Drivers of Woody Plant Form and Function in Relation to Water Acquisition and Use in Seasonal Forests

Drivers of woody plant form and function in relation to water acquisition and use in seasonal forests

A Dissertation SUBMITTED TO THE FACULTY OF UNIVERSITY OF MINNESOTA BY

Christina Marie Smith

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

Jennifer S. Powers, Advisor

June 2019

Acknowledgements

First and foremost, I thank my advisor, Dr. Jennifer Powers, for her outstanding guidance, patience, and support throughout this whole process. Jennifer’s passion for understanding how tropical dry forests function is inspiring and motivated me to conduct the research in this dissertation. I will always be incredibly grateful for the countless steps Jennifer has taken to aid my development as a scientist and I could not have wished for a better advisor. I also thank, Dr. Tim Brodribb and Dr. Stefan Schnitzer, who at times served as my honorary advisors, for their advice and help. My present committee members Dr. Rebecca Montgomery, Dr. Jeannine Cavender-Bares, and Dr. Walid Sadok, and past members, Dr. David Moeller, and Dr. Peter Kennedy have provided support and helpful feedback over the years. I am thankful for many past and present members of the Powers’ lab. In particular, I think Dr. Leland Werden for all his support and advice throughout these years and for all the fun times we had while doing fieldwork. I am also very grateful to Laura Toro for all her help and friendship. It has also been a pleasure to get to know and collaborate with German Vargas, Dr. Naomi Schwartz, Dr. Bonnie Waring, Dr. Maga Gei, Dr. Kristen Becklund, Dan Du, Erick Calderon, Ariadna Mondragon, and Julián Tijerin. This work could not have been possible without the help of Daniel Perez and Damaris Pereira. I also appreciate the advice and interesting discussions provided by Dr. Géraldine Derroire, Dr. Catherine Hulshof, and Dr. Cristina Chinchilla. I have spent many months in the field in Costa Rica and in Panama and I have many wonderful people to thank. A large portion of this research was conducted at Área de Conservación Guanacaste and I am thankful for the help and support provided by the staff at Estación Experimental Forestal Horizontes, in particular, Milena Gutiérrez for supporting my research and Pedro Alvarado for logistical support. During my time in Panama, Jeremy La Che spent many hours helping me in the field also Dr. Omar Lopez, Dr. Joe Wright, and Dr. Helene Muller-Landau provided me with advice and support. I am also very thankful for the wonderful friends I made in Panama: Carolina Bastos, Lourdes Hernández, Anita Weißflog, and in particular María García. María, thank you for being such an amazing friend and housemate. I have received support from the University of Minnesota with a Doctoral Dissertation Fellowship, thesis research travel grant, Carolyn Crosby grant, Dayton natural history grants, and Plant and Microbial Biology travel grants. My research was also supported by a National Science Foundation Doctoral Dissertation Improvement Grant and I received a Smithsonian Institution Big Ten Academic Alliance Fellowship. In the Department of Plant and Microbial Biology, I have been surrounded by many wonderful people including Dr. Mohamed Yakub, Derek Nedveck, and in particular, Dr. Beth Fallon, Dr. Allison Haaning, and Dr. Diana Trujillo have made this process so much more fun. I would also like to thank Elisabeth Anderson, and in particular, Thierry Degehet for the love and support. Finally, I thank my mother, Patricia Martin, and my father, Timothy Smith, for all the love and support throughout the years.

Dedication This dissertation is dedicated to my parents, Patricia and Timothy, for their unconditional love and support.

Abstract

Understanding plant-water relations is of central interest to ecology, as patterns in water availability drive distribution, composition, structure, and function of plant communities worldwide. Plant-water relations are particularly important in forests that experience water limitation and periods of seasonal drought. However, little is known about the initial post-germination response of multiple species from seasonal forests to water availability, or how species with different growth forms respond over time to distinct water availability patterns. Moreover, few studies have examined woody plant drought resistance across entire communities or key traits that are difficult to measure, such as biomass allocation and rooting depth, but which may mediate responses to seasonal water shortage. In Chapter 1, I determined if legumes have a different regeneration niche than non-legumes. I found that legumes may indeed have a different regeneration niche, in that they germinate rapidly and grow taller than other species immediately after germination, maximizing their performance when light and belowground resources are readily available, and potentially permitting them to take advantage of high light, nutrient, and water availability at the beginning of the wet season. For Chapter 2, I measured drought resistance in all dominant woody species within a dry sclerophyll woodland community to determine community level drought resistance. I found that the leaf and stem xylem of all species was highly drought resistant, with two species being slightly more vulnerable to drought than the others. The overall high resistance to drought and similar levels of resistance between leaves and stems suggest that canopy and understory species in this evergreen forest have evolved similar strategies of drought resistance. However, because two species were more vulnerable to drought than the others, this could lead to shifts in

iii

community-level species composition caused by future increases in drought frequency and severity. For Chapter 3, to shed light on rooting depth and biomass allocation, I harvested above- and belowground biomass and measured maximum rooting depth for juvenile and adult lianas, deciduous trees, and evergreen trees. This was the first study to show that lianas had the shallowest roots, followed by deciduous trees, then evergreen trees, which had the deepest roots. In addition, my study was the first to find that contrary to predictions lianas and trees have similar allocation patterns to stems. To test the hypothesis that lianas perform better than trees during seasonal drought, for Chapter 4, I used a common garden experiment with lianas and trees. My results are the first experimentally support the hypothesis that lianas perform better and experience less physiological stress than trees during seasonal drought. Moreover, only trees responded to dry season irrigation with increased growth, whereas lianas did not, suggesting clear differences between growth forms in response to altered rainfall regimes. Ultimately, better dry-season performance may explain why liana abundance peaks in seasonal forests compared to trees.

Collectively, these studies show that dry forest tree species differ in their strategies used to cope with seasonal drought, and how they respond to water availability.

Table of Contents

Acknowledgements...... i Abstract ...... iii Table of Contents ...... v List of Tables ...... viii List of Figures ...... ix Introduction ...... 1 Chapter 1: Effects of soil type and light on height growth, biomass partitioning, and nitrogen dynamics of 22 species of tropical dry forest tree seedlings: comparisons between legumes and non-legumes ...... 6 Synopsis ...... 6 Premise of the study ...... 6 Methods ...... 6 Key Results ...... 7 Conclusions ...... 7 Introduction ...... 7 Materials and Methods ...... 10 Site description ...... 10 Soils ...... 11 Light treatments...... 12 Tree species, planting, and growing conditions...... 13 Stem height and internodes ...... 14 Harvests ...... 14 Relative Growth Rate ...... 15 Foliar elements and isotopes ...... 15 Estimation of nitrogen fixation ...... 16 Data analysis ...... 17 Results ...... 19 Soil variation ...... 19 Seed mass and height growth ...... 19 Total biomass and relative growth rate ...... 21

Biomass partitioning and morphology ...... 22 Leaf water use efficiency, and nitrogen concentration ...... 23 Nodule mass fraction and nitrogen derived from fixation ...... 24 Discussion ...... 25 Seed mass, legume traits and distinct regeneration niches ...... 25 Conclusions ...... 30 Acknowledgements...... 30 Tables and Figures ...... 31 Supporting Information ...... 40 Chapter 2: Loss of leaf hydraulic conductance in response to water deficit is associated with stem xylem embolism formation in diverse canopy and understory species ...... 59 Synopsis ...... 59 Introduction ...... 62 Materials and Methods ...... 65 Study site and species ...... 65 Optical vulnerability curves ...... 66 Leaf vulnerability to water deficit curves ...... 68 Hydraulic safety margins ...... 68 Data Analysis ...... 68 Results ...... 69 Leaf hydraulic conductance and embolism ...... 69 Vulnerability segmentation between leaves and stems ...... 70

Hydraulic safety margins to P12 , P50, and P88 stem embolism ...... 71 Discussion ...... 71

Loss of leaf hydraulic conductance correlates with shoot xylem embolism ...... 72 Little evidence for vulnerability segmentation in seven out of eight species ...... 73 Community-level variation to drought resistance ...... 74 Conclusions ...... 75 Acknowledgements...... 76 Tables and Figures ...... 77 Chapter 3: Allometric scaling laws linking biomass and rooting depth vary across ontogeny and functional groups in tropical dry forest lianas and trees ...... 85 Synopsis ...... 85 Introduction ...... 86

Materials and Methods ...... 89 Site description ...... 89 Species selection ...... 90 Common garden ...... 90 Mature plants...... 91 Biomass harvest ...... 91 Canopy leaf area ...... 92 Data analysis ...... 92 Results ...... 94 Biomass allocation patterns in juvenile lianas, deciduous tree, and evergreen trees ...... 94 Biomass allocation patterns in mature lianas, deciduous tree, and evergreen trees ...... 94 Trends in maximum rooting depth in relation to biomass and canopy area in mature individuals ...... 95 Differences in total root length among functional group ...... 96 Discussion ...... 96 Allocation patterns of mature individuals of lianas and trees ...... 96 Juvenile versus adult biomass allocation patterns ...... 97 Rooting depths of adult lianas and trees ...... 98 Implications of rooting depths for models of canopy phenology ...... 100 Conclusion ...... 101 Acknowledgements...... 102 Tables and Figures ...... 103 Supporting Information ...... 111 Chapter 4: Effects of dry-season irrigation on leaf physiology and biomass allocation in tropical lianas and trees ...... 118 Synopsis ...... 118 Introduction ...... 119 Materials and Methods ...... 122 Experimental design...... 122 Leaf water potential, soil moisture, and gas exchange ...... 123 Diameter, biomass, and rooting depth and length ...... 124 Leaf chemistry, specific leaf are, and hydraulic traits ...... 125 Data analysis ...... 126 Results ...... 128

vii

Soil moisture ...... 128 Leaf water potential ...... 128 Plant diameter, biomass, and rooting dynamics ...... 129 Photosynthesis and water use efficiency...... 131 Specific leaf area, leaf chemistry, and hydraulic traits ...... 131 Discussion ...... 132 Liana and tree response to seasonal drought...... 133 Root dynamics of lianas and trees ...... 135 Liana growth patterns and forest carbon storage ...... 136 Conclusions ...... 139 Acknowledgements...... 140 Figures ...... 141 Supporting Information ...... 145 Concluding remarks ...... 157 Bibliography ...... 161

List of Tables

Table 1-1. F-ratios from a mixed model with species as a random effect and legume or non-legume status (Group), light level (Light), soil type treatment (Soil), and seed mass (Seed) as fixed effects...... 31

viii

Table 1-2. F-ratios from a mixed model with species as a random effect and legume or non-legume status (Group), light level (Light), soil type treatment (Soil), and seed mass (Seed) as fixed effects...... 32

Table 1-3. F-ratios from a linear model for nitrogen fixing legume species with species (N =22), light level (Light) and soil type treatment (Soil) as fixed effects...... 33

Table 2-1. List of eight dominant woody study species from an evergreen dry sclerophyll woodland on the east coast of Tasmania, Australia. The selected species were four canopy trees species (that included one gymnosperm and three angiosperms) and four understory woody shrubs (all angiosperms). All species are evergreen...... 77

Table 2-2. Pairwise comparison with Tukey’s test of 12% (P12) and 50% (P50) loss leaf hydraulic conductance and percent formation of embolism in stems...... 78

Table 3-1. Family, species used in the study, age of individuals at the time of harvest— juveniles harvested in common garden experiment, and mature individuals harvested from secondary forest (mature)—growth form, and leaf habit...... 103

List of Figures

Figure 1-1. Mean height measured every two weeks of seedlings of legume and non- legume species growing in low-nutrient soil and high-nutrient and 50% full sun and 25% full sun...... 34

Figure 1-2. Mass fractions at the time of harvest of 8 legumes species and 14 non-legumes grown in low-nutrient and high-nutrient soil and 50% full sun and 25% full sun treatment combinations...... 35

Figure 1-3. Mean total biomass (g), foliar δ 13C (%), foliar N (g), and foliar N (%) at the time of harvest of 8 legume species and 14 non-legume species grown in low-nutrient and high nutrient soil and 50% full sun and 25% full sun treatment combinations...... 36

Figure 1-4. Estimates of the percentage of plant nitrogen derived from the air at the time of harvest for six N-fixing legumes grown in low-nutrient and high-nutrient soil and 25% full sun and 50% full sun treatment combinations...... 38

Figure 2-1. The progression of embolism in the stems of C. rhomboidea, E. pulchella, and M. hexandrum at xylem water potentials of –4, –6, –8, and –10 MPa...... 79

Figure 2-2. The relationship between the critical water potential points extracted from the vulnerability curves generated using the rehydration kinetics method for measuring kleaf and the optical vulnerability method for measuring stem embolisms for the eight sampled species...... 80

Figure 3-3. Relationship between stem optical vulnerability curves, leaf hydraulic conductance, and xylem water potential for the four sampled tree species...... 82

Figure 2-4. Relationship between stem optical vulnerability curves, leaf hydraulic conductance, and xylem water potential for the four sampled shrub species...... 84

Figure 2-5. Boxplots of hydraulic safety margins calculated as the difference between minimum in situ leaf water potential and the leaf water potential associated with 12%, 50%, and 88% of observed stem embolism extracted from optical vulnerability curves for the four sampled tree species of and the four sampled shrub species...... 84

Figure 3-1. Image of above ground biomass of liana Combretum farinosum and a root of evergreen tree Manilkara chicle in the process of being excavated...... 104

Figure 3-2. Relationship between leaf and stem biomass, leaf and root biomass, and stem and root biomass of juvenile, and mature deciduous trees, evergreen trees, and lianas.. 105

Figure 3-3. Maximum rooting depth mature deciduous trees, evergreen trees, and lianas...... 107

Figure 3-4. Relationship between maximum rooting depth and leaf biomass, total canopy area, stem biomass, root biomass, total biomass, and diameter at breast height of mature deciduous trees, evergreen trees, and lianas...... 108

Figure 3-5. The simulated dry season deciduousness (from 2009 to 2014) for coexisting deciduous and evergreen functional types in a Costa Rican tropical dry forests under different parameterization for rooting depth allometry based on Xu et al. (2016)...... 110

Figure 4-1. Predawn, midday, and the difference between predawn and midday leaf water potential of lianas and trees in irrigated and control plots measured on the same individuals two times during the dry season (March and April), once during the transition from dry to wet (May), and two times during the wet season (September and October)...... 142

Figure 4-2. Diameter measured at 20 cm from the base, total biomass of lianas and trees grown in irrigated and control plots, and irrigation enhancement of biomass...... 143

Figure 4-3. Photosynthesis rates, stomatal conductance, and intrinsic water use efficiency of lianas and trees in control plots during the dry season and the wet season...... 144

Introduction

Understanding plant-water relations is of central interest to ecology, as patterns in water availability drive distribution, composition, structure, and function of plant communities worldwide. Moreover, climate change is leading to increasing variability in rainfall regimes (Feng et al. 2013). Plant-water relations are particularly important in forests that experience water limitation and periods of seasonal drought. During periods of drought, perennial woody plants tend to downregulate photosynthesis and growth

(Brodribb et al. 2002, Restrepo-Coupe et al. 2013, Schnitzer and van der Heijden 2019).

However, not all water-limited forests are equal, and rainfall regimes can differ dramatically among seasonal forests. Some forests experience a predictable distinct wet season and a distinct dry season, whereas other forest experience more sporadic and unpredictable rainfall throughout the year. For example, the tropical dry forest in

Northwestern Costa Rica has a mean annual rainfall of ~ 1,500 mm, with a distinct 5- to

6-month dry season, or period of less than 100 mm of rain per month, from December to

May (Gillespie 2000, Powers et al. 2009b, Powers and Perez-Aviles 2013, Becknell and

Powers 2014). Furthermore, rainfall patterns are relatively predictable in this particular tropical dry forest (Allen et al. 2017), which is a potential explanation for why many woody species are drought deciduous as they can ‘predict’ when water will be available and when not (Borchert 1994, Borchert et al. 2002, Xu et al. 2016). By contrast, many seasonal forests experience unpredictable rainfall. In these forests, vegetation tends to be evergreen. Thus, rainfall regime clearly affects the structure and function of seasonally dry forests, but there are many questions that remain. Little is known about the initial post-germination response of species from seasonal forest to water availability nor how species with different growth

forms respond over time to distinct water availability patterns. Also, few studies have examined woody plant drought resistance across entire communities or how biomass allocation and rooting depth varies among species with different leaf phenology and growth form. This dissertation investigates plant-water relations in seasonally dry forests, across lifeforms and growth stages. In particular, I use tools from physiological ecology to investigate community-wide variation in traits in species-rich forests.

The seedling stage in dry forests is a particularly critical life history stage where conditions of water limitation may have important consequences for community composition if species vary in response to initial rainfall. Despite tropical forests being known for their high plant biodiversity, they are also remarkable for the high abundance and diversity of trees from the legume family, Fabaceae (Gentry 1988). Legumes differ from many other plant families in their ability to host symbiotic nitrogen (N) fixing bacteria

(Corby 1988). Aside from N fixation, tropical legume trees may also have distinct suites of traits including rapid germination in response to initial rainfall (Vargas G et al. 2015).

However, few studies have worked with a large number of dry forest tree species to simultaneously evaluate the effects of above- and belowground resources on legume and non-legume seedling performance. For my first chapter, I investigated early growth of

22 common tropical dry forest tree species, including eight legumes, under common conditions in a shadehouse. This study allowed me to test the hypothesis that belowground resources are more important to early growth and performance than light and that legumes have a distinct “regeneration niche”.

In dry forests in which there are no defined rainfall patterns, perennial woody plants tend to be evergreen as they cannot ‘predict’ when water will be available and thus retain

their leaves to take advantage of any sporadic rainfall. For example, in dry sclerophyll woodland on the east coast of Tasmania the mean annual rainfall at is ~650 mm (Australian

Bureau of Meteorology). However, despite rainfall being lower at this location than in some seasonally deciduous tropical forests (Gillespie 2000), because rainfall is sporadic throughout the year, all woody plants in this community are evergreen. To maintain their leaves through the year under extreme drought; many species are predicted to have xylem that is highly resistant to drought-induced damage (Bouche et al. 2016, Skelton et al.

2017b, Klepsch et al. 2018, Skelton et al. 2018). However, few studies have examined an entire community of woody plants to determine interspecific variation in drought resistance. For my second chapter, I investigated community-wide drought resistance by comparing leaf/shoot and stem xylem vulnerability to drought in all eight dominant canopy and understory woody species in a diverse evergreen dry sclerophyll woodland forest.

Another strategy that evergreen species may employ, whether it be in forests with distinct wet and dry seasons throughout the year or sporadic rainfall, is deep-rootedness.

Markesteijn and Poorter (2009) found that evergreen drought-tolerant tree seedlings have high biomass investment in root systems compared to deciduous species, yet it is unknown whether adult trees follow the same allocation pattern as juveniles. Furthermore, in simulation models, assuming that evergreen trees are more deeply rooted than co-occurring deciduous trees is critical for simulating different phenology patterns (Xu et al. 2016).

However, due to the difficulty of excavating whole root systems, there have been few studies to attempt to determine variation in rooting depths among species with distinct phenology and growth form. Additionally, there are two major theories about how perennial woody plants allocate biomass to leaves, stems, and roots (Davidson 1969,

Thornley 1972, Bloom et al. 1985, Enquist and Niklas 2002, Niklas and Enquist 2002,

Niklas and Spatz 2004, McCarthy et al. 2007, Chave et al. 2014). One theory suggests that all plants follow common allometric scaling rules (Enquist and Niklas 2002, Niklas and

Enquist 2002, Niklas and Spatz 2004, McCarthy et al. 2007). An alternative perspective predicts that plants allocate more biomass to the organ capturing the most limiting resource

(Davidson 1969, Thornley 1972, Bloom et al. 1985, Chave et al. 2014). However, few studies have attempted to harvest mature tropical trees to evaluate which biomass allocation patterns evergreen and deciduous species follow. Furthermore, lianas (woody vines) are thought to differ from trees in that they do not invest in stems for support, because they use trees for mechanical support (Givnish and Vermeij 1976, Putz 1983,

Stevens 1987, Castellanos et al. 1989, Gehring et al. 2004, Selaya et al. 2007, van der

Heijden et al. 2013). Thus, it is believed that lianas allocate more biomass to leaves and maybe roots, and less to stems in comparison to trees (Putz 1983, Castellanos et al. 1989, van der Heijden and Phillips 2009, van der Heijden et al. 2013, Schnitzer et al. 2014, van

Der Heijden et al. 2015); however, this has never been tested. In my third chapter, I used whole-plant harvests of adult deciduous trees, evergreen trees, and deciduous lianas in a seasonally dry tropical forest to address if leaf habit can predict rooting depth and how these three functional groups allocate biomass to leaves, stems, and roots.

One of the main drivers of species distributions in tropical biomes are rainfall gradients, and plant species composition and forest structure shift considerably across broad precipitation gradients (Beard 1944, Engelbrecht et al. 2007). In general, the abundance of most tropical vascular plant groups (e.g., epiphytes, herbs, palms, and trees) increases with annual rainfall (Clinebell et al. 1995, Gentry 1995). However, lianas seem

to follow the opposite distribution pattern. Lianas reach maximum relative abundance in seasonally dry tropical forests and decrease in abundance as mean annual rainfall increases

(Gentry 1995, Schnitzer 2005, DeWalt et al. 2015). It is thought that lianas follow this distribution pattern because they can perform better and grow more than trees when water is limited (Schnitzer 2018, Schnitzer and van der Heijden 2019). A potential explanation for this pattern is that lianas are better at accessing water through deep roots and using water during seasonal drought than co-occurring trees (Restom and Nepstad 2004, Andrade et al. 2005, Schnitzer 2005, Cai et al. 2009, Zhu and Cao 2009, Chen et al. 2015). In my fourth chapter, I tested the hypothesis that lianas perform better than trees during seasonal drought and are deeper rooted, using a common-garden irrigation experiment in a seasonally moist forest in Panama.

Chapter 1: Effects of soil type and light on height growth,

biomass partitioning, and nitrogen dynamics of 22 species of

tropical dry forest tree seedlings: comparisons between

legumes and non-legumes

Smith-Martin, C. M., M. G. Gei, E. Bergstrom, K. K. Becklund, J. M. Becknell, B. G.

Waring, L. K. Werden, and J. S. Powers. 2017. Effects of soil type and light on height growth, biomass partitioning, and nitrogen dynamics of 22 species of tropical dry forest tree seedlings: comparisons between legumes and non-legumes. American Journal of

Botany 104:399-410. DOI: 10.3732/ajb.1600276.

Synopsis

Premise of the study

The seedling stage is particularly vulnerable to resource limitation with potential consequences for community composition. We investigated how light and soil variation affected early growth, biomass partitioning, morphology, and physiology of 22 common tropical dry forest tree species, including eight legumes. Our hypothesis was that legume seedlings are better at taking advantage of increased resource availability, which contributes to their successful regeneration in tropical dry forests.

Methods

We grew seedlings in a full-factorial design under two light levels in two soil types that differed in nutrient concentrations and soil moisture. We measure height bi-weekly and, at

final harvest, biomass partitioning, internode segments, leaf carbon, nitrogen, δ13C, and

δ15N.

Key Results

Legumes initially grew taller and maintained that height advantage over time under all experimental conditions. Legumes also had highest final total biomass and water use efficiency in the high light and high resource soil. For nitrogen-fixing legumes, the amount of nitrogen derived from fixation was highest in the richer soil. Although seed mass tended to be larger in legumes, seed size alone did not account for all of the differences between legumes and non-legumes. Both below- and aboveground resources were limiting to early tree seedling growth and function.

Conclusions

Legumes may have a different regeneration niche, in that they germinate rapidly and grow taller than other species immediately after germination, maximizing their performance when light and belowground resources are readily available, and potentially permitting them to take advantage of high light, nutrient, and water availability at the beginning of the wet season.

Key Words: legumes; light availability; nitrogen fixation; soil resources; tree seedlings

Introduction

The seedling stage is a particularly critical life history stage where conditions of resource limitation may have important consequences for community composition if species vary in response to resource availability. In wet tropical forests, the vast majority of studies investigating the effects of resource availability on seedling performance have

focused on light as a critical resource, because as little as 0.5 to 5% of sunlight reaches the forest understory in closed canopies of wet tropical forests (Augspurger 1984, Brokaw

1985, Sanford Jr 1989, Chazdon and Pearcy 1991, Hubbell et al. 1999, Nicotra and

Chazdon 1999). Many experimental studies have demonstrated interspecific trade-offs between rapid growth in high light conditions and survivorship under low light (Kitajima

1994), and indeed this is a cornerstone of the theory of gap phase succession (Denslow

1987). By contrast, tropical dry forests are characterized by lower annual rainfall, a dry season lasting between 3-8 months, and many deciduous species (Murphy and Lugo 1986), which collectively suggest that patterns of understory light availability in tropical dry forest contrast strongly with those of wet forests (Parker et al. 2005). Several seedling studies in tropical dry forest have shown that light may not be as limiting resource as in wet tropical forests (Gerhardt 1996, Markesteijn et al. 2007, Chaturvedi et al. 2013). For example,

Gerhardt (1996) found that increasing light levels during the wet season improved seedling growth, but that the higher light levels during the dry season increased desiccation, reducing seedling survival in some species. Thus, in tropical dry forest (TDF) water may be a more limiting resource than light in controlling seedling growth and survivorship

(Gerhardt, 1996; Chaturvedi et al., 2013).

Plants take up both water and nutrients from the soil. Recent studies have shown that tropical tree species may partition soil nutrient gradients as well as light gradients

(John et al. 2007, Baribault et al. 2012, Condit et al. 2013), and that seedlings of most species respond to fertilization with increased growth (Lawrence 2003). Some studies have shown that higher soil moisture and nutrient availability are more important during early stages of growth in comparison with later growth stages (Breugel et al., 2011; Chaturvedi

et al., 2013). Moreover, resources can interact such that the availability of nutrients modifies the responses to variation in another resource. For example, low soil nutrient availability may constrain seedlings’ ability to respond to high light availability (Huante et al. 1998), but different species may vary in their responses to diverse combinations of resources. Ceccon et al. (2003) found that seedlings of different species responded distinctly to full-factorial fertilization with nitrogen (N) and phosphorus (P) depending on light availability and bulk density of the topsoil; some of the species responded positively to the nutrient addition whereas in others this treatment had no effect or had a negative effect.

Although tropical forests are known for their high plant biodiversity, they are also remarkable for the large abundance and diversity of trees from the family Fabaceae, i.e. the legumes (Gentry 1988). Legumes differ from most other plant families in their ability to host symbiotic N fixing bacteria (Corby 1988). Beyond N fixation, tropical legume trees may also have distinct suites of traits including high water use efficiency, high foliar N and carbon, rapid germination rates and high photosynthetic rates (Adams et al. 2010b, Powers and Tiffin 2010b, Powers and Tiffin 2010a, Sui et al. 2011, Reyes-Garcia et al. 2012,

Vargas-G. et al. 2015, Adams et al. 2016), leading some researchers to incorporate them as a separate plant functional type in ecosystem models (Wang et al. 2007). However, field experiments underscore that, even for the legume tree species within the TDF biome, there is a great deal of interspecific variation of functional traits and the responses of N fixation to variation in resources (Gei and Powers 2015, Gei et al. in review).

There are very few studies that have worked with multiple species (>20) in the tropics to simultaneously evaluate the effects of above- and belowground resources on

seedling performance. In this study we investigated how light and soil affect early growth, biomass partitioning, morphology, and physiology (including N fixation) of 22 common tropical dry forest tree species, including eight legumes, under common conditions in a shadehouse. We grew seedlings in a full-factorial design under two light conditions in two soil types that differed in nutrient concentrations and water holding capacity. The tree species chosen reflected the composition of the plant communities in our study area in the

TDF of Northwestern Costa Rica, (Powers et al., 2009), with eight species in Fabaceae and

14 species of trees belonging to 11 other families. This sampling allowed us to test the hypothesis that legume seedlings are better able to take advantage of increased resource availability, which contributes to their apparently successful regeneration niche in tropical dry forests compared to other taxa, as some studies have suggested (Vargas-G. et al. 2015,

Gei et al. in review)We were particularly interested in whether legumes differ in their responses to resource availability (i.e. behave as a distinct plant functional group) or whether large interspecific variation within the legume family prevents broad generalizations. In particular, if N fixation is strongly limited by energy availability

(Vitousek and Howarth 1991) or if it is up-regulated in nitrogen-poor soils (Barron et al.

2011), we would expect all nitrogen-fixing legumes to have highest investment in nodules and reliance on N fixation under conditions of high light and low nutrient availability.

Materials and Methods

Site description

The study was conducted in Sector Santa Elena (referred to throughout the text as

Santa Elena) of Área de Conservación Guanacaste (ACG). Santa Elena is located in

Northwestern Costa Rica and has a mean annual temperature of 25C, mean annual precipitation of 1765 mm that ranges from 880–3030 mm and on average a 6 month dry season (Gillespie 2000, Powers et al. 2009b, Powers and Perez-Aviles 2013, Becknell and

Powers 2014). Soil, seeds, and seedlings for the experiment were collected both at Santa

Elena and at Parque Nacional Palo Verde (referred to throughout the text as Palo Verde), which is also located in Northwestern Costa Rica (10.35° N, 85.35° W). Palo Verde has a mean annual temperature of 25C, mean annual precipitation of 1444 mm, with a range from 714 to 2130 mm and on average a 5 month dry season (Becknell and Powers 2014).

Both Santa Elena and Palo Verde are protected areas predominantly covered by regenerating tropical dry forests and span much of the extent of what was originally Costa

Rica’s dry forest biome (Gillespie 2000, Powers et al. 2009b, Powers and Perez-Aviles

2013).

Soils

Soils throughout the Costa Rican tropical dry forest vary in parent materials and weathering stage, and hence in physical properties and chemical composition (Leiva et al.

2009, Becknell and Powers 2014). We collected soils from both Palo Verde and Santa

Elena to examine effects of high vs. low belowground resource availability on seedling performance. We analyzed samples of these two soils to confirm nutrient concentrations.

Although we do not have taxonomic descriptions of these soils, the soils from Santa Elena are likely Inceptisols, developed from volcanic ash deposits (Ulate 2001). Based on the known occurrence and dark color, the soils from Palo Verde would likely be classified as

Vertisols or Mollisols. To increase soil aeration and drainage of the pot environment, we

combined 1/3 parts sand to 2/3 parts sieved soil in ~4 L black plastic bags (~25 x 30 cm).

Soil characteristics were measured on unreplicated bulk samples collected at the time of planting. We measured clay, sand, and silt concentrations using the hydrometer method on soil type before and after we added sand (Appendix S1: see the Supplementary Data with this article). We measured Bray extractable phosphorus (P) by extracting samples of 3 g of air-dried soil in 25 mL of a 0.03N NH4F and 0.025N HCl solution. Soil extracts were

3- analyzed for PO4 with spectrophotometry following Lajtha et al. (1999). Total carbon

(C) and N were measured via dry combustion on a PDZ Europa ANCA-GSL elemental analyzer, and cation exchange capacity was quantified via the summation method at the

Research Analytical Lab, University of Minnesota. To minimize the differences between microbial communities in the different soils, we combined equal weights of Santa Elena and Palo Verde soil in 1L of water, shook it, and added 10 ml of this solution to each seedling bag two weeks before planting. Volumetric soil moisture was measured in each bag at 0-5 cm depth four times over the experiment with a SM150 Soil Moisture Sensor

(Delta-T Devices, Ltd, Cambridge, England). We acknowledge that soil moisture may not be the best comparative metric of soil water availability on soils that vary in texture. The

Palo Verde soil (high resource soil) had ~10-fold higher concentrations of exchangeable cations and extractable P than the Santa Elena soil (low resource soil), although total N did not differ appreciably.

Light treatments

Plants were grown under two different light treatments in a ~7 x 4 m shade house in a full-factorial manipulation of soil type and light environment. One side of the shade

house was covered with a standard shade cloth that transmitted 50% (high light) of photosynthetically active radiation (PAR) compared to open sky (quantified with an

Apogee MQ-100 Quantum sensor, Logan, Utah) and the other side had a shade cloth which allowed for ~25% (low light) of full sunlight. The low light treatment falls in between the average below-canopy dry season percentage of visible sky (33.6%) and wet season values

(13.4%) for successional forests in this region (Derroire, unpublished data).

Tree species, planting, and growing conditions

We selected 22 common species of trees present in Costa Rican dry forest (Powers et al. 2009a), including 8 species of legumes and 14 other species from 11 families

(Appendix S2). The legume species we used include the three legume subfamilies

(Caesalpinioideae, Mimosoideae, Papilionoideae). Average seed mass was measured on three representative seeds per species after drying at ~60°C for >48 hours. For very small seeds, between 5 and 10 seeds were weighed together, and thus average mass was calculated. We also compiled a seed mass database for 87 common woody species from the tropical dry forest of Northwester Costa Rica (Appendix S3).

Of the total 22 species, 19 were grown from seeds, but for three species (Bursera simaruba (L.) Sarg., Licania arborea Seem., Spondias mombin L.) we used ~three-week- old seedlings collected from the forest and transplanted into bags. Some of the species

(Crescentia alata Kunth, Dalbergia retusa Hemsl., Enterolobium cyclocarpum (Jacq.)

Griseb., Guazuma ulmifolia Lam., Hymenaea courbaril L. and Albizia saman (Jacq.) F.

Muell.) received a scarification treatment before they were planted to accelerate germination by placing them in 100 °C water for 30s and then in ice water for 1 min, and

repeating this process three times. Then seeds from all species were soaked in water overnight including seeds that received the scarification treatment. Two seeds of each species were planted per bag, with eight replicate bags per species per treatment, for a total of 704 plants.

After planting, seedling position within each light treatment was randomized. One week after germination, seedlings were thinned to one per bag. Germination was recorded once a week for the first two weeks and then every two weeks after that. Because we obtained low germination rates for some species, more seeds were added to those bags after scarification. When rainfall was scarce (i.e. lack of rain during 48 h), seedlings were watered via an overhead sprinkler system every other day to prevent desiccation. The experiment was conducted during the wet season, so environmental conditions were relatively similar throughout the growing period.

Stem height and internodes

Stem height of each seedling was measured to the nearest 0.5 cm approximately every two weeks throughout the experiment. Following the final harvest, we counted and measured all internode segments on each stem.

Harvests

One month after the first planting we did an early harvest of three individuals of each species by treatment combination. Four months after the first planting, the remaining seedlings were harvested over a three-day period. For each harvest, the aboveground portion of the seedlings was removed and divided into leaves and stems, then the bags were

cut open and their contents were carefully washed to extract roots. In the case of nodulating legume species, for the final harvest nodules were separated from the roots. All leaves, nodules, roots and stems of each seedling were dried at ~60 °C for >48 h and weighed.

Relative Growth Rate

We used the difference in biomass between the first harvest and the final harvest to be able to calculate relative growth rate (RGR) as follow:

RGR= (ln W2 - ln W1)/ (t2 - t1)

where W1 and W2 are the average biomass from the initial and final harvest, respectively, per species per treatment, and t2 - t1 are the number of days between the initial and final harvest and the units are g/day. Unfortunately, because seedlings did not all germinate at the same time, they were different ages when we did the initial and final harvest, which could have had an influence on growth rate, but even with this drawback these results can still shed light on initial seedling performance.

Foliar elements and isotopes

After drying and weighing the leaves, one leaf, including the petiole, from each individual of all legumes and, because of limited funds, a subset of the other species that grew for comparable periods of time (Guazuma ulmifolia, Luehea candida, Pachira quinata (Jacq.) W.S. Alverson and Tabebuia rosea (Bertol.) DC.) were dried at 65°C for

48 h and ground to a fine powder. Leaf C and N concentrations and their stable isotopes

δ13C and δ15N were measured on a PDZ Europa ANCA-GSL elemental analyzer interfaced

to a PDZ Europa 20–20 isotope ratio mass spectrometer (Sercon Ltd., Cheshire, UK) at the

Stable Isotope Facility at the University of California, Davis.

Estimation of nitrogen fixation

Once seed reserves are exhausted, the N in legume seedlings comes from two main sources, uptake of mineral N from the soil or nitrogen fixation, or perhaps in some cases through ectomycorrhizal fungi as some legumes associate with them (Frioni et al.,1999).

To estimate the percentage of plant N derived from fixation (%Ndfa), we used a simple mixing model based on stable N isotope signatures, i.e. δ15N abundances, as follows

(Shearer and Kohl, 1986):

15 15 15 %Ndfa = (δ Nreference – δ NN2-fixing) / (δ Nreference – B) x 100

15 15 where δ Nreference represents the mean value of the δ N of non-fixing reference

15 species grown in the same soil type and light environment as the N2-fixing species, δ NN2-

15 15 fixing is the N abundance for each N2-fixing legume, and B is the average δ N for each species grown in sand, i.e. when plants are relying only on N2 fixation to meet all N requirements. This model takes advantage of the difference in N isotopic composition of fixed N coming from air, and assumes that the δ15N composition of reference plants reflects that of soil mineral sources. We used this equation to estimate %Ndfa of the six potential

N2-fixing legume species. The two non-fixing legume species, Caesalpinia eriostachys

Benth. and Hymenaea courbaril, were used as reference plants. The average difference in

δ15N abundance between N derived from the soil (as indexed by the reference plants) and

that of the air (which, by convention equals 0‰) for our data was 2.95‰, indicating the suitability of this method to our data. The B value of four of the N2-fixing legumes

(Dalbergia retusa, Enterolobium cyclocarpum, Gliricidia sepium (Jacq.) Kunth ex

Walp., Lysiloma divaricatum Benth.) was calculated from seedlings of the same species grown in pure sand in a previous study (Gei et al. in review). For Lonchocarpus felipei N.

Zamora and Albizia saman, we used the B value of the most closely related legume for which we knew its 15N abundance in sand (Dalbergia retusa and Enterolobium cyclocarpum, respectively).

Data analysis

We used t-tests to compare average seed mass for the 22 species used in the experiment and for the 87 common tropical dry forest species from our seed mass database.

We log transformed seed mass in both instances to account for the fact that there were orders of magnitude of difference in the data. Because seed mass and legume/non-legume status (hereafter referred to as group) were key variables driving the underlying performance of seedlings independent of the applied treatments we conducted analysis of covariance (ANCOVA). We ran the ANCOVAs with group and seed mass and the interaction between these two variables as the explanatory variables and initial height, height at a common age, the total biomass at the time of the final harvest, and the relative growth rate of biomass as the response variables. We log transformed seed mass in all the analysis in which it was included. Because seeds germinated at different times (or were planted as ~3 week old seedlings), we compared seedling height among individuals after

approximately the same number of weeks of growth (10-11 weeks). These analysis where done in R (Version 3.3.1).

Our main goal was to determine if legumes responded differently to the two light and soil treatments compared to the other species. The responses we were interested in included: initial height of the seedlings after two weeks post-germination, height at a common age (after ~10-11 weeks of growth, hereafter referred to as final height), stem internode length and number, leaf mass fraction (LMF), stem mass fraction (SMF), root mass fraction (RMF), nodule mass fraction (NMF), total biomass (biomass), δ13C (as an integrated index of water use efficiency), total foliar N (total N), concentration of foliar N

(%N), and relative daily growth rate of biomass (RGR) between the initial to the final harvest. Response variables of interest were analyzed using a mixed model, with light environment, soil type, group, and seed mass as the main, fixed effects and species as a random effect nested within group. We included species as a random variable in our model to account for the differences among species. These analyses were conducted using the nlme R package. We log-transformed the following variables to meet assumptions of normally distributed residuals: final height, LMF, SMF, RMF, stem internode length and

%N. To understand how NMF and %Ndfa varied under different resource conditions for the N-fixing legumes, we used a full factorial ANOVA with species, light level and soil type as main effects and NMF and %Ndfa as response variables. In this case we included species as one of the main effects as we wanted to determine if all legumes responded to the treatments in a similar fashion. With conducted Tukey HSD post hoc analyses for means separation among legume species to understand %Ndfa response to treatment combination. All analyses were done in R (Version 3.2.1).

Results

Soil variation

Even though seedlings were watered frequently, there were large and persistent differences in soil moisture between soil types over the experiment that are likely due to the higher clay concentration in the high resource soil (Appendix S4; Appendix S1). Soil moisture in the high resource soil averaged 24.8% water (with a range of 12.8 to 38.9%), whereas in the low resource soil the average was 12.7% water (range of 6.1 to 28.5% soil moisture). All measures of nutrient availability (e.g. Bray P, CEC, etc) were higher in the

Palo Verde (hereafter referred to as high-nutrient soil) compared to Santa Elena (hereafter referred to as low-nutrient soil) soils (Appendix S1).

Seed mass and height growth

Seed mass varied amongst species for the 19 species grown from seed; however, on average it was over one third higher in legumes than non-legumes (0.72 g compared to

0.45 g), but it was only marginally statistically significant (T18 = 1.861, P = 0.08;

Appendix S5; S6; S7). Furthermore, when we compared legume and non-legume seed mass for 87 common dry forest species we found the same trend. Legume seed mass was approximately one third higher than that of non-legumes (0.28 g compered to 0.20 g) and was marginally statistically significant (T86 = 1.928, P = 0.06; Appendix S3; S7).

Following germination for the species grown from seed, legumes grew more rapidly than non-legumes (T252 = 5.1032, P< 0.001): mean height of legumes at first measurement was

12.6 cm and ranged from 4.0 to 33.5 cm, whereas the mean for non-legumes was 7.6 cm and ranged from 1.0 to 25.0 cm (Appendix S6). Initial height differed between legumes

and non-legumes (F1, 213 = 46.70, P<0.001), but part of this was explained by seed mass

(F1, 213 = 269.51, P<0.001); however, the significant interaction between group and seed mass (F1, 213 = 60.12, P<0.001) indicates that seed mass affected initial height in a different manner for legumes/non-legumes. In general species with larger seed mass were taller than species with smaller seed mass (Appendix S8), underscoring the importance of seed size for initial height growth.

When final height was analyzed, the main effects of light level and soil type and the two way interaction between group by light level and seed mass by soil type affected seedling height, and the four way interaction between all the fixed effects was marginally significant (Table 1), where the tallest seedlings were the legumes in the low light and high nutrient level, suggesting that legumes’ initial height advantage was maintained through time. This pattern was clear in the biweekly height measurements, in which legumes were consistently taller than non-legumes (Fig. 1) Furthermore, when comparing only the effect of seed mass (F1, 361 = 81.78, P<0.001) and soil type (F1, 361 = 8.03, P<0.01) on final height, both of these variables affected seedling growth, but not the interaction between these two variables. However, even though the interaction between soil type and seed mass was not statistically significant, the seedlings with larger seeds seemed to be less affected by soil type. Whereas, the seedlings with smaller seed mass grew taller in the high-nutrient soil than the ones in the low-nutrient soil (Appendix S9). When we examined the effect of only seed mass and group on final height, we found that there was a significant affect of group (F1, 361 = 35.21, P<0.001) and seed mass (F1,361 = 19.36, P<0.001), but not of the interaction between these two variable (Appendix S10). Seedlings with larger seed masses

continued to be taller after 10-11 weeks of growth, but on average legumes were taller than non-legumes.

Total biomass and relative growth rate

Group, soil type, light level, seed mass, the two way interaction between group by soil type and soil type by seed mass, and the three way interaction between group, soil type and seed mass, and soil type, light level and seed mass caused a significant difference on average total biomass at final harvest. The seedlings grown in high-light and high-nutrient soil had larger final biomass, and the legumes had the highest biomass (mean of 9.81 g and a range from 3.40 to 23.40 g), whereas non-legumes in the low-light and low-nutrient soil had the lowest biomass (mean of 1.41 g and a range from 0.04 to 3.36 g) (Table 1, Fig. 3).

When we examined the effect of only seed mass and group on total biomass, group caused the greatest difference (F1, 361 = 48.48, P<0.001). However, seed mass (F1, 361 = 5.93,

P<0.05) and the interaction between seed mass and group also caused a significant difference (F1, 361 = 4.02, P<0.05). Legumes had higher overall biomass than the non- legumes; however the legumes with smaller seeds had slightly higher biomass than the legumes with larger seeds, whereas non-legume biomass was the same independent of seed mass (Appendix S11).

In the full factorial mixed model, only seed mass caused a significant effect on RGR

(Table 1). Furthermore, when we compared RGR between seed mass and group we obtained the same results, only seed mass (F1,63 = 18.13, P<0.001) had a significant effect on daily plant biomass increments, although group was marginally significant (F1,63 =

3.40, P=0.069); between the initial and final harvest the seedlings with smaller seed masses

had higher RGRs, of which, legumes on average had slightly higher values than non- legumes (Appendix S12).

Biomass partitioning and morphology

Biomass partitioning, i.e. the allocation of photosynthate to the structures of root, stem, and leaves, for the final harvest was most strongly affected by light (Table 1), whereas group also had an affect in SMF, and LMF and SMF were affected by the interaction between seed mass and soil type. When we compared allocation between the initial and final harvest, in general, allocation to leaves increased under low light availability, whereas allocation to roots increased under high light conditions and to stems in legumes (Fig. 2). Also between the initial and final harvest, legumes initially allocated more to leaves and non-legumes to roots, but by the final harvest non-legumes had on average larger LMF and RMF (Appendix S13). Legumes had larger stem mass fractions both in the initial and final harvest, consistent with their greater heights. The three way interactions between group, soil type and seed mass and between soil type, light level and seed mass had a significant effect on the number of internodes in seedling stems, but in general the highest number of internodes was found in legumes grown in high-nutrient soil

(Appendix S14, Table 1). Light level, soil type, and the interaction between group and light level and the one between soil type and seed mass also affected the length of the internodes.

In general seedlings with larger seed mass had longer internodes and were also taller. The seedlings with the longest internodes were legumes grown in low light in high-nutrient soil

(Appendix S14, Table 1)

Leaf water use efficiency, and nitrogen concentration

For a subset of species, WUE, as indexed by δ 13C, differed significantly between light and soil conditions, with a significant interaction between soil type and group (Table

1, Fig. 3). In general, seedlings growing in low light had more negative δ 13C, and the lowest value of δ 13C was in non-legumes grown in high-nutrient soil in low light with a mean δ

13 C of -31.16‰ and a range of -29.70 to -32.74 ‰. By contrast, legumes grown in high- resource soil in high light had the least negative values, with a mean δ 13C of -29.33 and a range of -27.13 to -31.39 ‰. In other words, while low light consistently decreased WUE across all taxa, the effects of soil resources differed for legumes compared to non-legumes:

WUE decreased with low soil resources in the legume taxa, but increased in this same soil for the non-legumes.

Leaf N concentrations varied by a factor of six and depended on light and soil type

(Table 1, Fig. 3). In general, seedlings growing in lower light and in low-nutrient soil had higher % foliar N, with a mean of 2.97 and a range of 1.64 to 6.86% (Fig. 3). Seedlings with the lowest % foliar N were the non-legumes grown in high-nutrient soil in high light, with a mean of 1.52 and a range of 1.09 to 2.10%. Total foliar N, i.e. foliar N concentration multiplied by leaf mass, varied by a factor of eight and followed a similar trend as %N with soil type and light level both affecting leaf total N (Table 1). The interaction between soil type and seed mass also had a lower, but significant affect on total N (Table 1). Legumes grown in the high-nutrient soil had much higher values compared to the legumes grown in low-nutrient soils ( Fig. 3). By contrast, there were no significant treatment effects of light or soil resources on total foliar N in the non-legumes (Fig. 3). Light conditions were less important, but legumes in low light had more total foliar N than in high light (mean 0.08

and a range of 0.02 to 0.19 g per plant vs. 0.07 and a range of 0.02 to 0.13 g per plant, respectively) (Table 1, Fig. 3).

Nodule mass fraction and nitrogen derived from fixation

On average, nodules comprised ~2% of the total biomass of N-fixing legumes

(Appendix S5), and differed among species and soils, but not light treatments (Appendix

S12, Table 2). Based on the 15N natural abundance method, seedlings acquired from 7.6 to

100% of their N from fixation and the average value of Ndfa was 72.75% (Appendix S5).

Furthermore, like NMF, Ndfa also differed among species and soils (Table 2 and Appendix

S15), and species responded differently to both soil type and light level. Percentage of nitrogen derived from fixation for D. retusa, E. cyclocarpum, L. felipei and L. divaricatum was affected by soil type, but for A. saman light level instead of soil had a significant effect, and for G. sepium there was no significant effect of soil or light on %Ndfa (Fig. 4, Table

2). Ndfa was consistently higher under high fertility and soil moisture, suggesting that belowground resources limited N fixation more than light in young seedlings (Fig. 4). The correspondence between NMF and Ndfa was low. For example, on average L. felipei had the highest NMF (0.024 g g-1) and had the second highest percent of N derived from fixation (84.0%), but G. sepium had the second lowest NMF (0.006 g g-1) and the highest

Ndfa (92.0%). By contrast, L. divaricatum had both the lowest average NMF (0.002 g g-1) and the lowest Ndfa (53.0%).

Discussion

Resource availability at the seedling stage is particularly important for determining seedling survival, and how different groups of plants respond to resources can constrain stand-level forest composition. Our study examined how variation in above- and belowground resources affected early seedling growth and performance for a large number of tropical dry forest tree species, with a special focus on comparing legumes, the most abundant and diverse family in the study region, to other taxa. Our hypothesis was that legume seedlings are better at taking advantage of increased resource availability, which contributes to their apparently successful regeneration niche in tropical dry forests compared to other taxa. We found that legumes had larger seed mass, grew faster immediately after germination and maintained this initial height advantage over species from other families. Below we discuss how seed mass affects initial seedling performance, whether variation in above- or belowground resources drives early seedling performance, whether legumes have a distinct regeneration niche compared to diverse non-legume taxa, and environmental controls on symbiotic N fixation.

Seed mass, legume traits and distinct regeneration niches

One of our objectives was to determine if legume tree species behave as a distinct functional group such as used in ecosystem modeling (Bhaskar et al., 2016). Some authors have suggested that TDF legumes have a distinct regeneration niche compared to other tree species because they germinate quicker than non-legumes (Vargas-G et al. 2015). In our study, legumes had higher seed mass and faster post-germination growth rates, gaining a significant height advantage over non-legume species during the first two weeks of growth.

Legumes had similar root and leaf mass fraction compared to other species, but differed in having greater stem mass fraction as a result of their greater height. Larger seed mass may have provided legumes with sufficient resources to grow faster immediately after germination (Slot et al. 2012). However, after two weeks, legume growth rates decelerated.

Independent of whether a seedling was a legume or not, seed mass played an important role on initial seedling performance. Seedlings with larger seeds had greater initial and final heights; however in general seed mass was not as relevant for total final biomass, and this could have been influenced by the fact that seedlings with smaller seeds had higher

RGR. Larger seed mass may play an important role in the initial establishment of seedlings

(Foster 1986, Jurado and Westoby 1992), in particular if the growth period is limited to the often unpredictable rainfall patterns, which is the case in our study site. Similar trends have been found in other studies, in which species with larger seeds were more successful in early stages of establishment in adverse environments (Moles and Westoby, 2004) and in environments that experience a dry season (Dalling and Hubbell 2002, Moles and Westoby

2004). Additionally, our findings that smaller seeded species had higher RGR also coincides with what previous studies have found in that species with smaller seeds have higher RGR than species with larger seeds (Fenner 1983, Shipley and Peters 1990).

By contrast, independent of seed mass, non-legume species did not grow as fast initially, but maintained a steadier increase in height over time. Similar trends have been found in other species. For example, Jurado and Westoby (1992) found that arid species with larger seeds had larger seedlings 10 days after germination in relation to species with smaller seeds, however seed mass alone does not explain our results as legumes responded differently depending on resource availability. Collectively, our findings that legumes

grow quickly after germination coupled with faster germination rates relative to other tropical dry forest species (Vargas-G et al 2015) suggest a possible mechanism to explain their high abundance in our study site where legumes represent 18% of the stems and basal area (Powers et al., 2009). This rapid germination and initial growth may allow legumes to take advantage of resources quickly, and grow for longer periods of time during the wet season. This may give legumes a competitive advantage over other species in highly seasonal environments such as TDFs, where water and nutrients are often limiting and the majority of annual germination and growth occurs during the wet months (Ceccon et al.

2006) and potentially increasing legume survival at the seedling stage similar to what

Menge and Chazdon (2016) found in a tropical wet forest. Although we only grew the seedlings for a short period of time and are unable to extrapolate our results to the sapling and adult stage, distinct regeneration niches is just one possible explanation for legume success in TDF, as other studies have suggested that adult trees of legumes in these forests have distinct traits beyond high foliar N, such as high WUE (Powers and Tiffin, 2010).

Similarly, in a TDF in Mexico, Pineda-García et al. (2015) found N-fixing legume tree seedlings were not only able to maintain leaf function under drought, but were also able to sustain high growth rates when water was readily available.

For many of the response variables we measured, group (legume/non-legume status) by light or group by soil type interactions were significant, suggesting that legume responses to variation in resource availability differ from those of non-legume tree species.

In general, all species in our study had higher WUE as indexed by 13C in the high light level, presumably because higher light permitted increased photosynthesis rates relative to water loss. By contrast, legumes growing in the high resource soil had higher WUE and

much higher biomass than legumes growing in the low resource soil, whereas non-legumes used water more efficiently in the poor soil compared to legumes. One possible explanation for higher WUE efficiency and biomass in legumes in nutrient-rich, presumably wetter soils is that legumes are better able to take advantage of additional belowground resources compared to other taxa. Consistent with this explanation, all N-fixing legumes up- regulated fixation in the high resource soil, suggesting that this process was limited by soil nutrients and/or water, and that legumes utilize additional nutrients like phosphorus to maximize N acquisition and growth. This translated into higher biomass production by almost three-fold in the high resource soil compared to low resource soil, and apparently higher WUE under conditions of high above and belowground resources. Also, even though the non-legumes had higher WUE in the low nutrient, low moisture soil, final seedling biomass was not dramatically different between the two soil types.

All of our results are consistent with a growing body of work that suggests that legumes have different functional trait values and trait correlations compared to non- legume taxa (Powers and Tiffin 2010; Reyes-Garcia et al 2012; Adams et al 2016; Menge and Chazdon, 2016). It is firmly established that legumes have higher foliar N concentrations (Powers and Tiffin, 2010; Adams et al, 2016), but at the same time they have higher WUE (Reyes-Garcia et al 2012; Adams et al 2016) and growth under drought

(Pineda-García et al. 2015). In our experiment, foliar N concentrations were affected by both light and soil (Table 1, Figure 3), but the only interactions that were significant with legume/non-status involved soil type. Higher WUE when belowground resources are more abundant and higher photosynthesis rates may give legume seedlings an advantage over non-legumes during the wet season in environments where water availability varies

strongly seasonally and water is often a limited resource even during the wet season.

Consistent with this, in a recent meta-analysis Adams et al. (2016) found that legumes had higher foliar N and WUE across a diversity of arid environments indicating general trend of increased performance of legumes in environments with limited water. Furthermore,

Pellegrini et al. (2016) also found that N-fixing legumes were more abundant in arid tropical environments than in wetter ones further supporting the that legumes have a competitive advantage in drier environments.

In addition to allowing us to compare a large number of legume species to non- legumes, our study investigates how symbiotic N fixation varies in response to resources and whether all N-fixing legume taxa responded in similar or individualistic ways. Both nodule mass fraction and the percentage of N derived from fixation (Ndfa) varied considerably among species. While several of the resource-by-species interactions were significant, all species growing in high-resource soils consistently fixed more N than individuals in low-resource soils, and there was no evidence that N fixation differed by light treatments (Table 2). Together these results suggest that: 1) N fixation in TDF seedlings seems to not be constrained by light limitation, 2) despite higher soil water and nutrients in the high resource soil treatment, availability of mineral N in either soil was not sufficient to down-regulate fixation, and 3), water and/or nutrients like P or micronutrients like molybdenum might limit N fixation in this landscape, as found elsewhere ((Barron et al. 2008, Batterman et al. (2013)). Thus, even though there is broad variation in Ndfa and allocation to nodules among the N-fixing legume species, as well as variation in functional traits such as wood density (Powers and Tiffin, 2010), legume species respond to changes in resources in remarkably similar ways.

Conclusions

Our study shows that seed mass is an important factor in initial seedling growth and our results suggest that species with larger seed mass will initially grow taller following germination, but species with smaller seeds will later have higher relative growth rates.

Our study also shows that both below- and aboveground resources were limiting to early tree seedling growth and function in this tropical dry forest and species differed in their responses to resource availability. In this forest, legumes have a different regeneration niche than taxa from other families, in that they germinate rapidly and grow taller than other species immediately after germination. Higher seed mass in legumes only partially explains these results. Legumes are capable of maximizing performance when light and belowground resources are readily available, which could permit them to take advantage of higher levels of light, nutrients and water available at the beginning of the wet season in a tropical dry forest.

Acknowledgements

The authors thank Rebecca Montgomery, David Moeller, and Peter Kennedy for comments that improved this manuscript. We also gratefully acknowledge excellent assistance from Daniel Perez-Aviles in the field, logistical assistance from Roger Blanco and Maria Marta Chavarria of the Area de Conservación Guanacaste, and financial support from US National Science Foundation CAREER Award DEB-1053237 to JSP. Finally, we would like to thank and the editor and two anonymous reviewers for comments that greatly improved this work.

Tables and Figures

Table 1-1. F-ratios from a mixed model with species as a random effect and legume or non-legume status (Group), light level (Light), soil type treatment (Soil), and seed mass

(Seed) as fixed effects.

Variable Den. Group Soil Light Seed Group Group Soil Group

df X X X X

Soil Light Light Seed

Final height 334, 15 2.44 4.12* 16.23*** 0.01 1.35 6.32* 0.05 0.36$

Biomass 334, 15 10.21** 16.67*** 5.49* 4.64* 5.19* 0.01 0.26 2.64

LMF 334, 15 0.05 0.18 10.90** 0.88 3.58$ 2.25 0.00 1.25

SMF 334, 15 5.69* 0.05 4.03* 0.01 1.70 2.49 0.05 0.80

RMF 334, 15 3.07 0.29 14.89*** 0.54 2.13 0.77 0.23 1.13

Internode 334, 15 0.15 0.37 1.25 4.25$ 0.48 0.37 8.77** 4.20$

number

Internode 331, 15 1.96 4.65* 31.44*** 0.24 0.02 6.54* 2.36 0.00

length

δ13C 201, 8 0.11 6.27* 14.57*** 0.17 3.33$ 1.56 0.07 0.57

%N 201, 8 0.50 7.23** 23.51*** 1.30 1.25 0.01 0.14 0.02

Total N 201, 8 0.55 25.53*** 6.82** 2.37 1.22 0.23 0.47 0.33

RGR 38, 13 0.78 0.01 0.27 4.85* 0.24 0.09 0.01 0.09

Num. df 1 1 1 1 1 1 1 1

$, *, **, *** indicate significance at the P < 0.10, 0.05, 0.01, and 0.001 levels, respectively. The first denominator df is for all the variables except group and seed mass and the second one is for group and seed mass.

Table 1-2. F-ratios from a mixed model with species as a random effect and legume or non-legume status (Group), light level (Light), soil type treatment (Soil), and seed mass (Seed) as fixed effects.

Variable Den Group Soil Light Seed Group Group Soil Group

df X X X X

Soil Light Light Seed

Final 334, 15 2.44 4.12* 16.23*** 0.01 1.35 6.32* 0.05 0.36$

height

Biomass 334, 15 10.21** 16.67*** 5.49* 4.64* 5.19* 0.01 0.26 2.64

LMF 334, 15 0.05 0.18 10.90** 0.88 3.58$ 2.25 0.00 1.25

SMF 334, 15 5.69* 0.05 4.03* 0.01 1.70 2.49 0.05 0.80

RMF 334, 15 3.07 0.29 14.89*** 0.54 2.13 0.77 0.23 1.13

Internode 334, 15 0.15 0.37 1.25 4.25$ 0.48 0.37 8.77** 4.20$

number

Internode 331, 15 1.96 4.65* 31.44*** 0.24 0.02 6.54* 2.36 0.00

length

δ13C 201, 8 0.11 6.27* 14.57*** 0.17 3.33$ 1.56 0.07 0.57

%N 201, 8 0.50 7.23** 23.51*** 1.30 1.25 0.01 0.14 0.02

Total N 201, 8 0.55 25.53*** 6.82** 2.37 1.22 0.23 0.47 0.33

RGR 38, 13 0.78 0.01 0.27 4.85* 0.24 0.09 0.01 0.09

Num. df 1 1 1 1 1 1 1 1

Table 1-3. F-ratios from a linear model for nitrogen fixing legume species with species

(N =22), light level (Light) and soil type treatment (Soil) as fixed effects.

Variable Species Light Soil Species X Species X Light X Species X Den. df

Light Soil Soil Light X

Soil

NMF 15.392*** 0.426 7.88** 1.80 2.45* 4.58* 0.45 84

Ndfa 18.60*** 1.21 116.13*** 5.71*** 4.89*** 2.98 2.58 87

Num. df 5 1 1 5 5 1 5

$, *, **, *** indicate significance at the P < 0.10, 0.05, 0.01, and 0.001 levels, respectively.

Figure 1-1. Mean height (cm) measured every two weeks of seedlings of legume and non- legume species growing in low-nutrient (LN) soil and high-nutrient (HN) and 50% full sun

(light) and 25% full sun (dark). Error bars are standard errors.

Figure 1-2. Mass fractions at the time of harvest of 8 legumes species (dark gray) and 14 non-legumes (light gray) grown in low-nutrient (LN) and high-nutrient (HN) soil and 50% full sun (light) and 25% full sun (dark) treatment combinations. (A) Leaf mass fraction (g g-1) was significantly effected by the light level (F1,334 = 10.90, P<0.01). (B) Stem mass fraction (g g-1) was significantly different weather they were a legume or a non-legume

(F1,15 = 5.69, P<0.05) and the light level (F1,334 = 4.03 ,P<0.05). (C) Root mass fraction

(g g-1) was significantly effected by the light level (F1,334 = 14.89, P<0.05).

Figure 1-3. Mean total biomass (g), foliar δ 13C (%), foliar N (g), and foliar N (%) at the time of harvest of 8 legume species (square) and 14 non-legume species (circle) grown in low-nutrient (LN) and high nutrient (HN) soil and 50% full sun (open symbol) and 25% full sun (filled symbol) treatment combinations. (A) Mean total biomass (g) at the time of harvest of 8 legume species (square) and 14 non-legume species (circle) grown in low- nutrient (LN) and high nutrient (HN) soil and 50% full sun (open symbol) and 25% full

sun (filled symbol) treatment combinations. There was a significant effect of legume/non- legume status (F1,15 = 10.21, P <0.001), light level (F1,15 = 5.49, P <0.001), nutrient level

(F1,15 = 16.69, P <0.001), and the legume/non-legume by nutrient level (F1,15 = 5.19,

P<0.001) interactions. (B) Mean foliar δ 13C (%) at the time of harvest of a subset of species

(N = 12). There was a significant effect of light (F1,201 = 14.57, P <0.001) and nutrient level (F1,201 = 6.27, P <0.001). (C) Mean total foliar N (g) at the time of harvest of a subset of species. There was a significant effect of, light level (F1,201 = 6.81, P<0.01) and nutrient level (F1,201 = 25.53, P<0.001). (D) Mean foliar N (%) at the time of harvest of a subset of species. There was a significant effect of light level (F1,201 = 23.51, P

<0.001)and nutrient level (F1,201 = 7.23, P <0.01). Error bars are standard errors.

Figure 1-4. Estimates of the percentage of plant nitrogen derived from the air (%Ndfa) at the time of harvest for six N-fixing legumes grown in low-nutrient (LN) and high-nutrient

(HN) soil and 25% full sun (gray) and 50% full sun (white) treatment combinations. There was a significant effect of species (F5,84 = 18.60, P<0.001), nutrient level (F1,84 = 116.13,

P<0.001), species by light level (F5,84 = 5.71, P<0.001) and species by nutrient level

(F5,84 = 4.89, P<0.001) interaction. Significant differences within each species as denoted by post hoc Tukey’s test (P < 0.05) are indicated by different letters. Error bars are standard errors.

Supporting Information

Appendix S1. Soil characteristics measured on representative bulked soil samples collected from two locations in Northwestern Costa Rica. One location was Sector Santa

Elena of Área de Conservación Guanacaste (the low resource soil) and the other one was

Parque Nacional Palo Verde (the high resource soil). The soil was mixed with1/3 sand to grow the seedlings for this experiment.

Santa Elena Santa Elena + Palo Verde Palo Verde +

soil 1/3 Sand 1/3

Total C (%) 1.42 0.88 1.38 0.85

Total N (%) 0.11 0.07 0.12 0.08

Bray’s P (g g-1) 0.49 1.10 7.27 12.59

CEC (meq 100-g) 6.55 5.74 65.99 40.95

Sand (%) 29.9 47.5 5.1 36.2

Silt (%) 33.8 18.8 17.5 8.8

Clay (%) 36.3 33.8 77.4 55.0

Appendix S2. Family, species names, planting source (established from seed or seedling), and number of weeks of growth (calculated from germination or planting and harvest dates) for the 22 species of tropical dry forest trees used in the experiment.

Family Genus and species Age at harvest (weeks) Anacardiaceae Spondias mombin* 13 Bignoniaceae Crescentia alata 10-16 Bignoniaceae Tabebuia rosea 16-17 Bombacaceae Pachira quinata 16-17 Burseraceae Bursera simaruba * 13 Chrysobalanaceae Licania arborea* 13 Fabaceae Caesalpinia eriostachys 16-17 Fabaceae Dalbergia retusa 16-17 Fabaceae Enterolobium cyclocarpum 14-17 Fabaceae Gliricidia sepium 16-17 Fabaceae Hymenaea courbaril 10-1 Fabaceae Lonchocarpus felipei 14-17 Fabaceae Lysiloma divaricatum 16-17 Fabaceae Albizia saman 14-16 Fagaceae Quercus oleoides 14-16 Meliaceae Cedrela odorata 14-17 Meliaceae Swietenia macrophylla 14-16 Moraceae Brosimum alicastrum 10 Rubiaceae Genipa americana 14 Malvaceae Guazuma ulmifolia 16-17 Tiliaceae Apeiba tibourbou 13-16 Tiliaceae Luehea candida 16-17 * Planted as seedling.

Appendix S3. Family, genus, species, average seed mass (g), and growth habit (liana, shrub, or tree) of 87 common woody plants specie from the tropical dry forest of

Northwester Costa Rica.

Family Genus Species Seed (g) Habit

Anacardiaceae Astronium graveolens 0.022 tree

Anacardiaceae Spondias mombin 1.238 tree

Anacardiaceae Spondias purpurea 1.036 tree

Apocynaceae Marsdenia macrophylla 0.024 liana

Bignoniaceae Amphilophium paniculatum 0.054 liana

Bignoniaceae Fridericia patellifera 0.066 liana

Bignoniaceae Crescentia alata 0.033 tree

Bignoniaceae Bignonia aequinoctialis 0.123 liana

Bignoniaceae Bignonia diversifolia 0.058 liana

Bignoniaceae Bignonia spp. 0.005 liana

Bignoniaceae Amphilophium crucigerum 0.050 liana

Bignoniaceae Tabebuia ochracea 0.011 tree

Bignoniaceae Tabebuia rosea 0.025 tree

Boraginaceae Cordia alliodora 0.009 tree

Boraginaceae Cordia guanacastensis 0.008 tree

Boraginaceae Cordia panamensis 0.073 tree

Burseraceae Bursera simaruba 0.439 tree

Chrysobalanaceae Hirtella racemosa 0.133 tree

Chrysobalanaceae Licania arborea 0.923 tree

Bixaceae Cochlospermum vitifolium 0.033 tree

Combretaceae Combretum farinosum 0.022 liana

Dilleniaceae Curatella americana 0.013 tree

Dilleniaceae Tetracera volubilis 0.012 liana

Ebenaceae Diospyros nicaraguensis 0.178 liana

Elaeocarpaceae Sloanea terniflora 0.103 liana

Euphorbiaceae Euphorbia schlechtendalii 0.002 tree

Euphorbiaceae Sebastiania pavoniana 0.020 tree

Fabaceae Acacia collinsii 0.036 tree

Fabaceae Acacia tenuifolia 0.053 tree

Fabaceae Acosmium panamense 0.033 liana

Fabaceae Albizia adinocephala 0.026 tree

Fabaceae Albizia niopoides 0.010 tree

Fabaceae Ateleia herbert-smithii 0.019 tree

Fabaceae Bauhinia guianensis 0.079 tree

Fabaceae Bauhinia ungulata 0.043 liana

Fabaceae Caesalpinia eriostachys 0.230 tree

Fabaceae Caesalpinia sp. 0.217 tree

Fabaceae Dalbergia retusa 0.069 tree

Fabaceae Enterolobium cyclocarpum 0.693 tree

Fabaceae Gliricidia sepium 0.102 tree

Fabaceae Hymenaea courbaril 3.493 tree

Fabaceae Lonchocarpus felipei 0.282 tree

Fabaceae Lonchocarpus minimiflorus 0.048 tree

Fabaceae Lonchocarpus rugosus 0.061 tree

Fabaceae Lysiloma divaricatum 0.020 tree

Fabaceae Machaerium biovolatum 0.127 tree

Fabaceae Myrospermum frutescens 0.087 tree

Fabaceae Albizia saman 0.179 tree

Fagaceae Quercus oleoides 2.052 tree

Hippocrateaceae Semialarium mexicanum 0.078 tree

Malpighiaceae Banisteriopsis sp 0.046 tree

Malpighiaceae Byrsonima crassifolia 0.105 liana

Malpighiaceae Hiraea reclinata 0.006 tree

Malvaceae Guazuma ulmifolia 0.005 liana

Malvaceae Helicteres baruensis 0.002 shrub

Malvaceae Helicteres guazumifolia 0.003 shrub

Malvaceae Luehea candida 0.014 tree

Malvaceae Luehea speciosa 0.006 tree

Malvaceae Pachira quinata 0.041 tree

Malvaceae Sterculia apetala 1.721 tree

Melastomataceae Miconia argentea 0.010 tree

Melastomataceae Mouriri myrtilloides 0.065 tree

Meliaceae Cedrela odorata 0.026 tree

Meliaceae Swietenia macrophylla 0.255 tree

Meliaceae Trichilia americana 0.106 tree

Meliaceae Trichilia trifolia 0.041 tree

Moraceae Brosimum alicastrum 0.790 tree

Myrsinaceae Ardisia revoluta 0.075 tree

Ochnaceae Ouratea lucens 0.163 tree

Phyllanthaceae Margaritaria nobilis 0.006 tree

Polygalaceae Securidaca diversifolia 0.028 tree

Polygonaceae Triplaris sp 0.093 liana

Rhamnaceae Gouania polygama 0.006 tree

Rubiaceae Calycophyllum candidissimum 0.009 liana

Rubiaceae Genipa americana 0.038 tree

Salicaceae Casearia sylvestris 0.004 tree

Sapindaceae Serjania sp. 0.004 tree

Sapindaceae Allophylus occidentalis 0.062 tree

Sapindaceae Cupania guatemalensis 0.154 tree

Sapindaceae Paullinia cururu 0.013 liana

Sapindaceae Thouinidium decandrum 0.042 tree

Sapotaceae Manilkara chicle 0.229 tree

Sapotaceae Sideroxylon capiri 0.901 tree

Simaroubaceae Simarouba glauca 0.942 tree

Tiliaceae Apeiba tibourbou 0.006 tree

Trigoniaceae Trigonia rugosa 0.007 liana

Verbenaceae Redhera trinervis 0.053 tree

Appendix S4. Mean soil moisture content (%) for each combination of high resourse (Palo

Verde-PV) and low resource (Santa Elena-SE) soil and 50% full sun (light) and 25% full sun (dark). Error bars are standard errors.

Appendix S5. Measurements, units, and range of values in a shade house experiment with

22 species of tropical dry forest tree seedlings. Summary statistics of means, minimum and maximum values, and coefficients of variance are pooled across species and treatments.

Variable Abbreviation Number of Mean Minimum Maximum CV (%)

species analyzed

Seed mass (g) ⎯ ⎯ ⎯ 19 0.56 0.01 3.97 204.8

First height Initial height 19 9.5 1.0 33.5 75.6

measurement (cm)

Height at a Final height 22 18.5 3.0 77.0 65.4

common age (cm)

Relative growth RGR 20 0.033 0.001 0.077 45.9

rate (g day-1)

Leaf mass fraction LMF 22 0.35 0.07 0.69 35.0

(g g-1)

Shoot mass SMF 22 0.21 0.01 0.67 47.3

fraction (g g-1)

Root mass fraction RMF 22 0.44 0.02 0.89 32.4

(g g-1)

Total biomass (g) Biomass 22 3.70 0.04 23.40 95.0

Nodule fraction NMF 6 0.01 0.00 0.07 103.1

mass (g g-1)

Nitrogen derived Ndfa 6 72.75 7.62 100.00 35.6

from fixation (%)

Integrated water δ13C 12 -30.24 -32.86 -27.13 3.8 use efficiency

(indexed by δ13C)

Internode number ⎯ ⎯ ⎯ 22 9 1 24 46.2

Internode length ⎯ ⎯ ⎯ 22 17.9 3.5 45.5 52.6

(mm)

Total foliar N (g) Total N 12 1.13 0.03 5.38 81.9

Foliar nitrogen % N% 12 2.55 0.90 6.86 37.2

Appendix S6. A) Mean seed mass (g) of legume and non-legume species used in the experiment. B) Mean height (cm) of legume and non-legume seedlings two weeks or less after germination. Error bars are standard errors.

Appendix S7. (A) Mean seed mass of the 19 species grown from seed for this experiment

(8 legumes and 11 non-legumes. Legumes had a higher mean seed mass, but it was not statistically significant (T18 = 1.861, P = 0.08). (B) Mean seed mass of 87 common woody plants specie (22 legumes and 65 non-legumes) from the tropical dry forest of Northwester

Costa Rica. Legumes have a higher mean seed mass but it is not statistically significant

(T86 = 1.928, P = 0.06).

Appendix S8. Mean first height (cm) of the 19 seedlings grown from seeds and log transformed mean average seed mass (g) of each species. There was a significant affect of legume/non-legume status (group) (F1,213 = 46.70, P<0.001), seed mass (F1,213 = 269.51,

P<0.001) and the interaction between group and seed mass (F1,213 = 60.12, P<0.001) on the height of the seedlings after 1-2 weeks of growth.

Appendix S9. Mean final height (cm) of the 19 seedlings grown from seeds and log transformed mean average seed mass (g) of each species. There was a significant effect of seed mass (F1, 361 = 81.78, P<0.001) and nutrient level (F1, 361 = 8.03, P<0.01) on the final height of seedlings, but no the interaction between the two variables.

Appendix S10. Mean final height (cm) of the 19 seedlings grown from seeds and log transformed mean average seed mass (g) of each species. There was a significant effect of legume/non-legume status (group) (F1,361 = 35.21, P<0.001) and seed mass (F1,361 =

19.36, P<0.001), but not of the interaction between these two variables.

Appendix S11. Mean final biomass (g) of the 19 seedlings grown from seeds and log transformed mean average seed mass (g) of each species. There was a significant effect of legume/non-legume status (group) (F1,361 = 48.48, P<0.001), seed mass (F1,361 = 5.93,

P<0.05), and the interaction between these two variables (F1,361 = 4.02, P<0.05).

Appendix S12. Mean relative growth rate (RGR-g/day) of the 19 seedlings grown from seeds and log transformed mean average seed mass (g) of each species. There was a significant affect of seed mass (F1,63 = 18.13, P<0.001) and a marginally significant effect of legume/non-legume status (group) (F1,63 = 3.40, P = 0.069), but not of the interaction between these two variables.

Appendix S13. Leaf, stem, and root mass fraction (g g-1) of all legumes (dark gray) and non-legumes (light gray) harvested during the early (initial) and the final harvest growing in high nutrient (HN) and low nutrient (LN) soil and 50% full sun (light) and 25% full sun

(dark) treatment combinations. The error bars are standard errors.

Appendix S14. A) Mean internode number at the time of harvest of 8 legumes species and

14 non-legumes grown in high nutrient level (Palo Verde-PV) and low nutrient level (Santa

Elena-SE) soil and 50% full sun (light) and 25% full sun (dark) treatment combinations.

There was a significant affect of the interaction of seed mass by nutrient level (F1,13 =

8.33, P <0.001), and of the three way interactions between legume/non-legume by nutrient level by seed mass (F1,15 = 11.48, P <0.001) and between nutrient level by light level by seed mass (F1,15, = 5.80, P <0.05). B) Mean internode length (mm) at the time of harvesting. There was a significant affect of nutrient level (F1,331 = 4.65, P<0.05), light level (F1,331 = 31.44, P<0.01), the two way interaction between legume/non-legume status by light level (F1,15 = 6.54, P <0.05) and nutrient level by seed mass (F1,15 = 3.84, P

<0.05). Error bars are standard errors.

Appendix S15. Nodule mass fraction (g g-1) at the time of harvest of six N-fixing legumes grown in high nutrient soil (Palo Verde-PV) and low nutrient soil (Santa Elena-SE) and

25% full sun (gray) and 50% full sun (white) treatment combinations. There was a significant effect of species (F5,84 = 15.39, P<0.001), nutrient level (F1,84 = 7.88,

P<0.01), species by nutrient level (F5,84 = 2.45, P<0.05) and light level by nutrient level

(F1,84 = 4.58, P<0.05) interaction. Error bars are standard errors.

Chapter 2: Loss of leaf hydraulic conductance in response to water deficit is associated with stem xylem embolism formation

in diverse canopy and understory species

Smith-Martin, C. M., R. P. Skelton, K. M. Johnson, C. Lucani , and T. J. Brodribb. 2019.

Loss of leaf hydraulic conductance in response to water deficit is associated with stem xylem embolism formation in diverse canopy and understory species. In review at

Functional Ecology.

Synopsis

1. Recent findings suggest that tree mortality and post-drought recovery of gas exchange can be predicted from loss of function within the water transport system. Consequently, understanding causes of declines in leaf hydraulic conductance (kleaf) and its variation among species could have implications for predicting lags in recovery of gas exchange following drought as well as future community composition. Furthermore, not all plant tissues are thought to be equally vulnerable to loss of conductivity and embolism.

Vulnerability segmentation between plant tissues is believed to protect more energetically

“costly” tissues, such as woody stems, by sacrificing “cheaper” leaves early under drought conditions.

2. Here we investigated whether loss of function in evergreen leaves is associated with embolism in stem xylem and whether the capacity to withstand air entry varies among all

coexistent dominant canopy and understory woody species in a diverse dry sclerophyll woodland community, including multiple angiosperms and one gymnosperm.

3. Previously published terminal leaf/shoot vulnerability to loss of water transport capacity was compared with stem xylem vulnerability to embolism measured on the same species at the same site. We calculated hydraulic safety margins for stems to determine variation in the risk of hydraulic failure during drought among species.

4. We found a significant 1:1 relationship between the extent of embolism formation and loss of kleaf indicating that declines in kleaf in the measured tree and shrub species were likely caused by embolism formation in xylem. We also found no evidence of vulnerability segmentation between leaves and stems in seven of the eight species. Two of the dominant canopy trees were significantly more vulnerable to xylem embolism than the other six species.

5. Phylogenetically diverse canopy and understory species in this evergreen sclerophyll forest appear to have evolved similar strategies of drought resistance, including low xylem vulnerability to embolism and general lack of segmentation. However, some species are more vulnerable to damage by embolism during drought than others which could lead to shifts in species composition caused by future increases in drought frequency and severity.

Keywords: community composition, drought resistance, evergreen dry sclerophyll woodland, hydraulic safety margins, vulnerability segmentation, optical vulnerability method, water deficit induced embolism.

Introduction

Drought causes major damage to plant communities (Adams et al. 2010a), and forests around the globe are at risk of suffering from increases in drought frequency and severity (Engelbrecht 2012). Although species’ drought-induced damage thresholds and post-drought recovery of forest productivity are not yet fully understood (Anderegg et al.

2012), recent findings indicate that post-drought recovery of gas exchange can be predicted by properties of the water transport system (Brodribb and Cochard 2009, Anderegg et al.

2015, Skelton et al. 2017b). Severe drought causes declines in leaf water potential along with reduced stomatal conductance and assimilation. Upon rehydration of drought-stressed plants, water potential has been shown to rapidly return to an unstressed level, but there is often a lag in photosynthetic recovery (Miyashita et al. 2005, Brodribb and Cochard 2009).

Species that fail to recover gas exchange fully are frequently those found to have surpassed critical water potentials associated with loss of hydraulic conductance (Skelton et al.

2017b, Johnson et al. 2018). If the water potential threshold associated with loss of whole plant hydraulic conductance is an index of damage and reduced recovery of function under natural drought conditions, it becomes important to understand the drivers of loss of hydraulic conductance within plants, and how these processes might vary among tissue types.

Loss of leaf and stem hydraulic conductance can be caused by several processes.

Previous studies have shown irreversible loss of xylem hydraulic conductance in leaves, stems, and roots at low water potential to be caused by xylem embolism—air blockage

(Tyree and Sperry 1988, Tyree and Sperry 1989, Brodribb et al. 2016, Brodribb et al. 2017,

Skelton et al. 2017a, Skelton et al. 2018). Reductions in the water transport capacity of

xylem conduits caused by air entry can result in dehydration of the downstream leaf mesophyll tissue, thereby inducing tissue damage and ultimately death (Johnson et al.

2018). Recent evidence suggests that air entry into the xylem during water stress is non reversible in some species (Charrier et al. 2016), but partially reversible in others

(Brodersen et al. 2010). Consequently, embolism is a strong candidate process for driving long term damage during drought.

Loss of leaf hydraulic conductance has also been reported due to changes in extra- xylary hydraulic pathways (Scoffoni et al. 2011, Scoffoni et al. 2017), although it is uncertain whether this process is fully reversible or not. A similar process has also been observed in roots, where drought-induced mechanical failure (i.e. lacunae formation) in fine root cortical cells is associated with reduced fine root hydraulic conductivity (Cuneo et al. 2016). Finally, collapse of xylem conduits (Hacke et al. 2001) or free ending xylem veinlets in leaves under conditions of water deficit restricts the flow of water to the mesophyll tissues (Cochard et al. 2004, Brodribb and Holbrook 2005, Zhang et al. 2016), also resulting in loss of leaf hydraulic conductance. However, plants appear to recover rapidly following collapse of xylem conduits (Zhang et al. 2016), suggesting that xylem collapse is not a major factor contributing to term lags in the recovery of plant gas exchange following drought events.

Not all plant tissues are thought to be equally vulnerable to loss of conductivity during water stress. The hypothesis that plant tissues differ in vulnerability to loss of conductivity is referred to as the “vulnerability segmentation hypothesis” Zimmermann

(1983). This hypothesis proposes that vulnerability segmentation between plant tissues could serve to protect more energetically “costly” tissues from embolisms, such as stems,

and sacrifice “less costly” leaves (Tyree and Ewers 1991, Pivovaroff et al. 2014, Johnson et al. 2016, Wolfe et al. 2016, Zhu et al. 2016, Klepsch et al. 2018, Skelton et al. 2018).

Tree leaves have often been shown to be more vulnerable to water deficit than branches in both deciduous and in evergreen species (Cochard et al. 1992, Tyree et al. 1993, Choat et al. 2005, Johnson et al. 2011, Pivovaroff et al. 2014, Johnson et al. 2016, Zhu et al. 2016,

Hochberg et al. 2017, Scoffoni and Sack 2017, Charrier et al. 2018, Rodriguez-Dominguez et al. 2018, Skelton et al. 2018). The degree of segmentation may vary across rainfall gradients, with species in more arid environments having greater levels of segmentation than those occurring in more mesic environments (Zhu et al. 2016, Skelton et al. 2018).

However, segmentation was not found in a herbaceous species (Skelton et al. 2017a) nor in several deciduous and evergreen species (Bouche et al. 2016, Klepsch et al. 2018,

Skelton et al. 2018), indicating that segmentation may not occur across all species.

To survive in very dry environments, all perennial plant species within a community must be able to withstand substantial periods of drought; however, the degree to which each individual species can cope with drought may vary (McDowell et al. 2008,

Allen et al. 2010, Benito-Garzón et al. 2013). Hydraulic safety margins, or the difference between minimum dry season xylem water potential and xylem water potential causing losses in hydraulic conductivity, provide a reliable indicator of species’ likely exposure to hydraulic damage during drought in a given environment (Meinzer et al. 2009, Brodribb

2017, Skelton et al. 2017b, Benito Garzón et al. 2018). Water potentials causing 12%, 50%, and 88% loss of hydraulic conductivity are typically used to represent hydraulic safety margins (Meinzer et al. 2009, Anderegg et al. 2016). Evaluating the range of species’ level of resistance to drought and risk of hydraulic failure within diverse plant communities is

important for understanding community-level responses to drought and predicting future community composition (Koepke et al. 2010, Sack et al. 2016, Benito Garzón et al. 2018).

Here we investigated the processes underlying loss of leaf hydraulic conductance during water deficit to gain insight into the range of drought tolerance in all dominant woody species in a diverse evergreen dry sclerophyll woodland community. Our aim was to test the hypothesis that xylem embolism causes loss of hydraulic function, and that capacity to resist embolism might vary among co-occurring species as well as between leaf and stem tissues. To do so, we examined: 1) whether loss of leaf hydraulic conductance during water stress is associated with xylem embolism formation, 2) whether there is vulnerability segmentation between leaves and stems; and 3) to what extent species vary in degree of drought resistance among a diverse sample of woody canopy tree and understory shrub species.

Materials and Methods

Study site and species

This study was conducted on the east coast of Tasmania, Australia (42° 02´ 52.54˝

S, 147° 57´ 24.47˝ E, 134 m elevation). The mean annual rainfall at a nearby location

(Swansea, <10 km away) is ~600-700 mm (Australian Bureau of Meteorology). The soils are podzolic soils formed from Jurassic Dolerite parent material (Skelton et al. 2017b).

Vegetation at our study site is evergreen dry sclerophyll woodland. Woodlands originally covered close to 30% of the Australian continent including Tasmania and are often dominated by Callitris spp., Casuarina spp., Eucalyptus spp., and Melaleuca spp.

(Lindenmayer et al. 2005). At the study site the woody vegetation consisted of two canopy

layers: a 5–20 m overstory and a 1–3m understory (Skelton et al. 2017b). We selected the eight dominant woody species from this site, which consisted of four canopy trees species that included one gymnosperm and three angiosperm and four angiosperm understory shrubs (Table 1).

Optical vulnerability curves

We measured xylem embolism formation in stems using the optical vulnerability technique (Brodribb et al. 2017). For each species, branches from three individuals were collected from the study site. We cut the branches early in the morning (~ 6:00 am) to ensure that they were hydrated, and once cut, we immediately sealed them in plastic bag with moist paper towels to make sure that they were kept hydrated while being transported back to the lab at the University of Tasmania. All branches were over 1m in length to reduce the likelihood of any open-vessels being present in the distal parts of the branch.

Small distal branches, approximately 0.3 cm in diameter, with healthy foliage were chosen and a length of bark (approximately 2 cm) was carefully removed to expose the xylem. An adhesive hydrogel (Tensive, Fairfield, NJ, U.S.A.) was then applied to the exposed stem section to reduce reflections on the surface and aid light penetration into the sample and the branch was secured under a camera lens in a dark room. Once secured, the camera was set to capture images once every two minutes, and the sample was left to dry slowly until embolism events were no longer recorded (approximately 72–168 h later depending on the species). As the branch dried, the target region of the stem was illuminated using LED lights during image capture. Simultaneously, stem water potential was logged every 10 minutes using a stem psychrometer (ICT International, Armidale, Australia). Leaf water

potential was measured periodically using a Pressure Chamber Instrument (PMS

Instrument Company, Portland, OR, U.S.A.) to ensure agreement with the stem psychrometers. Leaves were excised, wrapped in moist paper towel and immediately measured.

Images of stem embolism were analyzed using ImageJ (Fig.1; National Institutes of Health, U.S.A.) following Brodribb et al. (2016) and Brodribb et al. (2017). Briefly, images are converted to 8-bit grayscale with pixel values ranging from black (0) to 255

(white). Successive images are subtracted to reveal embolism as subtle changes in light intensity (differences in pixel values). Differences in pixels due to noise or artefacts (e.g. sample movement or shrinkage) were removed using outlier removal in ImageJ or manually where it was clear that pixels were unrelated to an embolism event. Embolism in each stem was quantified as a cumulative total of embolized pixels in each image divided by the total number of embolized pixels in the fully dried sample (cumulative percentage xylem embolism). To determine the water potential at the time of image capture (Ψx), a regression model (typically linear) was fitted to the water potential measurements over time. Full details of the procedure, including an overview of the technique, image processing, as well as ImageJ scripts, are available at http://www.opensourceov.org.

Percentage embolism was plotted against Ψx and fitted with a sigmoidal relationship, from which we extracted the Ψx at 12% (P12 embolism, MPa) and 50% (P50 embolism, MPa) embolism:

Percent Embolism = 100 - 100/(1 + ea(Ψ – P50)), (Eqn.

where ‘a’ is a fitted parameter related to the slope of the curve.

Leaf vulnerability to water deficit curves

Leaf vulnerability to water deficit curves were extracted from Skelton et al.

(2017b), where the same plant community of trees and shrubs was measured in 2016. In this study, leaf hydraulic conductance (kleaf ) was measured using the rehydration kinetics method (RKM) on whole leaves or in the case of microphyllous leaves or needle-leaves small terminal shoots were measured (Skelton et al. 2017b). We extracted the xylem water potentials associated with 12% (P12 kleaf, MPa) and 50% (P50 kleaf, MPa) loss of kleaf.

Hydraulic safety margins

Hydraulic safety margins (HSM) were calculated as the difference between Ψmin and P12 (P12 HSM), P50 (P50 HSM), and P88 (P88 HSM) embolism extracted from optical vulnerability curves. Midday leaf water potential (Ψmin) was taken from Skelton et al.

(2017b) when water potential was measured on the same plant community during the end of drought caused by the 2015-2016 El Niño Pacific southern oscillation.

Data Analysis

All analyses were done in R (version 3.5.1). To determine the relationship between embolism formation and loss of leaf hydraulic conductivity, we conducted a regression with P12 embolism as a function of P12 loss of kleaf and a second regression of P50 embolism as a function of P50 loss of kleaf.

ANOVAs were fit to determine if there was segmentation between leaves and stems. We used P12 and P50 loss of kleaf as critical values for loss of function in leaves and

P12 and P50 embolism as critical values for loss of function in stems. We fit independent

ANOVAs with P12 and P50 for leaves and stems as the response variables and tissue type

(leaf or stem) and species as the predictors. Then we performed post-hoc Tukey tests with the emmeans package to interpret significant differences between leaf and stem loss of function.

To establish if there was a difference in hydraulic safety margins among species, independent ANOVSs were fit with P12 HSM, P50 HSM, and P88 HSM as the response variables and species as the predictor. Then post-hoc Tukey tests were performed with the emmeans package to interpret significant differences among species.

Results

Leaf hydraulic conductance and embolism

When comparing all species, the water potential associated with 12% (P12) and 50%

(P50) embolism in stems were strong predictors of the water potential associated with 12 %

2 2 (R =0.55, p=0.036; Fig. 2) and 50% (R =0.56, p=0.033; Fig. 2) loss of kleaf explaining 55% and 56% of the variation respectively with a 1:1 relationship.

The four tree species showed differences in stem vulnerability according to the timing of embolism during drying (Fig. 3). Below −4 MPa the frequency of embolism events started to increase for A. mearnsii and E. pulchella with most events occurring between −4 MPa and −6 MPa. However, C. rhomboidea and A. verticillata were more

resistant to embolisms and the frequency of events did not start to increase until water potentials fell below −6 MPa with most events occurring between −7 MPa and −10 MPa

(Fig. 3). Stems of all four tree species were completely embolized with no more events occurring at water potentials below −11 MPa (Fig. 3). Similarly, in leaves/shoots there was no decline in kleaf above approximately −3 MPa. However, kleaf declined sharply below water potentials of −4 MPa in A. mearnsii and E. pulchella and −6 MPa in C. rhomboidea and A. verticillata, coinciding with the increase in stem xylem embolism events (Fig. 3).

All four species of understory shrubs showed similar vulnerability to stem cavitation as the more resistant canopy species (i.e. C. rhomboidea and A. verticillata), with few embolism events occurring at water potentials higher than −4 MPa (Fig. 4). Below

−6 MPa the frequency of events started to increase, with most embolisms occurring between −7 MPa and −10 MPa, except for M. hexandrum, which was slightly less resistant, with most events occurring between −6 MPa and −7.5 MPa. Similarly to the canopy species, the stems of all four shrub species were completely embolized at water potentials below −11 MPa (Fig. 4). For the four shrub species, kleaf did not decline above approximately −4 MPa, however, below −5 MPa it declined sharply.

Vulnerability segmentation between leaves and stems

We only found significant differences between the water potential associated with embolism formation in stems and the water potential associated with loss of hydraulic conductance in leaves in one species of understory shrub, L. divaricata, in which 12%

(F7,28=2.07, p=0.029; Tukey T = 5.09, p=0.002) and 50% (F7,30=3.00, p=0.016; Tukey T =

4.46,p=0.008) loss of kleaf occurred at higher water potential than 50% embolism in stems;

suggesting that there is vulnerability segmentation between leaves and stems in this species. In all other species, there were not significant differences between P12 kleaf and P12 embolism and P50 kleaf and P50 embolism (Table 2), indicating no evidence of vulnerability segmentation between leaves and stems.

Hydraulic safety margins to P12 , P50, and P88 stem embolism

There was a significant difference in P12 HSMs among species (F7,12= 5.32, p=0.006; Fig. 5A). Two species (A. mearnsii and E. pulchella) had HSMs close to zero indicating that they function very close to the threshold of initial embolism formation; whereas other species (C. rhomboidea and L. divaricata) had much wider HSMs of around

3.5 MPa, thus functioning at minimum water potentials far from those associated with the onset of xylem damage. There was a clear difference in P50 HSMs among species (F7,12=

40.09, P<0.001; Fig. 5B). Mean P50 HSM for each species ranged from close to 1 MPa in the species with more vulnerable xylem (A. mearnsii and E. pulchella) up to 4.7 MPa in the more resistant species (C. rhomboidea and B. spinosa). There was also variation in P88

HSMs among species (F7,12= 12.14, P<0.001; Fig. 5C), in which case A. mearnsii and E. pulchella also had the narrowest HSMs (1.5 to 2.3 MPa) and L. divaricata, C. rhomboidea,

Al. verticillata and B. spinosa had the widest HSMs (6.3 to 7.9 MPa).

Discussion

We found a close relationship between the initial onset of stem xylem embolisms and loss of leaf hydraulic conductivity for all eight species. In seven of our eight dominant focal species, there was no evidence of any vulnerability segmentation between leaves and

stems, with only one species showing loss of leaf hydraulic conductance prior to embolism formation in the stem. These results provide support for our main hypothesis that xylem embolism is a primary cause of loss of function within natural plant communities.

Furthermore, there were clear differences among species in their resistance to drought- induced embolism within this phylogenetically and functionally diverse community. These findings have significant implications for our understanding of the drivers of drought- induced damage and gas exchange recovery and will help to inform predictions of future community composition and functionality in Australian woodlands following severe drought events.

Loss of leaf hydraulic conductance correlates with shoot xylem embolism

The strong correlation between the water potential causing xylem embolism in stems and the decline of kleaf suggest that loss of leaf hydraulic conductivity is likely to be driven by embolism formation in the xylem of the stem and leaf xylem in the studied species. Previous studies have also found a close relationship between embolism formation in leaves and the decline in leaf hydraulic conductance (Brodribb et al. 2016), and between embolism formation in leaves and stems in oaks (Skelton et al. 2018). Here we show that the decline of leaf hydraulic conductance is associated with embolism formation in stems in a diverse group of woody species including angiosperms and a gymnosperm.

Little evidence for vulnerability segmentation in seven out of eight species

We found little evidence for vulnerability segmentation between leaves and stems in seven of the eight species. In such a dry environment, with annual rainfall of ~600-700 mm, reaching as low as < 200 mm in over six month periods during drought events (Skelton et al. 2017b), it might be expected that vulnerability segmentation would provide species with a mechanism to reduce water loss during drought events by sacrificing “cheaper” leaves to protect more “expensive” stems (Wolfe et al. 2016). Instead, our results showing that stem embolism is coincident with the loss of kleaf for most of our study species indicates that segmentation is not a common strategy used to cope with drought in species from the evergreen woodland community examined here. Our results also provide evidence that individual plants may not be able to easily recover from drought-induced damage to leaf hydraulic function. A possible explanation of our results is that all our focal species are evergreen sclerophylls, which means that leaves are expected to have a long life span and be more “expensive” to construct (Ian et al. 2004, Reich 2014).

Overall our findings demonstrate that the presence or absence of vulnerability segmentation seems to be species-specific, even among species with similar leaf phenological behaviour/strategies. This conclusion is supported by a recent study in evergreen and deciduous oak species where only one deciduous oak showed evidence of vulnerability segmentation whereas all other species in the study had equally vulnerable leaves and stem (Skelton et al. 2018). Furthermore, in another recent study with two deciduous species and one evergreen species, leaves and stems had similar levels of xylem vulnerability to embolism in one deciduous species and the evergreen, while the other deciduous species actually had more vulnerable stems than leaves (Klepsch et al. 2018).

Community-level variation to drought resistance

Examining variation in vulnerability to embolism among coexisting, but unrelated species within a community can inform potential future changes in community composition and provide an important supplement to investigations into differences in vulnerability to embolism among phylogenetically related individuals (Pittermann et al. 2010, Larter et al.

2017, Skelton et al. 2018). Diverse levels of drought resistance between species will potentially lead to shifts in species composition as global-change-type drought events — droughts attributed to changes to global climate—become more frequent (Breshears et al.

2005, Koepke et al. 2010). Our results show that, at the community level, some of the studied species were more likely to experience hydraulic damage during droughts than others, as indicated by the substantial variation in P12, P50 and P88 hydraulic safety margins among our eight sample species. Two of the dominant tree species in our study (A. mearnsii and E. pulchella) were significantly more vulnerable to embolism and thus experienced greater risk of hydraulic failure during drought than the other two tree species. This finding is supported by previous observations showing that E. pulchella and A. mearnsii sustained more damage than other species at the same field site during drought (Skelton et al. 2017b).

Taken together, these findings indicate that over time, and with increases of global-change- type drought events, there will be a shift towards a canopy dominated by drought resistant

C. rhomboidea and A. verticillata in these evergreen dry sclerophyll woodlands, leading to a decrease in diversity. Similar outcomes have been seen in piñon–juniper woodlands in the south-western US, where one of the co-dominant species, pinyon pine (Pinus edulis), has suffered from more drought-induced mortality than the other dominant species, juniper

(Juniperus monosperma), and it is expected that with time juniper will become the dominant species in this region (Mueller et al. 2005). Since E. pulchella and A. mearnsii are both fast growing, ecologically important overstory tree species, this could have potentially significant impacts on the ecology of Australian woodlands including changes in fire regime and vegetation water use.

It is interesting to note that variation in the risk of hydraulic failure in our study community was driven primarily by variation in intrinsic capacity to withstand embolism between species rather than variation in minimum water potential. This finding is different to that observed in the Great Basin desert in the USA, where minimum water potential was strongly positively correlated with capacity to withstand embolism (Hacke et al. 2000) and in South Africa’s Cape Floristic Region, where minimum water potential varies considerably among species with different growth forms, leading to large differences in drought resistance (West et al. 2012). Taken together these results indicate that studies examining variation in vulnerability to embolism and minimum seasonal water potential among co-occurring species should be a future research priority.

Conclusions

We found that loss of leaf hydraulic conductive capacity was correlated with embolism formation in stem xylem, except in the understory species, L. divaricata, in which leaves lost hydraulic function before a significant percent of xylem had embolized.

Furthermore, the observed high resistance to embolism and general lack of segmentation suggests that canopy and understory species in the phylogenetically diverse evergreen dry forest community used for this study have evolved similar strategies of drought resistance.

However, the xylem in two of the dominant tree species, A. mearnsii and E. pulchella was shown to be more vulnerable, causing narrower hydraulic safety margins in these species than the other six species. These differences in vulnerability among species could lead to shifts in community level species composition caused by future increases in drought frequency and severity.

Authors' Contributions

C.M.S.M., R.P.S., and T. J. B. conceived the ideas and designed the methods; C.L.,

C.M.S.M., K. J., R.P.S., and T. J. B. collected the data; C.M.S.M., K. J., and R.P.S. analysed the data; C.M.S.M., R.P.S., and T. J. B. led the writing of the manuscript; C.L. and K. J. contributed to revisions.

Acknowledgements

The authors wish to thank Robert and Lisa Brodribb for use of the land on the east coast of

Tasmania. C.M.S.M. received a National Science Foundation Doctoral Dissertation

Improvement Grant (1700855), and a Doctoral Dissertation Fellowship and Thesis

Research Travel Grant from the University of Minnesota. R.P.S. received a postdoctoral research abroad scholarship from the South African National Research Foundation. This study was also supported by funds from the Australian Research Council to T.J.B.

(DP170100761).

Tables and Figures

Family Species Growth form

Cupressaceae Callitris rhomboidea R.Br. ex Rich. & A.Rich. tree

Fabaceae Acacia mearnsii De Wild. tree

Casuarinaceae Allocasuarina verticillata (Lam.) L.A.S.Johnson tree

Myrtaceae Eucalyptus pulchella Desf. tree

Pittosporaceae Bursaria spinosa Cav. shrub

Ericaceae Leptecophylla divaricata (Hook.f.) C.M.Weiller shrub

Myrtaceae Melaleuca pustulata Hook.f. shrub

Picrodendraceae Micrantheum hexandrum Hook.f. shrub

Table 2-2. Pairwise comparison with Tukey’s test of 12% (P12) and 50% (P50) loss leaf hydraulic conductance and percent formation of embolism in stems.

Species Threshold t ratio p-value

Acacia mearnsii P12 0.55 1

Allocasuarina verticillata P12 0.64 1

Bursaria spinosa P12 2.76 0.343

Callitris rhomboidea P12 1.26 0.994

Eucalyptus pulchella P12 0.62 1

Leptecophylla divaricata P12 5.09 0.002*

Melaleuca pustulata P12 2.01 0.804

Micrantheum hexandrum P12 3.19 0.164

Acacia mearnsii P50 0.36 1

Allocasuarina verticillata P50 1.98 0.815

Bursaria spinosa P50 3.02 0.218

Callitris rhomboidea P50 0.76 1

Eucalyptus pulchella P50 0.61 1

Leptecophylla divaricata P50 4.46 0.008*

Melaleuca pustulata P50 1.71 0.929

Micrantheum hexandrum P50 1.19 1

Figure 2-1. The progression of embolism (coloured area) in the stems of C. rhomboidea

(A), E. pulchella (B), and M. hexandrum (C) at xylem water potentials of –4, –6, –8, and

–10 MPa. Colour scale denotes the water potential.

Figure 2-2. The relationship between the critical water potential points extracted from the vulnerability curves generated using the rehydration kinetics method (RKM) for measuring kleaf and the optical vulnerability method (OVC) for measuring stem embolisms for the eight sampled species (Am—A. mearnsii, Av—A. verticillata, Bs—B. spinose, Cr— C. rhomboidea, Ep—E. pulchella, Ld—L. divaricata, Mp—M. pustulata, Mh—Micrantheum hexandrum). P12 (tringles) is the xylem water potential associated with 12% loss of leaf hydraulic conductance (kleaf ) and 12% observed embolism, P50 (circles) is the water potential associated with 50% loss of kleaf and 50% of observed embolism. Dashed lines represent the independent regressions for P12 and P50 and the dotted line represents a 1:1 relationship.

Figure 3-3. Relationship between stem optical vulnerability curves (OVC; A, C, E, G), leaf hydraulic conductance (kleaf; B, D, F, H), and xylem water potential (ψx) for the four sampled tree species. Black vertical lines and shading indicate mean water potentials associated with 50% embolized xylem and 50% loss of kleaf ± SE. In the OVC light lines indicate raw data for each individual (n=3).

Figure 2-4. Relationship between stem optical vulnerability curves (A, C, E, G), leaf hydraulic conductance (kleaf; B, D, F, H), and xylem water potential (ψx) for the four sampled shrub species. Black vertical lines and shading indicate mean water potentials associated with 50% embolized xylem and 50% loss of kleaf ± SE. In the OVC light lines indicate raw data for each individual (n=3 except for M. pustulata for which n=1).

Figure 2-5. Boxplots of hydraulic safety margins calculated as the difference between minimum in situ leaf water potential and the leaf water potential associated with 12% (P12;

A), 50% (P50; B), and 88% (P88; C) of observed stem embolism extracted from optical vulnerability curves (OVC) for the four sampled tree species of and the four sampled shrub species. Letters at the top of each bar indicate significant differences (P < 0.05) among species by pairwise comparison with Tukey’s test.

Chapter 3: Allometric scaling laws linking biomass and rooting

depth vary across ontogeny and functional groups in tropical

dry forest lianas and trees

Smith-Martin, C. M., X. Xu, D. Medvigy, S. A. Schnitzer, and J. S. Powers. 2019.

Allometric scaling laws linking biomass and rooting depth vary across ontogeny and functional groups in tropical dry forest lianas and trees. Submitted.

Synopsis

There are two theories about how allocation of metabolic products occurs. The allometric biomass partitioning theory (APT) suggests that all plants follow common allometric scaling rules. An alternative perspective is the optimal partitioning theory

(OPT), which predicts that plants allocate more biomass to the organ capturing the most limiting resource. In this study, we used whole-plant harvests of mature and juvenile deciduous trees, evergreen trees, and lianas in a tropical forest to address the following knowledge gaps: 1) Do mature lianas comply with the APT scaling laws or do they invest less biomass in stems compared to mature trees? 2) Do juveniles follow the same allocation patterns as mature individuals? 3) Is either leaf habit or life form a predictor of rooting depth? We found: 1) mature lianas followed the same allometric scaling laws as trees. 2)

Juveniles and mature individuals do not follow the same allocation patterns. 3) Lianas have shallower roots than trees in this dry forest.

Introduction

The allocation patterns of metabolic products in plants have been studied by ecologists from at least the 19th century (Kny 1894) until today (Poorter et al. 2015).

Allocation to different organs is a zero-sum dynamic, and is thought to reflect strategies to optimize performance. Allometric biomass partitioning theory (APT) suggests that all plants follow common allometric scaling rules that constrain allocation to different plant organs (Enquist and Niklas 2002, Niklas and Enquist 2002, Niklas and Spatz 2004,

McCarthy et al. 2007). Enquist and Niklas (2002) proposed a universal allocation rule for non-juvenile seed plants in which the scaling of leaf vs stem and leaf vs root is to the ¾ power and stem vs root mass is to the 1 power (i.e., an isometric relationship). However, in the case of perennial plants, for juveniles younger than one year, it was proposed that the scaling between above and belowground biomass will have a slope of 1, possibly due to the effect of the seed mass in young plants (Enquist and Niklas 2002, Niklas 2005).

An alternative perspective is the optimal partitioning theory (OPT), which allows some flexibility within fixed allocation rules. This flexibility allows plants to allocate biomass dynamically to optimize resource capture, depending on which resource most constrains productivity (Davidson 1969, Thornley 1972, Bloom et al. 1985, Chave et al.

2014). Thus, if water and/or nutrients are limiting, then plants should allocate more biomass to roots; if light is limiting, allocation to shoots and/or leaves should increase

(Bloom et al. 1985, Roa-Fuentes et al. 2012). The most limiting resource is tightly linked to specific lifeforms and life-history strategies (Banin et al. 2012, Poorter et al. 2012). For example, drought-deciduous tropical trees that avoid water stress by dropping leaves

during the dry season may be expected to invest less in roots compared to evergreen trees that maintain growth during periods of low water availability.

Whether either of these theories apply to lianas, which are woody vines, is unresolved Lianas are thought to differ from trees in that they do not invest in stems for support, because they use trees for mechanical support (Givnish and Vermeij 1976, Putz

1983, Stevens 1987, Castellanos et al. 1989, Gehring et al. 2004, Selaya et al. 2007, van der Heijden et al. 2013). Thus, it is believed that lianas allocate more biomass to leaves and maybe roots, and less to stems in comparison to trees (Putz 1983, Castellanos et al. 1989, van der Heijden and Phillips 2009, van der Heijden et al. 2013, Schnitzer et al. 2014, van

Der Heijden et al. 2015). For example, in a study comparing diameter growth between lianas and trees in an Amazonian forest in Peru, van der Heijden and Phillips (2009) found that lianas made up only one-third of the biomass that they displaced in trees due to competition because lianas grew less in diameter than trees. In two other multi-year liana removal studies conducted in a seasonally moist lowland tropical forest Panama, lianas were found to reduce tree stem growth in tree fall gaps (Schnitzer et al. 2014) and in the forest (van Der Heijden et al. 2015); in both experiments lianas did not make up for all the stem biomass that they displaced. However, these studies did not account for potential differences between lianas and trees in terms of stem extension. Niklas (1994) found that temperate vines partitioned their leaf and stem biomass following similar patterns as self- supporting gymnosperms. In another study with tree and liana seedlings from a single genus, Bauhinia, (Cai et al. 2007) found that lianas allocated more biomass to leaves and stems and less to roots than the trees.

Given a particular belowground biomass, lianas and trees may differ in belowground morphology, i.e. how belowground biomass is deployed. For example, many have postulated that lianas are deeper-rooted than co-occurring trees, permitting lianas to access deeper sources of soil water than trees (Holbrook and Putz 1996, Andrade et al.

2005, Schnitzer 2005, Schnitzer et al. 2005, Toledo-Aceves and Swaine 2008, Schnitzer and Bongers 2011, Chen et al. 2015). Evidence for this idea, however, is mixed. In support of the hypothesis that lianas are relatively deep-rooted, one study showed that 15-year-old liana saplings of one species had roots that reached depths of up to 10 m; however, co- occurring tree rooting depths were not examined in this study (Restom and Nepstad 2004).

By contrast, in a common garden experiment with four-year-old lianas and trees, rooting depths of lianas and trees were similar (Smith-Martin et al. in review). Additionally, De

Deurwaerder et al. (2018) found that in a rain forest in French Guiana, lianas used more superficial water than trees during the dry season based on stable isotopes of hydrogen and oxygen. Johnson et al. (2013) also found that two species of lianas did not have deeper roots than their host tree species in a seasonally dry tropical forest in Panama. Although these findings suggest that lianas may not be more deeply rooted than trees, the lack of in situ studies investigating rooting depths of lianas and co-occurring trees is conspicuous (Powers 2014).

Lianas are increasing in abundance in tropical forests (Schnitzer and Bongers 2011) and decrease forest-wide carbon sequestration (Schnitzer et al. 2014, van Der Heijden et al.

2015); thus, information on biomass allocation and maximum rooting depth is essential to fully understand their roles in forest ecosystems.

In seasonal tropical forests, belowground allocation and root distribution may be linked to leaf phenology. Markesteijn and Poorter (2009) found that evergreen drought-

tolerant tree seedlings have high biomass investment in root systems, yet it is unknown whether adult trees follow the same allocation pattern as juveniles. In simulation models, assuming that evergreen trees are more deeply rooted than co-occurring deciduous trees is critical (Xu et al. 2016). However, due to the difficulty of excavating whole root systems, there have been few studies to attempt to determine variation in belowground allocation and rooting depths among species with distinct growth forms and/or phenology.

In this study, we used whole-plant harvests in a seasonally dry tropical forest to address the following three knowledge gaps: 1) Do mature lianas comply with the common scaling laws proposed by Enquist and Niklas (2002), or do they invest less biomass in stems compared to mature trees? 2) Do juveniles follow the same allocation patterns as mature individuals? 3) Is either leaf habit or life form a predictor of rooting depth?

Materials and Methods

Site description

We conducted this study in a tropical dry forest in Northwestern Costa Rica at

Estación Experimental Forestal Horizontes (Horizontes; 10.718 N, 85.594 W), which is part of Área de Conservación Guanacaste (ACG). Mean annual precipitation is 1,730 mm, mean annual temperate is 25 °C, and there is a distinct 5-month dry season from December to May (www.investigadoresacg.org) during which there is little to no monthly rainfall.

This experimental station was formerly a farm with mixed uses including cotton, sorghum, and pasture (Milena Gutierrez, pers. comm.). Current land uses include young secondary forest, tree plantations, and abandoned fields. The naturally regenerated forest

communities at Horizontes are taxonomically and functionally diverse with over 60 species of deciduous, evergreen and semi-deciduous trees and lianas (Waring et al., in press).

Species selection

We sampled two groups of plants that differed in ontonogenic stage: juveniles and adults. Phylogenetically diverse species of deciduous trees, evergreen trees, and lianas were planted in a common garden to be harvested as juveniles, and mature individuals were selected in the surrounding forest for harvesting. We strived for as much overlap in species identity between the two groups as possible. The common garden experiment included seven species of deciduous trees (Bauhinia ungulata, Dalbergia retusa, Gliricidia sepium,

Luehea speciosa, Lysiloma divaricatum, Swietenia macrophylla, Tabebuia rosea), two species of evergreen trees (Crescentia alata, Simarouba glauca), and four species of deciduous lianas (Acacia tenuifolia, Amphilophium crucigerum, Dioclea violacea,

Tanaecium tetragonolobum; Table 1). For the forest harvest of mature individuals, we selected four species of deciduous trees (G. sepium, Guazuma ulmifolia, L. speciose, T. rosea), two species of evergreen tree (Manilkara chicle, S. glauca), and five species of deciduous lianas (A. tenuifolia, A. crucigerum, Combretum farinosum, Securidaca diversifolia, Serjania schiedeana; Table 1).

Common garden

For the common garden, we collected seeds in ACG in 2014 and in 2016 and planted them in 5 x 8 cm black polyethylene bags with a 3:1 mix of locally collected soil and sand in May 2014 and May 2016. We grew the seedlings for three months under 90%

polyethylene shade cloth and then planted them in August in a 1 x 1 m grid in an open field. One month before planting, we removed the shade cloth to acclimate the seedlings to higher light level before planting them. Prior to planting, the field was cleared of all vegetation and fence posts were installed for the lianas to grow on. All individuals from the common garden planted in 2014 were harvested from June-July 2016, and the ones planted in 2016 were harvested in June 2017. Thus, they were either ~12 or 24 months old at harvests, depending on planting date (2014 or 2016).

Mature plants

In addition to the common garden with juveniles, we harvested mature trees and lianas, which were located in the surrounding forest matrix or close to single-lane unpaved roads on the edge of the forest. All mature trees had a diameter at breast height greater than 12 cm, had reached the canopy, and were not over topped by other trees. Mature lianas were those that had reached the canopy and were growing across the top of host trees.

Biomass harvest

In the common garden, we harvested two to three individuals per species from June-

July in 2016 and in June 2017 for a total of 47 individuals (Table 1). Harvest of all the mature individuals occurred between August 2016–June 2017. We harvested three individuals per species for a total of 33 mature individuals (Table 1). Each individual was cut at ground level using a machete or chainsaw and aboveground liana biomass was pulled down from the canopy with a tractor (Fig.1a). The stem diameter was measured at 20 from the base and at breast height (1.3 m from the base), and then separated into stems and

leaves. Following stem removal, we dug up all belowground biomass with shovels and picks. We excavated all coarse roots down to diameters of ~2-5 mm (Fig. 1b) and measured maximum rooting depth and the length of all of the roots.

All leaf, stem, and root biomass of juveniles was dried and weighed. For the mature individuals, we weighed all the fresh biomass using a spring-loaded balance, and then dried and weighed a representative subsample of leaves, stems, roots to convert fresh weights to oven dried biomass estimates. We refer to all excavated belowground biomass as root biomass. Because some juveniles were shorter than 130 cm, we report diameters at 20 cm above the base for juveniles. The range of diameters at 20 cm for the base of juveniles was

4–83 mm for deciduous trees, 3–13 mm for evergreen trees, and 4–44 mm for lianas. For mature individuals the range in diameters of the stems at breast height for deciduous trees was 132–307 mm, of evergreen trees it was 137–219 mm, and sum of diameters of the stems at breast height for lianas was 19–290 mm.

Canopy leaf area

To determine canopy leaf area we measured specific leaf area (SLA, cm2 g-1) including petioles on 3–6 leaves per individual, and extrapolated total canopy leaf area from SLA and total leaf biomass.

Data analysis

We used a common framework for data analysis to answer all of our questions. We fit linear mixed models with functional group as a categorical variable, i.e. deciduous trees,

evergreen trees, and lianas (hereafter referred to as functional group) and species as a random effect. All continuous variables were log transformed in all linear mixed models.

Our first question was whether biomass allocation patterns to stems, leaves, and roots of juvenile and mature individuals from the different functional groups followed the same allocation patterns proposed by Enquist and Niklas (2002). To address this question, we fit linear mixed models with one fraction of biomass as a function of the other one, i.e. leaf biomass as a function of stem biomass and root biomass and stem biomass as a function of root biomass. Analyses were conducted separately for juvenile individuals and mature ones to address our second question, i.e. whether juveniles follow the same allocation patterns as adults. We also fit a model with shoot biomass (leaves plus stems) as a function of root biomass for the juveniles because Enquist and Niklas (2002) proposed that juveniles have an isometric relationship between above and belowground biomass. The third question that we addressed was whether rooting depths differed predictably among functional groups of mature individuals, with respect to biomass allocation and canopy area. We fit linear mixed models with maximum rooting depth as a function of leaf biomass, total canopy leaf area, stem biomass, root biomass, and total biomass and functional group as a grouping factor. Using mixed models, we also explored whether there was a difference among liana, deciduous tree, and evergreen tree total root length and DBH, shoot biomass, and total biomass. All statistical analyses were conducted in R (Version

3.5.1) using the lme4 package. When necessary, to be able to interpret significant differences, we conducted post-hoc Tukey tests with the emmeans package.

Results

Biomass allocation patterns in juvenile lianas, deciduous tree, and evergreen trees

In our whole-plant harvests of trees and lianas, allocation patterns differed among functional group (lianas, evergreen trees, deciduous trees) in juvenile plants. In general lianas allocated more biomass to leaves and stems and the deciduous and evergreen trees allocated more to roots. When comparing the allocation to leaves versus allocation to stems, lianas allocated proportionally more to leaves and deciduous trees to stems (Fig 2a; linear mixed model χ2=10.07, P=0.007; Tukey T=-2.95, P=0.045). Furthermore, in the tradeoff between the allocation to leaves versus roots, lianas also allocated more to leaves and deciduous and evergreen trees to roots (Fig 2b; linear mixed model χ2=15.15, P<0.001;

Tukey deciduous–liana T=-3.64, P= 0.010; evergreen – liana T=-2.49, P=0.051). Finally, when comparing allocation to stem versus allocation to roots, lianas allocated more biomass to stems and deciduous and evergreen trees to roots (Fig 2c; linear mixed model

χ2=13.39, P=0.001; Tukey deciduous–liana T=-2.90, P=0.036; evergreen–liana T=-2.65,

P= 0.036). We also found that juveniles did not scale isometrically between shoot and root biomass. Instead, as they became larger lianas allocated more biomass to shoots and deciduous and evergreen trees to roots (Sup. Fig. 1; linear mixed model χ2=16.80, P<0.001;

Tukey deciduous–liana T=-3.60, P=0.011; evergreen–liana T=-2.96, P= 0.018).

Biomass allocation patterns in mature lianas, deciduous tree, and evergreen trees

Mature lianas, deciduous trees, and evergreen trees all followed the same allocation patterns to leaves, stems and roots. These patterns were consistent when comparing allocation to leaves versus stems (Fig 2d; linear mixed model χ2=5.26, P=0.0720), leaves

versus roots (Fig 2e; linear mixed model χ2=4.21, P=0.122), and stems versus roots (Fig

2f; linear mixed model χ2=2.89, P=0.236).

Trends in maximum rooting depth in relation to biomass and canopy area in mature individuals

Overall, we found that lianas had the shallowest roots followed by deciduous trees, and evergreen trees had the deepest roots (Fig. 3; linear mixed model χ2=34.33, P<0.001; Tukey deciduous– evergreen T=-3.205, P= 0.026; deciduous–liana T=3.14, P= 0.034; evergreen

– liana T=5.70, P<0.001). Deciduous and evergreen trees had deeper roots than lianas independent of leaf biomass (Fig. 4a; linear mixed model χ2=13.70, P<0.001; Tukey deciduous–liana T=2.77, P=0.042; evergreen–liana T=3.47, P=0.011) and total leaf canopy leaf area (Fig. 4b; linear mixed model χ2=21.45, P<0.001; Tukey deciduous–liana T=3.36,

P=0.019; evergreen–liana T=4.24, P=0.005). As predicted, evergreen trees had deeper roots than deciduous trees and also lianas, when we controlled for stem biomass (Fig. 4c; linear mixed model χ2=16.81, P<0.001; Tukey evergreen—deciduous T=-3.11, P=0.033; evergreen—liana T=3.88, P=0.008), root biomass (Fig. 4d; linear mixed model χ2=10.50,

P=0.005; Tukey evergreen—deciduous T=-2.78, P=0.054; evergreen—liana T=2.62,

P=0.055), and total biomass (Fig. 4e; linear mixed model χ2=15.18, P<0.001; Tukey evergreen—deciduous T=-3.09, P=0.034; evergreen—liana T=3.58, P=0.012). When we took DBH into account as a covariate, evergreen trees still had deeper roots than lianas

(Fig. 4f; linear mixed model χ2=11.05, P=0.004; Tukey T=3.23, P=0.019).

Differences in total root length among functional group

We found no difference among total root length of deciduous trees, evergreen trees and lianas when accounting for DBH (Sup. Fig. 2a; linear mixed model χ2=2.81, P=0.246), shoot biomass (Sup. Fig. 2a; linear mixed model χ2=2.71, P=0.259), and total biomass

(Sup. Fig. 2a; linear mixed model χ2=0.267, P=0.263).

Discussion

This is the first study to use whole plant harvest of juveniles and mature individuals to address allometric scaling laws in deciduous trees, evergreen trees, and lianas, and to explore maximum rooting depth of these groups. Using the simple technology of shovels and chainsaws, we made three key discoveries. First, contrary to expectations, mature lianas did not allocate less biomass to stems than trees but instead followed the same allometric scaling laws. Second, juveniles and mature individuals do not follow the same allocation patterns. Our third discovery, also opposite to the general belief, was that lianas have shallower roots than trees.

Allocation patterns of mature individuals of lianas and trees

We found that mature lianas followed the allocation patterns predicted by the allometric biomass partitioning theory (APT). Moreover, by excavating whole coarse root systems and harvesting all aboveground biomass of mature lianas and trees, we found that lianas allocated more biomass to stems than we expected. There are no other whole- biomass harvests of mature tropical lianas and co-occurring trees that we are aware of for comparison; however, it has long been thought that lianas do not invest as much biomass

in stems relative to trees, because lianas use the stem architecture of trees to reach the forest canopy (Givnish and Vermeij 1976, Putz 1983, Castellanos et al. 1989, van der Heijden and Phillips 2009, van der Heijden et al. 2013). Here we show, at least for the individuals that we harvested, that lianas follow the same allocation patterns as mature co-occurring evergreen and deciduous trees. Stem length helps to explain this finding. Lianas and trees have similar biomass allocation patterns for stems, because even though lianas have smaller stem diameters at breast height than trees, lianas have very long stems and many small branches in the canopy of the host trees. Cumulatively, when expressed as a proportion of total biomass, the “longer, skinnier” liana stems compare similarly with the “shorter, fatter” tree stems. This pattern has also been observed by Niklas (1994) for aboveground biomass

(leaves and stems) of temperate vines and gymnosperms; however, that study did not harvest roots.

Juvenile versus adult biomass allocation patterns

Juveniles and adults did not follow the same allocation patterns. The adult individuals in our study all followed the same global allocation rules proposed by Enquist and Niklas (2002) in their allometric biomass partitioning theory (APT). However, the juveniles had biomass allocation patterns that complied better with the optimal partitioning theory (OPT). Perhaps while juvenile individuals are smaller they optimize resource acquisition to survive and grow; however, as individuals become larger, they become constrained by maximum resource harvesting capacity and transport forcing them to all to follow the same allocation rules (Enquist and Niklas 2002, Niklas and Enquist 2002).

Consistent with the findings of Cai et al. (2008), we found that juvenile lianas allocated

more biomass to leaves and stems and less to roots than trees. In a shade house experiment conducted in the same region and with some of the same species, Smith-Martin et al. (2017) found that under lower light levels seedlings allocated more biomass to leaves, whereas under higher light levels they allocated more to roots. Thus, juveniles may need to be more flexible in their allocation patterns to be able to respond to the environment and dynamically allocate more biomass depending on whether above- versus belowground resource availability is scarcest. Irrespective of the causes of different allocation patterns between juveniles and adults, these results imply that allocation in mature plants cannot be inferred by studying juvenile plants.

Rooting depths of adult lianas and trees

In the seasonally dry tropical forest we studied, lianas had the shallowest maximum rooting depths, followed by deciduous trees, and evergreen trees tended to have the deepest maximum rooting depths. This overall pattern was consistent, even after taking into account total biomass, diameter at breast height, allocation to leaves, stems, and roots, and total canopy area. Whether these patterns hold in other types of tropical forests is not known. Our findings that lianas do not have deeper roots than co-occurring trees are consistent with harvests of 4 yr old lianas and trees in a common garden in Panama (Smith-

Martin et al in review). Furthermore, a recent study using indirect methods to assess this question for wet forest in French Guiana showed that lianas do not use deeper sources of water than trees, even during the dry season (De Deurwaerder et al. 2018). As all the species of lianas that we studied are deciduous, it would be logical to expect that deciduous trees and lianas would have had similar rooting depth. One potential explanation for the

shallower rooting depth found in the lianas is that lianas only need their roots for resource uptake, whereas trees also need their roots for support, potentially explaining why deciduous trees had deeper roots despite having the same leaf phenology.

Our findings are contrary to the hypothesis that lianas are deep-rooted (Holbrook and Putz 1996, Andrade et al. 2005, Schnitzer 2005, Schnitzer et al. 2005, Toledo-Aceves and Swaine 2008, Schnitzer and Bongers 2011, Chen et al. 2015). We acknowledge the limitations of our study: it is difficult to harvest entire rooting systems of mature plants, the study of coarse roots does not allow us to distinguish roots designed for anchoring and support versus those designed for water uptake, and we have a limited sample size from only one location where lianas were deciduous and thus there may not have been the need for exceptionally deep roots. Nevertheless, our data suggest no evidence that lianas have deeper rooting depths than co-occurring trees in this dry forest, consistent with findings from another common garden study (Smith-Martin et al. in review), and thus the assumption of deeper liana rooting depths is not supported.

Our findings leave many important questions unresolved. How lianas maintain similar or better water status during the dry season (Smith-Martin et al. in review) and are able to grow more during the dry season while co-occurring trees grow more during the wet season in semi-moist forest still remains unknown (Schnitzer & van der Heijden 2019).

Smith-Martin et al. (in review) found that lianas appeared to explore more soil volume per stem diameter than did co-occurring trees, however in this study we did not find that lianas had greater root extensions per stem diameter, shoot biomass, or total biomass. Moreover, our measurements did not quantify fine root biomass (typically defined as roots < 2mm diameter), which may play a role in differentiating trees and lianas, if lianas have more fine

root mass. Clearly, there is more to be learned about comparative morphology and physiology of trees and lianas.

Implications of rooting depths for models of canopy phenology

Evergreen trees had the deepest roots of all the individuals that we harvested. One explanation for this pattern is that plants coordinate their belowground allocation with their aboveground water demand. Thus, to sustain leaves during the extended dry season at our study site, evergreen trees may be deeper rooted to access deeper sources of water during times when this resource is limited. To better illustrate this explanation, we used the

Ecosystem Demography model 2 (ED2) described by Xu et al. (2016) as a case study. We focused on the differences between deciduous and evergreen trees, because lianas are currently not represented in ED2 or any other ecosystem models. Specifically, we conducted simulations for six 0.1 ha plots in a Costa Rican tropical dry forest that differed parameterizations of rooting depth.

In our default (“DEFAULT”) simulation, we used the same rooting depth parameters as Xu et al. (2016). Specifically, we assigned a maximum rooting depth to the evergreen plant functional type (PFT E) that was about twice the rooting depth of the deciduous plant functional type (PFT D). In our “OBSERVATION” simulation, we assigned rooting depth parameters to the model’s PFT E and PFT D based on this study’s measurements (Sup. Fig.

3). In our “SWITCH” simulation, we assigned the measured PFT D rooting depth to the model’s PFT E, and vice-versa. In our “SAME” simulation, PFT E and PFT D were both assigned the same rooting depth parameter, calculated as the mean value of the

OBSERVATION PFT E and PFT D parameters (Fig. S2).

100

The degree of deciduousness simulated by the model indicates that PFT E can retain high enough water potential in the dry season and remain evergreen only by developing deeper rooting profiles relative to other PFTs (Fig. 5). This result holds despite the fact that

PFT E was parameterized to have the most drought-resistant aboveground traits (e.g. water potential at which 50% of hydraulic conductivity has been lost—P50, turgor loss point—

TLP) of any PFT. Thus, the simulations suggest that the differences between evergreen and deciduous trees might result from an ecological coordination between aboveground and below-ground processes. Moreover, the lianas that we measured in Costa Rica are deciduous in the dry season, which is again consistent with their observed shallow rooting depths. Our results provide evidence for such ecological coordination as well as valuable data sets for terrestrial biosphere models that consider vegetation demography (Fisher et al. 2018).

Conclusion

Taken together, our findings indicate that juvenile lianas, deciduous trees, and evergreen trees differ in allocation patterns, potentially consistent with the optimal partitioning theory, while mature individuals follow allometric biomass partitioning theory allocation patterns. Thus, it is not possible to infer the allometry and allocation patterns of mature plants from studies of seedlings and/or saplings. For the first time, we demonstrate that mature lianas follow the same biomass allocation patterns as co-occurring trees, and invest proportionally similar biomass to stems. Last, our results also show that lianas, which are deciduous at this site, had the shallowest rooting depths and evergreen tropical trees have the deepest roots, which are most likely needed to sustain their leaves during the

101

extended periods of seasonal drought as we show with the model simulation. There is an acute need to better understand how increasing liana abundances are impacting carbon storage and cycling in tropical forests (van Der Heijden et al. 2015). Studies such as ours that clarify the structure and implications of belowground biomass patterns in tropical forests are needed to inform both conceptual and simulation models of tropical forest dynamics and quantify the implications of global environmental change.

Acknowledgements

We thank NSF CAREER Grant Division of Environmental Biology 1053237 (to J.S.P.) and US Department of Energy, Office of Science, Office of Biological and Environmental

Research, Terrestrial Ecosystem Science Program under award number DESC0014363, and the University of Minnesota Natural History Award from the Dayton Bell Museum Fund (to

C.M.S.M.) for funding. We thank Milena Gutierrez L for logistical support at Estación

Experimetal Forestal Horizontes, and Damaris Pereira, David Pereira, Mario Blandon,

Pedro Alvarado, Jose Antonio, Tatiana Pereira, Maria José Gomez, and José Guevara for excellent help in the field and the lab. We also thank Daniel Pérez for helping with the species identification and Justin Touchon for statistical advice.

102

Tables and Figures

Species Family Age Growth form Leaf habit

Acacia tenuifolia Fabaceae juveniles, mature liana deciduous

Amphilophium Bignoniaceae juveniles, mature liana deciduous

crucigerum

Bauhinia ungulata Fabaceae juveniles tree deciduous

Combretum farinosum Combretaceae mature liana deciduous

Crescentia alata Bignoniaceae juveniles tree evergreen

Dalbergia retusa Fabaceae juveniles tree deciduous

Dioclea violacea Fabaceae juveniles liana deciduous

Gliricidia sepium Fabaceae juveniles, mature tree deciduous

Guazuma ulmifolia Malvaceae mature tree deciduous

Luehea speciosa Malvaceae juveniles, tree deciduous

mature

Lysiloma divaricatum Fabaceae juveniles tree deciduous

Manilkara chicle Sapotaceae mature tree evergreen

Securidaca diversifolia Polygalaceae mature liana deciduous

Serjania schiedeana Sapindaceae mature liana deciduous

103

Simarouba glauca Simaroubaceae juveniles, mature tree evergreen

Swietenia macrophylla Meliaceae juveniles tree deciduous

Tabebuia rosea Bignoniaceae juveniles, mature tree deciduous

Tanaecium Bignoniaceae juveniles liana deciduous

tetragonolobum

Figure 3-1. Image of above ground biomass of liana Combretum farinosum (a) and a root of evergreen tree Manilkara chicle in the process of being excavated (b).

104

Figure 3-2. Relationship between leaf and stem biomass, leaf and root biomass, and stem and root biomass of juvenile (a-c), and mature (d-f) deciduous trees (light green circles),

105

evergreen trees (dark green triangles), and lianas (brown squares). Juvenile lianas allocated proportionally more biomass to leaves, deciduous trees stems (a; χ2=10.07, P=0.007; Tukey deciduous – liana T=-2.95, P=0.045), and deciduous and evergreen trees to roots (b;

χ2=15.15, P=<0.001; Tukey deciduous–liana T=-3.64, P= 0.010; evergreen – liana T=-

2.49, P=0.051). Juvenile lianas also allocated more biomass to stems and deciduous and evergreen trees to roots (c; χ2=13.39, P=0.001; Tukey deciduous–liana T=-2.90, P=0.036; evergreen–liana T=-2.65, P= 0.036). Mature individuals allocated similar proportions of biomass to leaves and stem (d; χ2=5.26, P=0.072), leaves and roots (e; χ2=4.21 , P=0.122), and stems and roots (f; χ2=2.89, P=0.236). Shaded areas represents 95% confidence intervals and all variables have been log transformed.

106

Figure 3-3. Maximum rooting depth of all mature individuals. Lianas had the shallowest roots followed by deciduous trees and evergreen trees had the deepest roots (χ2=34.33,

P<0.001; Tukey deciduous– evergreen T=-3.205, P= 0.026; deciduous–liana T=3.14, P=

0.034; evergreen – liana T=5.70, P<0.001).

107

Figure 3-4. Relationship between maximum rooting depth and leaf biomass (a), total canopy area (b), stem biomass (c), root biomass (d), total biomass (e), and diameter at breast height (f) of mature deciduous trees (light green circles), evergreen trees (dark green

108

triangles), and lianas (brown squares). Deciduous and evergreen trees had deeper roots than lianas independent of leaf biomass (a; χ2=13.70, P<0.001; Tukey deciduous–liana T=2.77,

P=0.042 evergreen–liana T=3.47, P=0.011), and total leaf canopy area (b; χ2=21.45,

P>0.001; Tukey deciduous–liana T=3.36, P=0.019; evergreen–liana T=4.24, P=0.005).

Evergreen trees had deeper roots than deciduous trees and lianas independent of stem biomass (c; χ2=16.81, P<0.001; Tukey evergreen–deciduous T=-3.11, P=0.033; evergreen– liana T=3.88, P=0.008), root biomass (d; χ2=10.50, P=0.005; Tukey evergreen—deciduous

T=-2.78, P=0.054; evergreen—liana T=2.62, P=0.055), and total biomass (e; χ2=15.18,

P<0.001; Tukey evergreen—deciduous T=-3.09, P=0.034; evergreen—liana T=3.58,

P=0.012). Evergreen trees still had deeper roots than lianas independent of diameter at breast height (f; χ2=11.05, P=0.004; Tukey T=3.23, P=0.019). Shaded areas represents 95% confidence intervals and all variables have been log transformed.

109

Figure 3-5. The simulated dry season deciduousness (from 2009 to 2014) for coexisting deciduous (PFT D) and evergreen (PFT E) functional types in a Costa Rican tropical dry forests under different parameterization for rooting depth allometry based on Xu et al.

(2016). The dry season deciduousness is defined as the average ratio of active leaf area over maximum leaf area in April, the peak of dry season. Higher leaf cover means more evergreenness.

110

Supporting Information

Supplemental Figure 1 Relationship between shoot biomass and root biomass of juvenile deciduous trees (light green circles), evergreen trees (dark green triangles), and lianas

(brown squares). As they became larger lianas allocated more biomass to shoots and deciduous and evergreen trees to roots (χ2=16.80, P<0.001; Tukey deciduous–liana T=-

3.60, P=0.011; evergreen–liana T=-2.96, P= 0.018). Shaded areas represents 95% confidence intervals and all variables have been log transformed.

111

112

Supplemental Figure 2 Relationship between root length and diameter at breast height

(a), shoot biomass (b), and total biomass (c) of mature deciduous trees (light green circles), evergreen trees (dark green triangles), and lianas (brown squares). We found no difference among total root length of deciduous trees, evergreen trees and lianas when accounting for

DBH (Sup. Fig. 2a; linear mixed model χ2=2.81, P=0.246), shoot biomass (Sup. Fig. 2a; linear mixed model χ2=2.71, P=0.259), and total biomass (Sup. Fig. 2a; linear mixed model

χ2=0.267, P=0.263).Shaded areas represents 95% confidence intervals and all variables have been log transformed.

Supplemental Figure 3 Parameterization of rooting depth allometry based on observation in this study. The ED2 model assumes an exponential distribution of root biomass and we assume the observed maximum rooting depth is equivalent to the depth, above which 99% of root biomass reside. Dots represent observations and solid lines represent parameters used in the model based on the observation.

113

Supplemental Table 1. Intercepts and slopes from the mixed models performed in the analysis.

Figure Variables Age Group Intercept Intercept SE Slope Slope SE

Fig. 2 (a) leaf biomass x stem juvenile deciduous -0.51 0.29 0.96 0.06

biomass

Fig. 2 (a) leaf biomass x stem juvenile evergreen 0.34 0.80 0.80 0.24

biomass

Fig. 2 (a) leaf biomass x stem juvenile liana 0.16 0.70 0.90 0.14

biomass

Fig. 2 (b) leaf biomass x root juvenile deciduous -0.12 0.38 1.02 0.09

biomass

Fig. 2 (b) leaf biomass x root juvenile evergreen 0.27 1.06 0.80 0.32

biomass

Fig. 2 (b) leaf biomass x root juvenile liana 0.44 0.91 1.12 0.21

biomass

Fig. 2 (c) stem biomass x root juvenile deciduous 0.56 0.31 1.03 0.07

biomass

Fig. 2 (c) stem biomass x root juvenile evergreen -0.07 0.88 1.00 0.26

biomass

Fig. 2 (c) stem biomass x root juvenile liana 0.44 0.75 1.20 0.18

biomass

Fig. 2 (d) leaf biomass x stem mature deciduous -1.76 2.29 0.89 0.24

biomass

Fig. 2 (d) leaf biomass x stem mature evergreen 5.42 6.45 0.33 0.53

biomass

114

Fig. 2 (d) leaf biomass x stem mature liana 4.55 5.94 0.34 0.49

biomass

Fig. 2 (e) leaf biomass x root mature deciduous 2.45 2.27 0.61 0.21

biomass

Fig. 2 (e) leaf biomass x root mature evergreen 2.83 8.62 0.64 0.81

biomass

Fig. 2 (e) leaf biomass x root mature liana 4.91 4.76 0.38 0.45

biomass

Fig. 2 (f) stem biomass x root mature deciduous 4.51 2.52 0.70 0.23

biomass

Fig. 2 (f) stem biomass x root mature evergreen -8.59 9.46 1.98 0.89

biomass

Fig. 2 (f) stem biomass x root mature liana 1.62 5.28 1.04 0.50

biomass

Fig. 4 (a) maximum rooting depth mature deciduous 5.07 2.05 -0.02 0.22

x leaf biomass

Fig. 4 (a) maximum rooting depth mature evergreen 5.95 8.11 -0.05 0.87

x leaf biomass

Fig. 4 (a) maximum rooting depth mature liana 1.79 4.69 0.26 0.52

x leaf biomass

Fig. 4 (b) maximum rooting depth mature deciduous 6.27 2.41 -0.10 0.17

x canopy area

Fig. 4 (b) maximum rooting depth mature evergreen 7.13 8.45 -0.12 0.61

x canopy area

Fig. 4 (b) maximum rooting depth mature liana 3.35 5.47 0.04 0.40

x canopy area

115

Fig. 4 (c) maximum rooting depth mature deciduous 3.27 2.30 0.13 0.19

x stem biomass

Fig. 4 (c) maximum rooting depth mature evergreen 1.34 6.67 0.35 0.56

x stem biomass

Fig. 4 (c) maximum rooting depth mature liana -0.29 6.67 0.41 0.39

x stem biomass

Fig. 4 (d) maximum rooting depth mature deciduous 3.66 1.92 0.11 0.18

x root biomass

Fig. 4 (d) maximum rooting depth mature evergreen -2.53 7.58 0.77 0.72

x root biomass

Fig. 4 (d) maximum rooting depth mature liana 0.85 4.02 0.37 0.38

x root biomass

Fig. 4 (e) maximum rooting depth mature deciduous 3.26 2.38 0.13 0.19

x total biomass

Fig. 4 (e) maximum rooting depth mature evergreen 0.23 7.44 0.43 0.61

x total biomass

Fig. 4 (e) maximum rooting depth mature liana -0.70 4.95 0.44 0.40

x total biomass

Fig. 4 (f) maximum rooting depth mature deciduous 3.07 0.72 0.59 0.31

x diameter at breast

height

Fig. 4 (f) maximum rooting depth mature evergreen 9.61 0.99 1.85 2.48

x diameter at breast

height

116

Fig. 4 (f) maximum rooting depth mature liana 2.48 6.26 0.34 1.22

x diameter at breast

height

117

Chapter 4: Effects of dry-season irrigation on leaf physiology

and biomass allocation in tropical lianas and trees

Smith-Martin, C. M., C. L. Bastos, O. R. Lopez, J. S. Powers, and S. A. Schnitzer. 2019.

Effects of dry-season irrigation on leaf physiology and biomass allocation in tropical lianas and trees. In review at Ecology.

Synopsis

Lianas are more abundant in seasonal forests than in wetter forests and are thought to perform better than trees when light is abundant and water is limited. We tested the hypothesis that lianas perform better than trees during seasonal drought using a common garden experiment with 12 taxonomically diverse species (6 liana and 6 tree species) in 12 replicated plots. We irrigated six of the plots during the dry season for four years, while the remaining 6 control plots received only ambient rainfall. In year 5, we measured stem diameters for all individuals and harvested above- and belowground biomass for a subset of individuals to quantify absolute growth and biomass allocation to roots, stems, and leaves, as well as total root length and maximum rooting depth. We also measured photosynthesis, intrinsic water use efficiency (iWUE), pre-dawn and midday water potential, and a set of functional and hydraulic traits. During the peak of the dry season, lianas in control plots had 54% higher predawn leaf water potentials (ΨPD), and 45% higher photosynthetic rates than trees in control plots. By contrast, during the peak of the wet season, these physiological differences between lianas and trees become less pronounced

118

and, in some cases, even disappeared. Trees had higher SLA than lianas; however, no other functional trait differed between growth forms. Trees responded to the irrigation treatment with 15% larger diameters and 119% greater biomass than trees in control plots. Liana growth, however, did not respond to irrigation; liana diameter and biomass were similar in control and irrigation plots, suggesting that lianas were far less limited by soil moisture than were trees. Contrary to previous hypotheses, lianas did not have deeper roots than trees; however, lianas had longer roots per stem diameter than did trees. Our results support the hypothesis that lianas perform better and experience less physiological stress than trees during seasonal drought, suggesting clear differences between growth forms in response to altered rainfall regimes. Ultimately, better dry-season performance may explain why liana abundance peaks in seasonal forests compared to trees, which peak in abundance in less seasonal, wetter forests.

Keywords: climate change; drought; liana increases; rooting depth; seasonality; trees; tropical forests; woody vine

Introduction

Understanding the factors that control the abundance and distribution patterns of plant species is a fundamental goal in ecology. One of the main drivers of species distributions in tropical biomes is rainfall gradients, and plant species composition and forest structure shift considerably across broad precipitation gradients (Beard 1944,

Engelbrecht et al. 2007). The abundance of most tropical vascular plant groups (e.g., epiphytes, herbs, palms, and trees) increases with annual rainfall (Gentry 1995; (Clinebell et al. 1995). Lianas (woody vines), however, seem to follow an opposite distribution

119

pattern. Lianas reach maximum relative abundance in seasonally dry tropical forests, and decrease in abundance as mean annual rainfall increases (Gentry 1995, Schnitzer 2005,

DeWalt et al. 2015). For example, Schnitzer (2005) found that there were five-times more lianas in the drier forest compared to the wet forest across the rainfall gradient of the isthmus of Panama. Schnitzer (2005) also reported that lianas were considerably more abundant in highly seasonal forests compared to wet forests across a large number of tropical forests, especially compared to trees (Gentry 1991, Swaine and Grace 2007,

Dewalt et al. 2010). An explanation for lianas being more abundant relative to trees in seasonal forests is that lianas perform better than trees when water is limited because light is plentiful during dry periods (Schnitzer 2018). This seasonal growth advantage, however, is thought to be lost in aseasonal wet forests, where light is low year-round, thus explaining pan-tropical liana distribution (Schnitzer 2005). For example, in a seasonal forest in

Panama, understory lianas were found to grow in height seven times-more than understory trees during the dry season, but only two times more during the wet season (Schnitzer

(2005). In a more recent and much larger study in Panama, canopy lianas had a far higher growth rate during the dry season than the wet season, whereas trees had the opposite pattern, with a higher growth rate during the wet season than the dry season (Schnitzer & van der Heijden 2019).

There are several potential explanations for the ability of lianas to perform better than trees during seasonal drought. Lianas may grow well during the dry season because they have higher dry season photosynthetic rates, hydraulic conductance, predawn leaf water potential, and more efficient water use compared to trees (Cai et al. 2009, Zhu and

Cao 2009, Chen et al. 2015). Lianas may also access and use water more efficiently than

120

trees (Schnitzer 2005) and may have better developed and deeper root systems than co- occurring trees (Restom and Nepstad 2004), thus, permitting lianas access to deeper sources of water (Andrade et al. 2005, Chen et al. 2015).

An alternative explanation is that lianas have a suite of functional traits that allow them to grow when light is abundant even though water is limited. Lianas tend to be on the fast end of the leaf economic spectrum, with higher specific leaf area (SLA) and foliar nutrient content (Zhu and Cao 2009, Asner and Martin 2012, Wyka et al. 2013, Werden et al. 2017). However, more conservative resource-use strategies, such as higher water use efficiency, denser wood, and lower leaf turgor loss points are associated with greater drought resistance (Bartlett et al. 2014, Reich 2014, Greenwood et al. 2017, O'Brien et al.

2017). Thus, resolving whether lianas and trees differ systematically with respect to both leaf economics and hydraulic traits may help shed light on their distribution patterns and potential responses to global changes such as altered rainfall regimes.

Here we tested the hypothesis that lianas perform better than trees during seasonal drought with a common-garden irrigation experiment. We measured tree and liana water status and photosynthesis throughout the year, as well as functional traits for both plant groups. We predicted that lianas and trees would perform similarly during the wet season, but that during the dry season, lianas would experience less water stress and have higher physiological performance than trees. By contrast, we predicted that in plots that received dry season irrigation, both lianas and trees would have a similar level of water stress, stomatal conductance, and photosynthesis, which would be more similar to what they experience during the wet season. We predicted that lianas would have greater rooting depth, greater ratio of root length to diameter, better water use efficiency, lower turgor loss

121

point, and less dense wood. We also predicted that dry season irrigation would increase the size of both lianas and trees, but that trees would have a stronger response than lianas, indicating that they suffer more water stress than co-occurring lianas.

Materials and Methods

Experimental design

In 2011, we established a common garden at Parque Municipal Summit, located near the middle of the rainfall gradient along the isthmus of Panama. The mean annual rainfall at this location is 2226 mm and there is a distinct 4-month dry season (months with a rainfall of <100 mm) from December through April (Condit et al. 2000, Condit et al.

2004, Wishnie et al. 2007). In 2011, we established twelve 9 x 9 m plots that were each separated by at least 5 meters in an open field, and planted replicate seedlings of the same six common species of native lianas (Callichlamys latifolia, Davilla kunthii, Doliocarpus major, Machaerium milleflorum, Maripa panamensis, Paullinia pinnata) and six common species of native trees (Crescentia cujete, Dipteryx oleifera, Hieronyma alchorneoides,

Hura crepitans, Swietenia macrophylla, Terminalia amazonia) in all 12 plots (Appendix

S1: Table S1). We planted on average five individuals per species per plot for a total of N

= 60 individuals per species. All focal species were evergreen or brevi-deciduous at this location. We thoroughly irrigated six of the plots with sprinklers at least two days a week throughout the dry season (from January through April), while the six control plots did not receive dry season supplemental irrigation. To irrigate the plots we used one sprinkler per plot with a spray that reached the entire plot and irrigated with municipal water for

122

approximately six hours per plot, and the soil in the plots was saturated after each watering.

We installed trellises for the lianas to climb, and lianas were prevented from growing on the trees. We also put trellises next to the trees so that both growth forms would experience the same environmental conditions. We routinely mowed the grass within and around the plots so that herbaceous vegetation would not interfere with the experiment. We allowed the plants to grow for four years, and in February 2015, when many of them had reached reproductive stage and all the lianas were twining and very few of them were freestanding, we started collecting measurements (Appendix S2: Table S2).

Leaf water potential, soil moisture, and gas exchange

We randomly selected ~24 individuals of each species (1–2 individuals per plot) on which we measured leaf water status throughout the year. We measured leaf water potential at pre-dawn (ΨPD) and mid-day (ΨMD) on a total of 127 lianas and 100 trees from both irrigated and control plots on five separate sampling times throughout 2015: twice during the dry season (March and April), once during the transition from dry to wet season (May), and twice during the wet season (September and October). We performed these measurements using a 1505D Pressure Chamber Instrument (PMS Instrument Company,

Portland, Oregon). On the same day that we measured leaf water potential, we also measured soil moisture in the top 5 cm of soil at the base of each individual (~24 measurements per plot) with a soil moisture sensor (SM150, Delta-T Devices, Burwell,

UK) in March, April, May, and September. Soils were uniform throughout the common garden and appeared to be oxisols. We measured leaf gas exchange with an LI-6400 portable photosynthesis system (Li-Cor, Lincoln, Nebraska) once during the dry season

123

(March) and once during the wet season (October). The CO2 concentration was set to 400

µmol mol-1, photosynthetic photon flux density was set to 1200 µmol m-2 s-1, and the temperature was set to 32 ºC. Due to sampling constraints, we were able to take these measurements only for plants in the control plots. Nonetheless, the dry and wet season comparisons between lianas and trees allowed us to test the hypothesis that lianas perform better than trees during the dry season but not the wet season. We calculated intrinsic water use efficiency (iWUE) as the ratio of photosynthetic rate to stomatal conductance (gs).

Diameter, biomass, and rooting depth and length

Four years after the treatment began, we measured the diameter (20 cm from the base) for all individuals in the common garden (12–42 individuals per species per treatment, depending on survival rates at the time of measuring). We harvested above- and belowground biomass of a subsample of individuals from both irrigation treatments, which included all of the species except for Hieronyma alchorneoides. Whole-plant biomass harvests included a total of 12 trees (six from the irrigated and six from the control plots) and 27 lianas (15 from the irrigated and 12 from the control plots). For all the harvested individuals, we weighed all leaves, stems, and roots, and measured root lengths. For each individual, we collected a subsample of leaves, stems, and roots, which we dried at 60 °C for >72 hours until they reached constant weight, and then were weighed. Oven-dried weights were used to estimate total dry biomass for each tissue type per individual. We calculated the fraction of total biomass allocated to leaves (leaf mass fraction — LMF), stems (stem mass fraction —SMF), and to roots (root mass fraction —RMF). For all of the harvested individuals, we calculated the sum of all root lengths (hereafter root length) and

124

the maximum rooting depth. We also calculated the ratio between root length to the stem diameter at 20 cm from the base.

Leaf chemistry, specific leaf are, and hydraulic traits

We chose one individual from three of the irrigated plots and one from three of the control plots for all trait measurement (six individuals per species). We measured leaf carbon (C), nitrogen (N), and C stable isotopic concentrations (δ13C) on an Isotope Ratio

Mass Spectrometer at the Smithsonian Tropical Research Institute Soils Laboratory in

Panama. We collected the leaves during the wet season and, depending on leaf size, we ground, bulked, and analyzed three to five leaves per individual. We used the stable isotope of carbon, δ13C, as a proxy for integrated water use efficiency (WUE). We obtained specific leaf area (SLA) by measuring leaf area including petioles on three leaves from each of the chosen individuals using a LI-3100C Area Meter. We then quantified leaf dry weight by drying each leaf at 60 °C until it reached constant weight. We calculated SLA (cm2g-1) as the leaf area divided by dry leaf mass, and we calculated the average of three measurements per individual plant.

We calculated leaf turgor loss point (TLP) using the bench dry method to measure pressure-volume curves as described in Tyree and Hammel (1972). For each species, we created TLP curve for three individuals from the irrigated plots and three from the control plots. We selected each individual from a different plot. From the pressure-volume curves we calculated leaf relative water content at TLP (RWCTLP), leaf capacitance at TLP (CTLP), and absolute capacitance at full turgor (ACFT) (Ewers et al. 1997, Sack et al. 2003). All

125

hydraulic trait measurements were made during the wet season on fully expanded healthy leaves.

We quantified stem and root wood density for all the harvested individuals following

Chave et al. (2006). Specifically, stem and root wood density (WD) were calculated as the ratio of dry weight to fresh volume g cm-3. We collected fresh wood samples immediately after harvesting and estimated their volume as the weight of water needed to displace the sample. We then dried the samples at 60 °C until they reached constant weight and weighed them.

Data analysis

Our experimental design allowed for comparisons of response variables that were collected multiple times (i.e. soil moisture, plant water status,) between both seasons and irrigation treatments. For soil moisture, we used a mixed model with irrigation treatment

(irrigated and control), and sampling time (month) as fixed effects and plot as a random effect (Appendix S2: Table S2). For Ψ, we calculated diurnal regulation as ∆Ψ (ΨMD –

ΨPD). ΨPD, ΨMD, and ∆Ψ were analyzed using mixed models with irrigation treatment, growth form (liana and tree), and sampling time as fixed effects and species and individual as random effects (Appendix S2: Table S2). We analyzed photosynthetic rates, gs, and iWUE using a mixed effects model with season (dry and wet) and growth form as fixed effects, and species and individual as random effects (Appendix S2: Table S2).

For variables that we measured once at the end of the experiment (i.e. stem diameter at 20 cm, total biomass, LMF, SMF, RMF, maximum rooting depth, root length, ratio of root length to stem diameter), we used mixed models with growth form and irrigation

126

treatment as fixed effects and species as a random effect (Appendix S2: Table S2). We log transformed stem diameter at 20 cm, total biomass, maximum rooting depth, sum of root length, and the ratio of root length to stem diameter to account for the orders of magnitude of difference in the data. Because of the large difference in biomass among focal species, we calculated the mean irrigation enhancement of biomass for each species as follows:

휇 푏𝑖표푚푎푠푠 𝑖푟푟𝑖푔푎푡푒푑 − 휇 푏𝑖표푚푎푠푠 푐표푛푡푟표푙 퐼푟푟𝑖푔푎푡𝑖표푛 푒푛ℎ푎푛푐푒푚푒푛푡 표푓 푏𝑖표푚푎푠푠 = 휇 푏𝑖표푚푎푠푠 푐표푛푡푟표푙 where “µ” is the species mean.

We compared the irrigation enhancement of biomass between lianas and trees

(hereafter “growth form”) using a mixed model with growth form as a fixed effect and species as a random effect. We log transformed the data because of the large differences in biomass among species. One of the species (Machaerium milleflorum) responded negatively to irrigation, so we added 1 to all of the data to yield positive values prior to log transformation. Irrigation treatment did not have a significant effect on the following response variables: SLA, percent N, C/N ratio, WUE, TLP, RWCTLP, CTLP, ACFT, stem

WD, and root WD, therefore, for these variables we used mixed models with only growth form as a fixed effect and species as a random effect (Appendix S2: Table S2). We log transformed SLA. All analyses were performed in R (Version 3.3.3). We conducted all the mixed models using the nlme R package (Pinheiro J. et al. 2018). We used least squares means pairwise comparisons with the lsmeans R package (Russell V. L. 2016) to interpret significant differences among main effects.

127

Results

Soil moisture

The supplemental irrigation treatment increased soil moisture in the irrigated plots

(F1, 10 = 8.56, P=0.015). During the dry season, mean percent soil moisture in the irrigated plots was 4.4% higher in March than the control plots (24.86 vs 20.51%; T = 2.93, DF=

10, P=0.015), and 2.7% higher in April (22.32 vs 19.60% ; T = 1.94, DF= 10, P=0.081).

There was dramatic seasonal variation in soil moisture throughout the year (F3, 885 = 742.26,

P<0.001), although this was less pronounced in the irrigated plots (Appendix S3: Fig. S1).

However, no irrigation treatment was applied during the wet season and there was no significant difference between the irrigated and control plots during May (27.00 vs 25.39% respectively; T = 1.24, DF= 10, P=0.24) and September (45.74 vs 44.04% respectively; T

= 1.30 DF= 10, P=0.22).

Leaf water potential

During the wet season (May, September, and October), lianas and trees had similar

ΨPD (Fig. 1a). During the dry season, however, lianas in the control plots had significantly higher (less negative) water status than trees (higher ΨPD, F4,781 = 9.26, P<0.001; Fig. 1a;

Appendix S4: Table S3). The largest difference between liana and tree water status occurred for the pre-dawn water potential at the end of the dry season (in April), where the mean ΨPD of lianas in control plots was 54% higher than trees in control plots (-0.92 vs. -

1.42 MPa respectively). Lianas in the irrigated plots were also less water stressed than the trees in the irrigated plots during the dry season but the difference was slightly lower than in the control plots (Fig. 1a; Appendix S4: Table S3), with the lianas in irrigated plots

128

having only 43% higher mean values than trees in irrigated plots in April (-0.83 vs -0.58

MPa respectively). Lianas also had higher (less negative) midday dry-season water potentials (ΨMD) than trees in the control plots, an effect that disappeared with the irrigation treatment. Both lianas and trees in irrigated plots had less negative ΨMD than the ones in the control plots during the dry season (F1,235 = 18.49, P<0.001; Fig. 1b; Appendix S5:

Table S4), indicating that the irrigation treatment affected plant water status. These findings suggest that lianas were far less water stressed than the tress in the control plots, and that trees responded more to dry season irrigation than did lianas. Indeed, mean ∆Ψ was higher for lianas than trees at each measurement period; however, these differences were not statistically significant (F1,235 = 0.76, P=0.383; Fig. 1c).

Plant diameter, biomass, and rooting dynamics

Trees grew more in response to irrigation than did lianas (F1, 597 = 9.84, P=0.002;

Fig. 2a), with trees in irrigated plots having on average 15% larger diameters than the trees in control plots (T =3.58, df = 597, P<0.001). By contrast, mean liana diameter did not differ between irrigated and control plots (T =0.87, df = 597, P=0.383). Trees in irrigated plots also had much greater biomass than trees in control plots, whereas lianas had similar biomass independent of irrigation treatment. Mean tree biomass in the irrigated plots was

119% greater than tree biomass in the control plots, whereas liana biomass in the irrigated plots was only 26% greater than liana biomass in the control plots (Fig. 2b). Variation in biomass was high among individuals and plots; nevertheless, the growth form by irrigation treatment interaction was marginally significant (F1, 26 = 3.51, P=0.072). The mean of the calculated irrigation enhancement of biomass was 14-times greater for trees than for lianas

129

(8.40 vs 0.60 g g-1 respectively; Fig. 2c) indicating that the trees grown in the irrigated plots were much larger compared to individuals of the same species in the control plots.

However, despite the clear trend, the variation among individuals and species within growth forms was high and thus result was not statistically significant (F1, 9 = 2.99, P=0.12).

Lianas in control plots allocated 33.5% more of their biomass to leaves than the lianas in irrigated plots (F 1, 26 = 7.91, P = 0.009; T = 2.81, df = 26, P =0.043; Appendix

S6: Fig. S2a), whereas the irrigated lianas allocated more to stems (although this trend was not statistically significant; Appendix S6: Fig. S2b). Independent of irrigation treatment, lianas allocated similar fractions of their biomass to roots (Appendix S6: Fig. S2c). By contrast, trees allocated similar proportions of their total biomass to leaves independent of irrigation treatment (Appendix S6: Fig. S2a). For allocation to roots, there was a significant interaction between growth form and irrigation treatment (F 1, 26 = 4.558, P = 0.042), with lianas allocating similar fractions of biomass to roots independent of irrigation treatment, whereas trees allocated more biomass to roots when they were unirrigated. The irrigation treatment revealed a trade-off between allocation to stems versus roots for trees; trees in the control treatment allocated on average 38% more to roots at the expense of stems (T =

2.75, df = 26, P =0.050; Appendix S6: Fig. S2c), while the opposite pattern was true for trees in the irrigated treatment, with 20% greater allocation to stems (although this trend was not statistically significant; Appendix S6: Fig. S2b).

Despite clear differences in biomass, growth, and allocation, rooting depth showed no significant difference between growth forms (F 1, 9= 0.86, P = 0.377) or irrigation treatment (F 1,26= 0.15, P = 0.700; Appendix S7: Fig. S3a). For root length, there was no significant effect of between growth form, irrigation treatment, and the interaction between

130

growth form and irrigation treatment (F 1, 19 = 3.48, P = 0.078; Appendix S7: Fig. S3b).

However, lianas had a greater ratio of root length to stem diameter in both control and irrigated plots (F 1, 9 = 17.14, P = 0.003; Appendix S7: Fig. S3c), indicating that lianas allocate more to root length for a given stem diameter compared to trees in the control plots.

Photosynthesis and water use efficiency

Both lianas and trees had higher photosynthetic rates during the dry season than the wet season (F1, 105= 31.30, P <0.001; Fig. 3a). Lianas had higher photosynthetic rates than trees during both seasons (F1, 105= 8.19, P =0.005); however, the magnitude of these differences varied seasonally. During the dry season, lianas had 45% higher photosynthetic rates than trees (T = 2.86, DF= 105, P=0.024), but only 17% higher photosynthetic rates during the wet season (T = 2.06, DF= 105, P=0.178). That is, photosynthetic rate decreased significantly more for trees than for lianas during the dry season (Fig. 3a). Both lianas and trees had higher gs (F1, 105=136.42, P <0.001; Fig. 3b) during the wet season, and higher iWUE (F1, 105= 81.06, P <0.001; Fig. 3c) during the dry season.

Specific leaf area, leaf chemistry, and hydraulic traits

Aside from SLA, there were no differences in functional traits between growth forms. On average, trees had higher SLA than lianas (122.53 vs. 91.54 cm2g-1 respectively;

F1, 12= 10.66, P=0.009; Appendix S8: Table S5). However, lianas and trees had similar foliar N concentrations (F1, 10= 0.03, P =0.870; Appendix S8: Table S5), foliar C/N ratios

(F1, 10= 0.01, P =0.938;. Appendix S8: Table S5), and WUE (F1, 10= 1.764, P =0.214;

131

Appendix S8: Table S5), turgor loss point (F1, 10= 0.001, P =0.983; Appendix S8: Table

S5), relative water content at turgor loss point (F1, 10= 4.367, P=0.063; Appendix S8: Table

S5), capacitance at turgor loss point (F1, 10 = 3.583, P=0.088; Appendix S8: Table S5), and absolute capacitance at turgor loss point (F1, 10= 1.824, P=0.207; Appendix S8: Table S5).

Lianas and trees also had similar wood densities for stems (F 1, 9= 1.77, P = 0.216; Appendix

S8: Table S5) and roots (F 1, 9= 0.01, P = 0.936; Appendix S8: Table S5). The lack of difference in foliar N, δ13C, TLP, and wood density between lianas and trees limits our ability to array trees and lianas along the acquisitive-conservative trait continuum, thus we do not discuss these results further. One potential reason why we did not find any differences among liana and tree TLP is because we conducted these measurements in the wet season, and not in the dry season, as has been done in other studies when a greater range of value among species has been found (Maréchaux et al. 2015, Maréchaux et al.

2017).

Discussion

Our experiment is the first to examine the long-term effects of dry season irrigation on lianas and trees in a highly controlled environment (e.g. all individuals had the same age and access to similar levels of light in the open-field common garden). Our study is also the first to examine how dry season irrigation affects plant growth with a whole plant harvest after four years of treatment. One of our key findings was that lianas were able to maintain higher photosynthetic levels while experiencing less drought stress than co- occurring trees during the dry season. Another key finding was that trees responded more to irrigation than the lianas by growing far more in terms of diameter and total biomass,

132

which also suggests that the trees were more drought stressed during the dry season than were lianas. Our final key findings were that lianas had greater ratio of root length to diameter than control trees; however contrary to what we expected, lianas were not deeper rooted, did not have better water use efficiency, lower turgor loss point, and less dense wood than trees.

Liana and tree response to seasonal drought

The response of trees to dry-season irrigation indicates that trees in this study were limited by soil moisture during the dry season. By contrast, lianas in this study did not respond to increased soil moisture, and yet were able to maintain higher water potential and photosynthesis than trees during the dry season without dry season irrigation. The combination of higher photosynthesis levels and greater physiological regulation (stomatal control) could allow lianas to function more efficiently than trees when water is limited during seasonal drought. The ability of lianas to perform better than competing trees during seasonal drought gives lianas a seasonal growth advantage over trees in seasonal forests that they lack in wet forests, thus explaining higher liana abundance in seasonal forests because they have an extra 3 to 5 months of growth that they lack in ever-wet forests

(Schnitzer 2005, Schnitzer 2018). Our results are consistent with Schnitzer (2005), who also found that lianas grew more than trees when water was limited. Similarly, in a recent study of > 1100 canopy trees and > 800 canopy lianas in central Panama, Schnitzer and van der Heijden (2019) reported that liana growth rate was higher during the dry season than the wet season in each of the five years of their study, and that lianas grew as much during the ~4-month dry season as they did during the ~8-month wet season. By contrast,

133

they found that tree growth rate was far higher during the wet season, and that trees grew far more during the 8-month wet season than the 4-month dry season.

Our results are also consistent with previous studies that reported that lianas maintained better plant water status and higher rates of photosynthesis during seasonal drought than trees. For example, Zhu and Cao (2009) found that lianas growing in a seasonal tropical rainforest in China had less negative predawn leaf water potential at the end of the dry season than co-occurring trees. Chen et al. (2015) also found that during the dry season, in the karst forest in Xishuangbanna in Southwest China, lianas had much higher predawn leaf water potential than trees. In our study, however, trees responded to the irrigation treatment whereas lianas did not, demonstrating that the trees were more drought stressed than the lianas, and that this drought stress was caused, in part, by limited soil moisture. Maintaining plant water status during seasonal drought, when light levels are high, could permit lianas to perform better than trees in seasonal forests. In our study, lianas had higher rates of photosynthesis than co-occurring trees during seasonal drought similar to Cai et al. (2009). However, based on our results, higher photosynthesis in lianas compared to trees appears to be diminished when water is readily available. The combination of maintaining healthy water status and being able to maximize photosynthesis during the dry season, when light is plentiful, could explain why lianas reach highest relative abundance in seasonally dry tropical forests compared to wetter forests (Schnitzer 2018).

134

Root dynamics of lianas and trees

Previous studies suggested that lianas may grow better during seasonal drought because of deep root systems (Schnitzer 2005). For example, Restom and Nepstad (2004) had found that 15-year-old lianas of one of the same species that we used in our experiment

(Davilla kunthii) had roots up to 10 m deep. However, Restom and Nepstad (2004) did not excavate co-occurring trees, thus, it is difficult to determine whether lianas had deeper rooting depths compared to trees in their study. In the karst forest of Southwest China,

Chen et al. (2015) found that during the dry season, lianas used deeper soil water than co- occurring trees based on hydrogen stable isotope concentrations in xylem water. By contrast, we found that lianas had similar rooting depths as trees, and were not deeper rooted as we had predicted. Johnson et al. (2013) also found that lianas did not have deeper roots than a co-occurring tree species in a nearby forest in Panama. Other studies using stable isotopes found evidence that lianas were not using deep sources of water. For example, two studies conducted in Panama, in which stable isotopes of hydrogen were used to track water use in lianas and trees, found that lianas shifted to deeper water sources during the dry season whereas large trees did not; however, deeper sources of water was only measured at ~60 cm (Meinzer et al. 1999, Andrade et al. 2005). De Deurwaerder et al. (2018) used stable isotopes of hydrogen and oxygen and found that lianas used more superficial water in than trees during the dry season in a rainforest in French Guiana.

Our findings were similar in that we found that lianas were able to maintain water status better than trees during the dry season without being deeper rooted. Apparently, extracting water from ~20-100 cm depth in the soil profile was deep enough to maintain relatively healthy water status and function throughout the dry season at this location. The

135

absolute rooting depth of plants may not be the most important predictor of the ability of plants to take up water, and rooting depth likely depends on where the extractable water is located at a given site, which can change considerably with soil type and the geology of the site (Manzané-Pinzón et al. 2018).

Root length may also contribute to water access and uptake when water is limited. In our study, lianas had greater ratios of root lengths to diameter than the trees in the control plots, independent of irrigation treatment, which suggests that lianas were able to access greater soil volumes and potentially more water relative to their size than the drought-stressed trees. Thus, lianas may rely on more extensive root systems with more lateral roots to maintain water status during seasonal drought. Lianas did not alter their root growth patterns in relation to the irrigation treatment, whereas trees did, providing additional evidence that the alleviation of soil water limitation has less of an effect on lianas than trees. By contrast, we found that trees in the control plots allocated greater proportions of their biomass to roots than the ones in the irrigated plots, likely to increase water uptake.

Our findings for tree allocation patterns in response to dry season irrigation is consistent with earlier theory (e.g.,) Bloom et al. (1985), which suggests that plants will trade the higher carbon cost of root production to increase water acquisition capacity when water is limited.

Liana growth patterns and forest carbon storage

The ability of lianas to grow well during drought may explain why lianas are increasing in abundance and biomass in tropical forests. The pattern of increasing liana abundance was first documented in non-fragmented old growth Amazonian forests, where

136

Phillips et al. (2002) discovered that, over a 20 year period, lianas had increased in density, basal area, and size, with the increase becoming greater over time. Ingwell et al. (2010) found that the proportion of tree canopies infested with lianas in the seasonal tropical moist forest on Barro Colorado Island in Panama had increased from 32% to 75% in the last 50 years. Laurance et al. (2014) also found that over a 14-year period in an undisturbed rainforest in central Amazon, liana abundance increased by 1% per year. Subsequent studies have also presented evidence to support increases in lianas abundance, biomass, or dominance across Neotropical forests (Wright et al. 2004b, Yorke et al. 2013), reviewed by Schnitzer and Bongers (2011) and (Schnitzer 2015).

One of the potential explanations for the increase in liana abundance in neotropical forests relative to trees is more intense and prolonged droughts (Schnitzer & Bongers 2011,

Schnitzer 2015). Rainfall patterns in some tropical forests have been shifting, leading to an increase in the length and severity of seasonal drought and a decrease in rainfall ((Feng et al. 2013), and these trends are predicted to continue (Phillips 2009, Lee and McPhaden

2010, Dai 2013). For example, in Panama, where liana abundance appears to be increasing

(Wright et al. 2004b, Schnitzer et al. 2012) annual precipitation has decreased by nearly

20% during the last century (Schnitzer 2005, Schnitzer and Bongers 2011). One hypothesis for an increase in lianas relative to trees is that a rise in evapotranspirative demand caused by increases in seasonality and temperature, coupled with a reduction of precipitation, can give a competitive advantage to lianas over trees (Schnitzer and Bongers 2011). Thus, because lianas are less water stressed than trees, rising seasonality together with decreases in rainfall could be more detrimental to trees, giving lianas a competitive advantage and permitting their numbers and size to increase.

137

Increases in liana abundance can affect forest-level carbon storage. For example, van Der Heijden et al. (2015) found that liana competition with trees reduced forest carbon uptake by ~76%, while other studies have documented 84% reductions in carbon uptake

(van der Heijden and Phillips 2009). Furthermore, because lianas use trees for support, it is thought that lianas are able to invest less biomass in their stems and therefore do not make up for the biomass that they displace in trees, leading to an overall reduction in forest carbon uptake and storage (Schnitzer and Bongers 2002, Ingwell et al. 2010, van der

Heijden et al. 2013, van Der Heijden et al. 2015). For example, in a liana removal experiment in treefall gaps in Panama, lianas contributed only 24% of the total aboveground biomass that they displaced in trees (Schnitzer et al. 2014). Thus, increases in seasonality could lead to an overall reduction in forest carbon storage caused by a reduction in tree biomass due to drought stress and coupled with further tree biomass reduction due to increased competition with lianas.

Differential patterns of biomass allocation to tissues like leaves, stems, and roots between lianas and trees may alter forest-level carbon stocks and fluxes. Previous studies reported that lianas allocate more biomass to leaves, while trees allocate more to stems (Cai et al. 2007, van der Heijden et al. 2013). However, our results show that increases in seasonality due to climate change could cause lianas to allocate more biomass to leaves, whereas trees to allocate less to stems in favor of roots. Any consequent increases in allocation to leaves—which have a more rapid turnover time compared to other tissues

(Powers et al. 2009c), and a decrease in allocation to stems, may reduce net tropical forest carbon storage capacity (Galbraith et al. 2013). For example, in a large-scale liana removal experiment, van Der Heijden et al. (2015) found that forest areas with lianas have higher

138

biomass allocation to leaves, whereas forest areas in which lianas had been experimentally removed had higher allocation to stems. Furthermore, as lianas increase investment into leaves, they may compete more intensely with trees for light and belowground resources, further suppressing carbon uptake by trees. Thus, decreases in rainfall and rising liana densities could shift forest biomass allocation toward a greater proportion of leaves instead of stems leading to a reduction in total forest carbon storage. These findings have broad implications for forest–level carbon storage in the face of increasing drought.

Conclusions

Our findings suggest that lianas capitalize on seasonal drought via higher photosynthetic performance while limiting water stress. By contrast, trees suffer more from water stress, which appears to limit their photosynthetic performance. The ability of lianas to grow exceptionally well (and better than trees) during seasonal drought gives lianas a seasonal growth advantage, which could explain why liana abundance peaks in highly seasonal forests. Furthermore, the ability of lianas to perform well during stronger droughts may explain increasing liana abundance in tropical forests. Interestingly, we found that lianas and trees have similar rooting depths; nonetheless, lianas were still able to maintain higher rates of photosynthesis and healthier water status compared to trees without tapping into exceptionally deeper sources of water. Collectively, our findings help shed light on how lianas and trees vary in their life-history strategies and respond differently to drought, which ultimately may explain the unique distribution patterns of lianas and trees.

139

Acknowledgements

We thank Maria M. Garcia-Leon, Boris Bernal, Abelino Valdez, Silfredo Tascon, Felipe

Mello, Jeremy La Che, and Dathan Lythgoe for their contributions to planting, irrigation, maintenance, and harvesting of the common garden. Financial support was provided by

NSF-DEB 0845071, NSF-DEB 1822473, NSF-IOS 1558093 (to S.A.S.), NSF-DDIG

1700855 (to C.M.S.M.), and UMN Carolyn Crosby Research Grant and Natural History

Award from the Dayton Bell Museum Fund (to C.M.S.M.).

140

Figures

141

Figure 4-1. Predawn (a), midday (b), and the difference between predawn and midday (c) leaf water potential of lianas and trees in irrigated and control plots measured on the same individuals two times during the dry season (March and April), once during the transition from dry to wet (May), and two times during the wet season (September and October).

Error bars are standard error.

142

Figure 4-2. Diameter measured at 20 cm from the base (a) and total biomass (b) of lianas and trees grown in irrigated and control plots. Irrigation enhancement of biomass of all harvested individuals calculated as = (mean biomass of individuals in irrigated plots – mean biomass of individuals in control plots)/ mean biomass of individuals in control plots (c).

Trees in irrigated plots were larger than trees in control plots (T =3.58, df = 597, P<0.001), but there was not difference between lianas. Mean tree biomass in the irrigated plots was

119% greater than tree biomass in the control plots, whereas for lianas there was only a

26% difference (F1, 26 = 3.51, P=0.072). The mean of the calculated irrigation enhancement of biomass was 14-times greater for trees than for lianas; however, there was no statically significant difference. All data have been log transformed. Letters indicate significant differences (P < 0.05).

143

Figure 4-3. Photosynthesis rates (a), stomatal conductance (b), and intrinsic water use efficiency (c) of lianas and trees in control plots during the dry season (dry) and the wet season (wet). Lianas and trees had higher photosynthetic rates during the wet season than the dry season (F1, 105= 31.30, P <0.001) and lianas had higher rates than trees during the dry season (T = 2.86, DF= 105, P=0.024). Both growth forms also had higher rates of stomatal conductance (F1, 105=136.42, P <0.001) and intrinsic water use efficiency (F1, 105=

81.06, P <0.001) during the wet season compared to the dry season. All data have been log transformed. Letters indicate significant differences (P < 0.05).

144

Supporting Information

Table S1. List of growth form (liana or tree), families, and species planted in the common garden.

Growth form Family Species

liana Apocynaceae Maripa panamensis

liana Bignoniaceae Callichlamys latifolia

liana Dilleniaceae Davilla kunthii

liana Dilleniaceae Doliocarpus major

liana Fabaceae Machaerium milleflorum

liana Sapindaceae Paullinia pinnata

tree Bignoniaceae Crescentia cujete

tree Combretaceae Terminalia amazonia

tree Euphorbiaceae Hieronyma alchorneoides

tree Euphorbiaceae Hura crepitans

tree Fabaceae Dipteryx oleifera

tree Meliaceae Swietenia macrophylla

145

Table S2 List of measurement, abbreviations, units, and timeframe of measurements and in which plots they were measured, and fixed effects, random effects, and data transformations used in the analysis.

Random Data Measurement Unit Abbreviation Time of measurements Plots Fixed effects effects transformation

gm- Absolute capacitance 2 Subsample of irrigated and MPa ACFT Wet season growth form species none at full turgor control plots -1

Carbon to nitrogen Subsample of irrigated and C:N C/N ratio Wet season growth form species none ratio control plots

Diameter at 20 cm Diameter at 20 irrigation treatment, cm Dry season All individuals in all plots species log from the base cm growth form

µmol Intrinsic water use CO2 species, iWUE Dry season, wet season Only control plots season, growth form log efficiency mol-1 individual H2O

Leaf capacitance at MPa- Subsample of irrigated and CTLP Wet season growth form species none TLP 1 control plots

Leaf relative water Subsample of irrigated and % RWCTLP Wet season growth form species none content at TLP control plots

Subsample of irrigated and Leaf turgor loss point MPa TLP Wet season growth form species none control plots

Dry season, transition irrigation treatment, Subsample of individuals species, Leaf water potential MPa Ψ from dry to wet, wet sampling time, growth none from all plots individual season form

Maximum rooting Subsample of irrigated and irrigation treatment, cm Rooting depth Wet season species log depth control plots growth form

146

Subsample of irrigated and Percent nitrogen % percent N Wet season growth form species none control plots

µmol CO2 Photosynthesi species, Photosynthesis rates Dry season, wet season Only control plots season, growth form log m-2 s- s rates individual 1

g cm- Subsample of irrigated and Root wood density root WD Wet season growth form species none 3 control plots

Dry season, transition Subsample of individuals irrigation treatment, Soil moisture % Soil moisture from dry to wet, wet plot none from all plots sampling time season

cm2g- Subsample of irrigated and Specific leaf area SLA Wet season growth form species log 1 control plots

g cm- Subsample of irrigated and Stem wood density stem WD Wet season growth form species none 3 control plots

mol H2O species, Stomatal conductance gs Dry season, wet season Only control plots season, growth form log m-2 s- individual 1

Sum of root Subsample of irrigated and irrigation treatment, Sum of root length cm Wet season species log length control plots growth form

Subsample of irrigated and irrigation treatment, Total biomass g Biomass Wet season species log control plots growth form

Water use efficiency Subsample of irrigated and % WUE Wet season growth form species none from δ13C control plots

147

Figure S1. Percent soil moisture and soil water potential estimated from soil moisture measured in irrigated and control plots two times during the dry season (March and April), once during the transition from dry to wet season (May), and one time during the wet season (September). Error bars are standard error.

148

Table S3. Least mean square comparisons at each sampling time (month) between irrigation treatment (irrigated, control) and growth form (liana, tree) for predawn leaf water potential. All measurements were conducted in 2015. Treatments and growth forms have been abbreviated as following: C = control, L=Liana, Ir=Irrigation, T=Tree. P values equal or smaller than 0.05 have been indicated with an asterisk (*).

Month Contrast Estimate SE DF T ratio P value

Mar. CL- IrL -0.24 0.06 235 -3.81 0.001*

Mar. CL - CT 0.33 0.08 781 4.13 0.000*

Mar. CL - IrT -0.01 0.08 235 -0.11 0.999

Mar. IrL - CT 0.57 0.08 235 7.12 0.000*

Mar. IrL - IrT 0.23 0.08 235 2.91 0.021*

Mar. CT - IrT -0.34 0.07 235 -4.84 0.000*

Apr. CL- IrL -0.34 0.06 235 -5.48 0.000*

Apr. CL - CT 0.50 0.08 781 6.18 0.000*

Apr. CL - IrT -0.09 0.08 235 -1.11 0.685

Apr. IrL - CT 0.84 0.08 235 10.49 0.000*

Apr. IrL - IrT 0.25 0.08 235 3.24 0.007*

Apr. CT - IrT -0.59 0.07 235 -8.31 0.000*

May CL- IrL -0.02 0.06 235 -0.37 0.982

May CL - CT 0.06 0.08 781 0.79 0.860

May CL - IrT 0.10 0.08 235 1.28 0.575

May IrL - CT 0.09 0.08 235 1.08 0.701

149

May IrL - IrT 0.12 0.08 235 1.59 0.388

May CT - IrT 0.04 0.07 235 0.53 0.951

Sep. CL- IrL 0.09 0.06 235 1.46 0.463

Sep. CL - CT 0.08 0.08 781 1.02 0.739

Sep. CL - IrT 0.14 0.08 235 1.82 0.268

Sep. IrL - CT -0.01 0.08 235 -0.12 0.999

Sep. IrL - IrT 0.05 0.08 235 0.67 0.910

Sep. CT - IrT 0.06 0.07 235 0.88 0.816

Oct. CL- IrL 0.02 0.08 235 0.25 0.994

Oct. CL - CT 0.03 0.09 781 0.33 0.988

Oct. CL - IrT 0.04 0.09 235 0.48 0.963

Oct. IrL - CT 0.01 0.09 235 0.11 1.000

Oct. IrL - IrT 0.02 0.09 235 0.26 0.994

Oct. CT - IrT 0.01 0.08 235 0.16 0.999

150

Table S4. Least mean square comparisons at each sampling time (month) between irrigation treatment (irrigated, control) and growth form (liana, tree) for midday leaf water potential. All measurements were conducted in 2015. Treatments and growth forms have been abbreviated as following: C = control, L=Liana, Ir=Irrigation, T=Tree. P values equal or smaller than 0.05 have been indicated with an asterisk (*).

Month Contrast Estimate SE DF T ratio P value

Mar. CL- IrL -0.39 0.09 235 -4.30 0.000*

Mar. CL - CT -0.01 0.19 781 -0.07 1.000

Mar. CL - IrT -0.40 0.19 235 -2.15 0.140

Mar. IrL - CT 0.38 0.19 235 2.05 0.173

Mar. IrL - IrT 0.00 0.18 235 -0.02 1.000

Mar. CT - IrT -0.39 0.10 235 -3.72 0.001*

Apr. CL- IrL -0.38 0.09 235 -4.10 0.000*

Apr. CL - CT 0.22 0.19 781 1.18 0.638

Apr. CL - IrT -0.44 0.19 235 -2.38 0.084

Apr. IrL - CT 0.60 0.19 235 3.21 0.008*

Apr. IrL - IrT -0.06 0.18 235 -0.34 0.986

Apr. CT - IrT -0.66 0.10 235 -6.36 0.000*

May CL- IrL 0.10 0.09 235 1.10 0.688

May CL - CT -0.06 0.19 781 -0.30 0.991

May CL - IrT 0.09 0.18 235 0.50 0.959

May IrL - CT -0.16 0.19 235 -0.84 0.834

151

May IrL - IrT -0.01 0.18 235 -0.05 1.000

May CT - IrT 0.15 0.10 235 1.43 0.480

Sep. CL- IrL 0.06 0.09 235 0.69 0.902

Sep. CL - CT -0.17 0.19 781 -0.92 0.795

Sep. CL - IrT -0.15 0.18 235 -0.79 0.859

Sep. IrL - CT -0.23 0.19 235 -1.26 0.591

Sep. IrL - IrT -0.21 0.18 235 -1.13 0.670

Sep. CT - IrT 0.03 0.10 235 0.25 0.995

Oct. CL- IrL 0.09 0.12 235 0.78 0.863

Oct. CL - CT 0.03 0.20 781 0.16 0.998

Oct. CL - IrT -0.07 0.19 235 -0.37 0.983

Oct. IrL - CT -0.06 0.20 235 -0.31 0.990

Oct. IrL - IrT -0.17 0.20 235 -0.84 0.837

Oct. CT - IrT -0.10 0.12 235 -0.85 0.833

152

Figure S2. Leaf (LMF; a), stem (SMF; b), and root (RMF; c) mass fractions of lianas and trees grown in irrigated and control plots. Irrigation treatment had a significant effect on allocation to LMF (F 1, 26 = 7.91, P = 0.009). Lianas in control plots had greater LMF that lianas in irrigated ones (T = 2.81, df = 26, P =0.043), while trees allocated similar proportions to LMF independent of watering treatment (T = 0.92, df = 26, P =0.366).

Irrigation treatment had no significant effect on allocation to SMF (F 1, 26 = 3.241, P = 0.

083). For RMF, there was a significant interaction between growth form and irrigation treatment (F 1, 26 = 4.558, P = 0.042). Trees in control plots had higher RMFs that trees irrigated ones (T = 2.75, df = 26, P =0.050), whereas lianas had similar RMF independence

153

of irrigation treatment (T = 0.25, df = 26, P =0.995). Letters at the top of each bar indicate significant differences (P < 0.05) among groups by pairwise comparison of least-squares means.

Figure S3. Maximum rooting depth (a), sum of root length (b), and ration of root length to stem diameter (c) of lianas and trees grown in irrigated and control plots. There was no significant difference between maximum rooting depths between growth forms (liana and tree; F 1, 9= 0.86, P = 0.377) or irrigation treatment (F 1,26= 0.15, P = 0.700). For sum of root length, there was no significant effect of the interaction between growth form and

154

irrigation treatment (F 1, 19 = 3.48, P = 0.078). For the ratio of the sum of root length to the diameter at 20 cm from the base there was a significant effect of growth form (F 1, 9 = 17.14,

P = 0.003) with lianas having greater values than the trees in control plots. All data have been log transformed. Letters at the top of each bar indicate significant differences (P <

0.05) among groups by pairwise comparison of least-squares means.

155

Table S5. Mean values and standard deviation (sd) of specific leaf area (SLA; cm2g-1), foliar nitrogen (N; %), ration of foliar carbon to nitrogen (C/N), foliar stable isotopes δ13 carbon (δ13C; %), leaf turgor loss point (TLP; MPa), leaf relative water content at TLP

-1 -2 -1 (RWCTLP; %), leaf capacitance at TLP (CTLP ; MPa ), leaf absolute capacitance at full turgor (ACFT; gm MPa ), root wood density

(RWD; g cm-3 ), and stem wood density (SWD; g cm-3).

Growth form Species SLA N C/N δ13C TLP RWCTLP CTLP ACFT SWD RWD liana Callichlamys latifolia 81.08 1.62 26.05 -29.89 -1.59 77.24 0.2 0.15 0.62 0.62 liana Davilla kunthii 72.2 1.18 34.35 -31.42 -1.4 86.13 0.2 0.1 0.79 0.69 liana Doliocarpus major 92.54 2.28 20.8 -32.53 -1.52 83.79 0.23 0.11 0.57 0.44 liana Machaerium milleflorum 102.21 1.66 27.39 -31.24 -1.53 74.32 0.29 0.16 0.56 0.56 liana Maripa panamensis 100.56 2.36 19.62 -30.76 -1.45 81.73 0.23 0.12 0.63 0.52 liana Paullinia pinnata 99.61 1.85 25.04 -30.86 -1.91 91.55 0.2 0.04 0.68 0.72 tree Crescentia cujete 152.58 1.71 24.8 -30.48 -1.33 88.91 0.4 0.08 0.54 0.54 tree Dipteryx oleifera 117.68 1.64 28.34 -31.21 -2.15 85.53 0.29 0.07 0.65 0.52 tree Hieronyma alchorneoides 115.68 1.68 27.78 -31.02 -1.18 89.92 0.55 0.07 - - tree Hura crepitans 123.82 2.73 15.09 -29.77 -1.62 95.48 0.13 0.03 0.43 0.36 tree Swietenia macrophylla 91.91 1.59 28.78 -30.81 -1.76 84.03 0.24 0.13 0.49 0.66 tree Terminalia amazonia 130.96 1.85 27.02 -30.08 -1.33 88.85 0.5 0.08 0.7 0.92

156

Concluding remarks

For over two decades, scientist have used plant characteristics that reflect their evolutionary history and function, referred to as functional traits, to draw broad conclusions across ecosystems (Reich et al. 1997, Wright et al. 2004a, Reich 2014, Werden et al. 2017,

Bruelheide et al. 2018, Derroire et al. 2018, Werden et al. 2018). In particular, traits that are easy to measure “soft traits” such as leaf specific area, leaf nutrient content, and wood density have been particularly common in studies that span a large number of species from many different biomes (Reich et al. 1997, Wright et al. 2004a, Asner and Martin 2012,

Reich 2014). Although this strategy may shed some light on general life history strategies and plant function, often more detailed studies are necessary to understand the nuance of community-level composition and function. This is particularly important in tropical biomes where, because of the high levels of species diversity, it is often necessary to conduct in-depth studies on large groups of species within a community to understand how different species respond to environmental condition and if there are ways to group species based on different life history strategies.

To shed light on plant community form and function, I took an experimental approach to be able to understand how a diverse group of tropical dry forest legume and non-legume tree seedlings respond to above and belowground resource availability

(Chapter 1) and how liana and tree saplings respond to irrigation during seasonal drought after multiple years of treatment (Chapter 4). I used an observational approach to determine community-level vulnerability to drought of evergreen trees and shrubs (Charter 2) and if there are differences between rooting depth and biomass allocation patterns of lianas, deciduous trees, and evergreen trees (Chapter 3). Bellow, I summarize the key finding from

157

my four dissertation chapters and how this advances our understanding of community-level responses to resource availability and drought.

In Chapter 1, I worked with a large number of common tropical dry forest tree species to show that legumes seem to have a different regeneration niche than the other species. Legumes germinate more rapidly and grow taller than other species immediately after germination, maximizing their performance when light and belowground resources are readily available, and potentially permitting them to take advantage of high light, nutrient, and water availability at the beginning of the wet season. Although seed mass tended to be larger in legumes, seed size alone did not account for all of the differences between legumes and non-legumes. This study shows the importance of working with a large number of species to be able to draw a clearer picture of potential functional groups such as in this case legumes and non-legumes.

Chapter 2, is novel because I measured drought resistance in all dominate woody species within a community. I found that the xylem of all species was highly drought resistant, with only two of the eight species being significantly more vulnerable to drought than the other ones. Also, there was little evidence of hydraulic segmentation between leaves and stems, indicating that these two organs were equally vulnerable to drought. The high resistance to embolism and general lack of segmentation suggest that canopy and understory species in this evergreen forest have evolved similar strategies of drought resistance. However, because two species were more vulnerable to damage by embolism during drought than the others, this could lead to shifts in community-level species composition caused by future increases in drought frequency and severity. Again, this study demonstrates the importance of in detail measurements of all dominant species from

158

a single location to be able to predict how a particular community will respond to future climate.

The whole plant harvest of mature deciduous trees, evergreen trees, and lianas I conducted for Chapter 3 was the first study to show that lianas had the shallowest roots, followed by deciduous trees, and evergreen trees had the deepest roots. In addition, my study was the first to find that, contrary to predicted, lianas and trees have similar allocation patterns to stems. I also found that juvenile individuals did not follow the same allocation patterns as the adults, showing that even though harvesting mature trees and lianas can be very challenging, these data are very valuable and cannot be replaced by extrapolating patterns shown in seedlings and saplings.

Finally, in Chapter 4, this was the first common garden experiment to support the hypothesis that lianas perform better and experience less physiological stress than trees during seasonal drought. Moreover, I found that only trees responded to dry season irrigation with increased growth, whereas lianas did not, suggesting clear differences between growth forms in response to altered rainfall regimes. Ultimately, better dry-season performance may explain why liana abundance peaks in seasonal forests compared to trees.

By using a four-year-old common garden in which all plants had a similar age and were grown under similar light and soil conditions, I was able to draw robust conclusions that lianas perform better than trees when they experience season drought. I also demonstrated that these performance differences were not due to deeper roots as had been previously been hypothesized, however, why lianas perform better than trees when they experience seasonal drought remains unknown.

159

A large number of studies have focused on trying to understand plant evolutionary history and function within and across communities and biomes by using ‘soft’ or easy to measure function traits (Reich et al. 1997, Wright et al. 2004a, Powers and Tiffin 2010a,

Asner and Martin 2012, Reich 2014, Kunstler et al. 2015, Funk et al. 2017, Bruelheide et al. 2018, Chavana-Bryant et al. 2019). While this approach can shed some light on how we can group species, my research has shown that often it is necessary to measure ‘hard’ traits or those that are more difficult to obtain such as hydraulic traits, biomass allocation, and rooting depth; to be able to gain a better understanding of the complexity of species community composition and potential response to shifts in rainfall patterns and drought caused by anthropogenic climate change.

160

Bibliography

Adams, H. D., A. K. Macalady, D. D. Breshears, C. D. Allen, N. L. Stephenson, S. R.

Saleska, T. E. Huxman, and N. G. McDowell. 2010a. Climate-Induced Tree

Mortality: Earth System Consequences. Eos, Transactions American Geophysical

Union 91:153-154.

Adams, M. A., J. Simon, and S. Pfautsch. 2010b. Woody legumes: a (re)view from the

South. Tree Physiology 30:1072-1082.

Adams, M. A., T. L. Turnbull, J. I. Sprent, and N. Buchmann. 2016. Legumes are different:

Leaf nitrogen, photosynthesis, and water use efficiency. Proceedings of the

National Academy of Sciences of the United States of America 113:4098–4103.

Allen, C. D., A. K. Macalady, H. Chenchouni, D. Bachelet, N. McDowell, M. Vennetier,

T. Kitzberger, A. Rigling, D. D. Breshears, E. H. Hogg, P. Gonzalez, R. Fensham,

Z. Zhang, J. Castro, N. Demidova, J.-H. Lim, G. Allard, S. W. Running, A.

Semerci, and N. Cobb. 2010. A global overview of drought and heat- induced tree

mortality reveals emerging climate change risks for forests. Forest Ecology and

Management 259:660-684.

Allen, K., J. M. Dupuy, M. G. Gei, C. Hulshof, D. Medvigy, C. Pizano, B. Salgado-Negret,

C. M. Smith, A. Trierweiler, S. J. Van Bloem, B. G. Waring, X. Xu, and J. S.

Powers. 2017. Will seasonally dry tropical forests be sensitive or resistant to future

changes in rainfall regimes? Will seasonally dry tropical forests be sensitive or

resistant to future changes in rainfall regimes? 12:023001.

Anderegg, W. R. L., J. A. Berry, and C. B. Field. 2012. Linking definitions, mechanisms,

and modeling of drought-induced tree death. Trends in Plant Science 17:693-700.

161

Anderegg, W. R. L., A. Flint, C.-y. Huang, L. Flint, J. A. Berry, F. W. Davis, J. S. Sperry,

and C. B. Field. 2015. Tree mortality predicted from drought-induced vascular

damage. Nature Geoscience 8:367-371.

Anderegg, W. R. L., T. Klein, M. Bartlett, L. Sack, A. F. A. Pellegrini, B. Choat, and S.

Jansen. 2016. Meta-analysis reveals that hydraulic traits explain cross-species

patterns of drought-induced tree mortality across the globe. Proceedings of the

National Academy of Sciences of the United States of America 113:5024-5029.

Andrade, J. L., F. C. Meinzer, G. Goldstein, and S. A. Schnitzer. 2005. Water uptake and

transport in lianas and co-occurring trees of a seasonally dry tropical forest. Trees

19:282-289.

Asner, G. P., and R. E. Martin. 2012. Contrasting leaf chemical traits in tropical lianas and

trees: implications for future forest composition. Ecol Lett 15:1001-1007.

Augspurger, C. K. 1984. Seedling survival of tropical tree species: interactions of dispersal

distance, light- gaps, and pathogens. Ecology 65:1705-1712.

Banin, L., T. R. Feldpausch, O. L. Phillips, T. R. Baker, J. Lloyd, K. Affum-Baffoe, E. J.

M. M. Arets, N. J. Berry, M. Bradford, R. J. W. Brienen, S. Davies, M. Drescher,

N. Higuchi, D. W. Hilbert, A. Hladik, Y. Iida, K. A. Salim, A. R. Kassim, D. A.

King, G. Lopez-Gonzalez, D. Metcalfe, R. Nilus, K. S.-H. Peh, J. M. Reitsma, B.

Sonké, H. Taedoumg, S. Tan, L. White, H. Wöll, and S. L. Lewis. 2012. What

controls tropical forest architecture? Testing environmental, structural and floristic

drivers. Global Ecology and Biogeography 21:1179-1190.

162

Baribault, T. W., R. K. Kobe, and A. O. Finley. 2012. Tropical tree growth is correlated

with soil phosphorus, potassium, and calcium, though not for legumes. Ecological

Monographs 82:189-203.

Barron, A., D. Purves, and L. Hedin. 2011. Facultative nitrogen fixation by canopy legumes

in a lowland tropical forest. Oecologia 165:511-520.

Barron, A. R., N. Wurzburger, J. P. Bellenger, S. J. Wright, A. M. L. Kraepiel, and L. O.

Hedin. 2008. Molybdenum limitation of asymbiotic nitrogen fixation in tropical

forest soils. Nature Geoscience 2:42.

Bartlett, M. K., Y. Zhang, N. Kreidler, S. Sun, R. Ardy, K. Cao, and L. Sack. 2014. Global

analysis of plasticity in turgor loss point, a key drought tolerance trait. Ecology

Letters 17:1580-1590.

Batterman, S. A., N. Wurzburger, and L. O. Hedin. 2013. Nitrogen and phosphorus interact

to control tropical symbiotic N 2 fixation: a test in Inga punctata. Journal of Ecology

101:1400-1408.

Beard, J. S. 1944. Climax Vegetation in Tropical America. Ecology 25:127-158.

Becknell, J. M., and J. S. Powers. 2014. Stand age and soils as drivers of plant functional

traits and aboveground biomass in secondary tropical dry forest. Canadian Journal

of Forest Research 44:603-614.

Benito-Garzón, M., P. Ruiz-Benito, and M. A. Zavala. 2013. Interspecific differences in

tree growth and mortality responses to environmental drivers determine potential

species distributional limits in Iberian forests. Global Ecology and Biogeography

22:1141-1151.

163

Benito Garzón, M., N. González Muñoz, J.-P. Wigneron, C. Moisy, J. Fernández-

Manjarrés, and S. Delzon. 2018. The legacy of water deficit on populations having

experienced negative hydraulic safety margin. Global Ecology and Biogeography

27:346-356.

Bloom, A. J., F. S. Chapin, and H. A. Mooney. 1985. Resource Limitation in Plants – An

Economic Analogy. Annual Review of Ecology and Systematics 16:363-392.

Borchert, R. 1994. Soil and Stem Water Storage Determine Phenology and Distribution of

Tropical Dry Forest Trees. Ecology 75:1437-1449.

Borchert, R., G. Rivera, and W. Hagnauer. 2002. Modification of Vegetative Phenology in

a Tropical Semi-Deciduous Forest by Abnormal Drought and Rain. Biotropica

34:27-39.

Bouche, P. S., S. Delzon, B. Choat, E. Badel, T. J. Brodribb, R. Burlett, H. Cochard, K.

Charra-Vaskou, B. Lavigne, S. Li, S. Mayr, H. Morris, J. M. Torres-Ruiz, V.

Zufferey, and S. Jansen. 2016. Are needles of Pinus pinaster more vulnerable to

xylem embolism than branches? New insights from X-ray computed tomography.

Plant, Cell & Environment 39:860-870.

Breshears, D. D., N. S. Cobb, P. M. Rich, K. P. Price, C. D. Allen, R. G. Balice, W. H.

Romme, J. H. Kastens, M. L. Floyd, J. Belnap, J. J. Anderson, O. B. Myers, and C.

W. Meyer. 2005. Regional vegetation die-off in response to global-change-type

drought. Proceedings of the National Academy of Sciences of the United States of

America 102:15144-15148.

164

Brodersen, C. R., A. J. McElrone, B. Choat, M. A. Matthews, and K. A. Shackel. 2010.

The Dynamics of Embolism Repair in Xylem: In Vivo Visualizations Using High-

Resolution Computed Tomography. Plant Physiology 154:1088-1095.

Brodribb, T. J. 2017. Progressing from ‘functional’ to mechanistic traits. New Phytologist

215:9-11.

Brodribb, T. J., M. Carriqui, S. Delzon, and C. Lucani. 2017. Optical Measurement of Stem

Xylem Vulnerability. Plant Physiology 174:2054-2061.

Brodribb, T. J., and H. Cochard. 2009. Hydraulic Failure Defines the Recovery and Point

of Death in Water- Stressed Conifers. Plant Physiology 149:575-584.

Brodribb, T. J., and N. M. Holbrook. 2005. Water Stress Deforms Tracheids Peripheral to

the Leaf Vein of a Tropical Conifer. Plant Physiology 137:1139-1146.

Brodribb, T. J., N. M. Holbrook, and M. V. Gutiérrez. 2002. Hydraulic and photosynthetic

co-ordination in seasonally dry tropical forest trees. Plant, Cell & Environment

25:1435-1444.

Brodribb, T. J., R. P. Skelton, S. A. M. McAdam, D. Bienaimé, C. J. Lucani, and P.

Marmottant. 2016. Visual quantification of embolism reveals leaf vulnerability to

hydraulic failure. New Phytologist 209:1403-1409.

Brokaw, N. V. L. 1985. Gap- phase regeneration in a tropical forest. Ecology 66:682-687.

Bruelheide, H., J. Dengler, O. Purschke, J. Lenoir, B. Jiménez-Alfaro, S. M. Hennekens,

Z. Botta-Dukát, M. Chytrý, R. Field, F. Jansen, J. Kattge, V. D. Pillar, F. Schrodt,

M. D. Mahecha, R. K. Peet, B. Sandel, P. van Bodegom, J. Altman, E. Alvarez-

Dávila, M. A. S. Arfin Khan, F. Attorre, I. Aubin, C. Baraloto, J. G. Barroso, M.

Bauters, E. Bergmeier, I. Biurrun, A. D. Bjorkman, B. Blonder, A. Čarni, L.

165

Cayuela, T. Černý, J. H. C. Cornelissen, D. Craven, M. Dainese, G. Derroire, M.

De Sanctis, S. Díaz, J. Doležal, W. Farfan-Rios, T. R. Feldpausch, N. J. Fenton, E.

Garnier, G. R. Guerin, A. G. Gutiérrez, S. Haider, T. Hattab, G. Henry, B. Hérault,

P. Higuchi, N. Hölzel, J. Homeier, A. Jentsch, N. Jürgens, Z. Kącki, D. N. Karger,

M. Kessler, M. Kleyer, I. Knollová, A. Y. Korolyuk, I. Kühn, D. C. Laughlin, F.

Lens, J. Loos, F. Louault, M. I. Lyubenova, Y. Malhi, C. Marcenò, M. Mencuccini,

J. V. Müller, J. Munzinger, I. H. Myers-Smith, D. A. Neill, Ü. Niinemets, K. H.

Orwin, W. A. Ozinga, J. Penuelas, A. Pérez-Haase, P. Petřík, O. L. Phillips, M.

Pärtel, P. B. Reich, C. Römermann, A. V. Rodrigues, F. M. Sabatini, J. Sardans, M.

Schmidt, G. Seidler, J. E. Silva Espejo, M. Silveira, A. Smyth, M. Sporbert, J.-C.

Svenning, Z. Tang, R. Thomas, I. Tsiripidis, K. Vassilev, C. Violle, R. Virtanen, E.

Weiher, E. Welk, K. Wesche, M. Winter, C. Wirth, and U. Jandt. 2018. Global

trait–environment relationships of plant communities. Nature Ecology & Evolution

2:1906-1917.

Cai, Z. Q., L. Poorter, K. F. Cao, and F. J. J. M. Bongers. 2007. Seedling growth strategies

in Bauhinia species: comparing lianas and trees. Ann Bot 100:831-7364.

Cai, Z. Q., L. Poorter, Q. Han, and F. Bongers. 2008. Effects of light and nutrients on

seedlings of tropical Bauhinia lianas and trees. Tree Physiology 28:1277-1285.

Cai, Z. Q., S. A. Schnitzer, and F. Bongers. 2009. Seasonal differences in leaf-level

physiology give lianas a competitive advantage over trees in a tropical seasonal

forest. Oecologia 161:25-33.

166

Castellanos, A. E., H. A. Mooney, S. H. Bullock, C. Jones, and R. Robichaux. 1989. Leaf,

Stem, and Metamer Characteristics of Vines in a Tropical Deciduous Forest in

Jalisco, Mexico. Biotropica 21:41-49.

Ceccon, E., P. Huante, and E. Rincón. 2006. Abiotic factors influencing tropical dry forests

regeneration. Brazilian Archives of Biology and Technology 49:305-312.

Charrier, G., S. Delzon, J.-C. Domec, L. Zhang, C. E. L. Delmas, I. Merlin, D. Corso, A.

King, H. Ojeda, N. Ollat, J. A. Prieto, T. Scholach, P. Skinner, C. van Leeuwen,

and G. A. Gambetta. 2018. Drought will not leave your glass empty: Low risk of

hydraulic failure revealed by long-term drought observations in world’s top wine

regions. Science Advances 4:1-9.

Charrier, G., J. M. Torres-Ruiz, E. Badel, R. Burlett, B. Choat, H. Cochard, C. E. L.

Delmas, J.-C. Domec, S. Jansen, A. King, N. Lenoir, N. Martin-StPaul, G. A.

Gambetta, and S. Delzon. 2016. Evidence for Hydraulic Vulnerability

Segmentation and Lack of Xylem Refilling under Tension. Plant Physiology

172:1657-1668.

Chaturvedi, R. K., A. S. Raghubanshi, and J. S. Singh. 2013. Growth of Tree Seedlings in

a Tropical Dry Forest in Relation to Soil Moisture and Leaf Traits. Journal of Plant

Ecology 6:158-170.

Chavana-Bryant, C., Y. Malhi, A. Anastasiou, B. J. Enquist, E. G. Cosio, T. F. Keenan,

and F. F. Gerard. 2019. Leaf age effects on the spectral predictability of leaf traits

in Amazonian canopy trees. Science of The Total Environment.

167

Chave, J., H. C. Muller-Landau, T. R. Baker, T. A. Easdale, H. t. Steege, and C. O. Webb.

2006. Regional and phylogenetic variation of wood density across 2456 neotropical

tree species. Ecological Applications 16:2356-2367.

Chave, J., M. Réjou-Méchain, A. Búrquez, E. Chidumayo, M. S. Colgan, W. B. C. Delitti,

A. Duque, T. Eid, P. M. Fearnside, R. C. Goodman, M. Henry, A. Martínez-Yrízar,

W. A. Mugasha, H. C. Muller-Landau, M. Mencuccini, B. W. Nelson, A.

Ngomanda, E. M. Nogueira, E. Ortiz-Malavassi, R. Pélissier, P. Ploton, C. M.

Ryan, J. G. Saldarriaga, and G. Vieilledent. 2014. Improved allometric models to

estimate the aboveground biomass of tropical trees. Global Change Biology

20:3177-3190.

Chazdon, R. L., and R. W. Pearcy. 1991. The Importance of Sunflecks for Forest

Understory Plants. BioScience 41:760-766.

Chen, Y.-J., K.-F. Cao, S. A. Schnitzer, Z.-X. Fan, J.-L. Zhang, and F. Bongers. 2015.

Water-use advantage for lianas over trees in tropical seasonal forests. New

Phytologist 205:128-136.

Choat, B., M. C. Ball, J. G. Luly, and J. A. M. Holtum. 2005. Hydraulic architecture of

deciduous and evergreen dry rainforest tree species from north-eastern Australia.

Trees 19:305-311.

Clinebell, R. R., O. L. Phillips, A. H. Gentry, N. Stark, and H. Zuuring. 1995. Prediction

of neotropical tree and liana species richness from soil and climatic data.

Biodiversity & Conservation 4:56-90.

168

Cochard, H., N. Bréda, A. Granier, and G. Aussenac. 1992. Vulnerability to air embolism

of three European oak species (Quercus petraea (Matt) Liebl, Q pubescens Willd,

Q robur L). Ann. For. Sci. 49:225-233.

Cochard, H., F. Froux, S. Mayr, and C. Coutand. 2004. Xylem Wall Collapse in Water-

Stressed Pine Needles. Plant Physiology 134:401-408.

Condit, R., S. Aguilar, A. Hernandez, R. Perez, S. Lao, G. Angehr, S. P. Hubbell, and R.

B. Foster. 2004. Tropical forest dynamics across a rainfall gradient and the impact

of an El Nio dry season. Journal of Tropical Ecology 20:51-72.

Condit, R., B. M. J. Engelbrecht, D. Pino, R. Perez, and B. L. Turner. 2013. Species

distributions in response to individual soil nutrients and seasonal drought across a

community of tropical trees. Proceedings of the National Academy of Sciences of

the United States of America 110:5064-5068.

Condit, R., K. Watts, S. A. Bohlman, R. Pérez, R. B. Foster, and S. P. Hubbell. 2000.

Quantifying the deciduousness of tropical forest canopies under varying climates.

Journal of Vegetation Science 11:649-658.

Corby, H. D. L. 1988. Types of rhizobial nodules and their distribution among the

Leguminosae. Kirkia 13:53-123.

Cuneo, I., T. Knipfer, C. Brodersen, and A. J. McElrone. 2016. Mechanical failure of fine

root cortical cells initiates plant hydraulic decline during drought. Plant Physiology

172:1669–1678.

Dai, A. 2013. Increasing drought under global warming in observations and models. Nature

Climate Change 3:171-171.

169

Dalling, J. W., and S. P. Hubbell. 2002. Seed size, growth rate and gap microsite conditions

as determinants of recruitment success for pioneer species. Journal of Ecology

90:557-568.

Davidson, R. L. 1969. Effect of Root/Leaf Temperature Differentials on Root/Shoot Ratios

in Some Pasture Grasses and Clover. Annals of Botany 33:561-569.

De Deurwaerder, H., P. Hervé-Fernández, C. Stahl, B. Burban, P. Petronelli, B. Hoffman,

D. Bonal, P. Boeckx, and H. Verbeeck. 2018. Liana and tree below-ground water

competition—evidence for water resource partitioning during the dry season. Tree

Physiology 38:1071-1083.

Denslow, J. S. 1987. Tropical Rainforest Gaps and Tree Species Diversity. Annual Review

of Ecology and Systematics 18:431-451

Derroire, G., J. S. Powers, C. M. Hulshof, L. E. Cárdenas Varela, and J. R. Healey. 2018.

Contrasting patterns of leaf trait variation among and within species during tropical

dry forest succession in Costa Rica. Scientific Reports 8:285.

DeWalt, S. J., S. A. Schnitzer, L. F. Alves, F. Bongers, R. J. Burnham, Z. Cai, W. P. Carson,

J. Chave, G. B. Chuyong, F. R. C. Costa, C. E. N. Ewango, R. V. Gallagher, J. J.

Gerwing, E. G. Amezcua, T. Hart, G. Ibarra-Manríquez, K. Ickes, D. Kenfack, S.

G. Letcher, M. J. Macía, J.-R. Makana, A. Malizia, M. Martínez-Ramos, J.

Mascaro, C. Muthumperumal, S. Muthuramkumar, A. Nogueira, M. P. E. Parren,

N. Parthasarathy, D. R. Pérez-Salicrup, F. E. Putz, H. G. Romero-Saltos, M. Sridhar

Reddy, M. N. Sainge, D. Thomas, and J. v. Melis. 2015. Biogeographical patterns

of liana abundance and diversity. Pages 131-146 in S. A. Schnitzer, F. Bongers, R.

J. Burnham, and F. E. Putz, editors. Ecology of Lianas. John Wiley & Sons, Ltd.

170

Dewalt, S. J., S. A. Schnitzer, J. Chave, F. Bongers, R. J. Burnham, Z. Cai, G. Chuyong,

D. B. Clark, C. E. N. Ewango, J. J. Gerwing, E. Gortaire, T. Hart, G. Ibarra‐

manríquez, K. Ickes, D. Kenfack, M. J. Macía, J. r. Makana, M. Martínez‐ramos, J.

Mascaro, S. Moses, H. C. Muller‐landau, M. P. E. Parren, N. Parthasarathy, D. R.

Pérez‐salicrup, F. E. Putz, H. Romero‐saltos, and D. Thomas. 2010. Annual

Rainfall and Seasonality Predict Pan‐tropical Patterns of Liana Density and Basal

Area. Biotropica 42:309-317.

Engelbrecht, B. M. J. 2012. Forests on the brink. NATURE 491:675-677.

Engelbrecht, B. M. J., L. S. Comita, R. Condit, T. A. Kursar, M. T. Tyree, B. L. Turner,

and S. P. Hubbell. 2007. Drought sensitivity shapes species distribution patterns in

tropical forests. Nature 447:80.

Enquist, B. J., and K. J. Niklas. 2002. Global Allocation Rules for Patterns of Biomass

Partitioning in Seed Plants. Science 295:1517-1520.

Ewers, F. W., M. R. Carlton, J. B. Fisher, K. J. Kolb, and M. T. Tyree. 1997. Vessel

Diameters in Roots Versus Stems of Tropical Lianas and Other Growth Forms.

IAWA Journal 18:261-279.

Feng, X., A. Porporato, and I. Rodriguez-Iturbe. 2013. Changes in rainfall seasonality in

the tropics. Nature Climate Change 3:811.

Fenner, M. 1983. Relationships between Seed Weight, Ash Content and Seedling Growth

in Twenty-Four Species of Compositae. New Phytologist 95:697-706.

Fisher, R. A., C. D. Koven, W. R. L. Anderegg, B. O. Christoffersen, M. C. Dietze, C. E.

Farrior, J. A. Holm, G. C. Hurtt, R. G. Knox, P. J. Lawrence, J. W. Lichstein, M.

Longo, A. M. Matheny, D. Medvigy, H. C. Muller-Landau, T. L. Powell, S. P.

171

Serbin, H. Sato, J. K. Shuman, B. Smith, A. T. Trugman, T. Viskari, H. Verbeeck,

E. Weng, C. Xu, X. Xu, T. Zhang, and P. R. Moorcroft. 2018. Vegetation

demographics in Earth System Models: A review of progress and priorities. Global

Change Biology 24:35-54.

Foster, S. A. 1986. On the adaptive value of large seeds for tropical moist forest trees: A

review and synthesis. The Botanical Review 52:260-299.

Funk, J. L., J. E. Larson, G. M. Ames, B. J. Butterfield, J. Cavender-Bares, J. Firn, D. C.

Laughlin, A. E. Sutton-Grier, L. Williams, and J. Wright. 2017. Revisiting the Holy

Grail: using plant functional traits to understand ecological processes. Biological

Reviews 92:1156-1173.

Galbraith, D., Y. Malhi, K. Affum-Baffoe, A. D. A. Castanho, C. E. Doughty, R. A. Fisher,

S. L. Lewis, K. S. H. Peh, O. L. Phillips, C. A. Quesada, B. Sonké, and J. Lloyd.

2013. Residence times of woody biomass in tropical forests. Plant Ecology &

Diversity 6:139-157.

Gehring, C., S. Park, and M. Denich. 2004. Liana allometric biomass equations for

Amazonian primary and secondary forest. Forest Ecology and Management

195:69-83.

Gei, M. G., and J. S. Powers. 2015. The influence of seasonality and species effects on

surface fine roots and nodulation in tropical legume tree plantations. Plant and Soil

388:187-196

Gentry, A. H. 1988. Changes in plant community diversity and floristic composition on

environmental and geographical gradients. Ann. Missouri Bot. Gard. 75:1-34.

172

Gentry, A. H. 1991. The distribution and evolution of climbing plants. Pages 3-50 in F. E.

Putz and H. A. Mooney, editors. The Biology of Vines. Cambridge University

Press, Cambridge.

Gentry, A. H. 1995. Diversity and floristic composition of Neotropical dry forests. Pages

146-194 in S. H. Bullock, H. A. Mooney, and E. Medina, editors. Seasonally Dry

Tropical Forests. Cambridge University Press, Cambridge.

Gerhardt, K. 1996. Effects of root competition and canopy openness on survival and

growth of tree seedlings in a tropical seasonal dry forest. Forest Ecology and

Management 82:33-48.

Gillespie, T. W. G., Alfredo & Farris, Christine N. . 2000. Diversity, composition, and

structure of tropical dry forests in

Central America. Plant Ecology 147:37–47.

Givnish, T. J., and G. J. Vermeij. 1976. Sizes and Shapes of Liane Leaves. The American

Naturalist 110:743-778.

Greenwood, S., P. Ruiz-Benito, J. Martínez-Vilalta, F. Lloret, T. Kitzberger, C. D. Allen,

R. Fensham, D. C. Laughlin, J. Kattge, G. Bönisch, N. J. B. Kraft, and A. S. Jump.

2017. Tree mortality across biomes is promoted by drought intensity, lower wood

density and higher specific leaf area. Ecology Letters 20:539-553.

Hacke, U. G., J. S. Sperry, and J. Pittermann. 2000. Drought experience and cavitation

resistance in six shrubs from the Great Basin, Utah. Basic and Applied Ecology

1:31-41.

173

Hacke, U. G., J. S. Sperry, W. T. Pockman, S. D. Davis, and K. A. McCulloh. 2001. Trends

in wood density and structure are linked to prevention of xylem implosion by

negative pressure. Oecologia 126:457-461.

Hochberg, U., C. W. Windt, A. Ponomarenko, Y.-J. Zhang, J. Gersony, F. E. Rockwell,

and N. M. Holbrook. 2017. Stomatal closure, basal leaf embolism and shedding

protect the hydraulic integrity of grape stems. Plant Physiology 174:764–775.

Holbrook, N. M., and F. E. Putz. 1996. Physiology of Tropical Vines and Hemiepiphytes:

Plants that Climb Up and Plants that Climb Down. Pages 363-394 in S. S. Mulkey,

R. L. Chazdon, and A. P. Smith, editors. Tropical Forest Plant Ecophysiology.

Springer US, Boston, MA.

Huante, P., E. Rincon, and I. F.S. Chapin. 1998. Foraging for nutrients, responses to

changes in light and competition in tropical deciduous tree seedlings. Oecologia

117:209-216.

Hubbell, S. P., R. B. Foster, S. T. O'Brien, K. E. Harms, R. Condit, B. Wechsler, S. J.

Wright, and S. Loo De Lao. 1999. Light- gap disturbances, recruitment limitation,

and tree diversity in a neotropical forest. Science 283:554-557.

Ian, J. W., B. R. Peter, W. Mark, D. A. David, B. Zdravko, B. Frans, C.-B. Jeannine, C.

Terry, H. C. C. Johannes, D. Matthias, F. Jaume, G. Eric, K. G. Philip, G. Javier,

H. Kouki, B. L. Byron, L. Tali, L. William, L. Christopher, J. M. Jeremy, N. Marie-

Laure, N. ‹lo, O. Jacek, O. Noriyuki, P. Hendrik, P. Pieter, P. Lynda, I. P. Vladimir,

R. Catherine, C. T. Sean, G. T. Mark, J. V. Erik, and V. Rafael. 2004. The

worldwide leaf economics spectrum. Nature 428:821-827.

174

Ingwell, L. L., S. Joseph Wright, K. K. Becklund, S. P. Hubbell, and S. A. Schnitzer. 2010.

The impact of lianas on 10 years of tree growth and mortality on Barro Colorado

Island, Panama. Journal of Ecology 98:879-887.

John, R., J. W. Dalling, K. E. Harms, J. B. Yavitt, R. F. Stallard, M. Mirabello, S. P.

Hubbell, R. Valencia, H. Navarrete, M. Vallejo, and R. B. Foster. 2007. Soil

nutrients influence spatial distributions of tropical tree species. Proceedings of the

National Academy of Sciences of the United States of America 104:864-869.

Johnson, D. M., J. C. Domec, D. R. Woodruff, K. A. McCulloh, and F. C. Meinzer. 2013.

Contrasting hydraulic strategies in two tropical lianas and their host trees. Am J Bot

100:374-383.

Johnson, D. M., K. A. McCulloh, F. C. Meinzer, D. R. Woodruff, D. M. Eissenstat, and N.

Phillips. 2011. Hydraulic patterns and safety margins, from stem to stomata, in

three eastern US tree species. Tree Physiology 31:659-668.

Johnson, D. M., R. Wortemann, K. A. McCulloh, L. Jordan-Meille, E. Ward, J. M. Warren,

S. Palmroth, and J.-C. Domec. 2016. A test of the hydraulic vulnerability

segmentation hypothesis in angiosperm and conifer tree species. Tree Physiology

36:983-993.

Johnson, K. M., G. J. Jordan, and T. J. Brodribb. 2018. Wheat leaves embolized by water

stress do not recover function upon rewatering. Plant, Cell & Environment 41:2704-

2714.

Jurado, E., and M. Westoby. 1992. Seedling growth in relation to seed size among species

of arid Australia. Journal of Ecology 80:407-416

175

Kitajima, K. 1994. Relative importance of photosynthetic traits and allocation patterns as

correlates of seedling shade tolerance of 13 tropical trees. Oecologia 98:419-428.

Klepsch, M., Y. Zhang, M. M. Kotowska, L. J. Lamarque, M. Nolf, B. Schuldt, J. M.

Torres-Ruiz, D. W. Qin, B. Choat, S. Delzon, C. Scoffoni, K. F. Cao, and S. Jansen.

2018. Is xylem of angiosperm leaves less resistant to embolism than branches?

Insights from microCT, hydraulics, and anatomy. Journal of Experimental Botany

69 5611–5623.

Kny, L. 1894. On correlation in the growth of roots and shoots. Ann Bot 8:265-280.

Koepke, D. F., T. E. Kolb, and H. D. Adams. 2010. Variation in woody plant mortality and

dieback from severe drought among soils, plant groups, and species within a

northern Arizona ecotone. Oecologia 163:1079-1090.

Kunstler, G., D. Falster, D. A. Coomes, F. Hui, R. M. Kooyman, D. C. Laughlin, L. Poorter,

M. Vanderwel, G. Vieilledent, S. J. Wright, M. Aiba, C. Baraloto, J. Caspersen, J.

H. C. Cornelissen, S. Gourlet-Fleury, M. Hanewinkel, B. Herault, J. Kattge, H.

Kurokawa, Y. Onoda, J. Peñuelas, H. Poorter, M. Uriarte, S. Richardson, P. Ruiz-

Benito, I. F. Sun, G. Ståhl, N. G. Swenson, J. Thompson, B. Westerlund, C. Wirth,

M. A. Zavala, H. Zeng, J. K. Zimmerman, N. E. Zimmermann, and M. Westoby.

2015. Plant functional traits have globally consistent effects on competition. Nature

529:204.

Larter, M., S. Pfautsch, J.-C. Domec, S. Trueba, N. Nagalingum, and S. Delzon. 2017.

Aridity drove the evolution of extreme embolism resistance and the radiation

of conifer genus Callitris. New Phytologist 215:97-112.

176

Laurance, W. F., A. S. Andrade, A. Magrach, J. L. C. Camargo, J. J. Valsko, M. Campbell,

P. M. Fearnside, W. Edwards, T. E. Lovejoy, and S. G. Laurance. 2014. Long- term

changes in liana abundance and forest dynamics in undisturbed Amazonian forests.

Ecology 95:1604-1611.

Lawrence, D. 2003. The response of tropical tree seedlings to nutrient supply: meta-

analysis for understanding a changing tropical landscape. J. of Tropical Ecology

19:239-250.

Lee, T., and M. J. McPhaden. 2010. Increasing intensity of El Niño in the central‐

equatorial Pacific. Geophysical Research Letters 37:n/a-n/a.

Leiva, J. A., R. Mata, O. J. Rocha, and M. V. Gutiérrez-Soto. 2009. Cronología de la

regeneración del bosque tropical seco en Santa Rosa, Guanacaste, Costa Rica. I.

Características edáficas. Revista de Biologia Tropical 57:801-815.

Lindenmayer, D., M. Crane, and D. Michael. 2005. Woodlands: a disappearing landscape.

CSIRO Publishing, Collingwood.

Manzané-Pinzón, E., G. Goldstein, and S. A. Schnitzer. 2018. Does soil moisture

availability explain liana seedling distribution across a tropical rainfall gradient?

Biotropica 50:215-224.

Maréchaux, I., M. K. Bartlett, A. Iribar, L. Sack, and J. Chave. 2017. Stronger seasonal

adjustment in leaf turgor loss point in lianas than trees in an Amazonian forest.

Biology Letters 13:1-4.

Maréchaux, I., M. K. Bartlett, L. Sack, C. Baraloto, J. Engel, E. Joetzjer, and J. Chave.

2015. Drought tolerance as predicted by leaf water potential at turgor loss point

177

varies strongly across species within an Amazonian forest. Functional Ecology:n/a-

n/a.

Markesteijn, L., and L. Poorter. 2009. Seedling root morphology and biomass allocation of

62 tropical tree species in relation to drought- and shade-tolerance. Journal of

Ecology 97:311-325.

Markesteijn, L., L. Poorter, and F. Bongers. 2007. Light-Dependent Leaf Trait Variation

in 43 Tropical Dry Forest Tree Species. American Journal of Botany 94:515-525.

McCarthy, M. C., B. J. Enquist, and A. J. Kerkhoff. 2007. Organ Partitioning and

Distribution across the Seed Plants: Assessing the Relative Importance of

Phylogeny and Function. International Journal of Plant Sciences 168:751-761.

McDowell, N., W. T. Pockman, C. D. Allen, D. D. Breshears, N. Cobb, T. Kolb, J. Plaut,

J. Sperry, A. West, D. G. Williams, and E. A. Yepez. 2008. Mechanisms of plant

survival and mortality during drought: why do some plants survive while others

succumb to drought? Pages 719-739, Oxford, UK.

Meinzer, F. C., J. L. Andrade, G. Goldstein, N. M. Holbrook, J. Cavelier, and S. J. Wright.

1999. Partitioning of soil water among canopy trees in a seasonally dry tropical

forest. Oecologia 121:293-301.

Meinzer, F. C., D. M. Johnson, B. Lachenbruch, K. A. McCulloh, and D. R. Woodruff.

2009. Xylem hydraulic safety margins in woody plants: coordination of stomatal

control of xylem tension with hydraulic capacitance. Functional Ecology 23:922-

930.

178

Miyashita, K., S. Tanakamaru, T. Maitani, and K. Kimura. 2005. Recovery responses of

photosynthesis, transpiration, and stomatal conductance in kidney bean following

drought stress. Environmental and Experimental Botany 53:205-214.

Moles, A. T., and M. Westoby. 2004. Seedling survival and seed size: a synthesis of the

literature. Journal of Ecology 92:372-383.

Mueller, R. C., C. M. Scudder, M. E. Porter, R. Talbot Trotter, C. A. Gehring, and T. G.

Whitham. 2005. Differential tree mortality in response to severe drought: evidence

for long-term vegetation shifts. Journal of Ecology 93:1085-1093.

Murphy, P. G., and A. E. Lugo. 1986. Ecology of Tropical Dry Forest. Annual Review of

Ecology and Systematics 17:67-88.

Nicotra, A. B., and R. L. Chazdon. 1999. Spatial Heterogeneity of Light and Woody

Seedling Regeneration in Tropical Wet Forests. Ecology 80:1908-1926.

Niklas, K. J. 1994. Comparisons Among Biomass Allocation and Spatial Distribution

Patterns of Some Vine, Pteridophyte, and Gymnosperm Shoots. American Journal

of Botany 81:1416-1421.

Niklas, K. J. 2005. Modelling Below- and Above-ground Biomass for Non-woody and

Woody Plants. Ann Bot 95:315-321.

Niklas, K. J., and B. J. Enquist. 2002. Canonical Rules for Plant Organ Biomass

Partitioning and Annual Allocation. American Journal of Botany 89:812-819.

Niklas, K. J., and H.-C. Spatz. 2004. Growth and hydraulic (not mechanical) constraints

govern the scaling of tree height and mass. Proceedings of the National Academy

of Sciences of the United States of America 101:15661-15663.

179

O'Brien, M. J., B. M. J. Engelbrecht, J. Joswig, G. Pereyra, B. Schuldt, S. Jansen, J. Kattge,

S. M. Landhäusser, S. R. Levick, Y. Preisler, P. Väänänen, and C. Macinnis-Ng.

2017. A synthesis of tree functional traits related to drought-induced mortality in

forests across climatic zones. Journal of Applied Ecology 54:1669-1686.

Parker, G., C. Tinoco-Ojanguren, A. Martinez-Yrizar, and M. Maass. 2005. Seasonal

balance and vertical pattern of photosynthetically active radiation within canopies

of a tropical dry deciduous forest ecosystem in Mexico. Journal of Tropical Ecology

21:283-295.

Pellegrini, A. F. A., A. C. Staver, L. O. Hedin, T. Charles‐Dominique, and A. Tourgee.

2016. Aridity, not fire, favors nitrogen‐ fixing plants across tropical savanna and

forest biomes. Ecology 97:2177-2183.

Phillips, O. L. 2009. Drought sensitivity of the Amazon rainforest. Science 323:1344-1347.

Phillips, O. L., R. V. Martinez, L. Arroyo, T. R. Baker, T. Killeen, S. L. Lewis, Y. Malhi,

A. M. Mendoza, D. Neill, P. N. Vargas, M. Alexiades, C. Ceron, A. Di Fiore, T.

Erwin, A. Jardim, W. Palacios, M. Saldias, and B. Vinceti. 2002. Increasing

dominance of large lianas in Amazonian forests. Nature 418:770-774.

Pineda-García, F., H. Paz, F. C. Meinzer, and G. Angeles. 2015. Exploiting water versus

tolerating drought: water-use strategies of trees in a secondary successional tropical

dry forest. Tree Physiology 36:208-217.

Pittermann, J., B. Choat, S. Jansen, S. A. Stuart, L. Lynn, and T. E. Dawson. 2010. The

relationships between xylem safety and hydraulic efficiency in the Cupressaceae:

the evolution of pit membrane form and function. Plant Physiology 153:1919-1931.

180

Pivovaroff, A. L., L. Sack, and L. S. Santiago. 2014. Coordination of stem and leaf

hydraulic conductance in southern California shrubs: a test of the hydraulic

segmentation hypothesis. New Phytologist 203:842-850.

Poorter, H., A. M. Jagodzinski, J. Oleksyn, R. Ruiz-Peinado, S. Kuyah, Y. Luo, P. B.

Reich, V. A. Usoltsev, T. N. Buckley, and L. Sack. 2015. How does biomass

distribution change with size and differ among species? An analysis for 1200 plant

species from five continents. New Phytologist 208:736-749.

Poorter, H., K. J. Niklas, P. B. Reich, J. Oleksyn, P. Poot, and L. Mommer. 2012. Biomass

allocation to leaves, stems and roots: meta-analyses of interspecific variation and

environmental control. New Phytologist 193:30-50.

Powers, J. S. 2014. Reciprocal interactions between lianas and forest soil.in S. A. Schnitzer,

F. Bongers, R. J. Burnham, and F. E. Putz, editors. Ecology of Lianas.

Powers, J. S., J. M. Becknell, J. Irving, and D. Perez-Aviles. 2009a. Diversity and structure

of regenerating tropical dry forests in Costa Rica: geographic patterns and

environmental drivers Forest Ecology and Management 258:959-970.

Powers, J. S., J. M. Becknell, J. Irving, and D. Pèrez-Aviles. 2009b. Diversity and structure

of regenerating tropical dry forests in Costa Rica: Geographic patterns and

environmental drivers. Forest Ecology and Management 258:959-970.

Powers, J. S., R. A. Montgomery, E. C. Adair, F. Q. Brearley, S. J. DeWalt, C. T. Castanho,

J. Chave, E. Deinert, J. U. Ganzhorn, M. E. Gilbert, J. A. González-Iturbe, S.

Bunyavejchewin, H. R. Grau, K. E. Harms, A. Hiremath, S. Iriarte-Vivar, E.

Manzane, A. A. De Oliveira, L. Poorter, J.-B. Ramanamanjato, C. Salk, A. Varela,

G. D. Weiblen, and M. T. Lerdau. 2009c. Decomposition in tropical forests: a pan-

181

tropical study of the effects of litter type, litter placement and mesofaunal exclusion

across a precipitation gradient. Journal of Ecology 97:801-811.

Powers, J. S., and D. Perez-Aviles. 2013. Edaphic Factors are a More Important Control

on Surface Fine Roots than Stand Age in Secondary Tropical Dry Forests.

Biotropica 45:1-9.

Powers, J. S., and P. Tiffin. 2010a. Plant functional type classifications in tropical dry

forests in Costa Rica: leaf habit versus taxonomic approaches. Functional Ecology

24:927-936.

Powers, J. S., and P. Tiffin. 2010b. Plant functional type classifications in tropical dry

forests in Costa Rica: leaf habit versus taxonomic approaches. Functional Ecology

24:927-936.

Putz, F. E. 1983. Liana biomass and leaf area of a “Tierra Firme” forest in the Rio Negro

Basin, Venezuela. Biotropica 15:185-189.

Reich, P. B. 2014. The world-wide ‘fast–slow’ plant economics spectrum: a traits

manifesto. Journal of Ecology 102:275–301.

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 94:13730-13734.

Restom, T. G., and D. C. Nepstad. 2004. Seedling growth dynamics of a deeply rooting

liana in a secondary forest in eastern Amazonia. Forest Ecology and Management

190:109-118.

Restrepo-Coupe, N., H. R. da Rocha, L. R. Hutyra, A. C. da Araujo, L. S. Borma, B.

Christoffersen, O. M. R. Cabral, P. B. de Camargo, F. L. Cardoso, A. C. L. da

182

Costa, D. R. Fitzjarrald, M. L. Goulden, B. Kruijt, J. M. F. Maia, Y. S. Malhi, A.

O. Manzi, S. D. Miller, A. D. Nobre, C. von Randow, L. D. A. Sá, R. K. Sakai, J.

Tota, S. C. Wofsy, F. B. Zanchi, and S. R. Saleska. 2013. What drives the

seasonality of photosynthesis across the Amazon basin? A cross-site analysis of

eddy flux tower measurements from the Brasil flux network. Agricultural and

Forest Meteorology 182-183:128-144.

Reyes-Garcia, C., J. L. Andrade, J. L. Sima, R. Us-Santamaria, and P. C. Jackson. 2012.

Sapwood to heartwood ratio affects whole-tree water use in dry forest legume and

non-legume trees. Tree Structure and Function 26:1317-1330.

Roa-Fuentes, L. L., J. Campo, and V. Parra-Tabla. 2012. Plant Biomass Allocation across

a Precipitation Gradient: An Approach to Seasonally Dry Tropical Forest at

Yucatán, Mexico. Ecosystems 15:1234-1244.

Rodriguez-Dominguez, C. M., M. R. Carins-Murphy, C. Lucani, and T. J. Brodribb. 2018.

Mapping xylem failure in disparate organs of whole plants reveals extreme

resistance in olive roots. New Phytologist 218:1025-1035.

Sack, L., M. C. Ball, C. Brodersen, S. D. Davis, D. L. Des Marais, L. A. Donovan, T. J.

Givnish, U. G. Hacke, T. Huxman, S. Jansen, A. L. Jacobsen, D. M. Johnson, G.

W. Koch, C. Maurel, K. A. McCulloh, N. G. McDowell, A. McElrone, F. C.

Meinzer, P. J. Melcher, G. North, M. Pellegrini, W. T. Pockman, R. B. Pratt, A.

Sala, L. S. Santiago, J. A. Savage, C. Scoffoni, S. Sevanto, J. Sperry, S. D.

Tyerman, D. Way, and N. M. Holbrook. 2016. Plant hydraulics as a central hub

integrating plant and ecosystem function: meeting report for ‘Emerging Frontiers

183

in Plant Hydraulics’ (Washington, DC, May 2015). Plant, Cell & Environment

39:2085-2094.

Sack, L., P. D. Cowan, N. Jaikumar, and N. M. Holbrook. 2003. The ‘ hydrology’ of leaves:

co‐ ordination of structure and function in temperate woody species. Plant, Cell &

Environment 26:1343-1356.

Sanford Jr, R. L. 1989. Fine root biomass under a tropical forest light gap opening in Costa

Rica. Journal of Tropical Ecology 5:251-256.

Schnitzer, S. A. 2005. A mechanistic explanation for global patterns of liana abundance

and distribution. Amarican Naturalist 166:262-276.

Schnitzer, S. A. 2015. The Contribution of Lianas to Forest Ecology, Diversity, and

Dynamics. Pages 149-160 in N. Parthasarathy, editor. Biodiversity of Lianas.

Springer International Publishing, Cham.

Schnitzer, S. A. 2018. Testing ecological theory with lianas. New Phytologist 220:366-

380.

Schnitzer, S. A., and F. Bongers. 2002. The ecology of lianas and their role in forests.

Trends in Ecology & Evolution 17:223-230.

Schnitzer, S. A., and F. Bongers. 2011. Increasing liana abundance and biomass in tropical

forests: emerging patterns and putative mechanisms. Ecology Letters 14:397-406.

Schnitzer, S. A., M. E. Kuzee, and F. Bongers. 2005. Disentangling above- and below-

ground competition between lianas and trees in a tropical forest. Journal of Ecology

93:1115-1125.

Schnitzer, S. A., S. A. Mangan, J. W. Dalling, C. A. Baldeck, S. P. Hubbell, A. Ledo, H.

Muller-Landau, M. F. Tobin, S. Aguilar, D. Brassfield, A. Hernandez, S. Lao, R.

184

Perez, O. Valdes, and S. R. Yorke. 2012. Liana abundance, diversity, and

distribution on Barro Colorado Island, Panama. PLoS One 7:e52114.

Schnitzer, S. A., and G. M. F. van der Heijden. 2019. Lianas have a seasonal growth

advantage over co-occurring trees. Ecology 0:e02655.

Schnitzer, S. A., G. M. F. van der Heijden, J. Mascaro, and W. P. Carson. 2014. Lianas in

Gaps Reduce Carbon Accumulation in a Tropical Forest. Ecology 95:3008-3017.

Scoffoni, C., C. Albuquerque, C. Brodersen, S. V. Townes, G. P. John, M. K. Bartlett, T.

N. Buckley, A. J. McElrone, and L. Sack. 2017. Outside-xylem vulnerability, not

xylem embolism, controls leaf hydraulic decline during dehydration. Plant

Physiology 173:1197-1210.

Scoffoni, C., M. Rawls, A. McKown, H. Cochard, and L. Sack. 2011. Decline of leaf

hydraulic conductance with dehydration: relationship to leaf size and venation

architecture. Plant Physiology 156:832.

Scoffoni, C., and L. Sack. 2017. The causes and consequences of leaf hydraulic decline

with dehydration. Journal of Experimental Botany 68:4479-4496.

Selaya, N. G., N. P. R. Anten, R. J. Oomen, M. Matthies, and M. J. A. Werger. 2007.

Above-ground Biomass Investments and Light Interception of Tropical Forest

Trees and Lianas Early in SuccessionSelaya et al. — Biomass Allocation and Light

Interception in a Tropical ForestSelaya et al. — Biomass Allocation and Light

Interception in a Tropical Forest. Ann Bot 99:141-151.

Shipley, B., and R. H. Peters. 1990. The Allometry of Seed Weight and Seedling Relative

Growth Rate. Functional Ecology 4:523-529.

185

Skelton, R. P., T. J. Brodribb, and B. Choat. 2017a. Casting light on xylem vulnerability

in an herbaceous species reveals a lack of segmentation. New Phytologist 214:561-

569.

Skelton, R. P., T. J. Brodribb, S. A. M. McAdam, and P. J. Mitchell. 2017b. Gas exchange

recovery following natural drought is rapid unless limited by loss of leaf hydraulic

conductance: evidence from an evergreen woodland. New Phytologist 215:1399-

1412

Skelton, R. P., T. E. Dawson, S. E. Thompson, Y. Shen, A. P. Weitz, and D. Ackerly. 2018.

Low vulnerability to xylem embolism in leaves and stems of North American oaks.

Plant Physiology 177:1066-1077.

Slot, M., D. T. Palow, and K. Kitajima. 2012. Seed reserve dependency of seedling growth

for nitrogen and phosphorus. Functional Plant Biology 40:244-250.

Smith-Martin, C. M., M. G. Gei, E. Bergstrom, K. K. Becklund, J. M. Becknell, B. G.

Waring, L. K. Werden, and J. S. Powers. 2017. Effects of soil type and light on

height growth, biomass partitioning, and nitrogen dynamics on 22 species of

tropical dry forest tree seedlings: Comparisons between legumes and nonlegumes.

American Journal of Botany 104:399-410.

Stevens, G. C. 1987. Lianas as Structural Parasites: The Bursera Simaruba Example.

Ecology 68:77-81.

Sui, Y., Q. Q. Cui, M. Dong, and W. M. He. 2011. Contrasting responses of legume versus

non-legume shrubs to soil water and nutrient shortages in the Mu Us Sandland.

Journal of Plant Ecology 4:268-274.

186

Swaine, M. D., and J. Grace. 2007. Lianas may be favoured by low rainfall: evidence from

Ghana. Plant Ecology 192:271-276.

Thornley, J. H. M. 1972. A Balanced Quantitative Model for Root: Shoot Ratios in

Vegetative Plants. Ann Bot 36:431-441.

Toledo-Aceves, T., and M. D. Swaine. 2008. Above- and Below-Ground Competition

between the Liana Acacia kamerunensis and Tree Seedlings in Contrasting Light

Environments. Plant Ecology 196:233-244.

Tyree, M. T., H. Cochard, P. Cruiziat, B. Sinclair, and T. Ameglio. 1993. Drought-induced

leaf shedding in walnut: evidence for vulnerability segmentation. Plant, Cell &

Environment 16:879-882.

Tyree, M. T., and F. W. Ewers. 1991. The hydraulic architecture of trees and other woody

plants. New Phytologist 119:345-360.

Tyree, M. T., and H. T. Hammel. 1972. The Measurement of the Turgor Pressure and the

Water Relations of Plants by the Pressure-bomb Technique. Journal of

Experimental Botany 23:267-282.

Tyree, M. T., and J. S. Sperry. 1988. Do Woody Plants Operate near the Point of

Catastrophic Xylem Dysfunction Caused by Dynamic Water Stress?: Answers

from a Model. Plant Physiology 88:574-580.

Tyree, M. T., and J. S. Sperry. 1989. Vulnerability of Xylem to Cavitation and Embolism.

Pages 19-36 Annual Review of Plant Physiology and Plant Molecular Biology.

Ulate, G. V. 2001. Fitogeografía de ecosistemas secos el la meseta de ignimbritas de

Guanacaste, Costa Rica. Revista de Biologia Tropical 49:227-238.

187

van der Heijden, G. M., S. A. Schnitzer, J. S. Powers, and O. L. Phillips. 2013. Liana

Impacts on Carbon Cycling, Storage and Sequestration in Tropical Forests.

Biotropica 45:682-692. van der Heijden, G. M. F., and O. L. Phillips. 2009. Liana infestation impacts tree growth

in a lowland tropical moist forest. Biogeosciences 6:2217-2226. van Der Heijden, G. M. F., J. S. Powers, and S. A. Schnitzer. 2015. Lianas reduce carbon

accumulation and storage in tropical forests. Proceedings of the National Academy

of Sciences of the United States of America 112:13267–13271.

Vargas-G., G., L. K. Werden, and J. S. Powers. 2015. Explaining legume success in tropical

dry forests based on seed germination niches: a new hypothesis. Biotropica 47:277-

280.

Vargas G, G., L. K. Werden, and J. S. Powers. 2015. Explaining Legume Success in

Tropical Dry Forests Based on Seed Germination Niches: A New Hypothesis.

Biotropica 47: 277-280.

Vitousek, P., and R. Howarth. 1991. Nitrogen limitation on land and in the sea: How can

it occur? Biogeochemistry 13:87-115.

Wang, Y. P., B. Z. Houlton, and C. B. Field. 2007. A model of biogeochemical cycles of

carbon, nitrogen, and phosphorus including symbiotic nitrogen fixation and

phosphatase production. Global Biogeochemical Cycles 21.

Werden, L. K., J. M. Becknell, and J. S. Powers. 2018. Edaphic factors, successional status

and functional traits drive habitat associations of trees in naturally regenerating

tropical dry forests. Functional Ecology 32: 2766– 2776.

188

Werden, L. K., B. G. Waring, C. M. Smith-Martin, and J. S. Powers. 2017. Tropical dry

forest trees and lianas differ in leaf economic spectrum traits but have overlapping

water-use strategies. Tress Physiology:1-14.

West, A. G., T. E. Dawson, E. C. February, G. F. Midgley, W. J. Bond, and T. L. Aston.

2012. Diverse functional responses to drought in a Mediterranean-type shrubland

in South Africa. New Phytologist 195:396-407.

Wishnie, M. H., D. H. Dent, E. Mariscal, J. Deago, N. Cedeño, D. Ibarra, R. Condit, and

P. M. S. Ashton. 2007. Initial performance and reforestation potential of 24 tropical

tree species planted across a precipitation gradient in the Republic of Panama.

Forest Ecology and Management 243:39-49.

Wolfe, B. T., J. S. Sperry, and T. A. Kursar. 2016. Does leaf shedding protect stems from

cavitation during seasonal droughts? A test of the hydraulic fuse hypothesis. New

Phytologist 212:1007-1018.

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. 2004a. The worldwide leaf economics spectrum. Nature 428:821-

827.

Wright, S. J., O. Calderon, A. Hernandez, and S. Paton. 2004b. Are lianas increasing in

importance in tropical forests? A 17-year record from Panama. Ecology 85:484-

489.

189

Wyka, T. P., J. Oleksyn, P. Karolewski, and S. A. Schnitzer. 2013. Phenotypic correlates

of the lianescent growth form: a review. Annals of Botany 112:1667-1681.

Xu, X., D. Medvigy, J. S. Powers, J. M. Becknell, and K. Guan. 2016. Diversity in plant

hydraulic traits explains seasonal and inter‐ annual variations of vegetation

dynamics in seasonally dry tropical forests. New Phytologist 212:80-95.

Yorke, S. R., S. A. Schnitzer, J. Mascaro, S. G. Letcher, and W. P. Carson. 2013. Increasing

Liana Abundance and Basal Area in a Tropical Forest: The Contribution of Long-

distance Clonal Colonization. Biotropica 45:317-324.

Zhang, Y.-J., F. E. Rockwell, A. C. Graham, T. Alexander, and N. M. Holbrook. 2016.

Reversible Leaf Xylem Collapse: A Potential “Circuit Breaker” against Cavitation.

Plant Physiology 172:2261-2274.

Zhu, S.-D., and K.-F. Cao. 2009. Hydraulic properties and photosynthetic rates in co-

occurring lianas and trees in a seasonal tropical rainforest in southwestern China.

Plant Ecology 204:295-304.

Zhu, S.-D., H. Liu, Q.-Y. Xu, K.-F. Cao, and Q. Ye. 2016. Are leaves more vulnerable to

cavitation than branches? Functional Ecology 30:1740-1744.

Zimmermann, M. H. 1983. Xylem Structure and the Ascent of Sap. Berlin, Heidelberg :

Springer Berlin Heidelberg : Imprint: Springer, Berlin, Heidelberg.

190