Environmental filtering, local site factors and landscape context drive changes in functional trait composition during tropical forest succession

Accepted Manuscript Title: Environmental filtering, local site factors and landscape context drive changes in functional trait composition during tropical forest succession Author: Vanessa K. Boukili Robin L. Chazdon PII: DOI: Reference:

S1433-8319(16)30143-3 http://dx.doi.org/doi:10.1016/j.ppees.2016.11.003 PPEES 25337

To appear in: Received date: Revised date: Accepted date:

5-5-2016 28-11-2016 29-11-2016

Please cite this article as: Boukili, Vanessa K., Chazdon, Robin L., Environmental filtering, local site factors and landscape context drive changes in functional trait composition during tropical forest succession.Perspectives in Plant Ecology, Evolution and Systematics http://dx.doi.org/10.1016/j.ppees.2016.11.003 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Environmental filtering, local site factors and landscape context drive changes in functional trait composition during tropical forest succession

Authors: Vanessa K. Boukili*, Robin L. Chazdon University of Connecticut, Department of Ecology and Evolutionary Biology, 75 N. Eagleville Rd., Unit 3043, Storrs, CT. 06269-3043

*

Corresponding Author: [email protected]

Running Headline: Successional functional trait composition

1

Highlights    

Five functional traits changed directionally along the successional gradient. Functional trait patterns were also influenced by the location of the plots within the matrix. The abundance of palms strongly influenced community-level functional trait patterns. The strongest driver of functional trends was plot identity.

Abstract Second-growth forests provide an important avenue for at least partially recuperating biodiversity and ecosystem services in tropical regions. Yet, factors affecting changes in species and functional trait composition of trees during forest succession remain poorly understood. The environmental filtering hypothesis of community assembly predicts gradual successional changes in functional trait composition. However, differences in landscape conditions and local site factors can lead to idiosyncratic or divergent changes in species and functional composition. We examine community assembly patterns of canopy trees and palms during natural regeneration using functional trait measurements and annual vegetation dynamics data over 18 years from six second-growth and two old-growth forest stands in northeast Costa Rica. We measured eight functional traits on 94 species, which cumulatively accounted for at least 80% of the abundance in each plot. Old-growth specialists had significantly lower leaf nitrogen and phosphorus content and significantly higher wood specific gravity than second-growth specialists. At the community level, specific leaf area, leaf nitrogen content, leaf dry matter content, and leaf toughness showed directional trends from more acquisitive trait values towards more conservative trait values. The landscape context was also an important driver of the successional trends within and among plots, as most functional trait patterns were influenced by the location of the plots within the matrix. Furthermore, the differential abundance of palms

2

among stands strongly influenced functional trait differences among plots. After accounting for the effects of stand basal area and stand location, random plot effects explained an additional 45– 90% of the variation in community-weighted mean functional traits. Local and landscape scale heterogeneity often influenced functional trait variation more strongly than stand basal area, suggesting that both stochasticity and environmental filtering influence species and functional turnover during succession.

Keywords: Community assembly; environmental filtering; La Selva Biological Station; natural regeneration; second-growth forests; stochasticity.

Introduction Regenerating forests globally comprise over half of the remaining tropical forests (FAO, 2010), providing a promising avenue for recuperating some of the biodiversity and ecosystem services that have largely been lost in human-modified landscapes (Chazdon et al., 2009). Tree species composition recovers relatively slowly during tropical succession and can vary considerably among similarly aged secondary forests depending upon soil conditions, land-use history and local seed dispersal (Brown and Lugo, 1990; Chazdon, 2003; Chazdon et al., 2007; Martin et al., 2013; Norden et al., 2015). These site-specific factors typically have long-term effects on the community structure and composition of regenerating forests (Mesquita et al., 2015; Norden et al., 2011).

3

Species composition is further influenced by the changing environmental conditions during forest development, including light availability (Denslow and Guzman, 2000; Guariguata and Ostertag, 2001; Nicotra et al., 1999) and soil fertility (Batterman et al., 2013; Guariguata and Ostertag, 2001; Lamb, 1980). Species’ responses to these environmental changes are mediated by their functional traits, which reveal species-specific patterns of survival, potential resource acquisition, and actual resource allocation to leaves, wood, roots and seeds (Cornelissen et al., 2003; Díaz et al., 2004; Lasky et al., 2014; Menge and Chazdon, 2015; Westoby et al., 2002; Wright et al., 2004). Theoretically, under a purely functional model of community assembly, community composition is expected to be constrained to species with the appropriate traits to overcome the abiotic and biotic filters necessary to first arrive at a site, and then to establish and grow (Rees et al., 2001; Vile et al., 2006; Weiher and Keddy, 1999). The stages of tropical succession are often described in terms of the composition of pioneers versus shade-tolerant species (Chazdon, 2008; Finegan, 1996), thus we expect plant functional traits to differ significantly among species with different successional affinities. Typically, early successional, short-lived pioneer species are classified as fast resource acquisition species, investing in cheap, short-lived leaves and low-density wood that provide a quick return on investment (Poorter and Bongers, 2006; Selaya and Anten, 2010; Westoby et al., 2002). Although the tissues of these acquisitive species are poorly protected from abiotic and biotic damage, their traits confer fast photosynthetic rates and high relative growth rates (Bazzaz and Pickett, 1980; Ellsworth and Reich, 1996). In contrast, shade-tolerant species often demonstrate resource conservation strategies. Species with resource conservation strategies have slow growth rates, but invest in long-lived tissues that are well defended against herbivores, pathogens and physical damage (Augspurger and Kelly, 1984; Coley, 1983; Reich et al., 2003).

4

This generalized dichotomy between pioneers and shade-tolerant species has led ecologists to place species into ecological groups based on field observations rather than statistically rigorous classifications of habitat affinities (Poorter, 2007; Swaine and Whitmore, 1988). Here, we explicitly test whether leaf and stem functional traits differ significantly among species with different successional affinities, using a multinomial model approach to robustly classify species as second-growth specialists, old-growth specialists, or successional generalists based on their estimated relative abundances in second-growth and old-growth forests (Chazdon et al., 2011; Letcher et al., 2015). Furthermore, we can determine the mechanisms that drive community assembly by assessing changes in community level functional trait values along successional gradients. When environmental filtering drives community assembly, the functional composition of forest communities is expected to shift directionally with stand age, in accordance with changing environmental conditions. Yet, local and landscape factors, such as topography and soils, proximity to forest patches, or stochastic factors such as dispersal limitation (Chazdon, 2008; Hubbell, 2001), can influence community assembly processes and cause significant site-specific variation in functional composition (Norden et al., 2015). The studies that have assessed standlevel functional trait dynamics of second-growth wet tropical forests (Craven et al., 2015; Dent et al., 2013; Lohbeck et al., 2013; Muscarella et al., 2015) have relied on a static chronosequence approach, in which temporal successional trends are inferred by measuring stands of different ages at a single time point. Although chronosequence studies provide useful insight into successional patterns (Prach and Walker, 2011), they often deviate from true vegetation dynamics because community reassembly patterns can be highly idiosyncratic, showing variation with land-use history and landscape factors that influence seed dispersal and establishment

5

(Chazdon et al., 2007; Johnson and Miyanishi, 2008; Norden et al., 2015). Here, for the first time, we combine successional chronosequence and temporal data with functional trait measurements to test the relative influences of environmental filtering, site factors and landscape context on community assembly during natural regeneration. We use 18 years of tree vegetation dynamics data in eight 1-ha forest monitoring plots to assess changes in functional trait composition during post-pasture succession in the lowland rainforest of northeastern Costa Rica. Our study provides a unique setting for testing the influence of site factors on functional composition, both because our long-term data allows us to examine trends within sites as well as across sites, and because our study sites are located within a heterogeneous landscape (Fagan et al., 2013). Although our second-growth stands shared similar land-use histories, they differ in landscape configuration and proximity to old-growth forests, which can influence species composition (Chazdon, 2008; Norden et al., 2015, 2009) and potentially functional composition. In our study area, proximity to old-growth forest favors abundant recruitment of canopy palms during succession (Sezen et al., 2009, 2007). Additionally, previous studies from our study location have shown that canopy palm abundance strongly drives patterns of compositional similarity among plots (Guariguata et al., 1997; Norden et al., 2009). Palms are often removed from community-level functional trait analyses because of the differences in leaf and stem construction between monocots and dicots (Dominy et al., 2008). However, palms are a dominant feature of tropical forest communities (Marín-Spiotta et al., 2007; ter Steege et al., 2013), and they play an important functional role in the forest. The unique columnar architecture of arborescent palms inhibits liana growth and limits the amount of light reaching the forest floor (Salm et al., 2005), and their fruits are an important food source for large mammals such as peccaries and primates, and large birds, such as toucans (Queenborough

6

et al., 2012). Palm abundance can directly influence functional composition if the differences in the construction of palms and dicots result in differences in functional traits. Furthermore, as the foraging activities around the base of palms can alter plant community composition (Queenborough et al., 2012), the abundance of palms in the landscape may have an indirect influence on functional composition. Our study focuses on the intermediate stages of succession, spanning the stem exclusion and understory reinitiation phases of tropical forest succession (Chazdon, 2008). Whereas pioneer trees with acquisitive functional traits dominate the initial, stand initiation phase of succession (lasting up to ~10–15 years after disturbance subsides), we focus on the two subsequent phases of succession, which contain an increasing diversity of species and functional types and provide deeper insight into the transition from second-growth forest to old-growth forest. Although the exact timeline of these stages varies regionally and is influenced by land-use history and other factors, in the Caribbean lowlands of Costa Rica the stem exclusion phase spans approximately 10–25 years after disturbance, and the understory reinitiation phase begins approximately 25 years after disturbance and can last for up to 200 years (Chazdon, 2008). Our study also includes old-growth forests, which serve as reference points for the final stages of succession. To understand the driving mechanisms of community assembly along this wet tropical forest successional gradient, we address the following questions: 1. Do tree species with statistically established successional habitat affinities exhibit the expected trends in ecophysiological functional traits? Specifically, do second-growth specialists have fast resource acquisition traits, old-growth specialists have resource conservation traits, and generalist species have intermediate traits? Do generalist species show more intraspecific trait variation than second-growth and old-growth specialists?

7

2. Does functional composition change directionally along the successional gradient, supporting a deterministic view of community assembly? What is the relative importance of stand age, local site factors, and landscape context in driving functional community assembly during succession? 3. How does the abundance of canopy palms influence the successional patterns of functional community assembly within and across plots?

Materials and methods Study location and stand characteristics This study was conducted within and around La Selva Biological Station (hereafter La Selva), in the province of Heredia in northeastern Costa Rica (Table 1, Appendix Map). This region is classified as tropical lowland wet forest, with an average annual temperature of 26.5C and ~3900 mm of rainfall (McDade et al., 1994). The elevation of the study plots ranges from 40 to 200 meters above sea level. The region is comprised of a mixture of second-growth and oldgrowth forests, pasture, agriculture and plantations (Fagan et al., 2013). Our study sites include six second-growth forest plots, which are naturally regenerating following the abandonment of pasture, and two old-growth forest plots with no recent record of human disturbance (Appendix Map; McDade and Hartshorn, 1994). At the time of the first annual census (1997 for four plots, 2005 for four plots), the second-growth forest plots ranged in age from 10 to 25 years post-disturbance (Table 1). In 2014, these second-growth forest plots ranged in age from 19 to 42 years old. Two of the second-growth plots and one old-growth plot are located within La Selva, whereas the other five plots are located on private farms approximately 6-20 km west of La Selva. All eight plots have similar soil nutrient stocks, soil

8

texture, and nutrient cycling rates, although one second-growth forest plot, LSUR, has higher soil phosphorus concentration than the other plots (Logan, 2007; Menge and Chazdon, 2015; Wood et al., 2009). However, the plot locations differ in landscape composition and other site factors. La Selva has been protected for over 50 years (McDade and Hartshorn, 1994), resulting in a higher proportion of intact forest surrounding the La Selva plots compared to the plots outside of La Selva (Fagan et al., 2013). Moreover, since hunting is banned within La Selva, the density of large vertebrates is also higher, with long-term impacts on seed predation and dispersal (Kuprewicz, 2013). The vegetation dynamics of stems ≥ 5 cm diameter at breast height (DBH) were monitored annually in all plots for 10–18 years (Table 1; Chazdon et al., 2007; Lasky et al., 2014); the most recent census included in this study was completed in 2014. We restricted our analyses to species classified as canopy trees or palms (species with mature adult height ≥ 15 m). We measured functional traits for the most common species, whose cumulative abundance comprised at least 80% of the canopy tree and palm communities of each plot for each census year (range: 81.197.7%; mean ± SE: 92.0 ± 0.004%), and whose cumulative basal area comprised at least 90% of the canopy tree and palm communities (range: 90.998.5%; mean ± SE: 95.8 ± 0.002%). In total, our dataset comprised 89 dicot tree species and 5 canopy palm species.

Species classifications We classified our 94 focal species into successional specialist categories based on their relative abundance in our old-growth and second-growth forest plots between 1997 and 2011. Species were classified as second-growth specialists, old-growth specialists, generalists, or too

9

rare to classify using a multinomial model that minimizes bias due to different sampling intensities or insufficient sampling of rare species (Chazdon et al., 2011). Full details of this statistical method can be found in Chazdon et al. (2011). In brief, true relative species abundances are estimated for each species in old-growth forests and second-growth forests, and then the ratio of estimated species relative abundance in one forest type to estimated species relative abundance in the other forest types is computed. This ratio is compared to a specialization threshold value to determine if the species is a habitat specialist or a generalist, or if the species is too rare to classify. We performed the classifications in CLAM (Chao and Lin, 2011), using the conservative supermajority rule threshold value (K=2/3) and an overall P = 0.01.

Functional Trait Measurements Functional trait measurements were conducted over a four-year period (2008–2012). For each species we measured 8 functional traits, which demonstrate tradeoffs in resource allocation and are commonly used to classify ecological strategies (see Table A1 in the Appendix). The traits we measured were: leaf size (LS), specific leaf area (SLA), leaf dry matter content (LDMC), leaf thickness (LT), leaf toughness (LTO), leaf nitrogen content (LNC), leaf phosphorus content (LPC), and area-weighted wood specific gravity (wWSG; Plourde et al., 2015). Functional traits were measured on mature trees using standardized protocols (Cornelissen et al., 2003; Williamson and Wiemann, 2010; for detailed descriptions of methods, see Supplemental Methods in Appendix A). Leaf traits were measured on sunlit leaves with little to no herbivory or epiphyll cover, whenever possible. We generally sampled functional traits for

10

each species in each plot where it was common (i.e. one of the species that contributed to the cumulative 80% abundance in that plot), sampling on average 6 individuals per species per plot (range = 1–17). For our analyses we generally used plot-specific trait values for each species, although we used species-mean values, calculated from individuals across all plots, when traits were measured on fewer than 5 individuals for LS, SLA, LDMC, LT, LTO, or fewer than 3 individuals for LNC, LPC, wWSG, in a given plot. Our plot-specific measurements account for environmental plasticity across study sites, but do not account for year-to-year variation within plots.

Statistical Analyses We used Pearson’s correlation coefficient to assess the correlation among species mean functional trait values. To test for differences in functional trait values among successional specialist categories, we used ANOVA analyses on species mean trait values, followed by Tukey HSD post-hoc tests. We natural log-transformed leaf size, specific leaf area, leaf toughness and leaf phosphorus content to improve normality. For each species and each trait we computed intraspecific trait variability as the coefficient of variation (CV) of all of the measured individual trait values for that species (Albert et al., 2011), including within-plot and among-plot sources of variation. For the traits that were normally distributed across all measured individuals (LDMC and wWSG), we calculated CV as the sample standard deviation divided by the sample mean. For the traits that were log-normally distributed across all measured individuals (LS, SLA, LT, LTO, LNC, LPC), we calculated the geometric CV as the square root of the exponentiated sample variance of natural log transformed data minus one. We used ANOVA analyses to compare intraspecific trait variation among successional specialist categories.

11

To assess community-level functional changes along the successional gradient, we calculated community weighted mean (CWM) values for each trait in each plot and each census year. Community weighted mean is the sum over all species of the species’ trait value weighted by its relative abundance or relative basal area in the community (Garnier et al., 2004), and represents the trait value of an average individual in the community. We calculated CWM values using population-level mean trait values and species relative abundances for all 94 focal species. We focus on abundance-weighted CWM patterns, but, as a comparison, we also calculated CWM values using species relative basal area (CWMBA). Canopy palms contributed unequally to the proportion of individuals among sites and years (Fig. A1). To understand the extent to which canopy palms influenced successional functional trait patterns, we compared species mean trait values among the five canopy palm species and 89 canopy tree species using Welch’s t-tests, and used the Holm-Bonferroni method to adjust the P-values for multiple comparisons (Holm, 1979). Leaf size was natural logtransformed prior to analysis. We then recalculated CWM values for each plot and census year based only the dicotyledonous tree communities (i.e. excluding palms). To determine how CWM trait values varied along the successional gradient, we used linear mixed effects models fit using Restricted Maximum Likelihood (REML) estimation (Zuur et al., 2009). The linear mixed effect model approach allows us to compare how functional trait values change along the successional gradient both among the different forest stands as well as within each stand over time. Each model includes fixed effects for stand basal area, which is a proxy for both age and complexity (Lebrija-Trejos et al., 2010), and stand location (i.e. within or outside of La Selva), which is a proxy for landscape factors that may influence species colonization and distribution. Among our second growth plots, stand age was strongly correlated

12

with stand basal area (r = 0.87, Fig. A2), and using stand basal area instead of stand age allowed us to include old-growth forests in our mixed effect models. To account for the temporal correlation within each plot, each mixed effect model included a within-plot autocorrelation structure. Each model also contained a random effects term (intercept and/or slope) for plot to account for the repeated measurements in each plot across years. The random plot effects (ascribed to plot identity) serve as a proxy for residual variation due to idiosyncratic local site factors that is not explained by landscape context or stand basal area. Separately for each CWM trait, we first determined the best-fit autocorrelation structure (in all cases the best-fit autocorrelation structure was an autoregressive model of order one), and then optimized the random effect structure (Zuur et al., 2009). For each set of models, we used small sample size corrected Akaike’s Information Criterion (AICc) to determine the most parsimonious model (i.e., the simplest model within 2 AICc units of the model with the lowest AICc (Burnham and Anderson, 2002). For each CWM trait calculation, we ran likelihood ratio tests for models fit with maximum likelihood to determine which of the fixed effects  basal area, stand location, and the interaction between basal area and stand location  were significant. We also used mixed effects models to assess variation in the proportion of individuals in each successional specialist category along the successional gradient. We used stand age instead of stand basal area as the predictor to determine if the proportion of individuals in each specialist category changes significantly as the second-growth forest stands age. As in the CWM mixed effects models, the models included a within-group autocorrelation structure (autoregressive model of order one), as well as plot-level random effects. We used stand age instead of stand basal area as To determine if stand age was a significant predictor of the proportion of

13

individuals in each of the specialist categories, we compared a model with a fixed effect for stand age to an intercept only model (no fixed effects) and used AICc to determine the best-fit model. Following Nakagawa and Schielzeth (2013), two correlation metrics were calculated for each mixed effects model. The marginal correlation metric for linear mixed effects models, R2m, measures the variance described only by the fixed effects, which in our case include stand basal area and/or stand location. The conditional correlation metric, R2c, expresses the variance explained by both the fixed and random effects. All statistical analyses were performed in the R 3.2.4 statistical platform (R Core Team, 2016). Community weighted means were calculated in the ‘FD’ package (Laliberté and Legendre, 2010; Laliberté and Shipley, 2011), linear mixed effects models were analyzed using the ‘nlme’ package (Pinheiro et al., 2015), and marginal and conditional correlations were analyzed using the ‘piecewiseSEM’ package (Lefcheck, 2016).

Results Functional traits and successional categories The eight functional traits measured showed a wide range of values across the 94 focal species (Table A2). Twelve species mean functional trait correlations were significant, the strongest of which was the positive correlation between leaf phosphorus and leaf nitrogen concentrations r = 0.66, Table A3). Each successional category was well represented in our focal species (21, 24, 25 species for second-growth specialists, generalists and old-growth specialists, respectively). Twenty-four species were too rare to be classified in any of these groups, but those individuals comprised no

14

more than 11% of the abundance of any plot in any census (Fig. A3d). Of the five focal palm species, two were classified as generalists, one was classified as a second-growth specialist, one was classified as an old-growth specialist, and one was too rare to be classified. Among the second-growth forests, the proportion of second-growth specialists and rare species decreased significantly with stand age, whereas the proportion of generalists and old-growth specialists increased with stand age (Fig. A3). Three functional traits differed significantly among specialist categories. Old-growth specialists had significantly lower leaf phosphorus content (LPC) than second-growth specialists, and significantly lower leaf nitrogen content (LNC) and higher area-weighted wood specific gravity (wWSG) than either second-growth specialists or generalists (Fig. 1). Similar results were found when palms were excluded from the analyses (data not shown). Intraspecific trait variability was not significantly different among successional categories for any trait (ANOVA, P > 0.05 for all tests), although old-growth specialists consistently had the lowest mean intraspecific variability across all traits (Table A4).

Community level shifts in functional trait values along the successional gradient Across the canopy tree and palm communities, abundance-weighted community mean (CWM) trait values changed directionally along the successional gradient with both stand basal area and stand location for five of the eight measured traits, with stand basal area only for one trait, and with stand location only for one trait (Fig. 2, Table A5). As predicted by the environmental filtering hypothesis, CWM values for specific leaf area (SLA) and leaf nitrogen content (LNC) declined significantly with stand basal area, and CWM values for leaf dry matter content (LDMC) and leaf toughness (LTO) increased significantly with stand basal area. Stand

15

location also significantly influenced the successional trends for three of these traits (SLA, LNC, and LTO), where the trends in CWM trait values with stand basal area were stronger among plots within La Selva compared to the plots outside of La Selva. The CWM trend for leaf thickness (LT) also demonstrated the expected increase with stand basal area among the plots at La Selva, but showed a decline with stand age among the plots in the surrounding region. CWM leaf size (LS) increased along the successional gradient, contrary to the expectations of the environmental filtering hypothesis, and this pattern was also stronger among the plots at La Selva. The proportion of species with compound leaves increased over time in four secondgrowth plots (Fig. A4), but this is unlikely to be driving the CWM LS pattern since our functional trait measurements of compound-leaved species were performed on leaflets instead of entire leaves (see Supplemental Methods in the Appendix). The successional trajectory of CWM leaf phosphorus content (LPC) did not change directionally with stand basal area, but the intercept value was higher among plots at La Selva compared to the plots in the surrounding region (Fig. 2; Table A5). Neither stand basal area nor stand location significantly influenced the CWM trends for area-weighted wood specific gravity (wWSG). Almost identical results were found when using stand age instead of stand basal area as the predictor of CWM trends (data not shown). The fixed effects, stand age and location, explained 9–54% of the variation in CWM values for the eight measured traits (R2m; Fig. 2). Random plot effects explained an additional 45–90% of the CWM variation (R2c - R2m; Fig. 2), indicating that additional site-specific local factors strongly affected community level trait values. Some of the patterns for community weighted mean values weighted by basal area (CWMBA) differed from the abundance-weighted CWM trends. Notably, whereas CWM LS

16

increased with stand basal area, CWMBA LS decreased (Fig. A5, Table A6). Whereas CWM LPC did not show a trend with stand basal area, CWMBA LPC declined with basal area (Fig. A5, Table A6). Although CWM LTO demonstrated a strong increase with basal area, basal area was not a significant predictor of CMWBA (Fig. A5, Table A6). The remaining CWMBA trends were generally similar to the abundance-weighted CWM trends, although in some cases stand basal area had a weaker effect on the trends and stand location had a stronger effect (Fig. A5, Table A6).

Canopy palms and trees Canopy palms comprised a considerable proportion (>10%) of the individuals in three of the six second-growth forests and in both of the old-growth forests (Fig. A1). Across all measured individuals, the palms had significantly larger and tougher leaves, with significantly lower specific leaf area than the dicotyledonous trees (Table 2). The exclusion of palms from the dataset altered the successional CWM patterns for six functional traits. Whereas the CWM trend for LS increased along the successional gradient when palms were included, CWM LS decreased with stand basal area when palms were excluded. This trend was stronger for plots within La Selva compared to plots outside of La Selva. Whereas the CWM trend for SLA varied with stand basal area and location when both trees and palms were considered, no predictors were significant when palms were excluded (Fig. 3; Table A5). The influence of stand basal area on CWM LNC, LDMC, LT and LTO disappeared when palms were excluded, although the tree communities at La Selva had higher leaf nitrogen content and dry matter content, and lower leaf thickness and toughness than the tree communities outside of La Selva. Regardless of whether palms were included, CWM LPC was

17

higher among the plots within La Selva, and CWM wWSG showed no trend across the successional gradient (Figs 2 and 3; Table A5). Canopy palms generally comprised less than 10% of the basal area in each plot (except for LEPS, where they comprised up to 13% of the basal area, data not shown). Thus, the exclusion of palms from the dataset had little impact on CMWBA trends (Figs A5 and A6; Table A6).

Discussion This study is the first to combine long-term stand dynamics with chronosequence patterns to assess changes in functional trait distributions along a tropical wet forest successional gradient, providing a deeper understanding of the role of functional trait variation in community assembly. As predicted by the environmental filtering hypothesis, the community weighted mean (CWM) values for four functional traitsspecific leaf area, leaf nitrogen content, leaf dry matter content and leaf toughnessshifted from more acquisitive to more conservative states with stand basal area (a proxy for stand age and complexity). The increasing abundance of palms during forest succession strongly influenced these directional trends. Landscape context and local site factors also strongly influenced the variation in species composition and functional trends, suggesting that other factors, such as local edaphic differences, landscape history, and the species composition of the surrounding landscape, also play a major role in the community assembly of tropical second-growth forests. Our prediction that second-growth specialists would have fast resource acquisition traits and old-growth specialists would have resource conservation traits was not universally supported. As predicted, second-growth specialists had significantly higher leaf nitrogen and

18

phosphorus content and lower wood specific gravity than old-growth specialists (Fig. 1). These three traits have been shown to strongly influence tree growth and survival rates in our plots (Lasky et al., 2014) and in other studies (Kraft et al., 2010; Poorter and Bongers, 2006; Raaimakers et al., 1995; Wright et al., 2010). Old-growth specialists spend most of their lives in shaded understory conditions, where dense wood provides them with structural support and protection from abiotic and biotic damage (Augspurger and Kelly, 1984; van Gelder et al., 2006), and leaves with higher concentrations of structural and/or defensive compounds, and consequently lower nutrient concentrations, tend to have longer leaf lifespans (Poorter and Bongers, 2006). Second-growth specialists, on the other hand, are often adapted for faster growth rates rather than survival, and generally prioritize short-term gains over long-term costs and risks. Higher concentrations of leaf nitrogen and phosphorus allow second-growth specialists to have higher photosynthetic rates (Ellsworth and Reich, 1996; Raaimakers et al., 1995; Wright et al., 2004). Moreover, as pathogen activity is reduced in high light environments, dense wood is less critical for second-growth specialists (Augspurger and Kelly, 1984). Second-growth specialists preferentially allocate carbon to growth (diameter and height), and by producing thicker trunks of lower density wood, they are able to achieve high mechanical strength with low construction costs (Chave et al., 2009; Clark and Clark, 2001; Larjavaara and Muller-Landau, 2010; Plourde et al., 2015). Radial increases in wood specific gravity are common among the trees in our young and intermediate aged study plots (Plourde et al., 2015). Using phylogenetic independent contrasts to relate functional traits to habitat specialization across a 14 successional sites in the Neotropics, Letcher et al. (2015) also found some evidence that old-growth specialists have denser wood and lower SLA than second-growth specialists.

19

Surprisingly, the other functional traits in this study did not differ among successional specialist categories, nor did we find any significant differences in intraspecific trait variation among the categories. Our robust classification scheme was based on each species’ relative abundance in second-growth and old-growth forests, and not on their functional characteristics or demographic rates (Chazdon et al., 2011). Using a similar approach to ours, Letcher et al. (2015) also found that few functional traits differed among successional habitat classification groups. Our results suggest that few species fit precisely into the pioneer and shade-tolerant functional trait dichotomy, and instead fall along a continuum between these two extremes (Wright et al., 2010). Each of the successional categories demonstrated a large range in functional trait values (Fig. 1), suggesting that the functional characteristics we measured were not critical factors driving successional changes in species relative abundance. Although Lasky et al. (2014) found that average survival rates were higher for species with higher LDMC, LTO, or WSG values, we demonstrate that these trait values were not limited to any successional specialist or generalist categories (Fig. 1). This result could be due to the fact that the second-growth stands used for the multinomial classifications were already fairly well developed, with closed canopies, and shortlived pioneer species were already declining or absent. Indeed, most of the early successional specialist species that remain are long-lived pioneers. The low abundance of short-lived pioneers may at least partially explain the lack of significant differences among the successional specialist categories for many of the functional traits. Irrespective of the variation among successional specialist categories, four communitylevel functional traits showed directional trends with stand basal area, consistent with the predictions of the environmental filtering hypothesis (Fig. 3). We found directional shifts from acquisitive traits in young second-growth forests towards a predominance of resource

20

conservation traits related to shade-tolerance in older forests for two fast-resource acquisition traits (high SLA and high LNC) and two resource conservation traits (high LDMC and high LTO). Environmental filtering has also been shown to influence community-level functional composition in successional tropical forests of Mexico (Lohbeck et al., 2013), Puerto Rico (Muscarella et al., 2015), Panama (Craven et al., 2015), and New Guinea (Whitfeld et al., 2014), and along an environmental gradient in the Amazon (Fortunel et al., 2014). The community level patterns we observed for SLA, LNC, LDMC and LTO support the expectation that species with fast-resource acquisition strategies predominate in younger forests, which are more open and have higher light availability than older forests (Lebrija-Trejos et al., 2011; Nicotra et al., 1999). When light conditions and other resources become more restricted, resource conservation strategies become more important. Among other studies that have looked at successional changes in functional traits, SLA (or its inverse, leaf mass per area, LMA) and LNC most consistently demonstrate directional trends with stand age or stand basal area (Lohbeck et al., 2013; Muscarella et al., 2015; Whitfeld et al., 2014), although in some cases the trends are the opposite of what is expected under environmental filtering (ex., Becknell and Powers, 2014; Craven et al., 2015). Community-weighted LTO was also found to increase along a successional gradient in Panama (Craven et al., 2015), and community-weighted LDMC increased along a successional gradient in Mexico (Lohbeck et al., 2013). Stand location was a significant factor in many of the CWM regressions, suggesting that landscape context also influences functional composition. Although explicit landscape factors can be difficult to quantify, we are aware of certain differences between the two stand locations in our study. First, the plots located within La Selva Biological Station are surrounded by a higher proportion of old-growth forest than the plots outside of La Selva (Fagan et al., 2013;

21

McDade and Hartshorn, 1994). Second, the higher density of vertebrate fauna within La Selva increases rates of both seed dispersal and seed predation (Kuprewicz, 2013). Third, tree species composition has been shown to differ among our plots at La Selva and our plots in the surrounding region (Norden et al., 2009), and here we explicitly show that among similar aged plots, canopy palm abundance is higher for plots within La Selva (Fig. A1). The abundance of palms in these forests was clearly an important driver of the successional changes in functional composition. Although canopy trees only differed significantly from palms for a few of the functional traits (Table 2), the inclusion of palms influenced successional trends for six of the eight measured traits (Figs 2 and 3; Table A5). In general, the observed successional trends disappeared when palms were removed from the analyses. We attribute the unexpected increase in CWM leaf size with stand basal area to the increasing abundance of palms in older plots, particularly those plots within La Selva (Fig. A1). The differential palm abundance among plots may reflect variation in soil conditions or other landscape factors. The four most abundant palm species included in our study have specific soil preferences (Clark et al., 1999), which may influence their distribution within and among plots. Although soil texture and nutrient concentrations are largely similar among our study plots (Logan, 2007; Wood et al., 2009), topographic differences among our plots may influence palm density (Clark et al., 1995). Moreover, although palms can experience occasional long-distance dispersal events, local seed dispersal is much more common (Sezen et al., 2007). Thus the abundance of palms in any given location is highly dependent upon their abundance in the surrounding landscape. Palms are an important component of Neotropical wet forests (Guariguata et al., 1997; Marín-Spiotta et al., 2007; Svenning, 1998; ter Steege et al., 2013). Not only are palms widespread, but they also provide an important food source for animals

22

(Queenborough et al., 2012), and they increase the phylogenetic diversity of tropical forest communities (Letcher, 2010; Muscarella et al., 2015). Due to their importance in the community, and their strong influence on functional trends, we suggest that palms be included in future studies examining tropical patterns of community assembly. Plot identity explained an even larger proportion of the variation in CWM traits than stand basal area and/or stand location for most traits (i.e., R2c - R2m, Figs 2 and 3), indicating that local site-specific factors have far stronger effects on functional composition. Observed site-level differences in CWM traits are largely driven by differences in species composition and speciesspecific mortality and recruitment rates over time (Rozendaal and Chazdon, 2015). For example, the increase in CWM wWSG in TIR can be primarily attributed to the mortality of secondgrowth specialists (Fig. A3), particularly species with low wWSG values, such as Vochysia ferruginea (wWSG = 0.34), Simarouba amara (wWSG = 0.36) and Hampea appendiculata (wWSG = 0.27). Similarly, the decline in wWSG in LEPS was largely influenced by the increasing abundance of palms with relatively low wWSG values, such as Socratea exorrhiza (wWSG = 0.23) and Iriartea deltoidea (wWSG = 0.29). Idiosyncratic differences in CWM trait values among plots likely reflect stochastic processes and unmeasured conditions. Norden et al. (2015) demonstrate that successional trajectories in stem density, basal area, and species density are best explained by an approximately equal combination of deterministic and stochastic factors. Our results suggest that the influence of these factors extends to the functional composition of successional forest communities.

Conclusions

23

Our study provides strong evidence that environmental filtering, landscape context, and plot-specific factors are all important in structuring tree communities and functional trait distributions over time during tropical forest succession. These functional trait changes are likely to influence ecosystem processes and services in successional landscapes (Díaz et al., 2004). We found clear successional trends across a second-growth forest gradient for some leaf-based functional traits (leaf size, specific leaf area, leaf nitrogen content, leaf dry matter content, and leaf toughness). Yet, other functional traits demonstrated inconsistent trends among stand locations and stochastic functional patterns among plots, suggesting that differences in site conditions such as landscape configuration, edaphic conditions, biotic interactions, initial colonization patterns and local palm abundance also significantly influence community assembly. The variability in functional composition among plots highlights the importance of applying long-term forest dynamics data to better understand successional patterns.

Acknowledgments Long-term tree monitoring was supported by grants to RLC from the Andrew Mellon Foundation, US NSF awards 0424767, 0639393, 1147429, and 1110722, and NASA ROSES Grant NNH08ZDA001N-TE. Functional trait research was supported by the following awards to VKB: NSF Graduate Research Fellowship, NSF DEB-1110722, OTS and the Christiane and Christopher Tyson Fellowship, the Garden Club of America Award in Tropical Botany, Lewis and Clark Fund for Exploration and Field Research, and the Ronald Bamford Endowment to the Department of Ecology and Evolutionary Biology. We are thankful to B. Plourde, F. Cervo, R. Gonzales, M. Gaitan, E. Salicetti, J. Paniagua, B. Paniagua, K. Felich, A. Schmidt, K.M. Weaver, D. Schmidt, T. Harvey, F. Mok, and the staff at La Selva Biological Station and the

24

Organization for Tropical Studies. We thank D. Rozendaal for advice and assistance in the performing statistical analyses and N. Norden and three anonymous reviewers for providing helpful comments on earlier drafts of the manuscript.

Appendix A Additional details on functional traits, methodology and results.

Data Accessibility Tree and canopy palm abundance data is archived in Ecological Applications (Rozendaal and Chazdon, 2015) at http://dx.doi.org/10.1890/14-0054.1.sm.

25

Literature Cited Albert, C.H., Grassein, F., Schurr, F.M., Vieilledent, G., Violle, C., 2011. When and how should intraspecific variability be considered in trait-based plant ecology? Perspect. Plant Ecol. Evol. Syst. 13, 217–225. doi:10.1016/j.ppees.2011.04.003 Augspurger, C.K., Kelly, C.K., 1984. Pathogen mortality of tropical tree seedlings: experimental studies of the effects of dispersal distance, seedling density, and light conditions. Oecologia 61, 211–217. Batterman, S.A., Hedin, L.O., van Breugel, M., Ransijn, J., Craven, D.J., Hall, J.S., 2013. Key role of symbiotic dinitrogen fixation in tropical forest secondary succession. Nature 502, 224–7. doi:10.1038/nature12525 Bazzaz, F.A., Pickett, S.T.A., 1980. Physiological ecology of tropical succession: a comparative review. Annu. Rev. Ecol. Syst. 11, 287–310. Becknell, J.M., Powers, J.S., 2014. Stand age and soils as drivers of plant functional traits and aboveground biomass in secondary tropical dry forest. Can. J. For. Res. 44, 604–613. Brown, S., Lugo, A.E., 1990. Tropical secondary forests. J. Trop. Ecol. 6, 1–32. Burnham, K.P., Anderson, D.R., 2002. Model selection and multimodel inference: a practical information-theoretic approach, 2nd ed. Springer, New York, USA. Chao, A., Lin, S.Y., 2011. Program CLAM (Classification Method). Program and users’s guide at http://purl.oclc.org/clam. Chave, J., Coomes, D., Jansen, S., Lewis, S.L., Swenson, N.G., Zanne, A.E., 2009. Towards a worldwide wood economics spectrum. Ecol. Lett. 12, 351–66. doi:10.1111/j.14610248.2009.01285.x Chazdon, R.L., 2008. Chance and determinism in tropical forest succession, in: Carson, W.P., Schnitzer, S.A. (Eds.), Tropical Forest Community Ecology. Blackwell Publishing Ltd., West Sussex, UK. Chazdon, R.L., 2003. Tropical forest recovery: legacies of human impact and natural disturbances. Perspect. Plant Ecol. Evol. Syst. 6, 51–71. Chazdon, R.L., Chao, A., Colwell, R.K., Lin, S.-Y., Norden, N., Letcher, S.G., Clark, D.B., Finegan, B., Arroyo, J.P., 2011. A novel statistical method for classifying habitat generalists and specialists. Ecology 92, 1332–43. Chazdon, R.L., Finegan, B., Capers, R.S., Salgado-Negret, B., Casanoves, F., Boukili, V., Norden, N., 2010. Composition and dynamics of functional groups of trees during tropical forest succession in northeastern Costa Rica. Biotropica 42, 31–40. Chazdon, R.L., Letcher, S.G., van Breugel, M., Martínez-Ramos, M., Bongers, F., Finegan, B., 2007. Rates of change in tree communities of secondary Neotropical forests following major disturbances. Philos. Trans. R. Soc. Lond. B. Biol. Sci. 362, 273–89. doi:10.1098/rstb.2006.1990 Chazdon, R.L., Peres, C.A., Dent, D., Sheil, D., Lugo, A.E., Lamb, D., Stork, N.E., Miller, S.E., 2009. The potential for species conservation in tropical secondary forests. Conserv. Biol. 23, 1406–17. doi:10.1111/j.1523-1739.2009.01338.x

26

Clark, D.A., Clark, D.B., 2001. Getting to the canopy: tree height growth in a neotropical rain forest. Ecology 82, 1460–1472. Clark, D.A., Clark, D.B., Sandoval M., R., Castro C., M.V., 1995. Edaphic and human effects on landscape-scale distributions of tropical rain forest palms. Ecology 76, 2581–2594. Clark, D.B., Palmer, M.W., Clark, D.A., 1999. Edaphic factors and the landscape-scale distributions of tropical rain forest trees. Ecology 80, 2662–2675. Coley, P.D., 1983. Herbivory and defensive characteristics of tree species in a lowland tropical forest. Ecol. Monogr. 53, 209–234. Cornelissen, J.H.C., Lavorel, S., Garnier, E., Díaz, S., Buchmann, N., Gurvich, D.E., Reich, P.B., ter Steege, H., Morgan, H.D., van der Heijden, M.G.A., Pausas, J.G., Poorter, H., 2003. A handbook of protocols for standardised and easy measurement of plant functional traits worldwide. Aust. J. Bot. 51, 335. doi:10.1071/BT02124 Craven, D., Hall, J.S., Berlyn, G.P., Ashton, M.S., van Breugel, M., 2015. Changing gears during succession: shifting functional strategies in young tropical secondary forests. Oecologia 179, 1–13. doi:10.1007/s00442-015-3339-x Denslow, J.S., Guzman, S., 2000. Variation in stand structure, light and seedling abundance across a tropical moist forest chronosequence, Panama. J. Veg. Sci. 11, 201–212. Dent, D.H., DeWalt, S.J., Denslow, J.S., 2013. Secondary forests of central Panama increase in similarity to old-growth forest over time in shade tolerance but not species composition. J. Veg. Sci. 24, 530–542. doi:10.1111/j.1654-1103.2012.01482.x Díaz, S., Hodgson, J.G., Thompson, K., Cabido, M., Cornelissen, J.H.C., Jalili, A., MontserratMartí, G., Grime, J.P., Zarrinkamar, F., Asri, Y., Band, S.R., Basconcelo, S., Castro-Díez, P., Funes, G., Hamzehee, B., Khoshnevi, M., Pérez-Harguindeguy, N., Pérez-Rontomé, M.C., Shirvany, A., Vendramini, F., Yazdani, S., Abbas-Azimi, R., Bogaard, A., Boustani, S., Charles, M., Dehghan, M., de Torres-Espuny, L., Falczuk, V., Guerrero-Campo, J., Hynd, A., Jones, G., Kowsary, E., Kazemi-Saeed, F., Maestro-Martínez, M., Romo-Díez, A., Shaw, S., Siavash, B., Villar-Salvador, P., Zak, M.R., 2004. The plant traits that drive ecosystems: evidence from three continents. J. Veg. Sci. 15, 295. doi:10.1658/11009233(2004)015[0295:TPTTDE]2.0.CO;2 Dominy, N.J., Grubb, P.J., Jackson, R. V, Lucas, P.W., Metcalfe, D.J., Svenning, J.-C., Turner, I.M., 2008. In tropical lowland rain forests monocots have tougher leaves than dicots, and include a new kind of tough leaf. Ann. Bot. 101, 1363–77. doi:10.1093/aob/mcn046 Ellsworth, D.., Reich, P.B., 1996. Photosynthesis and leaf nitrogen in five Amazonian tree species during early secondary succession. Ecology 77, 581–594. Fagan, M.E., DeFries, R.S., Sesnie, S.E., Arroyo, J.P., Walker, W., Soto, C., Chazdon, R.L., Sanchun, A., 2013. Land cover dynamics following a deforestation ban in northern Costa Rica. Environ. Res. Lett. 8, 34017. doi:10.1088/1748-9326/8/3/034017 FAO (Food and Agriculture Organization), 2010. Global Forest Resources Assessment 2010. FAO, Rome, Italy. Finegan, B., 1996. Pattern and process in Neotropical secondary rain forests: the first 100 years of succession. Trends Ecol. Evol. 11, 119–24. doi:10.1016/0169-5347(96)81090-1

27

Fortunel, C., Paine, C.E.T., Fine, P.V.A., Kraft, N.J.B., Baraloto, C., 2014. Environmental factors predict community functional composition in Amazonian forests. J. Ecol. 102, 145– 155. doi:10.1111/1365-2745.12160 Garnier, E., Cortez, J., Billês, G., Navas, M.-L., Roumet, C., Debussche, M., Laurent, G., Blanchard, A., Aubry, D., Bellmann, A., Neill, C., Toussaint, J.-P., 2004. Plant functional markers capture ecosystem properties during secondary succession. Ecology 85, 2630– 2637. Guariguata, M.R., Chazdon, R.L., Denslow, J.S., Dupuy, J.M., Anderson, L., 1997. Structure and floristics of secondary and old-growth forest stands in lowland Costa Rica. Plant Ecol. 132, 107–120. Guariguata, M.R., Ostertag, R., 2001. Neotropical secondary forest succession: changes in structural and functional characteristics. For. Ecol. Manage. 148, 185–206. doi:10.1016/S0378-1127(00)00535-1 Holm, S., 1979. A simple sequentially rejective multiple test procedure. Scand. J. Stat. 6, 65–70. Hubbell, S.P., 2001. The unified neutral theory of biodiversity and biogeography. Princeton University Press, Princeton, New Jersey, USA. Johnson, E.A., Miyanishi, K., 2008. Testing the assumptions of chronosequences in succession. Ecol. Lett. 11, 419–31. doi:10.1111/j.1461-0248.2008.01173.x Kraft, N.J.B., Metz, M.R., Condit, R.S., Chave, J., 2010. The relationship between wood density and mortality in a global tropical forest data set. New Phytol. 188, 1124–36. doi:10.1111/j.1469-8137.2010.03444.x Kuprewicz, E.K., 2013. Mammal abundances and seed traits control the seed dispersal and predation roles of terrestrial mammals in a Costa Rican forest. Biotropica 45, 333–342. Laliberté, E., Legendre, P., 2010. A distance-based framework for measuring functional diversity from multiple traits. Ecology 91, 299–305. Laliberté, E., Shipley, B., 2011. FD: measuring functional diversity from multiple traits, and other tools for functional ecology. R package version 1.0-11. Lamb, D., 1980. Soil nitrogen mineralisation in a secondary rainforest succession. Oecologia 47, 257–263. Larjavaara, M., Muller-Landau, H.C., 2010. Rethinking the value of high wood density. Funct. Ecol. 24, 701–705. doi:10.1111/j.1365-2435.2010.01698.x Lasky, J.R., Uriarte, M., Boukili, V.K., Chazdon, R.L., 2014. Trait-mediated assembly processes predict successional changes in community diversity of tropical forests. PNAS 111, 5616– 5621. Lebrija-Trejos, E., Pérez-García, E.A., Meave, J.A., Bongers, F., Poorter, L., 2010. Functional traits and environmental filtering drive community assembly in a species-rich tropical system. Ecology 91, 386–98. Lebrija-Trejos, E., Pérez-García, E.A., Meave, J.A., Poorter, L., Bongers, F., 2011. Environmental changes during secondary succession in a tropical dry forest in Mexico. J. Trop. Ecol. 27, 477–489. doi:10.1017/S0266467411000253 Lefcheck, J.S., 2016. piecewiseSEM: piecewise structural equation modeling in R for ecology, 28

evolution, and systematics. Methods Ecol. Evol. 7, 573–579. Letcher, S.G., 2010. Phylogenetic structure of angiosperm communities during tropical forest succession. Proc. Biol. Sci. 277, 97–104. doi:10.1098/rspb.2009.0865 Letcher, S.G., Lasky, J.R., Chazdon, R.L., Norden, N., Wright, S.J., Meave, J.A., Pérez-García, E.A., Muñoz, R., Romero-Pérez, E., Andrade, A., Andrade, J.L., Balvanera, P., Becknell, J.M., Bentos, T. V., Bhaskar, R., Bongers, F., Boukili, V., Brancalion, P.H.S., César, R.G., Clark, D.A., Clark, D.B., Craven, D., DeFrancesco, A., Dupuy, J.M., Finegan, B., González-Jiménez, E., Hall, J.S., Harms, K.E., Hernández-Stefanoni, J.L., Hietz, P., Kennard, D., Killeen, T.J., Laurance, S.G., Lebrija-Trejos, E.E., Lohbeck, M., MartínezRamos, M., Massoca, P.E.S., Mesquita, R.C.G., Mora, F., Muscarella, R., Paz, H., PinedaGarcía, F., Powers, J.S., Quesada-Monge, R., Rodrigues, R.R., Sandor, M.E., SanaphreVillanueva, L., Schüller, E., Swenson, N.G., Tauro, A., Uriarte, M., van Breugel, M., Vargas-Ramírez, O., Viani, R.A.G., Wendt, A.L., Bruce Williamson, G., 2015. Environmental gradients and the evolution of successional habitat specialization: a test case with 14 Neotropical forest sites. J. Ecol. 103, 1276–1290. doi:10.1111/1365-2745.12435 Logan, C., 2007. Effect of forest age on soil nutrient stocks of a lowland tropical rain forest. Honor's Thesis, University of Virginia, Charlottesville, VA, USA. Lohbeck, M., Poorter, L., Lebrija-Trejos, E., Martínez-Ramos, M., Meave, J.A., Paz, H., PérezGarcía, Eduardo, A., Romero-Pérez, I.E., Tauro, A., Bongers, F., 2013. Successional changes in functional composition contrast for dry and wet tropical forest. Ecology 94, 1211–1216. Marín-Spiotta, E., Ostertag, R., Silver, W.L., 2007. Long-term patterns in tropical reforestation: plant community composition and aboveground biomass accumulation. Ecol. Appl. 17, 828–39. Martin, P.A., Newton, A.C., Bullock, J.M., 2013. Carbon pools recover more quickly than plant biodiversity in tropical secondary forests. Proc. R. Soc. B 280, 20132236. McDade, L., Bawa, K.S., Hespenheide, H., Hartshorn, G.S. (Eds.), 1994. La Selva: ecology and natural history of a Neotropical rain forest. University of Chicago Press, Chicago, Illinois, USA. McDade, L., Hartshorn, G.S., 1994. La Selva Biological Station, in: McDade, L., Bawa, K.S., Hespenheide, H., Hartshorn, G.S. (Eds.), La Selva: Ecology and Natural History of a Neotropical Rain Forest. University of Chicago Press, Chicago, Illinois, USA., pp. 6–14. Menge, D.N.L., Chazdon, R.L., 2015. Higher survival drives the success of nitrogen-fixing trees through succession in Costa Rican rainforests. New Phytol. doi:10.1111/nph.13734 Mesquita, R.C.G., Eduardo, P., Santos, D.O.S., Jakovac, C.C., Bentos, T.V., Williamson, G.B., 2015. Amazon rain forest succession: stochasticity or land-use legacy? Bioscience 65, 849– 861. doi:10.1093/biosci/biv108 Muscarella, R., Uriarte, M., Aide, T.M., Erickson, D.L., Forero-Montaña, J., Kress, W.J., Swenson, N.G., Zimmerman, J.K., 2015. Functional convergence and phylogenetic divergence during secondary succession of subtropical wet forests in Puerto Rico. J. Veg. Sci. 27, 283–294. doi:10.1111/jvs.12354 Nakagawa, S., Schielzeth, H., 2013. A general and simple method for obtaining R2 from

29

generalized linear mixed-effects models. doi:10.1111/j.2041-210x.2012.00261.x

Methods

Ecol.

Evol.

4,

133–142.

Nicotra, A.B., Chazdon, R.L., Iriarte, S.V.B., 1999. Spatial heterogeneity of light and woody seedling regeneration in tropical wet forests. Ecology 80, 1908–1926. Norden, N., Angarita, H.A., Bongers, F., Martínez-Ramos, M., Granzow-de la Cerda, I., van Breugel, M., Lebrija-Trejos, E., Meave, J.A., Vandermeer, J., Williamson, G.B., Finegan, B., Mesquita, R., Chazdon, R.L., 2015. Successional dynamics in Neotropical forests are as uncertain as they are predictable. Proc. Natl. Acad. Sci. U. S. A. 112, 8013–8. doi:10.1073/pnas.1500403112 Norden, N., Chazdon, R.L., Chao, A., Jiang, Y.-H., Vílchez-Alvarado, B., 2009. Resilience of tropical rain forests: tree community reassembly in secondary forests. Ecol. Lett. 12, 385– 94. doi:10.1111/j.1461-0248.2009.01292.x Norden, N., Mesquita, R.C.G., Bentos, T. V., Chazdon, R.L., Williamson, G.B., 2011. Contrasting community compensatory trends in alternative successional pathways in central Amazonia. Oikos 120, 143–151. doi:10.1111/j.1600-0706.2010.18335.x Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., Team, R.C., 2015. nlme: linear and nonlinear mixed effects models. R package version 3.1–120. Plourde, B.T., Boukili, V.K., Chazdon, R.L., 2015. Radial changes in wood specific gravity of tropical trees: inter- and intra-specific variation during secondary succession. Funct. Ecol. 29, 111–120. doi:10.1111/1365-2435.12305 Poorter, L., 2007. Are species adapted to their regeneration niche, adult niche, or both? Am. Nat. 169, 433–42. doi:10.1086/512045 Poorter, L., Bongers, F., 2006. Leaf traits are good predictors of plant performance across 53 rain forest species. Ecology 87, 1733–43. Prach, K., Walker, L.R., 2011. Four opportunities for studies of ecological succession. Trends Ecol. Evol. 26, 119–123. doi:10.1016/j.tree.2010.12.007 Queenborough, S.A., Metz, M.R., Wiegand, T., Valencia, R., 2012. Palms, peccaries and perturbations: widespread effects of small-scale disturbance in tropical forests. BMC Ecol. 12, 3. doi:10.1186/1472-6785-12-3 Raaimakers, D., Boot, R.G.A., Dijkstra, P., Pot, S., Pons, T., 1995. Photosynthetic rates in relation to leaf phosphorus content in pioneer versus climax tropical rainforest trees. Oecologia 102, 120–125. R Core Team, 2016. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Rees, M., Condit, R., Crawley, M., Pacala, S., Tilman, D., 2001. Long-term studies of vegetation dynamics. Science 293, 650–655. doi:10.1126/science.1062586 Reich, P.B., Wright, I.J., Cavender-Bares, J., Craine, J.M., Oleksyn, J., Westoby, M., Walters, M.B., 2003. The evolution of plant functional variation: traits, spectra, and strategies. Int. J. Plant Sci. 164, S143–S164. Rozendaal, D.M.A., Chazdon, R.L., 2015. Demographic drivers of tree biomass change during secondary succession in northeastern Costa Rica. Ecolgical Appl. 25, 506–516.

30

doi:10.1890/14-0054.1 Salm, R., Jalles-Filho, E., Schuck-Paim, C., 2005. A model for the importance of large arborescent palms in the dynamics of seasonally-dry amazonian forests. Biota Neotrop. 5, 1–6. doi:10.1590/S1676-06032005000300011 Selaya, N.G., Anten, N.P.R., 2010. Leaves of pioneer and later-successional trees have similar lifetime carbon gain in tropical secondary forest. Ecology 91, 1102–13. Sezen, U.U., Chazdon, R.L., Holsinger, K.E., 2009. Proximity is not a proxy for parentage in an animal-dispersed Neotropical canopy palm. Proc. R. Soc. B 276, 2037–2044. doi:10.1098/rspb.2008.1793 Sezen, U.U., Chazdon, R.L., Holsinger, K.E., 2007. Multigenerational genetic analysis of tropical secondary regeneration in a canopy palm. Ecology 88, 3065–3075. Svenning, J., 1998. The effect of land-use on the local distribution of palm species in an Andean rain forest fragment in northwestern Ecuador. Biodivers. Conserv. 7, 1529–1537. Swaine, M., Whitmore, T., 1988. On the definition of ecological species groups in tropical rain forests. Vegetatio 75, 81–86. ter Steege, H., Pitman, N.C.A., Sabatier, D., Baraloto, C., Salomao, R.P., Guevara, J.E., Phillips, O.L., Castilho, C. V., Magnusson, W.E., Molino, J.-F., Monteagudo, A., Nunez Vargas, P., Montero, J.C., Feldpausch, T.R., Coronado, E.N.H., Killeen, T.J., Mostacedo, B., Vasquez, R., Assis, R.L., Terborgh, J., Wittmann, F., Andrade, A., Laurance, W.F., Laurance, S.G.W., Marimon, B.S., Marimon, B.-H., Guimaraes Vieira, I.C., Amaral, I.L., Brienen, R., Castellanos, H., Cardenas Lopez, D., Duivenvoorden, J.F., Mogollon, H.F., Matos, F.D.D.A., Davila, N., Garcia-Villacorta, R., Stevenson Diaz, P.R., Costa, F., Emilio, T., Levis, C., Schietti, J., Souza, P., Alonso, A., Dallmeier, F., Montoya, A.J.D., Fernandez Piedade, M.T., Araujo-Murakami, A., Arroyo, L., Gribel, R., Fine, P.V.A., Peres, C.A., Toledo, M., Aymard C., G.A., Baker, T.R., Ceron, C., Engel, J., Henkel, T.W., Maas, P., Petronelli, P., Stropp, J., Zartman, C.E., Daly, D., Neill, D., Silveira, M., Paredes, M.R., Chave, J., Lima Filho, D.D.A., Jorgensen, P.M., Fuentes, A., Schongart, J., Cornejo Valverde, F., Di Fiore, A., Jimenez, E.M., Penuela Mora, M.C., Phillips, J.F., Rivas, G., van Andel, T.R., von Hildebrand, P., Hoffman, B., Zent, E.L., Malhi, Y., Prieto, A., Rudas, A., Ruschell, A.R., Silva, N., Vos, V., Zent, S., Oliveira, A.A., Schutz, A.C., Gonzales, T., Trindade Nascimento, M., Ramirez-Angulo, H., Sierra, R., Tirado, M., Umana Medina, M.N., van der Heijden, G., Vela, C.I.A., Vilanova Torre, E., Vriesendorp, C., Wang, O., Young, K.R., Baider, C., Balslev, H., Ferreira, C., Mesones, I., Torres-Lezama, A., Urrego Giraldo, L.E., Zagt, R., Alexiades, M.N., Hernandez, L., Huamantupa-Chuquimaco, I., Milliken, W., Palacios Cuenca, W., Pauletto, D., Valderrama Sandoval, E., Valenzuela Gamarra, L., Dexter, K.G., Feeley, K., Lopez-Gonzalez, G., Silman, M.R., 2013. Hyperdominance in the Amazonian tree flora. Science 342, 1243092. doi:10.1126/science.1243092 van Gelder, H.A., Poorter, L., Sterck, F.J., 2006. Wood mechanics, allometry, and life-history variation in a tropical rain forest tree community. New Phytol. 171, 367–78. doi:10.1111/j.1469-8137.2006.01757.x Vile, D., Shipley, B., Garnier, E., 2006. A structural equation model to integrate changes in functional strategies during old-field succession. Ecology 87, 504–517. doi:10.1890/05-

31

0822 Weiher, E., Keddy, P.A., 1999. Assembly rules as general constraints on community composition, in: Weiher, E., Keddy, P.A. (Eds.), Ecological Assembly Rules; Perspectives, Advances, Retreats. Cambridge University Press, Cambridge, pp. 257–271. Westoby, M., Falster, D.S., Moles, A.T., Vesk, P.A., Wright, I.J., 2002. Plant ecological strategies: some leading dimensions of variation between species. Annu. Rev. Ecol. Syst. 33, 125–159. doi:10.1146/annurev.ecolsys.33.010802.150452 Whitfeld, T.J.S., Lasky, J.R., Damas, K., Sosanika, G., Molem, K., Montgomery, R.A., 2014. Species Richness, Forest Structure, and Functional Diversity During Succession in the New Guinea Lowlands 46, 538–548. Williamson, G.B., Wiemann, M.C., 2010. Measuring wood specific gravity...Correctly. Am. J. Bot. 97, 519–24. doi:10.3732/ajb.0900243 Wood, T.E., Lawrence, D., Clark, D. a, Chazdon, R.L., 2009. Rain forest nutrient cycling and productivity in response to large-scale litter manipulation. Ecology 90, 109–21. Wright, I.J., Reich, P.B., Westoby, M., Ackerly, D.D., Baruch, Z., Bongers, F., Cavender-Bares, J., Chapin, T., Cornelissen, J.H.C., Diemer, M., Flexas, J., Garnier, E., Groom, P.K., Gulias, J., Hikosaka, K., Lamont, B.B., Lee, T., Lee, W., Lusk, C., Midgley, J.J., Navas, M.-L., Niinemets, U., Oleksyn, J., Osada, N., Poorter, H., Poot, P., Prior, L., Pyankov, V.I., Roumet, C., Thomas, S.C., Tjoelker, M.G., Veneklaas, E.J., Villar, R., 2004. The worldwide leaf economics spectrum. Nature 428, 821–7. doi:10.1038/nature02403 Wright, S.J., Kitajima, K., Kraft, N.J.B., Reich, P.B., Wright, I.J., Bunker, D.E., Condit, R., Dalling, J.W., Davies, S.J., Díaz, S., Engelbrecht, B.M.J., Harms, K.E., Hubbell, S.P., Marks, C.O., Ruiz-Jaen, M.C., Salvador, C.M., Zanne, A.E., 2010. Functional traits and the growth-mortality trade-off in tropical trees. Ecology 91, 3664–74. Zuur, A.F., Ieno, E.N., Walker, N.J., Saveliev, A.A., Smith, G.M., 2009. Mixed effects models and extensions in ecology with R. Springer, New York, USA.

32

Figure Legends

Fig. 1. Functional trait values of successional classification categories for canopy trees and palms ≥ 5 cm DBH. Categories include second-growth specialists (SG), generalists (Gen), and old-growth specialists (OG). Significant differences based on ANOVA analyses and Tukey HSD post-hoc tests are shown with different letters. Boxplots show median (thick horizontal line), first and third quartiles (lower and upper fences, respectively), and minimum and maximum values (lower and upper whiskers, respectively). Trait abbreviations: LS = leaf size, SLA = specific leaf area, LNC = leaf nitrogen content, LPC = leaf phosphorus content, LDMC = leaf dry matter content, LT = leaf thickness, LTO = leaf toughness, wWSG = area-weighted wood specific gravity. Although untransformed values are shown, LS, SLA, LTO, and LPC were natural logtransformed prior to analyses.

Fig. 2. Changes in community weighted mean trait values along the successional gradient. Closed symbols represent the plots inside of La Selva Biological Station and the open symbols represent the plots outside of La Selva. Linear mixed effects model predictions are demonstrated with solid line when plot basal area was a significant factor in the model. The trends for the plots within La Selva are shown with a dark grey line, and the trends for the plots outside of La Selva are shown with a light grey line. The marginal correlation coefficient, R2m, considers the model fit based on fixed effects only (i.e. plot basal area and stand location), while the conditional correlation coefficient, R2c, incorporates the variation explained by both the fixed and random effects. Trait abbreviations as in Fig. 1.

33

Fig. 3. Patterns of successional community weighted mean trait values for dicotyledonous canopy species only (i.e., excluding the five common palms species in our study sites). The closed symbols represent the plots inside of La Selva Biological Station and the open symbols represent the plots outside of La Selva. Linear mixed effects model predictions are demonstrated with solid line when plot basal area was a significant factor in the model. The trends for the plots within La Selva are shown with a dark grey line, and the trends for the plots outside of La Selva are shown with a light grey line. The marginal correlation coefficient, R2m, considers the model fit based on fixed effects, while the conditional correlation coefficient, R2c, incorporates the variation explained by both the fixed and random effects. Trait abbreviations as in Fig. 1.

34

400 200

0.4

600

0.3

800

0.2

-1

LDMC (mg g )

LS (cm2)

1000

0.5

1200

30

0.35

25

0.30

LT (mm)

SLA (mm2 mg-1)

0

20 15 10

a

ab

b

1.2

LTO (N mm )

LNC (mg g-1)

0.20 0.15

5

40

0.25

-1

35 30 25 20

1.0 0.8 0.6 0.4 0.2

15

0.0

a

b

wWSG (unitless)

LPC (mg g-1)

a 2.0 1.5 1.0 0.5

a

a

b

SG

Gen

OG

0.8 0.6 0.4 0.2

SG

Gen

OG

Fig. 1

35

JE FEB LSUR TIR LEPS CR LEPP SV

200

0.42

2

Rm=0.54

LDMC (g g )

R2c =0.99

●

●

● ●

●

●

-1

300

2

LS (cm )

●

● ● ●●

●●

●

●● ●●

100

0.40

2

Rm=0.24 R2c >0.99 ● ●

●

●

●

●

●

● ●

● ●●●

● ● ●● ●

0.38 0.36

15

20

25

30

35

15

20

25

30

●

18

●

●

●

●

2

Rm=0.33

R2m=0.09 ● ●

●

●● ●

●

16

R2c >0.99

●● ● ● ●

14

LT (mm)

2

-1

SLA (mm mg )

0.30 ●

R2c >0.99

0.25

0.20 ●

12

●

●

●

● ●

-1

LNC (mg g )

●

28

25

● ●

●

●● ●

30

35

●

26

●● ●● ●

24 2

Rm=0.16

22

2

Rc >0.99

15

20

25

30

0.5

●

●

●

● ●

●●

● ●●●

●

●●●●

1.2 1.1 1.0 0.9

2

Rm=0.41 2

Rc >0.99

15

● ●

●●

20

●● ● ●●

● ●● ●

25

30

35

30

35

30

35

R2m=0.11 ● ● ● ● ● ●

●

●●

●

0.3 ● ●

0.2

●

15

wWSG (unitless)

-1

LPC (mg g )

●

●

R2c >0.99

1.5

1.3

●

0.4

35

1.4

●

15

-1

20

LTO (N mm )

15

35

●

●

● ●

●

20

25

R2m=0.17

0.50

2

Rc >0.99

0.45 ●

●

●

●

● ●

●●

● ●● ●●

●●●●

0.40 20

25

30

Basal Area (m2 ha-1)

35

15

20

25

Basal Area (m2 ha-1)

Fig. 2

36

JE FEB LSUR TIR LEPS CR LEPP SV

200

0.42 -1

R2m=0.16 R2c =0.98

●

100

●

15

●

● ●

● ●

●

●

● ●●● ●

25

●

● ●

30

● ●●● ●

18

R2c >0.99

20

●

●

25

●

● ●

●●

● ●●● ●

30

25

30

●●●●

2

R2c >0.99

0.20

0.5

-1

LTO (N mm )

28 26 24 2

Rm=0.28 2

15

20

25

30

●

●

●

● ●

●●

● ●●● ●

●●●●

1.2 2

Rm=0.43 2

Rc =0.999

15

●

● ●

●●

20

● ●●●●

●●●●

25

30

35

30

35

30

35

R2m=0.33

0.3 0.2

●

●

15

wWSG (unitless)

●

●

R2c =0.99

1.5

1.3

●

0.4

35

1.4

35

Rm=0.41

15

Rc >0.99

-1

●●●●

0.25

35

-1

LNC (mg g )

●

LPC (mg g )

● ●●●●

0.30

●

15

0.9

●●

20

12

1.0

● ●

2

Rc >0.99

14

1.1

●

●

0.38

15

Rm=0.23

●● ● ● ●

16

22

●

●

35

2

●●

0.40

2

Rm=0.10

0.36

●●●●

LT (mm)

2

-1

SLA (mm mg )

●

20

●

LDMC (g g )

300

2

LS (cm )

●

●

●

● ●

●●

20

● ●●● ●

●●●●

25

R2m=0.01

0.50

2

Rc >0.99

0.45 ●

●

●

●

● ●

●●

●● ● ●●

●●●●

0.40 20

25

30

Basal Area (m2 ha-1)

35

15

20

25

Basal Area (m2 ha-1)

Fig. 3.

37

Table 1. Study sites, located in northeastern Costa Rica. Each stand is 1-ha. Vegetation dynamics have been monitored annually for all stems ≥ 5 cm DBH since the initial census year. Percent basal area of remnant trees is calculated from total basal area of stems ≥ 5 cm DBH in the initial census year. Adapted from Table 1 of Chazdon et al. (2010). Plot (abbreviation)

Finca el Bejuco (FEB)

Year

Year

Forest ages % Basal area Location

Latitude,

abandoned

census

during

of remnants

longitude

initiated

censuses

(# individuals)

2005

10–19

26.55 (22)

1995

Chilamate

10.46°N,

Surrounding landscape

Pasture, old-growth, and

-84.06°W second-growth forest Juan Enriquez (JE)

1995

2005

10–19

0.29 (1)

Chilamate

10.46°N,

Pasture, old-growth, and

-84.07°W second-growth forest Lindero Sur (LSUR)

1985

1997

12–29

16.93 (10)

La Selva

10.41°N,

Old-growth and second-

-84.03°W growth forest Tirimbina (TIR)

1982

1997

15–32

11.41 (6)

Tirimbina

10.40°N,

Pasture, plantations, and

-84.11°W second-growth forest Lindero El Peje Secondary (LEPS)

1977

1997

20–37

3.20 (3)

La Selva

10.43°N,

Old-growth and second-

-84.03°W growth forest Cuatro Rios (CR)

1972

1997

25–42

2.17 (2)

Tirimbina

10.39°N,

Pasture, second-growth,

-84.13°W and old-growth forest

38

Lindero El Peje Primary (LEPP)

Old-growth

2005

Old-growth

NA

La Selva

10.42°N,

Old-growth forest

-84.04°W Selva Verde (SV)

Old-growth

2005

Old-growth

NA

Chilamate

10.44°N,

Pasture, old-growth, and

-84.07°W second-growth forest

39

Table 2. Comparisons of functional trait values for 5 canopy palm species and 89 canopy tree species. Results from Welch’s t-tests, with Holm-Bonferroni adjusted P-values for the 8 measured functional traits. Traits with significant differences between palms and trees are shown in bold. Functional trait

Palm mean ± SE

Tree mean ± SE

df

t

P-value

log [Leaf size]

6.03 ± 0.32

4.25 ± 0.10

4.87

5.28

0.032

Specific leaf area

8.58 ± 0.48

14.23 ± 0.49

16.13

-8.22

<0.001

Leaf nitrogen content

19.70 ± 2.51

23.80 ± 0.63

4.51

-1.58

0.722

Leaf phosphorus content

1.23 ± 0.16

1.10 ± 0.04

4.56

0.81

0.763

Leaf dry matter content

0.44 ± 0.02

0.37 ± 0.01

5.12

3.81

0.091

Leaf thickness

0.25 ± 0.02

0.21 ± 0.01

4.71

2.07

0.483

Leaf toughness

0.98 ± 0.09

0.32 ± 0.01

4.22

7.45

0.014

Area-weighted wood specific gravity

0.35 ± 0.04

0.52 ± 0.01

4.90

-3.94

0.091

40