The evolution of bark mechanics and storage across habitats in a clade of tropical trees^†

Bark is a very morphologically diverse and functionally important part of stems (Romero, 2013). Ranging from smooth to rough, scaly, spiny, furrowed, and tesselated, bark is known to play roles in protection from desiccation, pathogens, and fire (Srivastava, 1964; Uhl and Kauffman, 1990; Pinard and Huffman, 1997; Romero and Bolker, 2008; Lawes et al., 2011b). Bark is also involved in photosynthate transport (Jensen et al., 2012), photosynthesis (Pfanz et al., 2002; Cernusak and Hutley, 2011), storage (Scholz et al., 2007), and stem mechanical support (Niklas, 1999), and it is even likely to intervene in repair of xylem embolism (Nardini et al., 2011).

Various comparative approaches can be applied to study the origin and functional significance of bark diversity across species. One approach involves sampling a wide array of plant lineages, within or across communities (Westoby et al., 2002). This community-based approach has allowed the discovery of fundamental trade-offs and functional coordination within the bark and with other parts of the plant (Poorter et al., 2013; Rosell et al., 2014). The main advantage of this approach is that it provides generality, with patterns potentially stretching across the woody plants at large. Another approach is a clade-centered one, which focuses on the descendants of a common ancestor. Because all of the species are closely related and share ancestry, any divergence in form or function between them has accrued since divergence from the same ancestral state (Martins, 2000). This approach therefore helps to identify features that potentially represent adaptive changes (Romero et al., 2009; Losos, 2011). It also allows for detailed comparisons of ontogeny to identify the developmental mechanisms underlying morphological and functional divergence between species (Raff, 2000). Here, we use a clade-centered approach to study the evolution of diversity in bark function across a clade of tropical trees with a high diversity in habitat, habit, and bark features. In particular, we focus on the roles of bark in the mechanical support of the stem, and in stem water storage, two potentially vital functions often neglected in favor of study of other bark functions such as fire protection.

We chose as our study system nine species within the simaruba clade of the neotropical genus Bursera (De-Nova et al., 2012). This clade has as its namesake the gumbo-limbo tree Bursera simaruba. With its large stature (to over 30 m tall) and its brilliant red bark, the gumbo-limbo is one of the most conspicuous trees in the lowland (sub)tropics of Florida, the West Indies, Mexico, and Central America. Other members of the clade range from small, water-storing trees of very dry tropical deciduous forests (Figs. 1A, B), to tall species of moister habitats (Rosell et al., 2010). The clade also includes two species of unusual habits, B. instabilis, a tree with long, lianescent branches in Mexican tropical dry forests (Fig. 1C), and B. standleyana, a primary hemiepiphyte (Zotz, 2013) growing high in rainforest canopies in Costa Rica. The simaruba clade is known for its “peeling bark” (Fig. 1E–G). However, only the outermost layer of bark, the phellem (Fig. 1D), peels. Produced by the cork cambium (Gómez-Vazquez and Engleman, 1984), the phellem comprises just 1–2% of the total bark thickness in main trunks. Although bark in the clade is always very thick, made up of mostly secondary phloem and phloem rays (Fig. 1D), there is wide variation in thickness across species (Fig. 1H, I). The extensive environmental and morphological range in the simaruba clade, all derived from a single common ancestor, allowed us to examine both the way that the environment is associated with bark functional variation and the developmental differences leading to functional diversity in the bark of these species. Moreover, the neotropical dry and wet forests that the simaruba clade inhabits do not experience natural fire as a significant selective factor. In these sites, fire is absent or returns at such wide intervals that it has never been recorded (Janzen, 1988; Murphy and Lugo, 1995; Maas et al., 2002; Pennington et al., 2009; Cantarello et al., 2011; Sánchez-Azofeifa and Portillo-Quintero, 2011). No species in these forests exhibits fire-adaptive traits (Pennington et al., 2009) such as serotiny, and there is no evidence of burning in the vegetation (e.g., burned stems, resprouting stumps; Maas et al., 2002; Rodríguez Trejo, 2008). Furthermore, the genus Bursera has been shown to have radiated within seasonally dry tropical forests, the same fire-free habitats they occupy today (De-Nova et al., 2012), ruling out the possibility that bark traits in the simaruba clade have been shaped evolutionarily by fire. As a result, we can in effect factor out fire as an important selective pressure on the bark of the simaruba clade, allowing us to focus on the role of mechanics and storage functions in the wide variation in bark traits of the clade.

We started by examining patterns in bark mechanics and water storage across habitats. Most explanations for variation in bark traits in the current ecological literature appeal to adaptation to environmental conditions. For example, variation in bark thickness is usually explained by differences in fire regime (Gill and Ashton, 1968; Catry et al., 2010; Lawes et al., 2011a). Although bark can be thick in large trees of areas that burn frequently, such as savannas, recent cross-community comparative studies have revealed that bark traits generally vary more within than across communities (Rosell et al., 2014). This diversity reflects the mixture of ecological strategies of the different lineages that make up any given community (Falster and Westoby, 2003). In contrast, members of a single lineage are likely to respond to selection through a similar set of strategies. In our case, all species of the simaruba clade have bark of generally similar structure, but very different proportions across species. For this reason, they represent an ideal system to examine the degree to which variation in bark amount and other bark functional traits might reflect adaptation to environmental conditions.

Here, we tested whether environmental variables predict bark traits associated with storage and mechanics, supporting the idea that environmental selective pressures do mold bark diversity across habitats within the simaruba clade. Stem water storage in dry forest species seems crucial to buffer daily and seasonal changes in water availability (Scholz et al., 2011) and to fuel leaf and fruit production (Borchert, 1994; Chapotin et al., 2006). Water stored in bark has been found to contribute in varying degrees to stem capacitance (Zweifel et al., 2000; Zweifel and Häsler, 2001; Scholz et al., 2007; Díaz-Espejo et al., 2013). Given the large volume of water stored in the bark of the simaruba clade species, we predicted bark amount to be associated negatively with precipitation and positively with seasonality. Higher water storage has been shown to trade off with mechanical rigidity in wood (Pratt et al., 2007; Chave et al., 2009). We predicted an analogous trade-off in bark. As a result, we expected a positive association between the rigidity of bark and wood with environmental variables reflecting greater water availability, and a negative one with seasonality variables. We tested these hypotheses with raw data and also taking into account phylogenetic relationships.

Bark plays a vital role in mechanical support and water storage (Romero, 2013), but it is not clear whether differences in performance reflect differences in tissue quantity, tissue performance, or both. As a tree grows, the resistance to bending of the wood in its trunk increases as a result of both its increase in girth (change in tissue quantity, Niklas, 1992), and the production of stiffer wood (change in tissue quality, Van Gelder et al., 2006). In contrast, previous studies have suggested that despite wide variation in bark tissue properties across species, an increase in mechanical performance in bark is achieved mainly by higher amounts of bark, and not so much by changes in tissue properties (Rosell et al., 2014). Here, we investigated the quality–quantity issue in bark storage and mechanics across the species of the simaruba clade. We also took an ontogenetic perspective on the issue, testing whether the mechanical and storage contribution of bark changed along branches via bark quantity, or whether tissue properties, as measured by bark Young's modulus and bark water content, also affected these contributions.

Bark must perform its mechanical support role not as a disarticulated tissue, but as a functionally and structurally integrated part of the stem, together with wood (Niklas, 1992). It has been shown that bark and wood covary strongly regarding density and water content (Poorter et al., 2013; Rosell et al., 2014), but it is unclear what this covariation implies for the mechanics of the branch. We compared the contribution of bark and wood to the stiffness of branches across species and habitats and also along branches within species. The simaruba clade is a particularly useful system for addressing the mechanical role of bark. Its wide range in environment and size, in addition to its growth habit, including lianescent and hemiephytic, allow us to examine how developmental differences in bark and wood traits can lead to morphological and functional diversity (Speck and Rowe, 2001; Rowe et al., 2006; Rosell et al., 2012). Together with the strong bark–environment associations across species, these developmental data illustrate how our clade-based approach provides an essential complement to community-based studies.

MATERIALS AND METHODS

RESULTS

We measured traits in 447 segments from 94 branches in nine species, making up the “ontogenetic” data set used for the analyses of trait change along branches. Branches ranged from 112 to 383 cm long in the self-supporting species and up to 603 cm in B. instabilis, the lianescent species. This variation in length resulted from the markedly different allometries across species (Rosell et al., 2012). In the species with shorter, thicker branches, older, consequently thicker, segments could not be tested due to limitations of our testing apparatus. Segments between 100 and 140 cm from the tip were grouped in the “mature segments” data set, which had 68 points and for which the descriptive statistics are reported in Table 1.

Correlations between bark and wood traits and the environment

Environmental variables indicated a wide ecological range in the habitats of the simaruba clade. For example, annual precipitation ranged from 680 in very dry, tropical forests to 3600 mm in the rainforest. Temperature seasonality also varied markedly, with annual temperature ranges covering from 25°C in a tropical, dry forest with markedly different cooler wet and hot, dry seasons, to 14°C in a relatively uniform rainforest (Appendix S1).

Bark and wood traits were closely associated with environmental variables, as suggested by correlations based on the “species means” data set. As predicted, traits reflecting the amount of bark, such as bark thickness and bark % area, were tightly and positively associated with temperature seasonality variables, and negatively with precipitation ones (Table 2; Appendix S2, see online Supplemental Data), suggesting greater allocation to water storage in dry areas. Bark thickness increased with the annual (r = 0.91, r_PIC = 0.89, P < 0.001, Fig. 2A) and diurnal ranges in temperature (r = 0.79, r_PIC = 0.75, P < 0.05), as well as bark % area (r = 0.95 and 0.89, r_PIC = 0.92 and 0.85 P < 0.01). Bark % area decreased with precipitation variables such as the precipitation of the wettest month (r = −0.76, r_PIC = −0.74, P < 0.05, Fig. 2B) or annual precipitation (r = −0.71, P < 0.05; r_PIC = −0.55, P > 0.05).

Table 2. Pearson correlations between bark and wood traits and environmental variables based on mean traits for nine species of Bursera (see Table 1).

Variable	BT	%BA	Ebark	Estruct	Ewood	%BMC	%BWC	%WWC
Ppt coldest qtra	−0.63	−0.71*	0.63	0.94***	0.82**	−0.72*	0.10	0.55
Ppt warmest qtra	−0.56	−0.71*	0.74*	0.75*	0.68*	−0.48	0.28	0.73*
Ppt driest qtra	−0.49	−0.58	0.34	0.77*	0.64	−0.69*	−0.15	0.21
Ppt wettest qtra	−0.60	−0.73*	0.73*	0.86***	0.73*	−0.51	0.14	0.69
Ppt seasonalitya	0.51	0.58	−0.35	−0.8*	−0.67*	0.73*	0.16	−0.17
Ppt driest montha	−0.32	−0.48	0.33	0.77*	0.70*	−0.76*	0.23	0.43
Ppt wettest montha	−0.62	−0.76*	0.76*	0.86***	0.74*	−0.53	0.15	0.71*
Annual ppta	−0.60	−0.71*	0.64	0.9***	0.76*	−0.62	0.06	0.57
Mean T coldest qtr	−0.67*	−0.5	0.36	0.31	0.12	0.00	−0.82*	−0.35
Mean T warmest qtr	−0.44	−0.3	0.13	0.1	−0.05	0.13	−0.82*	−0.4
Mean T driest qtr	−0.58	−0.47	0.26	0.32	0.1	−0.05	−0.76*	−0.34
Mean T wettest qtr	−0.5	−0.43	0.28	0.15	0.02	0.07	−0.60	−0.14
Annual range in T	0.91***	0.95***	−0.83**	−0.90***	−0.75*	0.64	0.08	−0.57
Min T coldest month	−0.83*	−0.76*	0.58	0.67*	0.47	−0.35	−0.68	−0.03
Max T warmest month	0.1	0.28	−0.38	−0.35	−0.44	0.45	−0.79*	−0.77*
T seasonality	0.73*	0.57	−0.57	−0.54	−0.34	0.24	0.42	0.15
Isothermality	−0.43	−0.23	0.41	0.16	0.05	0.05	0.03	−0.21
Mean diurnal range	0.79*	0.89***	−0.72*	−0.86***	−0.75*	0.66*	0.09	−0.57
Mean annual T	−0.59	−0.45	0.29	0.23	0.05	0.05	−0.8*	−0.34

• Notes: Ppt = precipitation, qtr = quarter, T = temperature, Max = maximum, Min = minimum; other abbreviations as in Table 1; *** P < 0.001; **P < 0.01; *P < 0.05; see Appendix S2 for the equivalent correlations based on phylogenetically independent contrasts.
• a Log₁₀-transformed.

Mechanical traits also changed across environments as predicted. The Young's modulus of the whole branch, the bark, and the wood increased with precipitation of the warmest and wettest quarters of the year (0.68 ≤ r ≤ 0.86, 0.70≤ r_PIC ≤ 0.87, P < 0.05) and decreased with annual and diurnal ranges in temperature (−0.90 ≤ r ≤ −0.72, −0.85 ≤ r_PIC ≤ −0.72, P < 0.05, Fig. 2C). The mechanical contribution of bark decreased in wetter sites (with precipitation of the driest month, r = −0.76, P < 0.05; r_PIC = −0.52, P > 0.05, Fig. 2D) but increased in more seasonal sites (with mean diurnal range r = 0.66, P < 0.05; r_PIC = 0.35, P > 0.05).

Contrary to our expectations, bark tended to have higher water content in sites with lower mean temperatures (with mean temperature of the coldest quarter, r = −0.82, r_PIC = −0.79, P < 0.05), and this content was not associated with precipitation (Table 2). Also in disagreement with our predictions, wood tended to have more water in conditions of higher precipitation (with precipitation in the warmest quarter and wettest month, r ≥ 0.71, P < 0.05; r_PIC ≥ 0.66, P > 0.05), and less water in warmer weather (with maximum temperature of the warmest month, r = −0.77, P < 0.05; r_PIC = −0.64, P > 0.05).

Bark quantity vs. quality in bark contribution to mechanics and water storage

Bark traits reflecting bark quantity, such as bark thickness and % area, were closely associated with one another (r = 0.81, P < 0.001, Table 3), and also with I_bark, a trait also reflecting the amount of bark from a mechanical perspective (r_Ibark-BT = 0.88, r_{Ibark-Bark%area} = 0.44, P < 0.001). The quantity of bark was negatively associated with bark mechanical quality (r_Ebark-BT = −0.70, Fig. 3, and r_Ebark-Ibark = −0.59, P < 0.001).

Table 3. Pearson correlations between bark and wood traits based on segments between 100 and 140 cm from the tip.

Variable	BTa	%BA	%WA	Ebark	Estruct	Ewood	Ibarka	Iwooda	%BMCa	%BWC
%BA	0.81***
%WA	−0.77***	−0.99***
Ebark	−0.70***	−0.66***	0.61***
Estruct	−0.60***	−0.62***	0.56***	0.66***
Ewood	−0.30*	−0.32*	0.27*	0.36**	0.88***
Ibarka	0.88***	0.44***	−0.40***	−0.59***	−0.45**	−0.25
Iwooda	0.61***	0.04	0.01	−0.40**	−0.24	−0.14	0.91***
%BMCa	0.14	0.24	−0.25	0.26*	−0.38**	−0.51***	0.05	−0.04
%BWC	0.15	0.06	−0.10	0.18	0.02	0.03	0.15	0.11	0.14
%WWC	−0.26	−0.27	0.19	0.47*	0.39**	0.30*	−0.12	0.01	0.03	0.24

• Notes: Abbreviations as in Table 1; *** P < 0.001; **P < 0.01; *P < 0.05; N = 47–68 segments depending on trait, due to missing data in some variables.
• a Log₁₀-transformed.

Across species, bark achieved a higher contribution to branch stiffness more through changes in bark quality than bark quantity. Across species, stiffness (quality) was more closely correlated with higher Young's modulus (r_Ebark-%BMC = 0.26, P > 0.05) than with the amount (thickness) of bark or its moment of area (r_%BMC-BT = 0.14, r_%BMC-Ibark = 0.24, P > 0.05, Table 3). In contrast to this across-species pattern, along the ontogeny of a branch, bark's contribution to stiffness depended more on bark quantity and not quality. The amount of bark (quantity) changed along branches, with bark thickness increasing and bark % area decreasing toward the older portions of the branch. Based on data for all species, bark thickness and distance from the tip were positively associated (r = 0.47, P < 0.001), whereas bark % area decreased with distance from the tip (r = −0.60, P < 0.001). These trends were stronger within species, with correlations ranging from 0.61 to 0.85 for bark thickness and −0.57 to −0.83 for bark % area (Table 4). The only exceptions were B. instabilis, the lianescent species, and B. longipes, a small species of very dry tropical forest, which showed no directional change in bark % area along branches. In contrast with bark quantity (thickness), bark Young's modulus (quality) did not show an ontogenetic increase with distance from the tip, with the exception of three of the stiffer-branched species (Table 4). In the case of water content (the quality trait reflecting water holding capacities of bark tissue), species had divergent behaviors. Bark water content did not covary with distance in most species; in three species, the water content decreased, whereas in another species this content increased (Table 4). In a majority of species, in both mechanics and water storage, ontogenetically the quantity of bark produced higher contributions than variation in bark quality.

Table 4. Pearson correlations between distance to the tip and bark and wood traits (log₁₀-transformed) to evaluate the ontogenetic change in these traits, and the distance from the tip of the branch where the percentage areas and mechanical contributions of bark and wood were the same (area and mechanical inversion points).

Taxon	N	BT	%BA	%WA	Ebark	Estruct	Ewood	%BMC	%BWC	%WWC	Area inversion point	Mech. inversion point
All Bursera spp.	250–363	0.47***	−0.60***	0.73***	0.24***	0.57***	0.41***	−0.46***	0.25***	−0.08	—	—
B. arborea	16–37	0.85***	−0.73***	0.79***	0.52**	0.68***	0.40*	−0.04	−0.28	−0.70***	39.8	No crossing
B. attenuata	40–50	0.74***	−0.69***	0.84***	0.53***	0.85***	0.76***	−0.62***	−0.38*	0.34*	67.7	1.6
B. cinerea	47–52	0.78***	−0.57***	0.82***	0.04	0.54***	0.34*	−0.46***	0.61***	−0.37**	77.7	0.6
B. grandifolia	9–14	0.61*	−0.72**	0.81***	−0.27	0.4	0.12	−0.61*	0.43	−0.98***	157.2	10.4
B. instabilis	27–38	0.79***	0.19	−0.06	−0.16	−0.25	−0.16	0.05	0.22	0.06	No crossing	No crossing
B. longipes	29–35	0.81***	−0.24	0.38*	0.01	0.3	0.37*	−0.35*	0.05	−0.66***	110.5	7.4
B. roseana	41–49	0.76***	−0.80***	0.83***	0.16	0.74***	0.61***	−0.58***	−0.34*	−0.1	55.8	7.4
B. simaruba	41–48	0.72***	−0.83***	0.89***	0.35*	0.76***	0.56***	−0.63***	−0.40**	−0.22	40.3	0.4
B. standleyana	37–40	0.61***	−0.69***	0.78***	−0.16	0.69***	0.55***	−0.62***	—	—	11.5	1.2

• Notes: Abbreviations as in Table 1; *** P < 0.001; **P < 0.01; *P < 0.05.

Regarding the predicted mechanics–storage trade-off in bark, we found that more water per unit of dry tissue did not seem to affect E_bark (r = 0.18, P > 0.05). In fact, the water content of bark did not show any significant association with other bark traits. The mechanics–storage trade-off was observed in wood (Table 3). Although wood water content was associated positively with E_wood across species (r = 0.30, P > 0.05), within species these traits had a negative association (correlations per species not shown).

Bark vs. wood contributions to branch stiffness

The simaruba clade had a very significant investment in bark as indicated by the large percentages of cross-sectional area occupied by this tissue. Percentage bark area ranged across species mean values from 30% to 52% at the base of branches 100–140 cm long (Table 1). This percentage decreased with distance (r = −0.60, P < 0.001, Table 4), indicating that although bark thickened as branches aged (r_BT-Distance = 0.47, P < 0.001), less bark was produced relative to wood. Figure 4 illustrates the two general ontogenetic trends in percentage areas of bark, wood, and pith observed in the clade. The first trend is exemplified by B. cinerea, a species of very dry sites with abundant bark representing 40–60% of transectional area along the first 3 m of branches (Fig. 4A). In this species, bark was the most abundant tissue up to 80 cm from the tip (Table 4). In contrast, thinner-barked rainforest species B. simaruba had more drastic ontogenetic changes in percentage tissue areas. Up to 40 cm from the tip, bark was the most abundant tissue, decreasing to 20% at 3 m from the tip (Fig. 4C). Although pith was abundant in the first 20 cm from the tip in all species, its relative area decreased quickly thereafter.

Regarding mechanical traits, we found that both the rigidity of bark and wood at the tissue level, as reflected by Young's modulus E and their relative contributions to branch stiffness, as reflected by the proportion that each tissue contributed to total segment flexural stiffness EI, changed with ontogeny. Young's modulus for wood was greater in older segments (based on all data, r_{Ewood-Distance} = 0.41, P < 0.001; within species, 0.34 ≤ r_{Ewood-Distance} ≤ 0.76, P < 0.05). In contrast, bark did not become stiffer with ontogeny, with the exception of three species (Table 4, Fig. 5A). In turn, the mechanical contribution of bark to branch stiffness usually decreased with ontogeny (r = −0.46, P < 0.001, Table 4), with wood quickly taking over the support of branches with stem age. The contribution of bark was larger than that of wood in just the first few centimeters of branches (up to 10 cm from the tip, Table 4), after which wood took over as the main contributor to branch stiffness. However, this does not mean that the mechanical contribution of bark was negligible later in ontogeny. For example, in B. cinerea, the contribution of bark was 40% in younger branches to 15% in older segments 3 m from the tip (Fig. 4B). In B. simaruba, the contribution of bark started at 30%, decreasing to around 10% in older segments (Fig. 4D). In the lianescent B. instabilis, percentage bark area and mechanical contribution varied widely (Fig. 4E, F). Mechanical contribution in this species ranged from 10% to 60%, with a mean of 27% and no steady decrease with ontogeny.

As was the case with mechanical traits, the content of water in bark tended to increase with distance from the tip when all species were pooled (r = 0.25, P < 0.05), but contrasting trends were observed across species. Water content did not change in four species (Fig. 5B), decreased in three species, and tended to increase in one. Similarly divergent trends were also observed in the water content of wood (Table 4).

Bark traits covaried with some wood traits (Table 3). E_bark was positively associated with E_wood, meaning that species with more flexible wood also had more flexible bark (r = 0.36, P < 0.01). Water content of bark and wood tended to be correlated, but not significantly (r = 0.24, P > 0.05). In contrast, bark and wood covaried strongly in their second moment of area (I, r = 0.91, P < 0.001) and in their % areas (r = −0.99, P < 0.001).

DISCUSSION

Our results highlight the complementary nature of the community- and clade-based approaches to study the significance of bark functional diversity. Community-based approaches have tended to highlight the variable nature of bark and, in general, seem to indicate that environment only weakly predicts important bark functional variables (Paine et al., 2010; Rosell et al., 2014). These perspectives highlight the diversity of bark functional strategies that can coexist in any single locality (Gignoux et al., 1997; Dantas and Pausas, 2013). The clade-based approach taken here, in contrast, reveals bark variation that is predicted very well by environmental variables, with stiffer bark appearing to be favored in the tall species of moister sites, and bark water storage being favored in drier habitats (Table 2, Fig. 2). We first discuss these coarse-scale, across-environment associations before turning to our developmental analyses.

Across the simaruba clade, the associations between bark traits and climate seemed congruent with the importance of bark as a site of water storage (Srivastava, 1964; Borchert, 1994; Scholz et al., 2007; Gillison, 2013). With both raw data and PICs, the absolute amount and the percentage of bark was negatively associated with moisture availability, as reflected not only by precipitation variables but also those indicating seasonality (Table 2). For example, annual temperature oscillation is much higher in the seasonally dry, hot, tropical forests, as opposed to lowland rainforests, which have more even temperatures and rainfall throughout the year (Appendix S1). The observed bark–environment correlations suggest that, compared with those of wetter areas, the species of drier areas with more seasonal precipitation had thicker bark and also a higher investment in bark as compared with wood (Fig. 2A,B). The species of dry areas store conspicuous amounts of water and also starch both in the abundant living fibers and ray cells in the xylem, as well as in the bark (Roth, 1981; Gómez-Vazquez and Engleman, 1984). Greater amounts of water-storing bark in dry areas is congruent with the notion that selection favors greater allocation to storage in areas with long dry seasons (see Boyce and Leslie, 2012). We found complementary results with regard to branch mechanics. The species of tall, moist forest tended to have stiffer bark tissue (higher E_bark), though the total mechanical contribution of bark was lower in these species than in dry forest species with their very thick bark (Table 1, Fig. 4B, D). The role of other bark functions such as photosynthesis or protection against nonfire agents in variation in bark amount and other bark traits remains to be understood for the simaruba clade and for plants in general (Paine et al., 2010). However, the patterns recovered here seem congruent with adaptive variation in water storage and mechanical support along a wet–dry environmental gradient.

The vegetations inhabited by the simaruba clade and the evolutionary history of this group allowed us to focus on the functions of water storage and mechanical support, factoring out fire as a relevant selective factor shaping bark traits. Seasonally dry tropical forests, the vegetation in which most of the species of the clade occur, do not experience natural fire (Murphy and Lugo, 1995; Maas et al., 2002; Pennington et al., 2009; Rodríguez Trejo, 2008). Although seasonally dry, these forests lack the grassy understory characterizing seasonal fire-prone environments such as savannas (Mooney et al., 1995). Furthermore, tropical dry forests completely lack species with fire-adaptive traits such as serotiny or requiring fire for seed germination. Moreover, they usually include fire-sensitive succulents such as columnar cacti or stem-photosynthetic bottle trees (Pennington et al., 2009), whose long-lived epidermis always lack the conspicuous scars they would bear if burned. As a result of this lack of fire adaptation, anthropogenic fire is considered a serious threat to tropical dry forests, given the large-scale damage it can produce (Cantarello et al., 2011; Sánchez-Azofeifa and Portillo-Quintero, 2011). The evolutionary history of Bursera unequivocally suggests that radiation of its species occurred within seasonally dry tropical forests, very similar to the ones the species of the clade inhabit today (De-Nova et al., 2012). The species in the genus have not been able to invade fire-prone habitats, but some of them have invaded wetter environments (e.g., B. simaruba and B. standleyana) and deserts (B. microphylla, B. hindsiana), areas that also do not experience natural fire on evolutionarily relevant intervals. The studied clade offers an excellent system to begin unraveling the associations between bark traits and environmental conditions distinct from fire. Our clade-based approach also allows for detection of ontogenetic differences of functional significance between species.

Ontogenetic studies confirmed that bark amount is a crucial aspect influencing bark performance across species, including not only mechanics but also storage. In six species, E_bark remained constant along the length of branches, even as E_wood and I_bark increased (Fig. 5A, Table 4). The trend for constant E_bark indicates that, although the absolute contribution of bark to branch mechanical support increased with ontogeny, it did so mainly via quantity—the addition of more bark—and not quality (increase in E_bark). In the other three species, E_bark did increase along the branch length (Table 4). These were all tall wet forest species, which leads us to our cross-species comparisons. Across species, the mechanical contribution of bark was slightly more associated with bark quality than quantity (Table 3). Stiffer barks appear to be associated with moister habitats, where the trees are larger and mechanical support needs presumably greater. With regard to bark water storage, bark water content increased in some species, decreased in others, and was constant along the branch in still others. Our results depict a situation in which both tissue quality and quantity vary. There are predictable increases in bark rigidity with increasing moisture and stature across species, but within species, along the branch the increase in bark contribution to mechanical support is largely due to more bark rather than bark of increasing rigidity.

Even though the mechanical properties of bark tissue vary relatively little in ontogeny, the relative mechanical contribution of bark to the total mechanical stiffness of a branch varies drastically, because the relative proportions of bark vs. wood vary from the stem tip to the base. Bark has a much lower Young's modulus than wood does (see Table 1; Onoda et al., 2010; Paine et al., 2010), so as secondary growth ensues and wood cross-sectional area increases, the proportional contribution of the wood to branch mechanical support increases and surpasses bark, even when I_bark is still greater than I_wood. However, the vital terminal organs of trees—leaves, flowers, and fruits—are likely generally supported, at least in the terminal 1–10 cm, mainly by bark. Even though it is surpassed in importance by wood after this point, its contribution can still be very important. In some species, bark relative contribution to branch mechanical support is 40% at 3 m from the branch tip. We were not able to measure rupture resistance of bark with our field-testing device. However, bark is known to resist tension and torsion well (Romero, 2013). For this reason, it would be expected that, in addition to contributing to the bending resistance of branches, bark would have an important contribution to the resistance to rupture. Bark is almost never considered an important mechanical tissue (Niklas, 1999), but our results underscore the ecological importance of bark in forest dynamics (Romero, 2013). They also highlight the role of bark ontogenetic changes potentially associated with habit diversification.

The inclusion of a lianescent species, B. instabilis, pointed to potential structural and developmental differences between self- and non-self-supporting plants (Speck and Rowe, 2001). In contrast to the divergent behavior of B. instabilis, the branches of the hemiepiphyte B. standleyana were entirely in line with the behavior of the other species with self-supporting branches. Bursera standleyana has long lianescent roots that reach to the forest floor, but stiff, self-supporting branches (Rosell et al., 2010). In the self-supporting species of the simaruba clade, the distance from the branch tip predicted percentage bark area and relative contribution of bark to mechanical support very well (Fig. 4C, D). In contrast, along the long and flexible stems of B. instabilis, distance from the branch tip was a poor predictor of these aspects (Table 4, Fig. 4E, F). In self-supporting plants, the base of a tall stem will be subject to greater moments (the tendency to produce rotation about an axis, proportional to the load applied and the distance between the load and the axis) than a small one, and the loads on the bark will be proportionately larger (Gere and Timoshenko, 1999). As a result, selection should favor greater mechanical stiffness and resistance to breakage in both wood and bark. The greater moments at the base of longer stems explain why stem length predicts mechanical properties so well in self-supporting plants (Rosell and Olson, 2007). However, non-self-supporting plants rest on other plants. As a result, neither distance from the branch tip nor branch diameter predicts the moments to which the branches are subjected in B. instabilis. This lack of clear mechanical change with distance from the tip is coupled with marked fluctuation in the relative amount of bark and wood along the branch (Fig. 4E). To our knowledge, this is the first time that the mechanical contribution of bark has been assessed in a non-self-supporting plant. It has been shown that ontogenetic changes in whole stem E are closely associated with diversification in plant habit (Speck and Rowe, 2001; Rowe et al., 2004). Given that bark tends to be very abundant in lianas and other non-self-supporting habits, and thus has a strong effect in whole stem E, the role of bark in plant habit diversification deserves further examination (see Rosell et al., 2012 for the role of E_wood).

Functional studies of bark also highlight that the emergent mechanical behavior and the storage capacity of a stem is the result of the interaction between bark and wood and that neither of these tissues ever functions alone (Niklas, 1999). Natural selection should act on the mechanical and storage performance of stems as a whole, making understanding the coordination of bark and wood essential (Rosell and Olson, 2007). Congruent with this prediction, we found tissue quality to covary between bark and wood in terms of the Young's modulus of the tissue; the water storage capacity of the bark also showed correlation in the expected direction with wood water storage but was not significant (Table 3). It has been previously shown that traits affecting these two functions such as density and water content per unit of dry tissue are closely associated between bark and wood (Poorter et al., 2013; Rosell et al., 2014). An unexpected result of the present study was the absence of a storage–mechanics trade-off within bark as reflected by nonsignificant covariation between E_bark and bark water content (r = 0.18, P > 0.05). This trade-off has been previously observed across species in bark (Rosell et al., 2014) and wood (Pratt et al., 2007; Poorter et al., 2010), and even in the wood of the simaruba clade (Rosell et al., 2012). It is possible that the range of %BWC and E_bark across the species studied was insufficient to detect the trade-off but that it would emerge with wider sampling.

Finally, our results raise the issue of the causes of variation in bark thickness. In the simaruba clade, both branch mechanical support and water storage increased with tissue amount. As a result, selection favoring increased bark contribution to branch mechanical support, as well as to bark water storage, would lead to increased bark thickness. For example, the highest bark transectional areas and the lowest Young's modulus for wood in the simaruba clade branches were found in the dryland species. These species grow in tropical dry forests that have relatively low rainfall (as low as 679 mm annual ppt), falling in just 4 mo of the year, leaving the landscape hot and dry the rest of the time. In addition to increasing water storage, selection could increase bark amount in these species by favoring an increase in mechanical support of low stiffness-wooded branches, especially in terminal portions. In general, variation in bark thickness has mainly been explained as an adaptive response to fire, and indeed, in tropical savannas, frequent fire seems to be a major selective factor favoring thick bark (Lawes et al., 2013). That mechanics and storage are likely important drivers of bark thickness in the fire-free habitats of the simaruba clade suggests that these factors might be also important even in highly seasonal fire-prone habitats. This idea underscores the multitasking nature of bark, with functions such as pathogen and fire protection as well as photosynthesis, water and photosynthate storage, and mechanical support, all representing likely targets of natural selection. A challenge for future studies of bark diversity across habitats is therefore to tie together this multiplicity.