and Louise Comas for guidance on root trait sampling methods in the field.

210 REFERENCES

1. Kattge J, et al. (2011) TRY - a global database of plant traits. Glob Chang Biol 17(9):2905–2935.

2. Weiher E, et al. (2011) Advances, challenges and a developing synthesis of ecological community assembly theory. Philos Trans R Soc Lond B Biol Sci 366(1576):2403–13.

3. Westoby M, Wright IJ (2006) Land-plant ecology on the basis of functional traits. Trends Ecol Evol 21(5):261–8.

4. Violle C, et al. (2012) The return of the variance: intraspecific variability in community ecology. Trends Ecol Evol 27(4):1–9.

5. Cornwell WK, Schwilk LDW, Ackerly DD (2006) A trait-based test for habitat filtering: convex hull volume. Ecology 87(6):1465–71.

6. Cornwell WK, Ackerly DD (2009) Community assembly and shifts in plant trait distributions across an environmental gradient in coastal California. Ecol Monogr 79(1):109–126.

7. Ackerly DD, Cornwell WK (2007) A trait-based approach to community assembly: partitioning of species trait values into within- and among-community components. Ecol Lett 10(2):135–45.

8. Kraft NJB, Valencia R, Ackerly DD (2008) Functional traits and niche-based tree community assembly in an Amazonian forest. Science 322(5901):580–2.

9. Jung V, Violle C, Mondy C, Hoffmann L, Muller S (2010) Intraspecific variability and trait-based community assembly. J Ecol 98(5):1134–1140.

10. Lavorel S, Garnier E (2002) Predicting changes in community composition and ecosystem functioning from plant traits: revisiting the Holy Grail. Funct Ecol 16(5):545–556.

11. Díaz S, et al. (2007) Incorporating plant functional diversity effects in ecosystem service assessments. Proc Natl Acad Sci U S A 104(52):20684–9.

12. Diaz S, Cabido M (2001) Vive la difference: plant functional diversity matters to ecosystem processes. Trends Ecol Evol 16(11):646–655.

13. Garnier E, et al. (2004) Plant functional markers capture ecosystem properties during secondary succession. Ecology 85(9):2630–2637.

14. Adler PB, Hillerislambers J, Levine JM (2007) A niche for neutrality. Ecol Lett 10(2):95–104.

211 15. Kraft NJB, Godoy O, Levine JM (2015) Plant functional traits and the multidimensional nature of species coexistence. Proc Natl Acad Sci 112(3):797–802.

16. Silvertown J, Dodd M, Gowing D (2001) Phylogeny and the niche structure of meadow plant communities. J Ecol 89(3):428–435.

17. Silvertown J, et al. (2006) Absence of phylogenetic signal in the niche structure of meadow plant communities. Proc Biol Sci 273(1582):39–44.

18. Mouillot D, Mason NWH, Wilson JB (2007) Is the abundance of species determined by their functional traits? A new method with a test using plant communities. Oecologia 152(4):729–37.

19. Violle C, Reich PB, Pacala SW, Enquist BJ, Kattge J (2014) The emergence and promise of functional biogeography. Proc Natl Acad Sci 11(38):13690–13696.

20. McGill BJ, Enquist BJ, Weiher E, Westoby M (2006) Rebuilding community ecology from functional traits. Trends Ecol Evol 21(4):178–85.

21. Shipley B (2007) Comparative plant ecology as a tool for integrating across scales. Ann Bot 99(5):965–966.

22. Violle C, et al. (2007) Let the concept of trait be functional! Oikos 116(5):882–892.

23. Adler PB, et al. (2014) Functional traits explain variation in plant life history strategies. Proc Natl Acad Sci 111(27):10019–10019.

24. Clark JS, et al. (2010) High-dimensional coexistence based on individual variation: a synthesis of evidence. Ecol Monogr 80(4):569–608.

25. Shipley B, et al. Reinforcing loose foundation stones in trait-based plant ecology. Oecologia (In Press).

26. Paine CET, et al. (2015) Globally, functional traits are weak predictors of juvenile tree growth, and we do not know why. J Ecol 103(4):978–989.

27. Herben T, Nováková Z, Klimešová J, Hrouda L (2012) Species traits and plant performance: functional trade-offs in a large set of species in a botanical garden. J Ecol 100(6):1522–1533.

28. Messier J, McGill BJ, Lechowicz MJ (2010) How do traits vary across ecological scales? A case for trait-based ecology. Ecol Lett 13(7):838–48.

29. Hulshof CM, Swenson NG (2010) Variation in leaf functional trait values within and across individuals and species: an example from a Costa Rican dry forest. Funct Ecol 24(1):217–223.

212 30. Albert CH, et al. (2010) A multi-trait approach reveals the structure and the relative importance of intra- vs. interspecific variability in plant traits. Funct Ecol 24(6):1192–1201.

31. Siefert A, et al. (2015) A global meta-analysis of the relative extent of intraspecific trait variation in plant communities. Ecol Lett 18:1406–1419.

32. Laughlin DC, Messier J (2015) Fitness of multidimensional phenotypes in dynamic adaptive landscapes. Trends Ecol Evol:1–10.

33. Albert CH, et al. (2012) On the importance of intraspecific variability for the quantification of functional diversity. Oikos 121(1):116–126.

34. Laughlin DC (2014) The intrinsic dimensionality of plant traits and its relevance to community assembly. J Ecol. doi:10.1111/1365-2745.12187.

35. Arnold SJ (1983) Morphology, performance and fitness. Am Zool 23(2):347–361.

36. Arnold SJ, Bürger R, Hohenlohe P a., Ajie BC, Jones AG (2008) Understanding the evolution and stability of the G-matrix. Evolution (N Y) 62(10):2451–2461.

37. Steppan SJ, Phillips PC, Houle D (2002) Comparative quantitative genetics: evolution of the G matrix. Trends Ecol Evol 17(7):320–327.

38. Blows MW (2007) A tale of two matrices: Multivariate approaches in evolutionary biology. J Evol Biol 20(1):1–8.

39. Kleyer M, Minden V (2015) Why functional ecology should consider all plant organs: An allocation-based perspective. Basic Appl Ecol 16(1):1–9.

40. Walsh B, Blows MW (2009) Abundant Genetic Variation plus Strong Selection = Multivariate Genetic Constraints: A Geometric View of Adaptation. Annu Rev Ecol Evol Syst 40:41–59.

41. Bonser SP (2006) Form defining function: interpreting leaf functional variability in integrated plant phenotypes. Oikos 114(1):187–190.

42. Levins R, Lewontin R (1985) The dialect biologist (Harvard University Press).

43. Shipley B, Vile D, Garnier E (2006) From plant traits to plant communities: a statistical mechanistic approach to biodiversity. Science 314(5800):812–4.

44. Grady KC, et al. (2013) Conservative leaf economic traits correlate with fast growth of genotypes of a foundation riparian species near the thermal maximum extent of its geographic range. Funct Ecol 27(2):428–438.

45. Milla R, Reich PB (2011) Multi-trait interactions, not phylogeny, fine-tune leaf size

213 reduction with increasing altitude. Ann Bot 107(3):455–65.

46. Pacala SW, et al. (1996) Forest models defined by field measurements : Estimation , error analysis and dynamics. Ecol Monogr 66(1):1–43.

47. Kobe RK, Pacala SW, Silander JA, Canham CD (1995) Juvenile tree survivorship as a component of shade tolerance. Ecol Appl 5(2):517–532.

48. Farris M a, Lechowicz MJ (1990) Functional interactions among traits that determine reproductive success in a native annual plant. Ecology 71(2):548–557.

49. Clark JS (2010) Individuals and the variation needed for high species diversity in forest trees. Science (80- ) 327(5969):1129–32.

50. Feininger T, Goodacre AK (1995) The eight classical Monteregian hills at depth and the mechanism of intrusion. Can J Earth Sci 8:440–446.

51. Maycock PF (1961) Botanical studies on Mont St.Hilaire, Rouville county, Quebec. Can J Bot 39:1293–1325.

52. Arii K, Lechowicz MJ (2002) The influence of overstory trees and abiotic factors on the sapling community in an old-growth Fagus-Acer forest. Ecoscience 9(3):386– 396.

53. Arii K, Hamel B, Lechowicz MJ (2005) Environmental correlates of canopy composition at Mont St. Hilaire, Quebec, Canada. J Torrey Bot Soc 132(1):90–102.

54. Elliott TL, Davies JT (2014) Challenges to barcoding an entire flora. Mol Ecol Resour 14(5):883–891.

55. Grubb PJ (1977) The maintenance of species-richness in plant communities: the importance of the regeneration niche. Biol Rev Camb Philos Soc.

56. Sterck FJ, Markesteijn L, Toledo M, Schieving F, Poorter L (2014) Sapling performance along resource gradients drives tree species distributions within and across tropical forests. Ecology (September):140311112851001.

57. Poorter L (2007) Are species adapted to their regeneration niche, adult niche, or both? Am Nat 169(4):433–42.

58. Palow DT, Nolting K, Kitajima K (2012) Functional trait divergence of juveniles and adults of nine Inga species with contrasting soil preference in a tropical rain forest. Funct Ecol 26(5):1144–1152.

59. Mason CM, McGaughey SE, Donovan L a (2013) Ontogeny strongly and differentially alters leaf economic and other key traits in three diverse Helianthus species. J Exp Bot 64(13):4089–4099.

214 60. Enright NJ, Lewis JE (1985) Aspect-related vegetation patterns and climate at Mont St. Hilaire, Quebec. Can Geogr 29(3):249–256.

61. Gilbert B, Lechowicz MJ (2004) Neutrality, niches, and dispersal in a temperate forest understory. Proc Natl Acad Sci U S A 101(20):7651–7656.

62. Karst J, Gilbert B, Lechowicz MJ (2005) Fern community assembly: The roles of chance and the environment at local and intermediate scales. Ecology 86(9):2473– 2486.

63. Radovski M (2010) Testing a trait-based model of fern community assembly. Dissertation (McGill, Montreal).

64. R Development Core Team R (2011) R: A Language and Environment for Statistical Computing. R Found Stat Comput 1(2.11.1):409.

65. Kerkhoff AJ, Enquist BJ (2009) Multiplicative by nature: Why logarithmic transformation is necessary in allometry. J Theor Biol 257(3):519–521.

66. Box G, Cox D (1964) An analysis of transformations. J R Statisistical Soc Ser B 26:211–46.

67. Farquhar GD, Ehleringer JR, Hubick KT (1989) Carbon discrimination and photosynthesis. Annu Rev Plant Physiol Plant Mol Biol 40:503–537.

68. Peres-Neto PR, Legendre P, Dray S, Borcard D (2006) Variation partitioning of species data matrices: estimation and comparison of fractions. Ecology 87:2614– 2625.

69. Crawley MJ (2007) The R book (John Wiley & Sons, Ltd, Chichester, England).

70. Sperry JS, Hacke UG, Pittermann J (2006) Size and function in conifer tracheids and angiosperm vessels. Am J Bot 93(10):1490–1500.

71. Hacke UG, Sperry JS (2001) Functional and ecological xylem anatomy. Perspect Plant Ecol Evol Syst 4(2):97–115.

72. Chave J, Coomes D, Lewis SL, Swenson NG, Zanne AE (2009) Towards a worldwide wood economic spectrum. Ecol Lett 12:351–366.

73. Niklas KJ (1992) Plant Biomechanics: an engineering approach to plant form and function (University of Chicago Press).

74. Niklas KJ (1994) Plant Allometry: The scaling of form and process (The University of Chicago Press, Chicago).

75. van Gelder H a, Poorter L, Sterck FJ (2006) Wood mechanics, allometry, and life-

215 history variation in a tropical rain forest tree community. New Phytol 171(2):367– 78.

76. White PS (1983) Corner’s rules in eastern deciduous trees : Allometry and its implications for the adaptive architecture of trees. Bull Torrey Bot Club 110(2):203–212.

77. White PS (1983) Evidence that temperate east North-American evergreen woody- plants follow Corner’s rules. New Phytol 95(1):139–145.

78. Ackerly DD, Donoghue MJ (1998) Leaf size, sapling allometry, and Corner’s rules: phylogeny and correlated evolution in maples (Acer). Am Nat 152(6):767–91.

79. Olson ME, Aguirre-Hernández R, Rosell J a (2009) Universal foliage-stem scaling across environments and species in dicot trees: plasticity, biomechanics and Corner’s Rules. Ecol Lett 12(3):210–9.

80. Preston K a, Cornwell WK, Denoyer JL (2006) Wood density and vessel traits as distinct correlates of ecological strategy in 51 California coast range angiosperms. New Phytol 170(4):807–18.

81. Tyree MT, Zimmermann MH (2002) Xylem Structure and the Ascent of Sap (Springer-Verlag, Berlin) doi:10.1007/978-3-662-04931-0.

82. Reich PB, et al. (1999) Generality of leaf trait relationships: A test across six biomes. Ecology 80(6):1955–1969.

83. Wright IJ, et al. (2004) The worldwide leaf economics spectrum. Nature 428(6985):821–827.

84. Enquist BJ, et al. (2007) A general integrative model for scaling plant growth, carbon flux, and functional trait spectra. Nature 449(7159):218–22.

85. Vile D, et al. (2005) Specific leaf area and dry matter content estimate thickness in laminar leaves. Ann Bot 96(6):1129–36.

86. Niinemets U, et al. (2007) Do we underestimate the importance of leaf size in plant economics? Disproportional scaling of support costs within the spectrum of leaf physiognomy. Ann Bot 100(2):283–303.

87. Lucas PW, Turner IM, Dominy NJ, Yamashita N (2000) Mechanical Defences to Herbivory. Ann Bot 86(5):913–920.

88. Lambers H, Chapin III FS, Pons TL (2008) Plant physiological ecology (Springer, New York). Second.

89. American Public Health Association (1999) Standard Methods for the Examination

216 of Water and Wastewater (Port City Press).

90. Cottam G, Curtis JT (1956) The Use of Distance Measures in Phytosociological Sampling. Ecology 37(3):451–460.

91. Frazer GW, Canham CD, Lertzman KP (1999) Gap Light Analyzer (GLA).

92. Davies BE (1974) Loss-on-ignition as an estimate of soil organic matter. Soil Sci Soc Am J 38(1):150–151.

93. Bremner JM (1960) Determination of nitrogen in soil by the Kjeldahl method. J Agric Sci 55:11.

94. Tran T Sen, Giroux M, Guilbeault J, Audesse P (1990) Evaluation of Mehlich‐III extractant to estimate the available P in Quebec soils. Commun Soil Sci Plant Anal 21(February 2015):1–28.

217 TABLES

Table 1. Eighteen ecophysiological study traits known to affect physiological functions. Functional Trait Units References Stem Wood Density g/cm3 (70–73) Stem Modulus of Elasticity (MOE) N/mm2 (74, 75) Branching Distance cm (76–78) Branch Wood Density g/cm3 (70–73, 79, 80) Branch Diameter mm (76–78) Stem Vessel Diameter (VD) μm (70, 71, 80) Lumen Fraction (% Lum) cm2/cm2 (70, 80, 81) Conductivity per Sapwood Area (KS) ml/s*cm2 (70, 81) Conductivity per Branch (KBRANCH) μm/μm (81) Vessel Thickness-to-Span Ratio ((t/d)2) μm/μm (70, 71, 80) Leaf Area cm2 (76–78) Leaf Mass per Area (LMA) g/cm2 (82–84) Leaf Thickness mm (85) Leaf Leaf Carbon Content (LCC) g/g (86, 87) Leaf Nitrogen Content (LNC) g/g (82–84) Leaf Phosphorus Content (LPC) g/g (82–84) δ13C δ13C‰ (67) Branch Leaf Area Ratio (LAR) cm2/g (88)

Table 2. Six environmental principal components summarizing the nineteen environmental variables measured (Table S2). These principal components were used in multivariate linear regression.

Environmental Principal Correlated Environmental Variables Components Soil Fertility - Total N; - Organic C; - Ca; – Mg; - K Light - Total Transmitted Light; - Canopy Opening; + Asymmetric Competition; + North Aspect Temperature + Average June-Dec Temperature; + Fall Minimum Temperature; – Elevation; + East Aspect; pH + pH; + Ca; - Asymmetric Competition; Soil Humidity & Depth + Soil Humidity; + Soil Depth; – Symmetric Competition Slope & P + P; – Slope; + North Aspect – East Aspect

218 Table 3. Fit of best models predicting Relative Growth Rate from different categories of predictor variables. E = 1st order environmental variables; E:E = environment- environment interaction terms; T = 1st order trait variables; T:T = trait-trait interaction terms; T:E = trait-environment interaction terms.

Model Categories of variables Adjusted p-value residual AIC retained in model R2 df All Variables T, T:T, E, E:E, T:E, Age 0.89 < 2.2e-16 176 253 Traits only T, T:T 0.64 < 2.2e-16 174 722 1st order traits only T 0.41 < 2.2e-16 252 794 Environment only E, E:E 0.26 1.1e-13 248 855 1st order environment only E 0.15 8.8e-09 255 885 Species only Species 0.14 3.2e-05 237 912 Age only Age 0.52 2.2e-16 259 729

219 FIGURES

T1

T2 Performance

T3 Metric 1

T 4 Fitness

T5 Performance Metric 2

T 6 Figure 1. Diagram modified from Arnold 1983 (35). Traits (solid arrows) and trait interactions (dashed arrows) affect plant performance metrics, such as mortality and seed output. Together performance indices in turn determine plant fitness. Usually, only the direct effects of traits on performance are considered in ecological studies and it is unknown how much ignoring trait-trait interactions affects our ability to predict performance.

Figure 2. Variance partitioning of RGR against ecophysiological traits, environmental variables and tree age. The variables included in each category are those selected by the best model built from all variables (Table 3 and S3). In both panels, first order trait effects (T) and trait-trait interactions (T:T) are included in the “Trait” category. In both panels, first order environmental effects (E), environment-environment interactions (E:E) and an environmental quadratic term are included in the “Environment” category. Trait- environment interaction terms (T:E) are left out of Panel B to emphasize how tree age interacts with traits and environmental variables. In Panel A, the intersecions without numbers indicate non-linear negative interactions.

220

Figure 3. Variance partitioning of RGR against (A) traits, parsing out first order and interaction terms, (B) species and the trait variables (T, T:T) retained by the best trait-species model, (C) environmental (E, E:E and E2) and trait variables (T and T:T) retained by the best trait-environment model without T:E interactions, and (D) environmental variables, parsing out first order and interaction terms.

Figure 4. Increase in variance in RGR explained by first order terms as the number of (A) traits and (B) environmental variables increase.

221 SUPPLEMENTARY MATERIAL– APPENDIX I. SUPPLEMENTAL TABLES AND FIGURES

Table S1. Twenty four study species

Xylem Species Family Orders Common name Leaf type Porosity Acer pensylvanicum Sapindaceae Sapindales Moosewood Simple Diffuse Acer rubrum Sapindaceae Sapindales Red Maple Simple Diffuse Acer saccharum Sapindaceae Sapindales Sugar Maple Simple Diffuse Acer spicatum Sapindaceae Sapindales Mountain Maple Simple Diffuse Amelanchier laevis Rosaceae Rosales Serviceberry Simple Diffuse Betula alleghaniensis Betulaceae Fagales Yellow birch Simple Diffuse Betula papyrifera Betulaceae Fagales White birch Simple Diffuse Betula populifolia Betulaceae Fagales Grey birch Simple Diffuse Carya cordiformis Juglandaceae Fagales Bitternut hickory Compound Semi or Ring Cornus alternifolia Cornaceae Cornales Alt. leaf dogwood Simple Diffuse Fagus grandifolia Fagaceae Fagales American beech Simple Diffuse Fraxinus americana Oleaceae Lamiales American ash Compound Ring Ostrya virginiana Betulaceae Fagales Ironwood Simple Diffuse Populus balsamifera Salicaceae Malpighiales Balsam Poplar Simple Diffuse Populus deltoides Salicaceae Malpighiales Deltoid Poplar Simple Diffuse Populus grandidentata Salicaceae Malpighiales Large-tooth Aspen Simple Diffuse Populus tremuloides Salicaceae Malpighiales Trembling Aspen Simple Diffuse Prunus pensylvanica Rosaceae Rosales Fire cherry Simple Diffuse Prunus serotina Rosaceae Rosales Black cherry Simple Diffuse Prunus virginiana Rosaceae Rosales Choke cherry Simple Diffuse Quercus rubra Fagaceae Fagales Red Oak Simple Ring Sorbus americana Rosaceae Rosales Mountain-Ash Compound Diffuse Tilia americana Malvaceae Malvales Linden, Basswood Simple Diffuse Ulmus americana Ulmaceae Rosales American Elm Simple Ring

222 Table S2. Environmental variables measured on each individual sapling, except for temperatures which were measured for each of the 23 general areas where the saplings were located.

Environmental Variables Abiotic Aboveground Elevation North and East Aspects Slope % Canopy Opening Total transmitted light Summer Maximum Temperature Fall Min Temperature Average June-Dec Temp Belowground Soil Humidity pH Organic C Total N P K Ca Mg Soil Depth Biotic Above & Asymmetric Competition Belowground Symmetric Competition

223 Table S3. Best model selected from all the categories of variables (T, E, T:T, E:E, E2, T:E, species, age). Note that species identity was not retained by model selection.

Variable Category Variable Estimate Std. Error t value Pr(>|t|) (Intercept) 0.222 0.056 3.99 0.000 *** 5 E Soil.Fertility 0.158 0.043 3.68 0.000 *** Temp 0.024 0.032 0.76 0.451 pH -0.115 0.036 -3.17 0.002 ** Light -0.022 0.048 -0.46 0.649 slope.P 0.090 0.029 3.05 0.003 ** 18 T Branch.Wood.Density -0.152 0.049 -3.12 0.002 ** Branch.Diameter -0.156 0.044 -3.51 0.001 *** LMA 0.292 0.069 4.26 4E-05 *** VD 0.144 0.066 2.18 0.031 * LNC -0.006 0.035 -0.16 0.870 Ks 0.049 0.106 0.46 0.644 KBRANCH 0.022 0.070 0.32 0.751 MOE -0.119 0.034 -3.52 0.001 *** LCC -0.143 0.036 -3.97 0.000 *** (t/d)2 -0.026 0.036 -0.72 0.474 % Lum -0.138 0.057 -2.43 0.016 * Stem.Wood.Density 0.199 0.050 4.00 1E-04 *** LPC -0.045 0.037 -1.21 0.228 δ13C 0.030 0.038 0.79 0.429 Leaf.Thickness -0.175 0.053 -3.33 0.001 ** Leaf.Area 0.245 0.055 4.49 1E-05 *** Branching.Distance 0.078 0.037 2.11 0.037 * LAR 0.167 0.040 4.22 4E-05 *** Age Age -0.662 0.038 -17.39 < 2e-16 *** 44 T:E Soil.Fertility:LMA -0.230 0.054 -4.29 3E-05 *** Soil.Fertility:Ks -0.116 0.038 -3.07 0.003 ** Soil.Fertility:LPC -0.146 0.034 -4.22 4E-05 *** Soil.Fertility:Leaf.Thickness 0.157 0.051 3.06 0.003 ** Soil.Fertility:Leaf.Area 0.243 0.048 5.08 1E-06 *** Soil.Fertility:Branching.Distance 0.111 0.035 3.22 0.002 ** Temp:Ks 0.236 0.045 5.26 5E-07 *** Temp:LCC 0.125 0.031 4.03 9E-05 *** Temp:(t/d)2 0.124 0.034 3.67 3E-04 *** Temp:% Lum -0.135 0.043 -3.14 0.002 ** Temp:Stem.Wood.Density -0.248 0.034 -7.18 3E-11 *** Temp:LPC -0.068 0.031 -2.19 0.030 *

224 Temp: δ13C 0.156 0.035 4.45 2E-05 *** Temp:Leaf.Thickness -0.236 0.038 -6.16 6E-09 *** Temp:Leaf.Area -0.111 0.038 -2.92 0.004 ** pH:LMA -0.133 0.043 -3.08 0.002 ** pH:Ks 0.127 0.043 2.95 0.004 ** pH: KBRANCH -0.141 0.043 -3.26 0.001 ** pH:(t/d)2 0.123 0.035 3.48 0.001 *** pH:Stem.Wood.Density -0.155 0.036 -4.32 3E-05 *** pH:LPC -0.155 0.036 -4.29 3E-05 *** pH: δ13C 0.286 0.044 6.47 1E-09 *** pH:Leaf.Area -0.127 0.040 -3.15 0.002 ** pH:Branching.Distance 0.123 0.035 3.51 0.001 *** Light:VD -0.216 0.073 -2.95 0.004 ** Light:Ks 0.224 0.085 2.62 0.010 ** Light:% Lum -0.182 0.055 -3.32 0.001 ** Light:Leaf.Area -0.097 0.049 -1.98 0.049 * Light:Branching.Distance 0.214 0.034 6.24 4E-09 *** Branch.Diameter:Soil.Humidity.Depth -0.095 0.036 -2.64 0.009 ** LMA:Soil.Humidity.Depth 0.145 0.034 4.25 4E-05 *** VD:Soil.Humidity.Depth 0.348 0.062 5.57 1E-07 *** Ks:Soil.Humidity.Depth -0.330 0.077 -4.28 3E-05 *** (t/d)2:Soil.Humidity.Depth -0.148 0.032 -4.57 1E-05 *** % Lum:Soil.Humidity.Depth 0.187 0.052 3.62 4E-04 *** Stem.Wood.Density:Soil.Humidity.Depth 0.115 0.031 3.71 3E-04 *** Leaf.Area:Soil.Humidity.Depth 0.121 0.036 3.38 0.001 *** slope.P:Branch.Wood.Density -0.109 0.042 -2.58 0.011 * slope.P:VD -0.290 0.075 -3.88 2E-04 *** slope.P:Ks 0.379 0.093 4.10 7E-05 *** slope.P:MOE -0.085 0.032 -2.64 0.009 ** slope.P:% Lum -0.294 0.053 -5.52 1E-07 *** slope.P:Stem.Wood.Density 0.088 0.039 2.24 0.027 * slope.P:Branching.Distance 0.079 0.031 2.59 0.011 * 34 T:T VD: KBRANCH 0.111 0.039 2.84 0.005 ** Branch.Wood.Density: (t/d)2 -0.180 0.036 -5.05 1E-06 *** LMA: (t/d)2 0.144 0.036 4.01 9E-05 *** (t/d)2:Leaf.Area 0.165 0.036 4.55 1E-05 *** Ks:% Lum -0.270 0.036 -7.52 4E-12 *** Ks:LCC -0.174 0.047 -3.72 3E-04 *** Ks: δ13C 0.128 0.051 2.51 0.013 * Branch.Diameter:Ks -0.151 0.043 -3.52 0.001 *** Ks:LAR -0.315 0.041 -7.69 2E-12 *** KBRANCH:% Lum 0.113 0.033 3.37 0.001 *** KBRANCH:LCC 0.219 0.038 5.83 3E-08 ***

225 Branch.Wood.Density:% Lum -0.149 0.040 -3.70 3E-04 *** % Lum:LPC -0.092 0.039 -2.36 0.019 * % Lum: δ13C -0.181 0.048 -3.81 2E-04 *** % Lum:Leaf.Thickness -0.123 0.040 -3.07 0.002 ** % Lum:Branching.Distance 0.131 0.033 3.99 1E-04 *** Branch.Wood.Density:Stem.Wood.Density 0.132 0.035 3.75 2E-04 *** Branch.Wood.Density:LPC 0.191 0.040 4.77 4E-06 *** MOE:Stem.Wood.Density -0.184 0.035 -5.23 5E-07 *** LMA:Stem.Wood.Density -0.162 0.042 -3.83 2E-04 *** LMA:MOE 0.132 0.035 3.75 2E-04 *** Branch.Diameter:MOE -0.267 0.038 -7.07 5E-11 *** MOE:Leaf.Area 0.187 0.039 4.80 4E-06 *** LCC:Leaf.Area 0.103 0.034 3.03 0.003 ** Branch.Diameter:LNC -0.111 0.032 -3.50 0.001 *** LPC:Branching.Distance 0.166 0.036 4.63 8E-06 *** δ13C:Leaf.Thickness 0.135 0.049 2.77 0.006 ** Branch.Diameter: δ13C -0.092 0.034 -2.69 0.008 ** LMA:Leaf.Thickness -0.133 0.057 -2.32 0.022 * LMA:Leaf.Area -0.188 0.061 -3.10 0.002 ** Branch.Diameter:Leaf.Thickness -0.117 0.046 -2.54 0.012 * Branch.Diameter:Branching.Distance 0.148 0.035 4.18 5E-05 *** Leaf.Area:Branching.Distance -0.240 0.036 -6.74 3E-10 *** Leaf.Area:LAR 0.073 0.035 2.08 0.039 * 3 E:E Soil.Fertility:pH -0.147 0.049 -2.96 0.004 ** Light:slope.P 0.089 0.035 2.53 0.013 * Temp:Soil.Humidity.Depth 0.079 0.031 2.55 0.012 *

226 Table S4. Best traits-only model

Variable category Variable Estimate Std. Error t value Pr(>|t|) (Intercept) -1.976 0.125 -15.75 2E-16 *** 18 T VD -0.074 0.158 -0.47 0.640 (t/d)2 -0.119 0.087 -1.36 0.174 Ks 0.308 0.255 1.21 0.228 KBRANCH 0.039 0.164 0.24 0.814 % Lum -0.291 0.132 -2.21 0.028 * Branch.Wood.Density -0.339 0.121 -2.80 0.006 ** Stem.Wood.Density -0.128 0.115 -1.11 0.269 MOE -0.345 0.077 -4.46 1E-05 *** LCC 0.060 0.081 0.75 0.457 LNC 0.057 0.084 0.68 0.497 LPC -0.187 0.086 -2.17 0.032 * δ13C -0.057 0.103 -0.56 0.578 LMA 0.780 0.160 4.86 3E-06 *** Leaf.Thickness -0.239 0.131 -1.83 0.069 . Branch.Diameter -0.392 0.108 -3.61 4E-04 *** Leaf.Area 0.313 0.139 2.25 0.025 * Branching.Distance 0.091 0.088 1.03 0.303 LAR 0.170 0.099 1.71 0.089 . 68 T:T VD: KBRANCH 0.295 0.098 3.00 0.003 ** VD: δ13C -0.511 0.215 -2.37 0.019 * VD:LMA 1.314 0.329 3.99 1E-04 *** VD:Leaf.Thickness -0.962 0.298 -3.23 0.001 ** VD:Branch.Diameter -0.451 0.149 -3.03 0.003 ** VD:Leaf.Area 0.423 0.120 3.53 0.001 *** (t/d)2:% Lum 0.150 0.074 2.01 0.045 * (t/d)2:Branch.Wood.Density 0.115 0.089 1.29 0.198 (t/d)2:MOE 0.132 0.088 1.51 0.133 (t/d)2:LMA 0.197 0.083 2.39 0.018 * (t/d)2:Leaf.Area 0.442 0.098 4.52 1E-05 *** (t/d)2:Branching.Distance -0.372 0.096 -3.87 2E-04 *** Ks:% Lum -0.300 0.093 -3.24 0.001 **

227 Ks:MOE -0.220 0.085 -2.61 0.010 ** Ks:LCC -0.180 0.096 -1.89 0.061 . Ks:LNC -0.155 0.092 -1.68 0.095 . Ks:δ13C 0.765 0.247 3.10 0.002 ** Ks:LMA -1.732 0.365 -4.74 4E-06 *** Ks:Leaf.Thickness 1.010 0.327 3.09 0.002 ** Ks:Branch.Diameter 0.237 0.157 1.51 0.133 Ks:LAR -0.342 0.146 -2.34 0.020 * KBRANCH:% Lum 0.140 0.081 1.73 0.085 . KBRANCH:LCC 0.391 0.082 4.78 4E-06 *** 13 KBRANCH:δ C 0.219 0.097 2.25 0.025 * KBRANCH:Leaf.Area -0.205 0.111 -1.85 0.065 . KBRANCH:LAR 0.286 0.125 2.28 0.024 * % Lum:Branch.Wood.Density -0.221 0.095 -2.33 0.021 * % Lum:LPC -0.233 0.089 -2.61 0.010 ** % Lum:δ13C -0.688 0.150 -4.59 8E-06 *** % Lum:LMA 0.714 0.237 3.02 0.003 ** % Lum:Leaf.Thickness -0.552 0.220 -2.51 0.013 * % Lum:Branching.Distance 0.245 0.078 3.16 0.002 ** Branch.Wood.Density:Stem.Wood.Density 0.147 0.082 1.78 0.076 . Branch.Wood.Density:LCC 0.270 0.085 3.18 0.002 ** Branch.Wood.Density:LNC -0.234 0.106 -2.21 0.029 * Branch.Wood.Density:LPC 0.428 0.100 4.30 3E-05 *** Branch.Wood.Density:δ13C 0.359 0.163 2.21 0.029 * Branch.Wood.Density:LMA -0.791 0.165 -4.79 4E-06 *** Branch.Wood.Density:Branch.Diameter 0.273 0.098 2.79 0.006 ** Stem.Wood.Density:MOE -0.322 0.090 -3.59 4E-04 *** Stem.Wood.Density:LNC 0.266 0.092 2.90 0.004 ** Stem.Wood.Density:δ13C -0.429 0.140 -3.07 0.002 ** Stem.Wood.Density:LMA 0.429 0.145 2.96 0.004 ** MOE:LPC -0.187 0.084 -2.21 0.028 * MOE:LMA 0.449 0.125 3.60 4E-04 *** MOE:Leaf.Thickness -0.416 0.112 -3.71 3E-04 *** MOE:Branch.Diameter -0.286 0.085 -3.35 0.001 *** MOE:Leaf.Area -0.274 0.102 -2.67 0.008 ** MOE:Branching.Distance 0.220 0.088 2.49 0.014 * LCC:LMA -0.155 0.075 -2.06 0.041 *

228 LCC:Leaf.Area 0.236 0.089 2.67 0.008 ** LNC:LPC 0.182 0.079 2.31 0.022 * LNC:δ13C 0.435 0.094 4.65 7E-06 *** LNC:Branch.Diameter -0.166 0.087 -1.90 0.059 . LPC:Branching.Distance 0.383 0.086 4.42 2E-05 *** δ13C:LMA 0.873 0.136 6.41 1E-09 *** δ13C:Leaf.Thickness -0.633 0.139 -4.57 9E-06 *** δ13C:Branch.Diameter -0.282 0.098 -2.87 0.005 ** δ13C:Leaf.Area 0.383 0.147 2.60 0.010 * δ13C:Branching.Distance -0.232 0.077 -3.03 0.003 ** LMA:Leaf.Thickness 0.165 0.108 1.53 0.128 LMA:Leaf.Area -0.652 0.208 -3.13 0.002 ** Leaf.Thickness:Branch.Diameter -0.247 0.113 -2.18 0.031 * Leaf.Thickness:Leaf.Area 0.698 0.164 4.25 3E-05 *** Branch.Diameter:Branching.Distance 0.327 0.104 3.14 0.002 ** Leaf.Area:Branching.Distance -0.240 0.103 -2.34 0.021 * Leaf.Area:LAR 0.339 0.095 3.56 0.000 *** Branching.Distance:LAR 0.218 0.092 2.38 0.018 *

229 Table S5. Best environment-only model

Variable Std. Variable category Estimate Error t value Pr(>|t|) (Intercept) -2.059 0.101 -20.48 < 2e-16 *** 6 E Soil.Fertility 0.458 0.085 5.36 0.000 *** Light -0.535 0.099 -5.40 0.000 *** Temp 0.292 0.080 3.66 0.000 *** pH -0.117 0.081 -1.44 0.150 Soil.Humidity.Depth 0.267 0.080 3.36 0.001 *** slope.P 0.091 0.082 1.11 0.266 1 E2 Temp2 0.230 0.067 3.44 0.001 *** 4 E:E Soil.Fertility:pH 0.349 0.101 3.44 0.001 *** Light:slope.P -0.251 0.080 -3.14 0.002 ** Temp:Soil.Humidity.Depth 0.200 0.079 2.54 0.012 * pH:slope.P 0.149 0.090 1.66 0.098 .

230

231

Figure S1. Distribution of trait values per species. Notice that most species have similar ranges for most traits.

232 233

Figure S2. Distribution of environmental variables per species. Most species are found across large environmental ranges and overlap with each other in their distribution. This large overlap is second best to a fully crossed design where all species are located in all environments. A fully crossed design was not possible in natural populations due to different environmental affinities of the different species.

234

Figure S3. Variance partitioning of RGR against trait and environment variables retained by the best trait-environment model with only first order variables (i.e. without trait:trait or environment:environment interactions).

235 SUPPLEMENTARY MATERIAL – APPENDIX III. DETAILED METHODS

Trait measurements Branch Collection. For each tree, the vertical height and DBH (diameter at breast height) were recorded. One healthy lateral branch of at least 50cm in length and 5mm in diameter, located in the top third of the tree crown (but not the leader) was randomly selected, pruned off and taken back to the lab for trait measurements the same day. From each branch a 50cm long segment (measured from the tip of the longest branch) was cut and all following measurements made on it. Leaf Traits. Total leaf area of the branch as well as average individual leaf area (Leaf Area) was measured using a Li-3100C leaf area meter at a resolution of 1mm2. Leaf Area measurements included the petiole on simple leaves and the rachis and petiolule on compound leaves. Leaves on each branch were divided sorted as healthy and mature or non- healthy and/or immature. A leaf with less than ca.10% of its area damaged was considered healthy. All individual leaf measurements were done on the healthy subset. The thickness of leaf blades (Leaf thickness) was measured by taking 4 micrometer measurements on each of five randomly selected healthy leaves. The measurements were taken halfway between the mid-vein and the edge of the leaf, avoiding major secondary veins. Total branch mass, total leaf mass, and average individual leaf mass of healthy leaves were measured after drying the material for 72hrs or more of at 60°C in a drying oven. Leaf mass fraction (LMF) was calculated as the ratio of the total leaf mass of a branch to its total branch mass, leaf mass per area (LMA) as the ratio of leaf mass per leaf area of individual healthy leaves. Stoichiometric composition was measured on the bulk sample of healthy leaves after removing petioles and mid-veins and grinding them using a Thomas Wiley mini-Mill. Leaf nitrogen content (LNC), carbon content (LCC) and carbon isotope ratio (δ13C) were conducted in the isotope lab of the University of Arizona. Leaf phosphorus content (LPC) was measured on triplicates by colorimetric assays after persulfate oxidation and reaction with acid molybdate (89) using a ThermoScientific Genesys20 spectrophotometer.

236 Branch Traits. The average branching distance of each branch was calculated by measuring the total branch length of the segments, including all lateral segments and dividing it by the number of bifurcations. Branch diameter at the base of the 50cm segments was recorded as the average of two perpendicular measurements taken with electronic calipers measuring to 2 decimals after the mm. Average branching angle (Branching Angle) of each branch was measured on pictures of the defoliated branches on a white background using ImageJ. Only angles lying flat against the plane were measured and for each branch, the smallest and largest angles were excluded. Branch wood density (Branch WD) was measured by volumetric replacement methods on a 12cm long segment of branch of ca.5mm diameter. This section was taken from the 50cm branch segment when possible or from further up along the branch when the diameter at the base of the 50cm segment was less than 5mm. A 10 mm segment with an average diameter of ca.5mm was cut and maintained frozen for xylem anatomy measurements. This segment was also taken on the 50cm segment when possible and further up the branch otherwise. Xylem Anatomy Traits. Safranin-stained cross sections of ca. 100 microns in thickness cut from each branch using a microtome and permanently mounted on microscope slides. Digital pictures were taken at 20x, 50x and 200x using an Olympus BX-50 microscope, an Optronics Microfire microscope camera and the PictureFrame software. The 20x pictures were used to calculate the % sapwood area of the branch with Image J. Bark and pith area were excluded from the sapwood area but since the branches were 1-10 yrs old, we assumed all xylem area to be conductive for diffuse-porous species and only calculated the last year of xylem area for ring-porous species. The 50x pictures were used to calculate the ratio of cell wall thickness to vessel diameter (t/d2) in twelve adjacent vessel pairs of similar size (area within 30% of each other) for each sample. The 200x pictures were used to calculate for each sample the vessel diameter (VD) of all the vessels within two to three pie-shaped xylem areas going from pith to bark. This was done using STEM_GUI, a software developed for ImageJ. Only conduits greater than 25microns in diameter were considered as vessels (70).

237 STEM_GUI calculates total conductivity from the individual vessel diameters using the Hagen-Poisseuille equation. We then calculate conductivity per sapwood area (Ks) and per

Branch (KBRANCH). Lumen Fraction was calculated as the sum of the vessel area over the pie- area. Root Traits. Within the two weeks before 2012 leaf flush, eight root growth bags per species were installed on 5-8mm diameter roots on 200 of the study saplings. The roots were identified by following the main stem, cut with pruners, placed in the root bag with a mixture of the original soil and ca.30% sand, then watered, relocated in their original location and covered with soil and litter. Cutting the coarse roots of this size before leaf flush stimulates growth of fine absorptive roots. The 20cm x 25cm root bags were made with landscape fabric and nylon thread in order to prevent the roots of the focal tree from growing roots outside the bag and neighboring roots from growing into the bag. Three months later the bags were harvested in the same order as they were installed. The roots were rinsed and the 1st and 2nd order roots were stained with Neutral Red and scanned using WinRhizo to calculate total fine-root length. They were then oven dried for >72hrs at 60°C and weighed. Specific root length (SRL) was calculated as the ratio of total root mass per total root length. Of the 200 root bags installed, 159 produced 1st and 2nd order absorptive roots. The 5-8mm coarse roots to which the fine roots were attached were used to calculate coarse root wood density (Coarse Root WD) using the same volumetric replacement methods described for Branch WD. Stem Traits. At the end of the 2012 growing season, the trees were cut to collect a section of the stem at the base. This section was used to age the trees, calculate yearly basal area increment and calculate stem wood density (Stem WD). After removing the outer bark, Stem WD was measured using the same volumetric replacement methods described for Branch WD. Young’s modulus elasticity of the stem (MOE) was calculated using Euler’s equation from Tree Height, DGH and Stem WD.

238 Environmental measurements All environmental variables described below were measured on each study sapling. Symmetric and asymmetric competition was assessed by quantifying the basal area occupied around the focal tree by competitors of similar size (1-10cm dbh) and larger size (>10cm dbh), respectively. This area was measured using the point quarter method (90). Soil humidity was measured for each tree with a ML2x Theta Probe in the top soil, after brushing off the leaf litter. All measurements were taken within 36 consecutive hours following 10 days without rain. For each tree, we calculated the average of three measurements taken 1 feet from the stem of the focal tree. Elevation was recorded using a Garmin GPS. The slope of each tree was measured using a clinometer on the 20m plane around the focal tree. The aspect of the slope was then measured with a compass and converted into easting using the sine of the angle and northing as the cosine of the angle. To measure canopy opening, hemispherical photos were taken below the crown of each tree 1m aboveground with a Nikon Fisheye lense and a Nikon CoolPix 4500 Digital camera. Gap Light Analyzer v2.0 was then used to calculate % canopy opening and total transmitted light (91). Soil depth around the tree was measured by taking the largest of three measurements taken with a 52cm soil probe ca. 30cm around the tree. Soil chemistry measurements were take on 12 individuals per species. For each sapling, four soil samples were taken in each cardinal direction from the top 15cm of soil. Each of the four samples was located at increasing distances from the tree. Those distances were adjusted to increase with tree size: from 1,2,3 and 4 ft from the tree for saplings 1cm in dbh up to 3, 6, 9 and 12 feet from the tree for saplings 5cm in dbh. The four soil samples were oven-dried, sieved through a 100 µm mesh and homogenized. Hydrogen concentration was determined in H2O, Organic carbon was determined by loss on ignition (92), total nitrogen was determined using the Kjeldahl method (93) and exchangeable ion concentrations (P, K, Ca, Mg) were determined using the Melich-3 method (94). All analyses were conducted at the “laboratoire de chimie et de pédologie” of Université Laval.

239 Thermochron iButtons (model DS1921G-F50, accuracy ±1°C from -30°C to +70°C) were located in each of the 23 general areas where the study trees were located. In each area, three iButtons were installed ca.1.7m aboveground on the North-facing side of trunks of three dominant neighboring trees. The iButtons were placed in small opaque plastic containers in order to create shading and recorded air temperature at 4hr intervals from mid- June to early December 2015. For each area, we compiled monthly minimum, maximum and average temperatures across the 3 iButtons for each months. We ran a principal component analyses on these values and based on the results included August Maximum, November Minimum and the Mean Temperature across months as environmental variables.

240 APPENDIX E:

Detailed protocols

SAMPLING DESIGN

Sapling sampling 15-20 saplings from 25 angiosperm tree species were sampled, for a total of 395 saplings. See species list in Table 1. In order to get post-establishment and pre-reproductive stages, the saplings studied varied in size from 1-5 cm at breast height (1.3 m) and up to 2/3 of the canopy height for saplings growing under a canopy, for a range of 2-8m in height. To collect saplings from the whole range of habitats existing in the area, the saplings were collected from 23 sites spread on and around mount Saint-Hilaire. 19 sites are located on the mountain proper and 6 sites are located around the base of the mountain in forest remnants or abandoned land (see table 2 and figure 1.). Within each site, saplings were chosen in order to represent the range of existing

1km microhabitat. When two saplings of the same species

Figure 1. My tree-based sampling design are sampled within a site, they were selected from on Mont Saint-Hilaire. Each black dot different microenvironments. represents an individual sapling.

Traits measured on the whole tree Tree Height – In both 2011 and 2012, the vertical height from the base of the tree to the end of the highest branch was recorded using a telescopic measuring stick for saplings up to 6m in height and using a clinometer and a measuring tape for saplings higher than 6m.

DBH:DGH – In 2011, the diameter at breast height (1.3m) and at ground height were measured using a DBH tape. The tape was moderately squeezed around the tree to account

241 for the effects of elastic and flaking barks. For ground height, diameters were measured above the bulge resulting from the root-stem junction. For breast height, bulges due to branches were avoided. In 2012, the DBH was re-measured on all trees.

Height:DBH Ratio – The 2011 height was divided by the 2011 DBH. This provides a tapering index of the main stem.

Branch sampling date – The date at which the height and DBH were measured and at which the branch was sampled were recorded. These dates were used as co-variates in order to remove the potential effect of date on traits.

Location - For each, UTM coordinates were recorded using a GPS. We took the coordinates of a point central to a number of trees. We then measured the distance and angle from the central point to each tree using a compass and measuring tape, and calculated the GPS coordinates of each tree based on their position relative to the central point and the UTM coordinates of the central point.

Traits measured on branches

Branch sampling design for branch and leaf traits Both leaf traits and architectural traits are known to vary significantly with their position along the main stem and with respect to their depth in the crown. For this reason, the location of the branch in the crown as well as the length of the branch sampled was standardized. In June and July of 2011, one branch from each sapling was collected. One of the healthy lateral 50cm branches with healthy, mature leaves was randomly selected from the top third of the crown. The branches were originally Picture 1. Example of cut to be at least 50cm in length from the tip and at least 5mm branch sample in diameter. This branch was cut down with pole pruners and

242 taken back to the lab. In the lab, a segment of 50cm in length, measured from the leading branch tip was re-cut and used for all further branch-level measurements (see picture1). All measurements on the branches were done on the same day the branches were harvested. Two pictures of each branch were taken: the first with all the leaves present on the branch and the second without leaves.

Branch sampling design for wood anatomy traits and wood density Wood and sap conductivity traits are known to vary with 10mm branch diameter. For this reason, two separate branch segments, standardized for diameter, were taken for each of the wood 12mm density and wood anatomy measurements. For wood density, a 5mm 12mm long branch segment was cut where the average diameter diameter of at least one end was 5mm. For wood anatomy measurements, a

Picture 2. Illustration of 10mm long branch segment was cut where the average diameter branch sampling method of at least one end is 5mm. The 10mm long segments for wood for wood anatomy and wood density anatomy were immidiately frozen to preserve them until further measurements processing. The average diameter was calculated with calipers by taking two perpendicular measurements on the branch. Branch tip diameters and tapering vary among species, so that sometimes the 5mm-diameter segment was taken from the 50cm- long branch and sometimes it was taken from below the 50cm-long branch (for example, on Betulaceae which have very fine branches at the tip). The distance from the tip to the 10mm and 12mm long segments were recorded. When the branch never reached 5mm (for example, for Rhus typhina which have coarse branches at the tip), segments with the smallest diameter was taken.

Branch age – The age of the 50cm branch was calculated by counting the number of terminal bud scars present on the branch.

243 Branching Distance – Average distance between branch bifurcations. The total length of the branch, including all lateral segments, was measured by carefully following all the branch segments using a seamstress tape. The number of bifurcations were counted on each branch and the total branch length was divided by the total number of bifurcations. For species with opposing branching patterns (example, for Acers), the two opposing bifurcations were counted as two bifurcations.

Twig Diameter – The diameter of the branch at its base (50cm from the tip) was measured by taking the average of two perpendicular diameters at this point with a caliper.

Branch Dry Mass -The total weight of the branch was measured by cutting up the branch in smaller segments, drying them for for >72hrs at 60C and weighing them.

Leaf Mass Fraction –The total dry mass of the leaves on each branch was divided by the total dry weight of the branch. See below for details on leaf dry mass measurements.

Leaf Area Ratio –The total area of the leaves on each branch was divided by the total dry weight of the branch. See below for details on leaf area measurements.

Branch Wood Density – Wood density of the branch, including the bark, at a diameter of 5mm. The 12mm long branch segment with a diameter of 5mm was used to measure branch wood density. To measure its volume, a volumetric cylinder with a downspout was filled with soapy water until it overflowed and the meniscus was flush with the bottom of the spout. An empty weighing boat was then placed underneath the spout and the branch was immersed under water for 5 seconds using a needle to submerge the branch. The weight of the water displaced was then recorded in grams using a scale with 3 decimals. This was repeated 3 times and the average of the 3 measurements was taken. The density of the soapy water was calculated and the room temperature was recorded for each measurements in order to adjust the water density to temperature. The branch segments were then rinsed,

244 dried in the oven at 60C for 72hrs or more and their dry mass was measured. The bark was not removed since it tightly adhered to the branch. Branching angle - Average bifurcation angle. Using ImageJ, the average branching angle was calculated in 2014 from pictures of the branches without twigs. The angles were only calculated for the bifurcations that layed perpendicular to the camera (flat against the background).The angle was drawn from the outside of the primary branch to the Picture 3. Example of middle point of the lateral branch (see picture 3). branching angle measurement

Traits measured on leaves Leaf Area – Average area of a leaf. The branches were defoliated by gently bending the petioles of each leaf backward at their point of attachment to the branch. The total number of leaves were counted and then separated into two categories: healthy and mature vs non- healthy or immature. Leaves were considered non-healthy if they had more than ~10% of its area damaged. Each category of leaves were scanned separately using a Li-3100C leaf area meter, with a resolution of 1mm2 . We flatten curled leaves and plucked off the leaflets on compound leaves to avoid leaflet overlap. Measurements included the petiole on simple leaves and the rachis and petiolule on compound leaves. The average leaf area was calculated by dividing the total healthy leaf area by the total number of healthy leaves.

LMA – Leaf dry mass per unit area. The total mass of all healthy and mature leaves and of unhealthy or immature leaves was calculated separately. The total mass of healthy and mature leaves was divided by their total area to calculate leaf mass per area.

Leaf Thickness – Leaf lamina thickness. Using a microtome, one measurement was taken per leaf on 5 randomly selected mature and healthy leaves. Measurements were taken on the leaf

245 lamina, halfway between the base and the tip and 1/3 of the way from the middle vein to the edge. We avoided the mid-vein and secondary veins.

Leaf Tissue Density – Leaf tissue density was calculated by dividing LMA by leaf thickness to provide a mass per volume measurement.

LCC – Leaf Carbon Content. All the healthy leaves of each branch were pooled to constitute a single leaf sample per sapling. The leaf samples were dried in an oven at 60C for at least 72hrs then ground with a Thomas Wiley mini-Mill, taking good care to fully clean the mill between each sample. The petiole and mid-vein were removed from the leaves to measure the chemical composition of the leaf lamina only. The leaves were then sent out to University of Arizona’s Environmental Isotope Lab for determination of chemical composition analysis (LCC, LNC d13C and d15N).

LNC – Leaf Nitrogen Content. Same as for LCC. d15N – Nitrogen isotope ratio (15 vs 14). Same as for LCC d13C – Carbon isotope ratio (13 vs 12). Same as for LCC

LPC – Leaf Phosphorus Content. After the leaf material was ground, leaf phosphorus content was determined in the Enquist lab by following the attached protocol. Three replicates of each sample were measured and additional replicates were added based on the standard deviation and coefficient of variation among replicates of a sample. Further, additional replicates were added when the sample mean was 25% larger than the species mean. See attached document for further details .

C.N – Ratio of leaf carbon content (LCC) to leaf nitrogen content (LNC)

C.P - Ratio of leaf carbon content (LCC) to leaf phosphorus content (LPC)

N.P– Ratio of leaf nitrogen content (LCC) to leaf phosphorus content (LNC)

246 Traits measured on twig cross-sections Branch cross-section preparation To measure the wood anatomy traits, the 10cm branch sections were kept frozen from the day of collection to the day of preparation of the cross-sections. The 5mm end of the 10cm twig was mounted on a microtome and ~100 µm cross-sections were shaved off using a microtome, stained with Safranine to highlight the cell walls and permanently mounted on slides using Permount. See the Wood Anatomy Mounting Protocol for details. 395 slides were prepared, one for each sample. Next, 24bit .tiff color digital pictures were taken at 20x, 50x and 200x using an Olympus BX-50 microscope, an Optronics Microfire microscope camera and the PictureFrame software. See Wood Anatomy Picture-taking Protocol for details. The 20x pictures were used to calculate the % sapwood area of the branch. The 50x pictures were used to analyze in STEM_GUI and calculate Vessel Diameter (VD), Lumen Area (%Lum), conductivity per sapwood (Ks) per lumen(Kl) and per cross-sectional area (Ktwig). Given heretoreneity of the wood in the cross-section, for each sample 3 different areas of the cross- section were used as replicates and a 50x picture. Within each 50x field of view, two 200x pictures were taken to calculate t/b2. A total of 3950 pictures were taken, 10 for each sample (1 at 20x, 3 at 50x and 6 at 200x).

KS – Conductivity per sapwood area. The 50x images were processed using the STEM_GUI image-analysis software to select the conduits and calculate wood properties from conduit properties. The software automatically does an original selection of the conduits based on parameters, but requires extensive manual corrections to be accurate. See the STEM_GUI protocol for more details. The output summary sheet of the STEM_GUI includes the total conductivity as well as the processed area. Conductivity per sapwood area was calculated by dividing the total conductivity by the total processed area.

247 %Lumen – Lumen fraction. The output summary sheet of the STEM_GUI includes the total lumen area. The lumen fraction is calculated as the ratio of the lumen area to the total sapwood area.

KL – Conductivity per lumen area. The output summary sheet of the STEM_GUI includes the total conductivity as well as the lumen area. Conductivity per sapwood area was calculated by dividing the total conductivity by the total lumen area.

VD –Vessel Diameter. The average vessel diameter of each conduit is calculated by the software and reported in a summary sheet. In order to exclude tracheids from vessel diameter averages, all output files were sorted to exclude any conduit with an area of 490micron2. This corresponds to a circular conduit of 25microns (see Sperry 2006). See Vessel Diameter Protocol for more details.

% sapwood – Sapwood fraction. The fraction of the cross-sectional area that is conducting sap was calculated from the 20x pictures using ImageJ For diffuse-porous species, all the xylem area was calculated as sapwood (i.e. the pith and primary xylem were excluded from the cross-sectional area). For ring-porous species, only the last year’s xylem area was calculated as sapwood area.

KTWIG - Conductivity per cross-sectional area. The conductivity per sapwood area (KS) was multiplied by the fraction of the cross-section that actively conducts sap (%sapwood)

(t/d)2 – Thickness to diameter ratio of vessels. This reflects the ratio of the cell wall thickness of two adjacent vessels of similar diameter to the average vessel diameter. The 200x pictures were used to calculate the ratio of vessel diameter to cell thickness from vessel pairs of similar diameters. Twelve measurements were calculated for each sample, except for some species where the vessels grow in a single fashion (surrounded by tracheids) and pairs are difficult to find. See Thickness to diameter ratio Protocol.

248 Traits measured on stems At the end of summer 2012, each sapling was cut at the base of the tree, above the bulge resulting from the root-stem junction. A 2-3 inch segment was kept to calculate annual biomass growth and tree age.

Stem Wood Density – The wood density of the stem. The stem segments were cut in half and the upper half was used to calculate stem wood density. Since part or all of the bark had come off on some samples, but none on others, the bark was sanded off from the stem sections in order to standardize the absence of bark across samples. Stem volume was calculated using immersion method (see Cornelissen et al. 2003 ) and the stem sections were dried in an oven for at least 72hrs at 60C and weighed. The mass of each stem was divided by its volume to give stem wood density.

Tree ring methods The bottom half of the stem segment was sanded in order to accurately read the tree rings on it. It was sanded with a belt sander down to 240 grit then manually with 600 grit microsanding paper.

The rings of most species are much too pale to be r2 r3 distinguished visually from a scan or picture so that the BAI was calculated from 4 perpendicular radii measured using a r4 binocular and a measuring system. A Velmex TA UniSlide r1 measurement system with an accuracy of 0.001mm (combining a TA 4030H1-S6 Unislide stage assembly, a ENCINDL1-24 Encoder, a VRO Digital Readout, and a TAB2 remote. The measurement system is located at the Picture 4. Location of the four radii used to calculate BAI on tree ring laboratory of the university of Arizona. a stem cross section for the year 2011.

249 The growth pattern of some trees around the pith was asymmetrical, leaving the pith off-center and the basal area increment concentrated in an area of the cross-section instead of being spread evenly around the stem. This leads to an underestimation of the basal area increment when the geometric average of the 4 radii is used ([r1+r2+r3+r4]/4) to calculate the area of a circle of equivalent radius. To correct for that bias, 3 methods of measuring the basal area from the radii were tested. Twelve stems with strongly asymmetrical piths and clear rings were scanned and processed in ImageJ. For each stem, the basal area increment of the last complete year (year 2011) was measured using the polygon tool. Then, 4 perpendicular radii were measured from the pith to the end of the year 2010 and of the year 2011 (see Picture 4). The first radius was located to go through the longest growth area and the other 3 radius were located 90 degrees from it in a clockwise fashion. The BAI of the year 2011 was calculated by subtracting the 2010 basal area (comprising the pith to the end of the year 2010) from the 2011 basal area (comprising the pith to the end of the year 2011). The 2010 and 2011 basal areas were calculated using 3 different methods: 1) calculating the area of a circle from the quadratic mean of the radii ((r1+r2+r3+r4)/4), 2) calculating the area of a circle from the geometric mean of the radii (sqrt([r12+r22+r32+r42]/4])) and 3) calculating the area of ellipse by summing up the 4 ellipse quarters formed by each adjacent radii pair. The area of an ellipse is π*r1*r2, so that the formula used was (r1*r2* π /4 + r2*r3* π /4 + r3*r4* π /4 + r4*r2* π /4). The true BAI was then compared with the BAI estimated using each of these three methods. The geometric method produced estimated BAI measurements ranging from 1.3 to 28% of the true BAI, with an average of 8.9% from the true BAI. The quadratic method produced estimated BAI ranging from 0.2 to 16.4% of the true BAI, with an average of 5.3% from the true BAI. The ellipse method gave produced estimated BAI ranging from 0.7 to 32% of the true BAI, with an average of 9.3% of the true BAI. Based on these results, all remaining BAIs were measured using the quadratic mean method.

Biomass5 – The basal area increment (BAI) of the last 5 complete years (2007 to 2011) were measured and multiplied by the stem’s wood density to get a measure of growth in g/cm.

250 Biomass1 - The basal area increment (BAI) of the last complete years (2011) was measured and multiplied by the stem’s wood density to get a measure of annual growth in g/cm.

Tree Age –Using a dissecting scope, the age was counted by two separate individuals and reconciled when there were differences.

Traits measured on roots SRL – Specific root length. In 2012, root samples were taken on 200 of the 395 saplings. Root growth bags were installed before leaf flush (from April 23rdth to May

11th) and harvested 3 months later in the same order as they were installed (from July

16th to 30th). Since the bags need to be installed after the soil has thawed but before leaf flush, this we only had the time to install

200 bags, or approximately 8 per species. For Picture 5. Installation of root growth bags. each tree, a coarse root of 5-10mm in diameter was found by following a main root from the trunk. It was then cut diagonally with a hand pruner and inserted in a root growth bag. The 20cmx25cm root growth bags were made from landscape fabric and stitched with nylon thread. The bags were filled with a mixture of the local soil and sand (25-30% by volume). The bags were then watered

Picture 6. Cleaning harvested roots. and re-covered with soil and litter (see picture 5). Cutting the coarse roots of

251 this size before leaf flush stimulates growth of fine absorptive roots. The landscape fabric minimizes the growth of tree roots outside of the bag and the growth of external roots into the bag while allowing water and nutrient circulation. Three months later, the root bags were harvested by cutting the coarse root at the entrance of the bag. The roots bags were then cut open and the roots gently washed free of soil and sand with a hose gun. (See Picture 6).The fine roots were clipped off from the coarse root and placed in a water bath under a magnifying glass. A selected subset of the 1st and 2nd order fine roots was harvested. These were then placed in scintillation vials and left overnight to dye in a dilution of Neutral Red (0.16 g/L) They were then rinsed twice in water. The staining procedure insures that the roots are clearly visible when scanning (see Picture 7). The roots were then scanned using the WinRhizo software and scanning system and total root length was calculated for the samples. They were then dried for >72hrs at 60C and weighed. The specific root length was then calculated by dividing the total mass of each root sample by its total length. Of the 200 root bags

st installed, 159 produced 1st Picture 7. Selection, clipping, staining and scanning of 1 and 2nd order fine roots.

252 and 2nd order absorptive roots. One species has only 3 fine-root samples (Populus grandifolia), four species have 4 fine-root samples (Amelanchier sp., Populus deltoides, Prunus pensylvanica and Tilia americana) and all other species have 5 to 8 fine-root samples.

Root wood density – Wood density of coarse roots. The coarse roots from which the fine roots were harvested were further washed and frozen. In 2014, they were air-dried and their volume was weighed twice using the immersion method described in Cornelissen et al.(2004). A third volume measurement was taken when the C.V. of the first two volumes was larger than 1%. The samples were left to air-dry between the volume measurements. The roots were then dried at 60C for >72hrs and weighed twice. A third mass was taken when the C.V. of the first two volumes was larger than 1%. The samples were re-dried at 60C for >72hrs between measurements.

Species-level traits collected from the literature Hmax – The maximum height that a species can reach was noted from the Trees In Canada textbook (Farrar. 2000. Trees in Canada. Fitzhenry & Whiteside Limited. 502pp.)

Tolerance Indices - Each species’ shade, drought and waterlogging tolerance indices was noted from Niinemets & Valladares. (2006).

Environmental variables collected Symmetric Competition – Symmetric competition is defined here as competition from trees of similar sizes to the focal trees. Here, this includes trees from 1-10cm dbh. The strength of symmetric competition from close neighbors was estimated using the point quarter method. This method consists of sampling the individual closest to the focal tree in each of four quadrants delineated by the cardinal points. The identity, dbh and distance to the focal tree is noted for each closest competitor in each quadrant. The average distance, daverage, to the

253 focal tree and average dbh, DBHaverage, I calculated from the four neighbors and a basal area index is calculated using this formula: Basal Area Index =π * (DBHaverage/2)2/daverage2.

Asymmetric Competition - Asymmetric competition is defined here as competition from trees significantly larger than the focal trees. Here, this includes trees from larger than 10cm dbh. The strength of asymmetric competition was estimated using the same method as above. Soil Humidity - Soil humidity was measured for each tree using a ML2x Theta Probe soil moisture sensor. All measurements were taken over a period of 36 consecutive hours following 10 days without rain. For each tree, 3 measurements were taken on the top soil, after brushing off the leaf litter. The measurement points were 1 foot away from the tree in three opposite directions. For each tree, the average soil humidity of the three points was calculated.

Slope – A mesoscale slope was measured over a 20m scale. A piece flagging tape was installed at eye level 10m upslope of the focal tree and the user walked 10m downslope from the focal tree to measure the angle to the flagging tape using a clinometer. This scale was used to reflect the slope on drainage and lighting.

North Aspect – The aspect of the slope was taken using a compass and the degree to which it was facing North was measured by taking the cosine of the angle. A value of -1 reflects a slope facing due South, a value of 1 reflects a slope facing due North and a value of 0 reflects a slope facing due West or East. The sites that are on flat ground but have are located at the base of the mountain were given an aspect equivalent to the direction of the mountain to account for the effect of shading form the mountain. Sites that were flat and away from the mountain were given an aspect of 180 to reflect full sun exposure.

East Aspect - The aspect of the slope was taken using a compass and the degree to which it was facing East was measured by taking the sine of the angle. A value of -1 reflects a slope facing due West, a value of 1 reflects a slope facing due East and a value of 0 reflects a slope

254 facing due North or South. The sites that are on flat ground but have are located at the base of the mountain were given an aspect equivalent to the direction of the mountain to account for the effect of shading form the mountain. Sites that were flat and away from the mountain were given an aspect of 180 to reflect full sun exposure.

Elevation – The elevation of neighboring focal trees was measured using a GPS. Data was recorded for 1 minute when the accuracy of measurements were within 6m and data averaging was used to get an accurate location from the average of the data collected over the minute.

% Canopy Opening – To characterize the light environment of the study saplings, hemispherical photos were taken at 1m from the ground at dawn and dusk or during fully overcast days. A Nikon Fisheye lense and a Nikon CoolPix 4500 Digital camera. The lens was levelled with a bullseye level and North was marked on the picture using masking tape on the lens. Percent gap opening was calculated from these images using the software Gap Light Analyzer V2 software. See the file “GLA protocol” for more details

Total Transmitted Light – The total transmitted light was calculated from the hemispherical images using Gap Light Analyzer. To calculate these values, the latitude and longitude of the site, the slope, aspect and elevation of each tree and the location of North on the picture were provided.

Soil Depth. Soil depth around the trees was measured by taking the largest of three measurements taken with a 52cm soil probe ca. 30cm around the tree. 52cm was recorded for depths of 52cm or more.

Air temperature. In June 2014, Thermocron iButtons (model DS1921G-F50, accuracy ±1°C from -30°C to +70°C) were installed in 23 general areas around which the trees were located as well as next to 4 weather stations (See Picture 10). The iButtons at the weather stations were installed in order to assess the accuracy of the iButton measurements. In each area,

255 three iButtons were installed at a height of ca.1.7m on the North- facing side of three dominant neighbooring trees. The iButtons were placed in small opaque plastic containers in order to create shading (Picture 8). The iButtons were Picutre 8. iButtons installed on the North side of the calibrated to record every 4 hours. trunk of dominant trees in small plastic containers. The setup minimizes direct sunlight. They were removed in early December and the daily maximum and minimum temperatures of each area was calculated. The monthly minimum, maximum and average temperatures were then calculated for each area.

Soil sampling design for soil chemistry Four samples from the top 15cm of soil were taken from around each study tree. The four samples were located North, South, East and West of the tree at increasing distance. Since larger trees have larger root zones, the distances from the trees were increased proportionally to the dbh of the focal trees (Picture 8). Trees with a 1cm dbh had soil samples taken at 1, 2, 3 and 4 feet from the trunk, trees with a 2cm dbh had soil samples taken at 1.5, 3.0, 4.5 and 6.0 feet from the trunk, trees with a 3cm dbh had soil samples taken at 2.0, 4.0, 6.0 and 8.0 feet from the trunk, trees with a 4cm dbh had soil samples taken from 2.5, 5, 7.5 and 10 feet from the trunk and tress with a 5cm dbh had soil samples taken from 3, 6, 9 and 12 feet from the trunk. The volume of the soil samples samples were of ca.6cm by 6cm by 15cm (Picture 9). All 4 soil samples were mixed together and left to dry in a drying oven for >48hrs at 30°C. They were then sieved through a 100µm mesh and 30g samples of thoroughly mixed soil were sent to the “laboratoire de chimie et de pédologie” of Université Laval for chemical composition analysis.

256 Picture 9. Top soil collection

Total Nitrogen. Total nitrogen in the soil he Kjeldahl method was used to measure total nitrogen available in the soil (Bremner 1960).

pH. Hydrogen concentration was determined in H2O,

Total organic Carbon. Total organic Carbon determined by loss on ignition (Davies 1974).

Exchangeable ions. The concentration of exchangeable ions (P, K, Ca Mg et Na) was determined using the Mehlich-3 method (Tran et al. 1990).

257 Courtesy of David Maneli

Picture 10. Location of iButton sensors at each of the sampling areas

258 PHOSPHORUS PROTOCOL

Prepared by Joy Shifflette, January 5, 2006, Modification from APHA 1992 & Pete Gaube’s 2002 protocol Updated Feb. 2009 by Vanessa Buzzard

Water Purification System:

Part One:

1. Push operate button to turn on system. 2. Make sure system reads at a resistivity of 18.2. If resistivity drops < 18.2 see pg. 29 of Operating and Maintenance Manual (a resistivity of 18.1 or 17.9 should be ok to continue using system. 1

Reagents:

Part One:

1. Check expiration dates on all chemicals before starting Phosphorus assay. 2 2. Make up the following chemicals if not already made: A) Dilute H2S04- (Wear goggles, lab coat and gloves) 3. Fill 500ml graduated cylinder with 500ml dd H20 and pour into 1000 ml glass container labeled dilute H2S04 4. Fill same 500ml graduated cylinder with 200ml dd H20 and pour into same 1000 ml glass container labeled dilute H2S04 5. Pour 300m1 concentrated sulfuric acid into a different 500ml graduated cylinder. 6. Slowly add 300m1 concentrated H2 SO4 into 700ml dd H20 for 1liter total volume. 7. Loosely cap and allow container to cool down some. 8. Once reagent has cooled down some tightly cap and mix. Solution can be used while warm. 9. Place label tape onto container with your initials and the date made. 10. Store under fume hood. B) 5M sulfuric acid- (Wear goggles, lab coat and gloves) 11. Fill 500ml graduated cylinder with 500m1 dd H20 and pour into 1000 ml glass container labeled 5M sulfuric acid(Repeat step a using the same 500ml graduated cylinder). 12. Pour 140ml concentrated H2 SO4 into a different 500ml graduated cylinder.

259 13. Slowly add 140m1 concentrated H2 SO4 into the 1000ml dd H20. 14. Loosely cap and allow container to cool down some. 15. Once reagent has cooled down some tightly cap and mix. Solution can be used while warm. 16. Place label tape onto container with your initials and the date made. 17. Store under fume hood. C) Potassium Antimony Titrate 18. Fill 500ml graduated cylinder with 500 ml dd H20 and pour into 1000 ml glass container labeled Potassium Antimony Titrate (Repeat step a using the same 500ml graduated cylinder). 19. Weigh out 2.743g of Antimony Potassium Tartrate Trihydrate powder and add to 1000ml dd H20. 20. Tightly cap and mix. 21. Place label tape onto container with your initials and the date made. 22. Store in refrigerator.

Part Two:

The following reagents are normally purchased pre-made from VWR but can be made up if necessary:

A) Phenolphthalein Indicator 0.5% w/v Alcoholic - 1g/l00ml 100% ethanol B) Sodium Hydroxide 2.000N- dissolve 8g into l00ml total volume in dd H20 C) Phosphate Standard- 0.050mg/ml D) Ammonium Molybdate Solution 40g/L- 40gm/1000ml dd H20

Sample preparation:

Part One: (Grinding plant samples)

23. Dry samples in oven at 50- 60 Celsius for a minimum of 72 hours before homogenizing and grounding to a fine powder. 24. Collect liquid nitrogen from Life Sciences south (rm. 218). 25. Use dd H20 to clean dust from mortar and pestle, wipe with paper towel and allow to air dry. 26. Place sample in mortar. 27. Cover sample with liquid nitrogen. 28. Allow liquid nitrogen to almost fully evaporate and grind sample with pestle. If sample will not grind to a fine powder, repeat steps 5 & 6. If sample will still not

260 grind to a fine powder, complete steps 7 – 10 and place red label tape on the cap of the 8-dram vial. Re-grind at a later date. 29. Scrape off as much sample as possible from mortar and pestle using the non-sticky side of sticky note. 30. Transfer as much sample as possible into a clean 8 dram vial using weighing paper.3 31. Place label tape onto vial and label with permanent marker. 32. Place vials into vial rack labeled (location name – your name). 33. Clean mortar and pestle with Millipore H20 between samples.4 34. Use a new sticky note and clean weighing paper for each new sample to avoid cross contamination. 35. Place caps loosely onto 8-dram vials and put in drying oven set at 50 to 60 Celsius for 24 hours. 36. Put date and your initials on check off sheet (clip board) for all samples ground and record in lab your notebook.

Procedure: (20ml preparation)

Part One: (Weighing and labeling – Takes approx. 2 1/2 hours for 24 samples & 6 controls)

37. Make a piece of Al foil to weigh samples on if one is not already made. 38. Turn microbalance on, make sure surface of scale is clean and that the scale is level (bubble should be inside the circle). 39. Tare microbalance without anything on the scale. 40. Enter your name, the location name, and start date into the P-Spreadsheet Template. xls and save as (location name)_(your name and a letter for that particular assay).xls 41. Enter sample ID information under the sample ID column in the P-assay data sheet. 42. Gather four vial racks. Make sure that one of the racks is labeled standard curve. 43. Place label tape across the front of four racks and label with permanent marker (location name-your name and a letter for that particular assay). 44. Using tweezers place Al foil piece on scale, weigh out between 1.5mg to 2.0mg of sample using micro sample spoon and place onto foil, let set for 60 seconds. 6 45. Record mass of foil and sample onto data sheet under net mass in micro-grams omitting decimal. 46. Using tweezers place sample into new 8-dram vial. 47. Place Al foil back onto scale and record the mass onto data sheet under residual mass in micro-grams omitting decimal. 48. Repeat steps 22-25 (Weighing out 3 sub samples per sample). 49. Label the row of the rack for the 3 sub samples.

261 50. Weigh between 1.5mg to 2.0mg of NIST 11 8437- hard red spring wheat powder (0.137% P) on Al foil piece. 7 51. Record mass of foil and sample of NIST 11 8437- hard red spring wheat powder (0.137% P) onto data sheet under net mass in micro-grams omitting decimal. 52. Put into new 8-dram vial. 53. Place Al foil back onto scale and record weight onto data sheet under residual mass in micro-grams omitting decimal. 54. Make 6 controls for every 24 samples (or 3 controls for every 15 samples) and place into rack labeled standard curve. 55. Place label tape along the side of the standard curve rack that is not going to be used for the standard curve and label A,B,C, D, E, F for the controls. Note: If there is enough racks controls can place in their own rack.

Part Two: (Standard Curve and preparation for autoclaving) 56. Sign up for the autoclave upstairs (Rm. 338). 8 57. Place 12 8-dram vials into rack labeled standard curve and labeled with concentrations 0,2,4,8,12 and 16. 58. Measure 7mL of ddH2O. Pour into 125mL glass container. Measure 3mL of new (.163mg Phosphate) P-Stock and add it to the 7mL ddH2O. This is the new Standard Curve dilution. 59. Make two standards per assay (one for trial A and one for trial B) using the table below. Add l of dd H20 first then l of p-stock. standard curve Concentration Final concentration l of P-stock l of dd H20 (M P) (G/ml) 0 0 0 1000 2 0.061 25 975 4 0.122 50 950 8 0.242 99 901 12 0.364 149 851 16 0.484 198 802

60. Add 1000l dd H2O to all samples and controls. Don’t add 1000 l dd H2O to standard curve. Samples can sit for a few days after this step. 61. Measure 750ml of dd H2O (enough for 24 samples, controls and standard curve, 90 vials total) or450ml dd H20 (enough for 12 samples, controls and standard curve, 54 vials total). 62. Pour into 1000 ml glass container labeled Persulfate Solution, loosely cap and place into water bath set to 60 ºC 63. Heat dd H20 in water bath while completing steps 41 and 42.

262 64. Add 400l of dilute Sulfuric acid to all samples, controls and standard curve (Wear gloves). 65. Mix samples, controls and standards. 66. Weigh out 35.5g of potassium persulfate for 24 samples, controls and standard curve, 90 vials total or weigh out 21.3g of potassium persulfate for 12 samples, controls and standard curve, 54 vials total. 67. Pour persulfate salt into heated dd H2O to make 5% Potassium Persulfate solution, cap, invert and swirl until completely dissolved. Use immediately while solution is still warm (wear gloves). 68. Using 10 ml pipet to add 8ml persulfate solution to samples, controls and standard curve. 69. Loosely cap samples, controls and standard curve and place four racks in a metal autoclave tray. 70. Place tray in autoclave immediately.9 Make sure door is tightly closed hit cycle one then cycle one again to start. This cycle takes about 40 mins. to complete. Sterilization time is 30mins at 249.98 F. After completion open door 1 inch and 3 centimeters, allow samples to cool down for 10 mins. Remove samples from autoclave immediately after the 10 min. cool down since samples will evaporate if left in too long. 71. Allow samples to cool to room temperature before starting step 10.

Part Three: (neutralization)

72. Add 20l phenolphthalein to all samples, controls and standard curve. 11 73. Measure the volume of 6 random samples using a new 10ml pipet for each sample. 74. Record the total volume of 6 random samples, average and record the average as MEAN A on raw data worksheet. 75. Pour some 2.000 N NaOH into a flask. 76. Select 6 random samples to titrate that are different from the ones selected above. 77. Use 5ml pipet fill up to zero mark with NaOH and titrate sample until it turns pink (Repeat 5 times). 78. Record the amount of 2.000 N NaOH used to titrate 6 samples, average and record on worksheet as MEAN B. 79. Add the average amount of 2.000 N NaOH to the rest of the samples. Note: If a sample does not turn pink it is ok. 80. Sum MEAN A and MEAN B and record on worksheet as TOTAL A. 81. Subtract TOTAL A from 20m1, record on worksheets as TOTAL B. 82. Sum TOTAL A and TOTAL B, should equal 20.0ml. 83. Add the amount of TOTAL B to each sample using dd H20. 84. Add the raw data worksheet to the raw data binder. 85. Put clean vial caps on samples tightly.

263 86. Wash used caps with tap water followed by dd H20 rinse and allow to air dry under hood.

Part Four: (Phosphate Reagent and reading absorbances)

A) Absorbic Acid 1. 1. Using 500ml graduated cylinder measure 120m1 dd H20 pour into Erlenmeyer flask. 2. 2. Weigh out 2.11g of absorbic acid power and pour into 120m1 dd H20. 3. 3. Swirl flask to mix and let sit for a few minutes while starting to make Phosphate Reagent. B) Phosphate Reagent 1. Pour chemicals in the same order as seen below into 1000ml glass container labeled Phosphate Reagent. a. 5M sulfuric acid 200ml b. Potassium Antimony Titrate 20ml c. Ammonium Molybdate 12 60ml d. Ascorbic acid 13 120ml 400ml total 2. 400 ml is enough for 35 samples, controls and standard curve, 123 vials total (200 ml enough for 14 samples, controls and standard curve, 60 vials total). 3. If mixed correctly Phosphate reagent should be light yellow.

Steps 65 -76 needs to be completed within 4hrs. 87. Add 3.2ml of Phosphate Reagent using 5ml pipet to standard curve first, then to samples and controls. 88. Cap vials tightly and invert. 89. Allow to develop for 10-30 minutes before reading absorbances. 90. Turn on Spec and make sure it is set at 880nm. 91. Use dd H20 to rinse cuvet 2-3 times. 92. Fill up cuvet with dd H20, put into spec. and hit the O ABS 100% T button (this will set the blank). 93. Empty the dd H20 in the cuvet into the absorbance waste bucket and tap cuvet upside down onto a stack of paper towels. Note: Rinse cuvet with dd H20 2-3 times then tap upside down onto stack of paper towels in between reading each new absorbance value. Invert vial of sample right before reading absorbance. 94. Read the absorbances for trial one of the standard curve and enter the absorbances into the standard curve table under absorbance. Save the blank (zero concentration) in case diluting samples is necessary. 95. Enter the absorbances and the R^2 from the Standard curve table into the

264 Standard curve absorbances table under trial one. 96. Read the absorbances for trial two of the Standard curve and enter the absorbances into the Standard Curve table under absorbance. Save the blank (zero concentration) in case samples need to be diluted. 97. Enter the absorbances and the R^2 from the standard curve table into the standard curve absorbances table under trial two. 98. Enter the absorbances of the trial with the R^2 closest to 100% into the standard curve table under absorbance.

Procedure for excessive absorbances: (threshold for low absorbance. ~ 0.02, for high absorbance ~0.325 Beer’s Law)

a. If an absorbance > 0.325 is read for a sub-sample save the sample. The sample will need to be diluted. i. Get a new 8-dram vial and 5ml pipet. ii. To dilute the sample by half pipet 5ml of the blank (zero absorbance) into new 8-dram vial and then pipet 5ml of the sample into the same 8-dram vial that contains the 5ml blank. iii. Cap and mix. iv. Read new absorbance and enter new absorbance in place of old absorbance under sample absorbance column. Record old absorbance in notebook. v. Change 20ml under the volume of sample ml to 40ml (doubling the 20 ml volume).

99. Read absorbances of samples and enter in data sheet under sample absorbance column. 100. Read absorbances of controls and enter into data sheet under sample absorbance column. (This region is highlighted in yellow at the end of the data sheet). 101. Enter the finish date and total number of samples into data sheet and save. 102. Coefficient of Variance range – make sure CV < or = 0.09, consult Brian if CV > 0.06. 103. Creae pivot table and print up hard copy. 104. Dump all waste in waste bucket in fume hood.

Cleaning Glassware: (Don’t use soap) 105. Rinse vials and caps with tap water, fill vials with dd H20, cap and store under hood. 106. Prior to starting new assay, empty vials, rinse with dd H20, place in racks

265 upside down to dry & place caps under hood in tray to air dry.

Reference Material: NIST 11 8437- hard red spring wheat powder (0.137% P) Markow, T. A. et al. Elemental stochiometry of Drosophila and their hosts. Functional Ecology 13 (1999).

Notes: 1 - Our Synergy Water Purification System doesn’t have a UV lamp 2 - Write the date received on all chemicals. 3- All glassware and plastic-ware were treated in 20%v/v concentrated HCL in dd H20 for at least two hours followed by a rinse of dd H20. All standards are run through same procedure as above. 4- If mortar and pestle doesn’t come clean with dd H20 use an ethanol rinse followed by a good rinse of dd H20 to make sure the carbon from the alcohol is removed. - Ethyl Alcohol USP Absolute – 200 Proof purchased from UA Stores. 5- If mortar and pestle doesn’t come clean with dd H20 use an ethanol rinse followed by a good rinse of dd H20 to make sure the carbon from the alcohol is removed. 6- Take samples out of the drying oven one at a time prior to weighing. 7- Reference material is dried (50-60C) and used fresh out of the oven. 8 - Use P4 cycle for 120 minutes. 9 - Persulfate breaks down over time. 10 - Before and after neutralization samples are stable enough to sit for days or weeks and the standards can sit for weeks/possibly months. 11 - Samples shouldn’t sit after adding Phenolphthalein (it doesn’t last very long) & needs to be added right before neutralization steps. 12- If solution starts forming a snowy white precipitate, it has gone bad and needs to be remade or repurchased. Store under hood at 4 C. 13- Make absorbic acid fresh, allow to cool to room temp and use on same day as made. It goes bad quickly.

266 Phosphorus Redo Decision tree

1 - Assess calibration curve validity

Corrects for suspiciously high or low values of entire batches

Look at the straightness of the calibration curve If calibration curve looks good (straight) -> Use corrected values If calibration curve looks bad -> Use "idealized" calibration curve - If standard value gets better (closer to 13.7) -> Use corrected values with that new curve (suggests there was a problem with measuring the standards to create calibration curve) - If standard value gets worse (further from 13.7) -> Redo the whole batch (suggests something went wrong somewhere along the processing of the whole batch)

2 - Replicate Outliers

Corrects for high variability among replicates of the same sample

Look at standard deviation (St.Dev) and coefficient of variation (C.V) of replicates from sample mean If St.Dev [2;5%] and C.V. <15% -> add one more replicate If St.Dev [2;5%] and C.V. [15;30] -> add two more replicates If St.Dev is >5% OR C.V. > 40% -> add three more replicates

3 - Assess Individual Outliers

Corrects for unreasonably high values of individuals relative to the species (*** if one expects constant values within a species***. If one sampled leaves from different environments, it might be normal to observe different values – ex. along a light or soil Phosphorus gradient)

Look at standard deviation of individuals from species mean Calculate the species mean trait value using values fixed in step 1 and ignoring the replicates that are obviously wrong (in step 2)

267 Calculate the “coefficient of variation” of each sample mean to the species mean ((value- mean)/mean). If > 17% [ P] and 25% above species average -> add two replicates for that sample - If [P] values of new replicates are similar to original [P] values -> Use average of all 5 replicates for [P] value of sample - If P values of new replicates are not similar to original P values -> Use average of the two new replicates for [P] value of sample

268 WOOD ANATOMY PROTOCOLS

1 - CROSS-SECTION MOUNTING PROTOCOL Safranin & Lugol Staining

Prepared by Julie Messier Updated Oct. 24th 2013

Organizing 0. If some slides have been drying for 48 hours, put them in the slide box and fill in the index sheet . Clean the glasses if dirty. Use a scrapper to remove any dried Permount on the glasses.

Slicing 1. Take samples out of the freezer (as much as possible, leave all samples in the freezer) 2. Write the names of the samples on one 1.5 inch wooden sticks with a pencil and cover the writing with PVC cement - Work on a piece of cardboard or brown paper towel under the fume hood if space allows - Wear gloves - Let it dry about 5mins 3. Measure the diameter of both ends of each sample with a caliper and record it on datasheet (if necessary) 4. Mount twig on microtome - Cut the twig to about 4.5 cm in length from the end with the dot (if necessary) so that it fits into the microtome holder. - The end with the sharpie dot is the one to cut - Make sure the twig is vertical (i.e. the slice is perpendicular to the long axis of the twig) - Don’t deform the twig by clamping it down too hard - Let the end of the twig stick out of the stage about 5mm or less 5. Adjust the stage with the coarse adjustment handle so that the twig touches the bottom of the blade 6. Flatten out the surface of the wood using the fine increment lever - Remove the wood material using a paintbrush or tweezers, not with your fingers

269 7. Once the wood surface is flat, wet the top of the twig by touching it with a wet finger and make slices of 90-120 µm (9-12 increments of 10 µm each), depending on the species. Shoot for slices of ~100 µm. Stay consistent within a species - Take 5 slices of each sample and put them in a pierced lid - Add the corresponding label into the lids - Record the thickness of the slices you are taking for the species - If it is difficult to get a good slice a) make sure it is not cracked – if so, cut the sample below the crack b) try soaking the twig in water for 5-15 minutes prior to taking slices 8. Let the cut slides sit in water so they do not desiccate while cutting the other samples.

Staining (in Fume hood!) Safranin 8. Pipette 30 ml of 0.25 % Safranin stain in the corresponding jar and 30ml of ethanol destain in the corresponding petri dishes (if necessary)

9. Pipette 30ml of Lugol stain in the corresponding jar and 30ml of deionized H2O in the corresponding petri dish (if necessary) 10. Put the lids with the samples into the Safranin stain and leave in for 25 mins 11. After 25 minutes, quickly place each lid on a dry paper towel to absorb excess Safranin stain 12. Transfer the Safranin samples into the ethanol destain and leave in for 2 mins (time will vary depending on species). 13. For Safranin, pipette 1) 20ml of 50/50% ethanol/toluene and 2) 20 ml of 100% toluene solutions into corresponding jars and place them in the fume hood. 14. Remove the samples from the destain solution and let them sit on a paper towel while preparing for mounting. Lugol (stain & mount samples 6 at a time max) 8a. For Lugol, pipette 1) 20ml of 50%/50% ethanol/ D.O. water, 2) 20ml of ethanol, 3) 20 ml of 50%/50% ethanol/toluene and 4) 100% toluene into corresponding jars. 9a. swirl a lid in the Lugol stain for 60 seconds 10a. place the lid on a dry paper towel to absorb excess Lugol stain 11a. swirl the lid in the water destain for 10 seconds 12a. place the lid on a paper towel while staining the other samples

Mounting

270 15. In pencil, write the sample ID, the stain name (Safra or Lugol/IKI) and your initials on one slide. Safranin 16. In the fume hood, place all the lids into 50/50 % ethanol/toluene solution. Let the prepared samples sit in this batch while mounting. 17. Swirl a lid into 100% toluene solution for 10 seconds and put it down on a paper towel Lugol 16a. In the fume hood, place all lids in jar 1 ( 50/50 Water/Ethanol) for 30 seconds 17a. Transfer all lids in jar 2 (100% Ethanol) for 30 seconds 18a. Transfer all the lids in jar 3 (50/50 Ethanol/Toluene) and gently swirl for 2minutes 19a. Transfer all the lids into jar 4 (100% Toluene) and let them sit there while mounting each sample.

18. With the designated (blue) automatic pipette, or the green pipette pump and a 5ml pipette, spread a small amount of Permount onto the slide in an area large enough to mount all your slices. 19. If necessary, gently remove any loose bark around the slice with tweezers and place the sample onto the Permount 20. Put a cover glass over the Permount by touching one side of the cover glass to the bottom of the slide and gently hinging it down towards the top. Keep the slide tilted forward as you do so in order for the extra permount to pool at the top of the slide. 21. Gently press down with your finger and gently wipe off, from the middle outward, any excess of Permount that may spill out from under the cover slip. Repeat steps 18-21 for all samples

22. When all samples from batch are mounted onto the slides, place the slides between four layers of paper on each side. Place the papers between two glasses and place the glass containers filled with water on top of the sheets. 23. Write the date on the top piece of paper and let the samples sit for 48 hrs or more.

Cleaning 24. Clean a. Vial lids b. Petri dishes – discard material in waste container in the fume hood c. Fume hood surface d. Microtome – remove wooden pieces

271 Wood Anatomy Picture-taking Protocol

General Info Location David Killick’s office: ______David Killick’s Lab Office phone ______Haury 411 - Anthropology dept. Cell phone #: ______Door Code :______e.mail: ______Building closes at around 7pm Lab phone #: ______Computer ID: IGERT PW: ______

Google Calendar Address : ______PW: ______Do make reservations Do make cancellations if necessary. Email the Killick lab if you are cancelling at the last minute so that people can take advantage of the free microscope time.

1. OLYMPUS BX51 MICROSCOPE

a. Turning on i. Turn on microscope light on the right side on the microscope 1. Turn light all the way up using the knob ii. Turn on microscope camera by turning on black box behind phone iii. Mount slide holder 1. Lower stage all the way down 2. Swivel the lens holder to the shortest lens. 3. Align screw (at 50 degrees) and use microscope screwdriver to gently tighten the screw IMG_1

b. Check default settings i. Light arm – halfway out: ½ of light to eyepieces ½ to camera IMG_2 ii. Polarized light filter: out IMG_3

272 iii. Bertrand lens arm: out IMG_4 iv. Black cards = Analyzer lenses -> out IMG_5 v. Disk: at position 4, 5 or 6 IMG_6 vi. Shutter: on or off - doesn’t matter IMG_7 vii. 530 nm light filter: on or off - doesn’t matter IMG_8 viii. Light filters: LBD – in . All the others: out. IMG_9 ix. AS/FS knobs: on or off - doesn’t matter IMG_10 x. Focus slider: doesn’t matter IMG_11 c. Adjust the condenser (insures light intensity is even across the image) i. Make sure there is a slide on the microscope 1. Leave the slider of the condenser to the left (past 0.8) IMG_12 ii. Close the diaphragm leaf ¾ of the way IMG_13 iii. Swivel up the second condenser IMG_14 iv. Sharpen the octagon shape of the light by moving it up and down IMG_14 1. The best is in between the blue and red fringe v. Center the light using the two condenser screws IMG_12 1. Look at cross hair only with your right eye vi. Swivel down the second condenser vii. Open leaf all the way d. Adjust eyepieces – do it by focusing your eyes at infinity (i.e. relax them) i. Adjust the width of eye pieces ii. Open right eye only -> focus view by raising the stage. iii. Open left eye only -> focus view by turning left eyepiece

IMPORTANT – DO NOT TOUCH/CLEAN THE FOCUS LENSES - DO NOT CHANGE THE LENSES. Ask David, me, or a graduate student in the Killick lab.

2. COMPUTER a. Open Picture Frame i. Settings: 1. Click “Color” 2. “Gain” – all the way left ii. Adjustments:

273 1. “Exposure” – use sliding bar 2. “Snap” – take picture

b. Saving i. set document type @ .tif & 24 bit ii. save in “Julie’s wood anatomy pictures” -> Safranine Samples file iii. Create a folder for each species iv. Create a folder for each sample

3. PICTURE TAKING Take 3 replicates per slide, i.e. 3 areas of good quality that span the pith to the bark. They can be on the same slice or on different slices. If you can’t have 3, take 2. If you don’t have 2, write “REDO” beside the sample ID in the index sheet in the slide box.

Each replicate will get 3 pictures: 1 picture at 50 x spanning the radius of the twig (pith to bark) and 2 pictures at 200x. We may need to take 2 pictures at 50x if the wood section is so large that the radius does not fit into one single picture.

With 3 replicates, this will result in 9-12 pictures per sample. With 2 replicates (exceptional), this will result in 6-8 pictures per sample)

a. At 20x i. Select three (or two at worse) good areas which will be the replicates. Label them on the slide with a fine sharpie 1 through 3  Good area = clear and sharp from pith to bark  Remember, the image is reversed, so that top right corner in eyepieces = bottom left corner on the slide.  Lower the stage all the way to write on the slide

b. At 50x i. Focus the image by looking at the computer screen. The eyepieces and camera are not confocal. ii. Adjust the exposure to get the sharpest delineation of the xylem cells. iii. Get as much of the slide cross section into the field of view by moving the slide with the stage.

274  The best results are usually obtained by aligning one of the wood rays with the bottom of the image. You can rotating the slide by rotating the stage. IMG_13 iv. Click “Snap” and save the image into the sample’s folder. Name it : “SAMPLEID_replicate#_magnification”. For example: “J17BEA14_replicate1_50x” v. Remember to set the document type @ .tif & 24 bit

c. At 200x i. Adjust the exposure ii. Take two pictures in that region that have at least two groups of xylem cells that are adequate for (t/b)2 measures.  (t/b)2 measures require two or more adjacent xylem cell of the same size.  You will need to refocus for each image since the slides are not perfectly flat. iii. Click “Snap” and save the image into the sample’s folder. Name it : “SAMPLEID_replicate#_magnification_Pic#”. For example: “J17BEA14_replicate1_200x_pic1”

d. When the pictures are taken, write “pic” in the Index Sheet by the sample e. Process the samples by species. Each species is processed by only one person

4. When you are done a. Log in your hours in the Excel spreadsheet b. Back up files on SanDisk USB key c. Turn off PC d. Turn off microscope camera (black box) e. Turn microscope light to lowest intensity and wait one minute before turning it off f. Put plastic cover over the scope

275

IMG_1 Slide holder on. Screw beside 50° mark. IMG_4.Bertrand lens arm: out

IMG_2. Light arm half way out IMG_5. Black cards = Analyzer lenses -> out

IMG_ 6. Disk: at position 4, 5 or 6 IMG_3.Polarized light filter: out 276

IMG_7. Shutter: does not matter for our purposes IMG_10. AS/FS knobs: doesn’t matter for our purposes

IMG_8. 530 nm light filter: does not matter for IMG_11. Focus slider: doesn’t matter for our purposes our purposes

IMG_12. Condenser slider and adjustment screws

IMG_9. Light filters: LBD in. All others out.

277

IMG_13. Diaphragm leaf IMG_ 15. Screw holding stage in place

IMG_14. Second condenser swivel arm and condenser height adjustment

278 3- STEM_GUI PROTOCOL

Prepared by Julie Messier Updated June 10th 2013 General directions

Each person will process 1 replicate per sample. Beside each sample that you process, write down your name and the date in the file “STEM_GUI Sample Processing list” located in the dropbox under Summer 2013. For problematic slides: 1- Talk to Julie about it bef.ore processing it 2- If you can’t talk to Julie, take note of the slide ID and problem and put it aside. Talk to Julie later. 3- Once a decision has been made regarding the issue, make good notes in the notebook of how to proceed with this type of circumstance. Communicate it to co-workers. 4- If the protocol is unclear or needs modification, make notes of the modification that need to be made. Preparation

1- Read through this protocol and briefly go through the steps with one image.

2- Read through the LEAF_GUI manual – take your own notes of what you think is most important. Rephrase in your own words. Make sure you understand what each button does.

3- Watch the LEAF_GUI videos 1,3,4,5 & 6 and go through the motions with STEM_GUI (www.leafgui.org)

4- Look at the “Slide Quality Check” protocol to know how to identify “good” and “bad” areas on the slides.

5- Before starting each new species

- look at the species’ wood features in the Wood Features document. See the Inside Wood image bank of the IAWA for more images (http://insidewood.lib.ncsu.edu).

Sample order Analyze the species from easy to hard: Easy: ACP, ACR, ACSA, ACSP, AMSP, CAC, COA, POD, PRS, PRV, ULA, POG, RHT Medium: BEA, BEPA, BEPO, OSV, POB, SOA, POT, FAG, Hard: FRA, PRP, QUR, TIA

279 Protocol for conductivity and lumen fraction measurements using STEM_GUI software (Safranine 50x pictures)

1. In the Dropbox folder, under “Wood Anatomy Summer 2013”, open the “STEM_GUI Sample Processing List” file and find the next sample to process (one with a threshold value that has not been done). Write down your name, and the date in the row corresponding to this sample.

CREATING THE WORKING AREA 2. Open Matlab & type “STEM_GUI” in the command window. 3. Upload image (File->Open Menu Item) a. The Images are located in the Desktop in Julie’s Files_YourName -> Julie’s wood anatomy pictures. 4. If you are working on the Mac, copy the image into the dropbox. On the virtual machine, you can retrieve then from the dropbox. 5. Enter scale: 0.138462 for 50x pictures. Hit ‘Enter’ 6. In the Dropbox, open the “Polygon” folder and open the “RoughPolygon” image of your sample to use it as a guide. These polygons were drawn roughly and should not be followed exactly. 7. Define your working area by clicking Polygon. a. Delineate an area following the ray parenchyma. When ray parenchyma is thick, include it on one side of the wedge but not the other b. Create many corners to closely delineate the xylem area. Mostly along the slide edge and along the pith. c. The cursor becomes a circle when you overlap the first corner of the polygon. Click it to close the polygon. d. When you are done, click Polygon Done. Tip: you can use the zoom and hand tool while you are drawing a polygon. Tip: Unselect the hand and zoom tools to resume tracing. Do not click “polygon” multiple times. 8. Save the Original Image in Julie’s Files_YourName -> STEM_GUI Working Images -> Original (Polygon) Images. a. Use the file name format: “samplename_replicate#_Polygon_YourName” b. Save as a .tif file with no compression, NOT a Matlab file

280 CREATING THE BINARY IMAGE

Thresholding 9. Enter the threshold value noted in the “Sample Processing List” next to the Regional Min button in the STEM_GUI main window 10. Save this first black and white image by clicking Send to Queue. 11. Click the Remove Spurs button under the “Clean Binary Image” drop-down menu (Ctrl+N). This removes convex single white pixels in the vessels. 12. Click ONCE on the Convex Hull button in the “Clean Binary Image” box. This takes the concavities out of the vessels. 13. Click the red Superimpose_GUI button. a. Click Superimpose Perimeter. This shows overlays the area perimeters over the color image. Image Clean-Up 14. In the SuperImpose_Gui Window, a. Between each modification, you will need to click on the Superimpose Perimeter and the Zoom In buttons, in this order. b. First, use Remove Bad Regions to manually remove the bad areas one by one. c. Use Polygon – to manually draw vessels. d. Divide Two Regions – Manually removes selected areas. Use to separate two vessels that have been merged. You need to keep the left mouse button depressed while you trace.

N.B. We can keep all the little tracheids! Make sure that no spurious regions are included.

15. Close the SuperImpose_GUI window when you are done clean up the image. 16. Press Image Complement twice to make the changes that you just made in the Superimpose_GUI window appear. This is very important because the B&W image in the main STEM_GUI window is not automatically updated 17. Save the final B&W image by clicking Send to Queue. 18. Save the B&W image in Julie’s Files_YourName -> STEM_GUI Working Images -> B&W Images. a. Use the file name format: “samplename_replicate#_B&W_YourName” b. Save as a .tif file with no compression

281 CALCULATING THE SUMMARY STATISTICS 19. Click Summary statistics. Choose the .xls format. a. Click “Save As” from the Edit dropdown menu. b. Save it in Julie’s Files_YourName -> “ConduitStats Files” folder. c. Use the file name format: “samplename_replicate#_YourName_ConduitStats”. i.e. replace the 50x with your name. 20. Unfortunately, the NonConduit Area value is not correct and we need this value. It gives the area of the whole window minus the conduit area, instead of the Non-Conduit Area within the polygon. To get the actual size of the working area: a. in the Global Threshold menu, set the left cursor all the way down and the right cursor all the way up and click the Global Threshold button. This makes the whole polygon white. b. Click “Binarize RBG”, then “Conduit Statistics” again. c. Select and copy all the values from this new excel file and paste them to the right of the original values in the original ConduitStats file. Click Save. d. You can now close the ConduitStats excel file of the Outline without saving. It is easy to recognize because it only has one column of statistics and the number of conduits is 1. 21. Save the Outline image in Julie’s Files_YourName -> STEM_GUI Working Images -> Outline Images. a. Use the file name format: “samplename_replicate#_Outline_YourName” b. Save as a .tif file with no compression

COMPILING THE SAMPLE’S STATISTICS 22. From the Dropbox folder, open the folder “All Sample Conduits” 23. Open the most recent “All Sample Conduits” file. If this is the first sample of the day, save it using today’s date right away, as it is very easy to forget to rename it when you save your changes later on. 24. Create a new row with the sample’s name, the date, your name and the Threshold and Area/Perim values used. Make sure to follow the file’s format for each column. 25. From the sample’s ConduitStats file, copy these values into their corresponding column. a. Cell B1 to the Total Conduit (Lumen) Area column b. Cell C1 to the Total (Sapwood) Area column c. Cell B12 to the Mean Equivalent Diameter column d. Cell B21 to the Mean Conductivity per Conduit column

282 e. Cell B3 to the # Conduit column f. Copy down the formulas in columns J, K and L 26. When you are sure everything is saved, click “Clear Images” in the main STEM_GUI window to work with next image.

TIPS 1) BE PATIENT ! Do not be clicker-happy. There is often a lag time when you click on buttons or draw on the images. 2) Press Ctrl A to refresh the zoom area after you have changed or moved it. 3) Save the modified image by clicking Send to Queue. Up to 10 images gets stored on the right hand-side of the window. Save your working image once in a while in case you do something irrevocable to the image. 4) The “undo” button does not undo everything. It will not undrag the polygon before you hit “polygon done” nor the “Remove Regions” command 5) Give a final close look to the “Superimpose Perimeter” view before calculating the statistics. 6) Click “Interrupt” if your computer is frozen or has been processing your “Calculate Statistics” command forever without any output. 7) When clicking “Calculate Statistics”, listen carefully for sounds. A beep means an error in Matlab which means that the computer will produce results, even if the STEM_GUI window still shows that it is thinking. 8) when selecting a Threshold, it is important to select a value that fills a larger fraction of less conduits (the larger ones) instead of a value that selects a smaller fraction of more conduits. i.e. chose quality over quantity. 9) When using the “Clean Binary Image” tools, you will often get the error message saying “This function only works on binary images”. Click the button “Binarize RBG” and ignore the next error message that appears saying that the image is already a binary image. This will allow you to click the button you wanted… once. 10) For the functions that have a sliding bar, typing a value and pressing “Enter” does not work. You have to use the sliding bar. 11) The buttons “full” and “zoom” do not work. What you always apply the Clean Binary functions to the whole image. 12) In Superimpose GUI, you need to click “Enter” after selecting regions to add or remove.

283 TROUBLES, TROUBLES File size  Files larger than 10 Mb cannot be processed in STEM_GUI. Take note of the problem in the Sample Processing list and move on to the next usable image.

STEM_GUI crashing  STEM_GUI is moody and difficult. It can and will crash. If you can’t move the zooming box when you open a new image, it is a good sign it will soon crash. You can close and restart STEM_GUI alone, or you can close and restart both STEM_GUI and Matlab. Too little area  If there is no or very little usable area on the slide, come and see me. Bad Image  If the replicate is too bad to use, do not use it and write it down on the ‘Poor - Picture taking Redos’ list in the Sample Processing List. Oversized cross-sections (Two 50x images)  Some cross-sections are so large that they do not fit from pith to bark into one single image. For those slides, two images have been taken. In this case, both images will need to be processed. Look at the overlap of both images before selecting the workable polygon and use reference points on the slides. Image Problems  Nodes: do not to select the area when you delineate the polygon  Cracks: if you have to work with an area that has small cracks or microfractures, they will get selected by the software as conduits. Remove them by hand. This will affect the “Conduit Area Fraction” measure, so make a note of it in the Sample Processing List.

284 4- VESSEL DIAMETER PROTOCOL

Prepared by Julie Messier January 10th 2013

1. Open the latest worksheet (“All Samples – Vessel Diameter_date”) and save a new file with today’s date 2. Open the conduit stats file of your sample in the original folder (“ConduitStats files”). Save it as a new file by changing the name of the file to “Samplename_replicate#_YourName_VD” and save this new file in the “Conduit Stats – Vessel Diameter” folder. 3. Go to the “ConduitDimensions” sheet. 4. Select the whole spreadsheet and sort the data by the column head “Area” (Use “Sort” in the “data” menu) 5. Select the rows with an Area < 490 micron2. (This is based on a minimum vessel diameter and maximum tracheid diameter of 25micron, as reported in Sperry et al. 2006. Am.J.Bot). Delete those rows and make a mental note of how many rows you selected. 6. In the “All Samples – Vessel Diameter” spreadsheet, write the name of the same, the date, your name, the species of the sample and note down the # of rows you deleted. 7. Add the titles “Average” and “Median” at the end of Column A. Calculate the Average and Median values of the Area column (column B) in the corresponding cells in column B using the “=average(B2:B#)” and “=median(B2:B#)” formulas. Make sure NOT to include the average value in your calculation of the median value. 8. Select these two values and copy them. Paste them in the “All samples-Vessel Diameters_date” file using copy special. Select “values” and “transpose”. 9. Save the modified sample file and close it. 10. Open a new sample and start over. (Steps 2-9) 11. At the end of your shift, copy the “All samples – Vessel Diameters_date” file in the dropbox folder.

285 5- (t/d)2 PROTOCOL

Prepared by Julie Messier August 23th 2013

1. Open new image (Click “Ctrl+O”, recent places) 2. Set scale at 270.17 pixels/mm (Analyze – Set Scale). o Check “Global” 3. Set measurement values (Analyze – Set measurements). o Only select the “Display Label” option. o Set the # of decimals to 4. 4. Take measurements for up to 3 pairs of vessels per image (most likely, 2) 5. Find pairs of vessels of similar size that are adjacent to one another o Similar size means with areas 30% similar (roughly, diameters 30% similar too) o Adjacent means that you can clearly see that their cell walls are really touching each other. Make sure there is no other cell in between. In doubt, do not measure the uncertain pair. 6. For each pair: o Measure two perpendicular diameters for each vessel first . Zoom in the image until the vessel fills the screen (about 300%) . Select the “line” tool. To start measuring double-click. . Take the first diameter measurement across the longest diameter. Take the second diameter perpendicular to it. . Click “Ctrl+m” to measure o Measure the thickness of the two adjacent cell walls . Take two measurements, at 1/3 and 2/3 of the length o Copy and paste the values in the measurement window into the Excel file. . Highlight the length values, copy them and paste them with the transpose option into columns I to J. . Select the last row across columns K to N and drag and drop them down to copy the formulas. . In the pair # columns, write the number of vessel pairs that have been measured so far for this replicate. 7. Close the image 8. When you open the next image, a window will pop-up informing you of a conflict between the current image scale and the global scale. Check only the 'Disable this message' option

286 9. Save the Excel file with today’s date at the end (that’s important!) and keep a copy on your desktop as well as on the shared Google Drives document.

Tips: ** You can adjust the position of the ends by selecting them and dragging them ** You can move the whole line by selecting the middle marker and dragging it. ** When measuring, go to the edge of cell wall. i.e. what is in sharp focus. ** “Ctrl +” and “Ctrl -” to zoom in and out. ** To find where the edge of the cell wall is, zoom in and out. Having both the close-up and far perspectives help in determining where the edge of the wall is. ** When the line tool is selected click the space bar to drag the image around. ** You can draw on the image to report issues. If you do so, save it under a different name. ** The best way to remember to save the excel file under a new name every day is to rename it as soon as you open it at the beginning of the day’s work.

287 TREE RING WORK

Basal Area Index (BAI) Protocol

Prepared by Julie Messier July 30, 2013

Look at each sample and see how well polished it is—if the surface is not ideal, sand/re-sand in advance

Sample Preparation

1. Draw two orthogonal lines across the diameters on the cookie, one going along the longest diameter with the other perpendicular to it. 2. Find the ring from 2007 and draw dots following all the way around the circumference of the ring. If it is too small to draw a dot on, put a dot on an earlier ring (e.g 2006 or before). a. Make sure that at least one dot is next to each of the four radius. b. If it fits, write the year (or the last number of each year) in each ring. 3. For each of the four radius, count backwards from 2012 to 2007, making sure that no rings are missing and that no partial or false rings are being counted. a. If there are missing rings, make a note of the partial rings the Excel worksheet and for missing year(s) in write “0” for the yearly growth rate in the corresponding radius. 4. Mark radii I-IV, starting at the longest and working clockwise around the cookie. 5. If possible, with a sharpened pencil make fine tick marks next to the four radii along each of the years that you will measure . Measuring the Basal Area Index

1. Make sure the cross hair of your eyepiece is parallel to the movement of the stage. 2. On the first radius, zoom out and align the microscope’s cross hair parallel to the first radius line. 3. Zoom all the way in and refocus on the last five years. 4. Without touching the focus or zoom, scroll the turntable to the pith—if it is out of focus, leave it as is. 5. Open the “J2K” program, select “Series,” then “New,” and start year = 0. 6. Turn on the monitor, click “Reset” and then “Print” on the clicker. A values of 0 should appear in the J2X table.

288 7. Hit “Start” on Measure J2K and, after scrolling to the end of year 2006, (turn handle to the right) hit print once (i.e. hit “Print” for the line right before the ring with dots). 8. Continue to scroll and hit “Print” at the end of each subsequent year (2006-2012). a. If there are years with no ring, hit “Print” without scrolling. A value of 0 will print in the yearly growth column 9. When done, highlight both columns for the seven measurement values and copy paste into the corresponding worksheet.

Notes

--There is one excel file per species, one worksheet per sample. --Save the worksheet with a new date everyday. --Write notes in the worksheet if there are missing years, partial rings, false rings, etc. --Do not turn the table backwards (to the left) while measuring—if this happens, you will need to restart measuring the sample --Do not lean into the eyepiece or otherwise touch the microscope head, as it will move the crosshair.

289 GAP LIGHT ANALYZER PROTOCOL

GLA (Gap Light Analyser) protocol

Prepared by Julie Messier, August 2012

Step 1) Open an Image process in order of site&name a. File -> Open -> Prêtes à Analyser File Type : choose ‘Other Graphics’

Step 2) Register Image a. Configure -> Register Image Select “Fix Registration for Next Image” if the image gives an obvious area of registration

Step 3) Load and Edit a Configuration a. Configure -> Load Configuration . In the “Photos hémisphériques” file, choose “MSH GLA configuration settings.scn” You should only have to do this the first time that you open the program

b. Configure -> Edit configuration i. “Site” Tab 1. Elevation : changes at each point (JX.X) 2. Orientation : changes for each tree Inclined: Slope in degrees (1-90) & aspect in degrees (1- 360) Horizontal, when aspect and angle are both 0° ii. “Image”, “Resolution” and “Radiation” Tabs should not be modified.

Step 4) Image Classification (Go from Color to B&W Image) critical step a. Threshold – Ultimately the function that transforms the color image into B&W i. Modify the threshold value until only the sky areas appear as white. Click “OK” if 3 or more sensitive areas of the photograph transfer properly into B&W. ii. If not, click “Cancel” and play with a combination of functions under the “Image” menu to modify the original image:

290 b. Colour planes c. Gamma correct d. Brightness e. Contrast f. Sharpen i. If that still doesn’t work great and only a given area of the image is problematic, click on Select Region under Edit. Select the appropriate tool and delineate the region of interest that needs a separate threshold value. In the toolbar, you can select whether you want to work with the inside or outside of the selected area.

Step 5) Compute Results a. Select “Run Calculation” under “Calculate” i. Select “Canopy Structure and Transmitted Gap Light” ii. Click “Calculate” iii. Under User Field, enter separately the tree ID and the image number. iv. Select ‘Append’. The information gets saved in the Output Summary Data window.

Step 6) Close your working image. a. Save the Working Image under the “Working Image” folder in the “Photos hémisphériques” folder. b. Do not save the Registered Image

Step 7) When you are done working, save the output Summary data in an Excel file. a. In the Calculation Output Summary Data Window, click “Save As”. Save the file in the “Photos hémisphériques” folder, under “GLA Output Summary” + today’s date. b. Open the file in a with Nopepad, select all the information and Copy-paste it in the Excel File named “GLA Outputs”. Click on “Use Text Import Wizard”. Select “Delimited” and “Semicolon” c. “Save as” and add today’s date.

Notes: - When two files have the same Tree ID, pick the one that offers the best contrast. - Calibrating your eye and learning your tools takes a while. When all the images are processed, you will therefore rerun the first images on which you trained yourself.

291