Functional Diversity in Woody Organs of Tropical Dry Forests and Implications for Restoration

Abstract

Tropical dry forests (TDFs) represent one of the most diverse and, at the same time, most threatened ecosystems on earth. Restoration of TDFs is thus crucial but is hindered by a limited understanding of the functional diversity (FD) of original communities. We examine the FD of TDFs based on wood (vessel diameter and wood density) and bark traits (total, inner, and outer bark thicknesses) measured on ~500 species from 24 plant communities and compare this diversity with that of seven other major vegetation types. Along with other seasonally dry sites, TDFs had the highest FD, as indicated by the widest ranges, highest variances, and largest trait hypervolumes. Warm temperatures and seasonal drought seem to drive diverse ecological strategies in these ecosystems, which include a continuum from deciduous species with low-density wood, thick bark, and wide vessels to evergreen species with high-density wood, thin bark, and narrow vessels. The very high FD of TDFs represents a challenge to restoring the likely widest trait ranges of any habitat on earth. Understanding this diversity is essential for monitoring successional changes in minimal intervention restoration and guiding species selection for resilient restoration plantings in the context of climate change.

1. Introduction

Restoration ecology aims to accelerate the recovery of the functions and services of ecosystems that have been degraded or destroyed [1]. This recovery is mediated by the distribution of the functional trait values of the plants that make up the landscape at any given moment along the regeneration trajectory [2]. The distribution of traits determines, to a large degree, ecosystem processes [3,4,5]. For example, in early successional moist tropical forests, there is a high proportion of fast-growing species with low tissue density [6]. These pioneer species reach tall statures quickly, and their tissues are short-lived, decomposing quickly [7]. The abundance of such traits results in faster nutrient cycling and carbon capture, both important ecosystem services, although there is also faster carbon release [8,9,10]. A clear understanding of the range and distinctness of functional trait values in natural communities is thus crucial for ecological restoration that aims to recover ecosystem function, ecosystem services, and ecosystem resilience.

Mimicking the natural range and distribution of trait values and, thus, services in restored systems is challenging in any vegetation type. However, this challenge reaches its apogee in tropical communities, especially dry tropical ones [11,12]. This is because tropical dry forests (TDFs) have long been recognized as having remarkably high functional diversity [12,13], i.e., the value, range, and relative abundance of functional traits in a community [14,15,16,17]. Here, by TDFs, we mean non-fire-dominated dry tropical lowland woody communities. The functional diversity of these communities is a combination of high beta diversity [18,19,20] and one of the widest ranges in life forms (Figure 1, Figure 2 and Figure 3) and ecological strategies documented in terrestrial ecosystems [21,22,23,24]. Although the remarkable functional diversity of TDFs has been previously discussed [12,13,25,26], studies focusing on TDFs are still very limited in number in comparison with those from moist biomes. Moreover, discussions regarding TDF functional diversity usually have a stronger focus on leaves in comparison with woody organs [13], the most massive component of plant biomass [27]. In addition, previous studies have not discussed the breadth of functional diversity in the context of the challenges it may represent for the ecological restoration of one of the most threatened tropical ecosystems on earth [28,29,30]. Because efforts to restore TDFs are confronted not only with daunting floristic lists but also remarkably high levels of functional diversity, it is crucial to understand the spans of plant trait variation present in TDF original communities.

Previous studies have shown that the high functional diversity in TDF species reflects their contrasting ecological strategies in the face of drought and the strong spatial heterogeneity in water availability that characterizes TDFs [13,31,32]. TDF species vary broadly along a continuum, with one extreme represented by species that are fugaciously deciduous, dropping their leaves at the first sign of drought [33]. These species are also characterized by low wood density, large leaves, thick twigs, rapid twig growth rates, and high specific leaf area [34,35]. All these traits are associated with an acquisitive ecological strategy in which plants acquire and process water, nutrients, and other resources rapidly and produce low-cost tissues that are short-lived and quickly replaced (Wright et al., 2004; Reich, 2014). Toward the other extreme of this phenology continuum, species can retain their leaves longer into drought (tardily deciduous) or even all year (evergreen). In both cases, species are characterized by having higher wood density, smaller leaves, thinner twigs, slower twig growth rates, and lower specific leaf areas than fugaciously deciduous ones. These traits are associated with a more conservative strategy in which plants produce longer-lived structures with higher carbon content [25,34,35]. Although different strategies along this continuum can coexist in TDFs, they vary in relative abundance along water availability gradients [25,36]. While fugaciously and tardily deciduous species occupy drier sites, evergreen species characterize moister areas [32,37,38].

Although woody organs are central to the diverse ecological strategies of TDF species, studies on these organs are limited in comparison with those on leaves and have mostly focused on wood density [12,26]. The data available to date seem to suggest that TDFs have extraordinary ranges of wood density and related traits, such as water storage [25]. For example, in their survey of variation in wood density across neotropical regions, Chave et al. [39] showed that the range of wood densities in the TDFs of Mexico and Central America rivaled that of the much larger and vastly more species-rich tropical wet parts of montane and lowland Amazonia. Wood density is not only associated with phenology and life-history traits [40] but also with ecosystem processes such as decomposition, with lower-density tissues decomposing more rapidly and affecting services such as nutrient cycling and carbon sequestration [40].

In addition to wood density, some studies have examined variation in vessel diameter in TDF species. Deciduous species have been regarded as carrying out efficient water transport during the limited rainy season of TDFs [31,41]. Higher efficiency is purportedly achieved through wider vessels in the sapwood [33]. Nevertheless, recent studies have shown that vessel diameter is mainly driven by plant size and, to a very small degree, by temperature and other environmental conditions [42]. As a result, vessels are wider in taller species as a result of natural selection compensating for the increased resistance of longer hydraulic pathlengths with wider conduits [43,44]. Evidence does suggest that drought-deciduous species (of any habitat) tend to have wider vessels for a given plant height [45], but the question remains whether low-density wooded TDF deciduous species have wider vessels than expected given their heights. All else being equal, wider vessels are likely associated with a greater risk of formation of gas embolisms in vessels under water stress and, therefore, vulnerability to climate-change-induced drought [46]. Hence, to the extent that it might be connected to drought vulnerability, for climate-resilient restoration purposes, it is important to document the range of variation in height-standardized vessel diameter in TDFs compared to other vegetation types.

While wood traits such as density and vessel diameter have been at least summarily examined in TDF species, a quantitative perspective on functional diversity is essentially lacking for bark. Despite the fact that bark carries out multiple functions in plants and represents a significant percentage of stem biomass [47], its structure and function are poorly known from an ecological perspective [48,49,50]. Bark can be divided into a mostly living inner region (hereafter inner bark, IB) carrying out photosynthate transport, water and non-structural carbohydrate storage, and, in many cases, photosynthesis, and an outer dead region mainly serving as protection (hereafter outer bark, OB). The most widely measured bark trait is total bark thickness, a characteristic that, as in the case of vessel diameter, changes with plant size, mainly stem diameter [51]. Once plant size is taken into account, variation in bark thickness reflects how tradeoffs and synergies between bark functions play out. As a result, size-standardized bark thickness can vary widely across species of a single community [50]. TDFs and other seasonal sites have been shown to have thick bark, especially IB, for a given diameter, variation that has been interpreted as reflecting storage demands [50,52,53]. Hence, while size-standardized bark thickness is likely to reflect important differences in species ecological performance, very little is known about the ranges of this variable in TDFs relative to other vegetation types.

Here, we aim to document the functional diversity of TDFs and discuss its implications for the restoration of these important social–ecological systems [11]. To this end, we provide empirical illustrations of the uni- and multi-dimensional trait spaces occupied by TDFs in comparison with seven other major vegetation types globally. We largely base our inferences on poorly understood wood and bark traits, most of which have not been adequately examined from a functional diversity perspective. Drawing on our own data on 24 communities from the northern and southern hemispheres, we show that functional diversity is higher in TDFs. We conclude by summarizing the restoration implications of these broad trait spans, including the need to select species based not merely on commonness or phylogeny or practical criteria such as availability in nurseries or ease of propagation but also, crucially, on trait values if restoration plantings are to succeed in delivering the services provided by natural systems in an acceptable time frame.

2. Materials and Methods

2.1. Species Selection

To examine how the wood and bark of species in TDFs compare with those of other biomes representative of global vegetation, we gathered previously published data, collected by us, for 504 to 537 species depending on the trait (

Table 1. Latitude (lat), longitude (lon), mean annual temperature (MAT), and mean annual precipitation (MAP) of the localities included grouped into eight major vegetation types. Number of species for each kind of trait set is included. Abbreviations used in figures are included next to major vegetation types.

Major Vegetation Type and Locality	Lat, lon	MAT (°C)	MAP (mm)	Species with Wood Traits	Species with Bark Traits
Deserts and xerophytic shrublands (Desert & xerophytic)
Baja California Cape, Mexico	23.03° N, 109.72° W	23.8	213	40	43
Mojave Desert, California, USA	34.1° N, 116.6° W	16.3	324	30	27
Pedregal de San Ángel Reserve, Mexico	19.31° N, 99.19° W	14.7	905	-	18
Tropical dry forests
Fazenda Almas, Paraíba, Brazil	7.5° S, 39.9° W	22.3	580	40	41
Chamela, Jalisco, Mexico	19.5° N, 105.04° W	26.2	795.7	33	57
Mediterranean woodlands and shrublands (Mediterranean)
Santa Monica Mountains, California, USA	34.1° N, 118.7° W	14.9	575	24	24
Migliarino San Rossore Park, Italy	43.8° N, 10.3° E	14.7	905	-	7
Temperate woodlands
Bothwell, Tasmania, Australia	42.4° S, 147° E	7.1	733	-	12
Yengo National Park, NSW, Australia	32.8° S, 150.9° E	16.3	792.3	-	23
Sydney area, Australia	33.8° S, 151.1° E	16.5	1162.2	34	62
Temperate treeline ecotones (Treeline ecotones)
San Pedro Mártir, Baja California, Mexico	31.0° N, 115.5° W	7.4	763	24	-
Coyhaique, Patagonia, Chile	45.5° S, 72.0° W	4.1	944	20	-
Pyrenees highlands, Spain	42.8° N, 0.3° W	3.2	1263	18	-
Mount Field, Tasmania, Australia	42.7° S, 146.6° E	4.5	1515	23	21
Temperate deciduous forests (Temp decid forests)
Yale Forest, Connecticut, USA	42.0°N, 72.1° W	8.1	1238	29	-
Pyrenee foothills, Spain	43.2° N, 1.6° E	12.2	1279	25	-
Pordenone, Italy	46.1° N, 12.5° E	11.7	1284	-	22
Tropical savannas (Savannas)
Botucatu cerrado, São Paulo, Brazil	22.9° S, 48.5° W	18.9	1331	31	40
Howard Springs, NT, Australia	12.5° S, 131.1° E	27.3	1570	24	24
Tropical and temperate rainforests (Trop & temp rainfor)
Loja, Ecuador	4.0° S, 79.2° W	14.8	1083	21	-
Atherton Tablelands, Australia	17.7° S, 145.5° E	19	1382	-	11
New South Wales temperate rainforests, Australia	34.1° S, 151.0° E	16.3	1419.5	26	18
Daintree, Queensland, Australia	16.1°S, 145.45° E	25.2	2081	29	35
Los Tuxtlas, Veracruz, Mexico	18.6°N, 95.1° W	24	3356.5	41	52

). These data are available from the TRY database [54]. Species were collected from a very wide range of habitats grouped into eight major vegetation types (Table 1). The sampled localities span deserts and xeric shrublands, Mediterranean forests and shrublands, savannas, temperate woodlands, tropical and temperate rainforests, treeline ecotones, temperate deciduous forests, and TDFs. Climatically, these habitats ranged from 213 to 3484 mm in mean annual precipitation and from 3.2 to 27.3 °C in mean annual temperature. The fire return interval of these sites ranged from >100 years in TDFs or deserts to 1–3 years in one of the tropical savannas (cerrado, Table 1). From a phylogenetic point of view, the species in our dataset belong to 151 families in 43 orders of non-monocot angiosperm trees, shrubs, and globular lifeforms (2 species of cacti). From a morphological point of view, our sampling covers most life forms in woody species, including bottle trees (Figure 1c), giant tree succulents (Figure 2b), water-storing trees (Figure 3c), tall emergent trees, as well as shrubs. This wide ecological, morphological, and phylogenetic coverage allows us to compare the traits of TDF species with most of the vegetation types occurring globally. The two localities representing TDFs are neotropical dry forests (Table 1). Located in Mexico and Brazil, these two sites have long dry seasons punctuated by a single, short rainy season and are characterized by very high taxonomic diversity [55,56].

2.2. Wood Functional Traits

We examined two wood traits: vessel diameter standardized by plant height or stem length and wood density. These traits were derived from Olson et al. [42]. For that study, we collected wedges of sapwood at the base of trunks or leading stems of shrubs above basal swellings using a saw and a screwdriver. We collected sapwood from three mature individuals per species, although, for a few species, sapwood from only one or two individuals were collected (see [42] for details). Wedge size was approximately 10 cm in width, 3 cm in depth, and 2 cm in height. Each sample was divided into a subsample that was preserved in 70% ethanol for processing in the laboratory, and another was used to measure wood density. From the preserved subsample, we cut sections of wood and stained them to measure vessel diameter under a light microscope. We selected 25 vessels randomly from the outermost region of stems to make sure we measured sapwood (hereafter wood). Given that vessel diameter increases as plants grow in height, measuring cells from the outermost region also ensured that vessel diameters reflected the current size of the measured individual. To take into account the effect of hydraulic pathlength in vessel diameter comparisons, we measured the stem length of each sampled individual using a tape measure or a TruePulse 200B laser rangefinder (Laser Technology). We measured wood density in the other subsample cutting blocks of 1 × 1 × 1 cm to measure fresh volume through the water displacement method [57]. We oven-dried these blocks at 95 °C until constant weight. Wood density was calculated as the ratio of oven-dry mass to fresh volume. We calculated mean values per species based on the three individuals collected.

2.3. Bark Functional Traits

The data for total, inner, and outer bark thicknesses were derived from Rosell [50]. As for wood sampling, we collected wedges of bark at the base of trunks or at the base of leading stems in shrubs above local swellings. We measured total and inner bark (IB) thicknesses using digital calipers and a hand lens or a light microscope when needed. We calculated outer bark (OB) thickness as the difference between total and IB thicknesses. We identified IB from OB based on color, texture, water content, and anatomical structure. To take into account plant size in bark thickness comparisons, we measured stem diameter at the point at which the bark wedge was sampled using a tape measure. We collected bark from three individuals per species in most cases and calculated a mean value for total bark, IB, and OB thicknesses. For 31 of the 537 species for which total bark thickness was measured, IB and OB thicknesses were not available.

2.4. Statistical Analyses

We based our analyses on mean values per species. Vessel diameter is known to be mainly driven by plant size. To take this into account, we calculated the residuals of a simple linear regression predicting log₁₀ vessel diameter based on log₁₀ plant height. In addition to meeting regression assumptions, the log₁₀ transformation of variables allowed us to examine proportional biological variation in variables spanning several orders of magnitude that are, thus, best modeled with multiplicative errors [58]. Log transformation is required when the functional consequence of a given unit of change in a variable of interest does not scale linearly. Take, for instance, a change in tube diameter of 10 µm. Because conductance scales as the fourth power of tube diameter [59], this increase has a very large impact on the change from 10 to 20 µm, a 16-fold increase in conductance. However, the same increase in diameter has very little impact when comparing a 100 µm tube and a 110 µm one; the difference in conductance is just 1.5 times greater. Accordingly, in biological systems, areas, pressures, conductances, and other variables measured in terms of dimensions raised to exponents or affecting them require log transformation [58]. The residuals of the vessel diameter–plant height regression were used instead of the raw vessel diameter data (hereafter residuals VD ~ height). Given that bark thickness is also driven by plant size, we also calculated the residuals of regressions predicting log₁₀ total, inner, or outer bark thicknesses based on log₁₀ stem diameter. These residuals (hereafter residuals total, inner, or outer bark thicknesses ~ stem diameter) were used in analyses instead of the raw total, inner, and outer bark thickness. We checked assumptions of linear regressions visually.

We analyzed functional diversity across major vegetation types using uni- and multi-dimensional approaches. Here, we define functional diversity as the amount of functional space that is occupied by the species in a vegetation type (functional richness sensu [17]). For the unidimensional approach to functional diversity, we calculated the range and variance for each functional trait separately. We examined whether the traits required transformation to avoid the effect of the variance increasing with the mean.

To include two or more traits in functional diversity comparisons, we calculated hypervolumes for each major vegetation type through convex hulls [60]. To this end, we used the R package hypervolume and the Gaussian kernel density estimation method [61,62]. Species with available data were different depending on the trait, so the calculations of hypervolumes were based on a different number of species, depending on the set of traits analyzed. Unlike other studies, we did not use phylogenetic imputation for missing data, given that previous studies have shown that the phylogenetic signal in residual vessel diameter and bark thickness is negligible [42,50]. As a result, the assumption that closely related species would tend to have similar trait values [63] would not hold for our dataset. Visual inspection showed that wood and bark traits had normal distributions. We scaled and centered all traits before calculating hypervolumes, aiming to have comparable units for all axes. To our knowledge, this represents the first study using functional diversity metrics for bark traits and residual vessel diameter, so we were interested in examining richness in more detail for these traits. To this end, we calculated hypervolumes for wood traits (residual vessel diameter and wood density), bark traits (residual inner and outer bark thicknesses), and wood and bark traits combined (four traits). We did not include total bark thickness in the calculation of hypervolumes because total bark thickness is the sum of inner and outer bark thicknesses and including it would introduce strong correlation. Species overlap for the datasets of wood and bark traits was around 50%. For this reason, the dataset with wood and bark traits had a small number of species of temperate deciduous forests, causing very biased estimations of the hypervolume of this vegetation type. Therefore, we excluded temperate deciduous forests from analyses based on the dataset of wood and bark traits. Hypervolumes are expressed in units of standard deviation to the power of the number of traits used to calculate these hypervolumes [61].

Comparison of hypervolumes can be influenced by sample size [60]. In our case, vegetation types with higher species richness could have larger hypervolumes as a result. To rule out such a bias, we calculated hypervolumes resampling with different sample sizes using 100 bootstrap replicates. We used plots to examine the behavior of hypervolumes and trait means under different sample sizes and calculated non-parametric confidence intervals using quantiles of the resampling process. Resampling and plots were generated using the functions hypervolume_resample and hypervolume_funnel in the R package hypervolume [62]. Finally, to compare the similarity of the functional diversity of TDFs with that of other major vegetation types, we carried out pairwise comparisons of hypervolumes using the Sørensen similarity index using the function hypervolume_overlap_statistics in the R package hypervolume [62]. Traits were not significantly correlated with one another, permitting this comparison. All analyses were carried out in R v.3.5.9 [64].

3. Results

Our sampling covered a wide range of sizes, from small globular cacti < 10 cm to rainforest emergent trees over 35 m tall (

Table 2. Stem diameter, plant height, and wood and bark traits for species included in the eight major vegetation types. Number of species (N) and means per vegetation are shown with minimum and maximum values in parentheses.

Major Vegetation Type and Locality	Stem Diameter (cm)	Plant Height (m)	Vessel Diameter (μm)	Wood Density (g cm−3)	Total Bark Thickness (mm)	Inner Bark Thickness (mm)	Outer Bark Thickness (mm)
Deserts and xerophytic shrublands (Desert & xerophytic)
Baja California Cape, Mexico	N = 50 7.7 (0.3, 34.6)	N = 50 2.6 (0.1, 7.7)	N = 40 49.8 (20.0, 104.2)	N = 38 0.52 (0.06, 0.83)	N = 43 5.9 (0.2, 31.4)	N = 31 6.1 (0.3, 30.6)	N = 31 0.8 (0.04, 4.8)
Mojave Desert, California, USA	N = 31 2.6 (0.4, 14.4)	N = 31 1.4 (0.1, 4.9)	N = 30 32.4 (13.1, 109.5)	N = 29 0.57 (0.31, 0.83)	N = 27 2.0 (0.5, 7.8)	N = 27 1.0 (0.2, 3.2)	N = 27 1.0 (0.03, 5.5)
Pedregal de San Ángel Reserve, Mexico	N = 18 13.2 (0.9, 32.6)	N = 18 4.9 (1.0, 19.0)	-	-	N = 18 8.2 (0.5, 22.1)	N = 17 5.7 (0.3, 15.5)	N = 17 2.3 (0.1, 9.3)
Tropical dry forests
Fazenda Almas, Paraíba, Brazil	N = 43 10.7 (0.8, 27.4)	N = 43 5.2 (0.9, 12.6)	N = 40 59.8 (19.5, 144.5)	N = 40 0.64 (0.18, 0.90)	N = 41 5.6 (0.3, 19.0)	N = 41 4.0 (0.1, 14.2)	N = 41 1.6 (0.02, 10.2)
Chamela, Jalisco, Mexico	N = 63 18.7 (0.1, 59.4)	N = 63 7.4 (0.2, 20.0)	N = 33 81.1 (20.3, 173.5)	N = 31 0.51 (0.21, 0.80)	N = 57 8.4 (0.3, 48.5)	N = 55 6.8 (0.6, 27.7)	N = 55 1.9 (0.1, 20.8)
Mediterranean woodlands and shrublands (Mediterranean)
Santa Monica Mountains, California, USA	N = 26 3.9 (0.4, 15.0)	N = 26 2.7 (0.6, 7.1)	N = 24 32.1 (19.8, 88.0)	N = 24 0.59 (0.30, 0.78)	N = 24 1.6 (0.3, 4.0)	N = 22 1.1 (0.2, 3.8)	N = 22 0.6 (0.1, 2.1)
Migliarino San Rossore Park, Italy	N = 7 5.7 (0.2, 11.9)	N = 7 2.3 (0.2, 4.8)	-	-	N = 7 2.4 (0.3, 6.2)	N = 7 1.5 (0.1, 3.7)	N = 7 0.9 (0.2, 2.6)
Temperate woodlands
Bothwell, Tasmania, Australia	N = 12 12.0 (0.4, 68.0)	N = 12 4.5 (1.1, 21.0)	-	-	N = 12 4.9 (0.4, 22.1)	N = 12 3.9 (0.3, 18.7)	N = 12 1.0 (0.03, 4.0)
Yengo National Park, NSW, Australia	N = 23 9.5 (0.3, 46.0)	N = 23 6.7 (0.3, 35.0)	N = 15 43.6 (15.2, 98.3)	N = 15 0.67 (0.51, 0.82)	N = 23 8.9 (0.4, 52.9)	N = 23 4.2 (0.1, 29.9)	N = 23 4.7 (0.1, 41.3)
Sydney area, Australia	N = 63 7.2 (0.2, 119.4)	N = 63 3.5 (0.2, 31.3)	N = 25 47.9 (11.2, 143.6)	N = 25 0.54 (0.31, 0.69)	N = 62 2.7 (0.2, 35.2)	N = 57 1.8 (0.2, 17.2)	N = 57 1.1 (0.05, 22.3)
Temperate treeline ecotones (Treeline ecotones)
San Pedro Mártir, Baja California, Mexico	N = 24 3.1 (0.5, 12.4)	N = 24 1.7 (0.1, 11.2)	N = 24 27.0 (11.8, 58.9)	N = 24 0.51 (0.24, 0.72)	-	-	-
Coyhaique, Patagonia, Chile	N = 20 2.4 (0.7, 7.3)	N = 20 1.7 (0.7, 7.7)	N = 20 26.1 (17.1, 39.6)	N = 20 0.54 (0.38, 0.69)	-	-	-
Pyrenees highlands, Spain	N = 17 5.7 (0.4, 16.8)	N = 17 3.1 (0.4, 7.3)	N = 17 37.4 (17.3, 88.6)	N = 17 0.49 (0.31, 0.63)	-	-	-
Mount Field, Tasmania, Australia	N = 23 3.9 (0.4, 33.0)	N = 23 1.8 (0.3, 13.7)	N = 23 29.4 (14.1, 118.0)	N = 21 0.60 (0.40, 0.83)	N = 21 1.8 (0.4, 10.6)	N = 19 1.5 (0.1, 9.8)	N = 19 0.4 (0.1, 1.5)
Temperate deciduous forest (Temp decid forests)
Yale Forest, Connecticut, USA	N = 29 17.4 (0.4, 60.6)	N = 29 10.1 (0.7, 30.2)	N = 29 51.2 (14.5, 136.2)	N = 29 0.52 (0.29, 0.71)	-	-	-
Pyrenee foothills, Spain	N = 21 11.2 (0.3, 47.3)	N = 21 5.9 (0.7, 18.0)	N = 21 45.6 (16.9, 106.5)	N = 21 0.56 (0.37, 0.77)	-	-	-
Pordenone, Italy	N = 22 6.4 (0.4, 16.9)	N = 22 5.6 (0.6, 10.0)	N = 5 35.0 (21.1, 47.7)	N = 5 0.57 (0.50, 0.70)	N = 22 2.4 (0.3, 11.0)	N = 22 1.6 (0.1, 5.2)	N = 22 0.8 (0.1, 5.8)
Tropical savannas (Savannas)
Botucatu cerrado, São Paulo, Brazil	N = 46 9.5 (0.8, 39.8)	N = 46 3.2 (0.8, 7.3)	N = 31 62.0 (14.1, 122.8)	N = 31 0.53 (0.32, 0.68)	N = 40 11.6 (0.4, 53.8)	N = 40 5.2 (0.2, 17.0)	N = 40 6.4 (0.04, 38.9)
Howard Springs, NT, Australia	N = 24 12.7 (0.3, 35.5)	N = 24 6.7 (0.1, 15.6)	N = 24 86.9 (20.1, 158.1)	N = 24 0.64 (0.32, 0.84)	N = 24 10.9 (0.3, 26.6)	N = 19 6.4 (0.5, 11.7)	N = 19 7.0 (0.3, 19.8)
Tropical and temperate rainforests (Trop & temp rainfor)
Loja, Ecuador	N = 21 2.7 (0.6, 12.7)	N = 21 1.2 (0.6, 2.9)	N = 21 23.6 (12.1, 41.2)	N = 20 0.48 (0.40, 0.60)	-	-	-
Atherton Tablelands, Australia	N = 11 10.2 (1.9, 22.9)	N = 11 5.7 (2.0, 12.0)	-	-	N = 11 3.6 (0.7, 9.9)	N = 11 3.2 (0.3, 9.7)	N = 11 0.5 (0.1, 1.7)
New South Wales temperate rainforests, Australia	N = 22 20.2 (1.3, 63.2)	N = 22 10.9 (1.3, 25.3)	N = 20 65.0 (29.6, 198.7)	N = 20 0.47 (0.18, 0.70)	N = 18 4.9 (0.4, 13.0)	N = 18 3.5 (0.3, 8.6)	N = 18 1.5 (0.1, 6.7)
Daintree, Queensland, Australia	N = 35 21.3 (0.7, 72.6)	N = 35 10.9 (0.5, 30.8)	N = 29 85.4 (20.2, 210.4)	N = 29 0.53 (0.33, 0.71)	N = 35 5.3 (0.6, 26.8)	N = 35 4.6 (0.4, 24.3)	N = 35 0.7 (0.1, 3.0)
Los Tuxtlas, Veracruz, Mexico	N = 55 36.4 (1.2, 130.2)	N = 55 15.0 (0.9, 32.5)	N = 41 103.2 (24.1, 244.3)	N = 41 0.45 (0.03, 1.03)	N = 52 8.0 (0.9, 21.1)	N = 50 6.8 (0.7, 17.1)	N = 50 1.3 (0.1, 8.6)

and Table S1, and Figure S1a in the Supplementary Materials). In terms of diameters, the sampled species included thin stems of 14 cm in basal diameter to very wide trunks of 130 cm (Table 2 and Table S1, Figure S1b).

3.1. Unidimensional Approach to Functional Diversity

All major vegetation types included species with vessels that are wider or narrower than expected for their hydraulic pathlength (plant height or length). This was translated into species appearing above or below the regression line in Figure S2 (Table S2) and, thus, having negative or positive residuals (

Table 3. Descriptive statistics for residual vessel diameter, wood density, and residual total, inner, and outer bark thicknesses across major vegetation types.

Vegetation Types	N	Median	Minimum	Maximum	Range	Variance
Residuals vessel diameter ~ height (total N = 512)
Desert & xerophytic	70	0.040	−0.244	0.508	0.752	0.019
Trop dry forests	73	0.010	−0.371	0.363	0.734	0.031
Mediterranean	24	−0.116	−0.313	0.245	0.558	0.018
Temp woodlands	40	−0.039	−0.324	0.260	0.583	0.018
Treeline ecotones	84	−0.068	−0.306	0.297	0.603	0.013
Temp decid forests	55	−0.139	−0.322	0.218	0.540	0.016
Savannas	55	0.149	−0.356	0.462	0.819	0.025
Trop & temp rainfor	111	−0.006	−0.361	0.360	0.721	0.028
Wood density (g/cm3) (total N = 504)
Desert & xerophytic	67	0.58	0.06	0.83	0.77	0.032
Trop dry forests	71	0.64	0.18	0.90	0.72	0.034
Mediterranean	24	0.60	0.30	0.78	0.48	0.011
Temp woodlands	40	0.59	0.31	0.82	0.51	0.015
Treeline ecotones	82	0.55	0.24	0.83	0.59	0.013
Temp decid forests	55	0.54	0.29	0.77	0.48	0.007
Savannas	55	0.58	0.32	0.84	0.53	0.016
Trop & temp rainfor	110	0.46	0.03	1.03	1.00	0.017
Residuals total bark thickness ~ stem diameter (total N = 537)
Desert & xerophytic	88	0.070	−0.538	0.839	1.377	0.067
Trop dry forests	98	0.043	−0.985	0.565	1.550	0.078
Mediterranean	31	−0.117	−0.874	0.453	1.328	0.055
Temp woodlands	97	−0.009	−0.401	0.648	1.050	0.047
Treeline ecotones	21	−0.040	−0.585	0.359	0.944	0.044
Temp decid forests	22	−0.154	−0.503	0.214	0.716	0.025
Savannas	64	0.311	−0.507	0.706	1.213	0.085
Trop & temp rainfor	116	−0.175	−0.719	0.383	1.103	0.039
Residuals inner bark thickness ~ stem diameter (total N = 506)
Desert & xerophytic	75	0.089	−0.821	0.968	1.789	0.103
Trop dry forests	96	0.111	−1.275	0.634	1.909	0.090
Mediterranean	29	−0.147	−0.804	0.268	1.071	0.083
Temp woodlands	92	0.003	−0.414	0.563	0.977	0.042
Treeline ecotones	19	0.040	−1.414	0.229	1.643	0.160
Temp decid forests	22	−0.121	−0.878	0.064	0.942	0.051
Savannas	59	0.149	−0.896	0.605	1.501	0.077
Trop & temp rainfor	114	−0.106	−0.878	0.536	1.414	0.056
Residuals outer bark thickness ~ stem diameter (total N = 506)
Desert & xerophytic	75	0.067	−1.031	0.901	1.932	0.201
Trop dry forests	96	−0.138	−1.264	1.068	2.332	0.289
Mediterranean	29	0.008	−0.901	0.793	1.694	0.192
Temp woodlands	92	−0.042	−1.546	1.389	2.935	0.268
Treeline ecotones	19	0.014	−0.720	0.713	1.433	0.102
Temp decid forests	22	−0.246	−1.189	0.733	1.922	0.215
Savannas	59	0.684	−0.905	1.356	2.262	0.309
Trop & temp rainfor	114	−0.208	−1.438	0.685	2.123	0.168

). Ranges were higher in seasonal sites, including savannas, deserts and xerophytic shrublands, and TDFs, but also in the tropical and temperate rainforests (Figure 4a). In terms of variances, TDFs had the highest values, followed by tropical and temperate rainforests and savannas (Table 3). A similar trend was observed with wood density. The highest ranges were observed in the rainforests, which included the extremes of wood density in our dataset (Figure 4b). However, the highest variances were recorded in the drier environments, especially in the TDFs (Table 3), indicating that even when dry forests do not include the extremes in wood density, they include higher levels of variation than other major vegetation types.

As was the case for residual vessel diameter, residuals for all bark thickness traits (total, inner, and outer bark thicknesses) included positive and negative values across all major vegetation types, suggesting that there are species with thicker and thinner than expected total, inner, and outer bark for a given stem diameter (Figure S3, Table S3). For total bark thickness residuals, ranges were widest in TDFs and other seasonal but fire-prone environments and narrowest in temperate deciduous forests and treeline ecotones (Table 3, Figure 5a). Variances indicated that the highest levels of variation were observed in savannas, followed by TDFs. Variance in residual total bark thickness was small in temperate deciduous forests (Table 3).

Within bark, inner and outer bark exhibited different patterns. Inner bark residuals had, again, the widest range in TDFs, followed by desert and xerophytic shrublands and treeline ecotones (Table 3, Figure 5b). In terms of variance, these three systems had the largest values, with the highest variance observed in the treeline ecotone (Table 3). In turn, residuals of outer bark thickness had large ranges in the fire-prone temperate woodlands and savannas but also in the non-fire-prone TDFs and rainforests (Table 3, Figure 5c). Regarding variances, savannas stood out for their high values, followed by TDFs and temperate woodlands (Table 3).

3.2. Multi-Dimensional Approach to Functional Diversity

There was a strong overlap in hypervolumes across major vegetation types for all sets of traits examined (Figure 6 and Figure S4). However, different patterns were recovered depending on the types of traits used to calculate hypervolumes. When only wood traits were used, hypervolumes were the largest for TDFs, followed by rainforests and dry or seasonal environments (

Table 4. Hypervolumes of major vegetation types based on different sets of traits. Units are standard deviations to the power of the number of traits used in calculations; 95% confidence intervals shown in parentheses.

Vegetation Type	Wood Traits (Vessel Diameter and Density)		Bark Traits (Inner and Outer Bark Thickness)		Wood and Bark Traits
	N	Hypervolume	N	Hypervolume	N	Hypervolume
Desert & xerophytic	67	19.09 (14.65, 21.45)	75	20.65 (14.99, 23.44)	49	314.08 (164.40, 342.38)
Trop dry forests	71	24.15 (19.73, 26.75)	96	20.62 (14.69, 24.45)	63	452.92 (234.38, 540.90)
Mediterranean	24	13.35 (6.51,16.66)	29	18.14 (9.79, 22.21)	21	232.11 (71.16, 294.24)
Temp woodlands	40	14.40 (9.99, 16.26)	92	16.35 (11.46, 17.92)	36	218.07 (103.39, 241.02)
Treeline ecotones	82	11.83 (8.50, 13.34)	19	17.82 (5.05, 24.68)	18	113.79 (24.62, 146.36)
Temp decid forests	55	9.73 (6.48, 11.27)	22	14.80 (6.29, 19.61)	5	12.69 (0.13, 12.90)
Savannas	55	16.70 (11.78, 19.23)	59	19.72 (13.77, 23.41)	44	288.41 (120.68, 346.02)
Trop & temp rainfor	110	18.53 (13.28, 22.00)	114	14.12 (10.41, 15.72)	82	238.00 (126.76, 269.17)

). Figure 6a shows the overlap in hypervolumes and how these major vegetation types occupy extensive areas of two-dimensional trait space, especially TDFs, deserts and xerophytic shrublands, and savannas. TDFs not only occupy larger areas of this space, but they also tend to occupy areas that are not occupied by other vegetation types. This can also be observed in Figure 7, which presents pairwise dissimilarity comparisons of hypervolumes. TDFs have the least overlap in the wood trait space with temperate deciduous forests, treeline ecotones, and Mediterranean systems (

Table 5. Similarity of hypervolumes between pairs of major vegetation types based on their overlap, quantified by the Sørensen statistic. Similarity based on wood traits (two traits) is shown below the diagonal; similarity based on bark traits (two traits) is shown above the diagonal.

	Desert & Xerophytic	Trop Dry Forests	Mediterr	Temp Woodlands	Treeline Ecotones	Temp Decid Forests	Savannas	Trop & Temp Rainfor
Desert & xerophytic	-	0.709	0.688	0.690	0.694	0.612	0.705	0.655
Trop dry forests	0.790	-	0.602	0.741	0.634	0.620	0.671	0.700
Mediterranean	0.620	0.653	-	0.659	0.705	0.731	0.630	0.698
Temp woodlands	0.742	0.711	0.767	-	0.638	0.621	0.723	0.690
Treeline ecotones	0.664	0.614	0.735	0.776	-	0.650	0.624	0.675
Temp decid forests	0.511	0.532	0.713	0.677	0.760	-	0.567	0.757
Savannas	0.652	0.673	0.548	0.649	0.565	0.472	-	0.596
Trop & temp rainfor	0.713	0.729	0.629	0.693	0.713	0.618	0.609	-

, below diagonal).

In the case of hypervolumes based on bark traits, again, TDFs stood out, along with the deserts and xerophytic vegetations and the savannas, as having the highest values (Table 4). However, in terms of the areas occupied in bark trait space, TDFs had a strong overlap with most n types (Figure 6b). Deserts and xerophytic shrublands occupied areas of thicker residual inner bark, whereas savannas showed the same trend with residual outer bark (Figure 6b). In contrast, species of treeline ecotones tended to have thinner inner bark for their size (Figure 6b). The hypervolume of TDFs differed more strongly from that of Mediterranean systems, treeline ecotones, and temperate deciduous forests (Figure 7b, Table 5, above diagonal).

Combining wood and bark traits for calculations, TDFs had a remarkably large hypervolume (Table 4). Extending beyond the other vegetation types in most bivariate plots, TDFs apparently have combinations of wood and bark traits not observed in other major vegetation types, explaining the considerably larger hypervolume of this vegetation (Figure S4). Based on wood and bark traits, TDFs were most different from treeline ecotones and most similar to deserts and xerophytic shrublands (Figure S4,

Table 6. Similarity of hypervolumes between pairs of major vegetation types based on their overlap, quantified by the Sørensen statistic. Similarity based on wood and bark traits (four traits) is shown below the diagonal. Temperate deciduous forests had fewer than 15 species and were not included in comparisons (cells with NA).

	Desert & Xerophytic	Trop Dry Forests	Mediterr	Temp Woodlands	Treeline Ecotones	Temp Decid Forests	Savannas	Trop & Temp Rainfor
Desert & xerophytic	-	-	-	-	-	-	-	-
Trop dry forests	0.661	-	-	-	-	-	-	-
Mediterranean	0.524	0.465	-	-	-	-	-	-
Temp woodlands	0.642	0.547	0.608	-	-	-	-	-
Treeline ecotones	0.412	0.308	0.506	0.473	-	-	-	-
Temp decid forests	NA	NA	NA	NA	NA	-	-	-
Savannas	0.489	0.529	0.453	0.524	0.395	NA	-	-
Trop & temp rainfor	0.599	0.523	0.524	0.587	0.409	NA	0.414	-

, above diagonal).

Analyses examining the effect of sample size on hypervolume estimations suggested that, in general, larger hypervolumes were not simply the result of larger sample sizes (larger number of species in a biome). Resampled hypervolume means were very stable across the different sample sizes tested in the bootstrap process. Moreover, in most cases, estimated hypervolumes converged with or were slightly larger than the hypervolumes estimated through the resampling process. To be sure, convergence between estimated hypervolumes and bootstrapped hypervolume means was lowest in biomes with smaller sample sizes (<30 species), especially when calculating hypervolumes with larger dimensions (i.e., more traits included in calculations), suggesting that increasing the sample size could lead to more stable estimates. This was the case for Mediterranean systems, temperate deciduous forests, and treeline ecotones. However, we did not detect a trend of increase in hypervolumes with larger sample sizes in bootstrap procedures, suggesting that larger hypervolumes were not the mere result of larger sample sizes but of a higher representation of functionally diverse species in these vegetation types.

4. Discussion

4.1. Patterns and Causes of High Functional Diversity in Wood and Bark of Tropical Dry Forests

Our analyses highlighted that across global vegetation types, TDFs had the highest functional diversity for most wood and bark traits. Although TDFs did not always have the widest absolute ranges, they tended to have the highest variances for most traits. For example, the minimum and maximum values of wood density were observed in tropical and temperate rainforests. This is perhaps not surprising given that these habitats have the highest species diversity per unit area on earth and, thus, have greater opportunities for the representation of extreme trait values. However, these extremes are relatively rare in rainforests compared to TDFs. Bottle trees such as Carica and Jacaratia (Caricaceae) have water-storing wood that is nearly entirely non-lignified [65] and maximally low wood density [25]. Although bottle trees of these genera can be present in rainforests, this life form is far more abundant in terms of the number of species and individuals in TDFs (Figure 1). So are tree succulents and trees with wood that, while of higher density than that of Caricaceae species, is still replete with water (Figure 2) [66]. The same goes for high wood density—there are, to be sure, occasional high-wood-density species in rainforests, but high wood density is much more common in TDFs. The greater trait variances of TDFs (Table 3) thus seem to suggest that TDFs have a greater density in the occupation of the full range of trait values in comparison with the other major vegetation types.

Seasonal drought in the context of lack of frost seems to be a central factor driving the very high functional diversity in TDFs and other dry vegetation [31,67]. Our data on wood density and height-standardized vessel diameter illustrate this point. Low wood density in dryland plants is associated with water storage [68,69]. It is therefore unsurprising that water-storing stems should be so common in TDFs and less frequently found in rainforests, an environment with lower selection pressure for water storage. At the same time, low wood density should be selected against in cold situations as a result of its vulnerability to the freezing-induced deformation of thin cell walls or the forcing of the contents of living cells into adjacent vessels [70,71,72]. TDFs and other dry environments should therefore permit the evolution of more low-wood-density species than in any other ecosystem. If storage is one extreme of a continuum of strategies permitting persistence in TDFs, extreme embolism resistance is another [73]. This resistance is favored by high wood density, a trait that is also commonly observed in TDF species [31]. In TDFs, very low wood density species are “fugaciously deciduous”, i.e., they tend to drop their leaves immediately when faced with drought [33], likely as a result of low wood density providing low resistance to embolism [45]. In contrast, species with high wood density maintain their leaves through drought (as evergreens) or drop them only tardily, well into the dry season [74]. Because very dense woods represent a large amount of photosynthates per unit volume, producing very high-density wood likely requires long growing seasons or at least highly productive ones, conditions absent in frost-prone environments. In this way, seasonal drought in the context of highly productive tropical frost-free habitats likely favors the evolution of a remarkably wide array of functional diversity.

An even wider range of factors might favor variation of size-standardized bark thickness in TDFs. The traditional explanation of thick bark as the result of frequent fires [75,76] would be congruent with savannas having the highest median and maximum values of standardized total bark thickness in our dataset (Table 3). However, inferences based on total bark do not take into account that the two major regions of bark, inner (IB) and outer bark (OB), carry out different functions and are thus shaped by different selective factors [47]. While IB carries out photosynthate transport and storage [52,53,77], OB protects stems against fire and other factors [78]. In our dataset, TDFs had the widest range of residual IB and high variance (Table 3), results likely associated with the wide range of drought survival strategies in TDF species, in which IB might have a central role. IB and wood covary in density suggesting functional coordination regarding water storage [49,79]. In addition, species with high stem water storage also tend to have thick bark [52]. The opposite, high-density wood and thin bark, is also commonly observed. As a result, the high variation in residual IB likely reflects that this living portion of bark might be aligned with the fugaciously, tardily deciduous, and evergreen syndromes in TDF species [33]. In contrast, and as would be expected, given its protective function [50,78], residual OB showed wider ranges and highest variances in fire-prone environments such as savannas (Table 3). TDFs typically do not burn or experience fire over such long intervals that fire is not a significant selective factor [80]. However, some TDF species can reach OBs as thick as in any savanna (e.g., Aralia mexicana or Moringa concanensis), suggesting an important protective role against boring insects or other herbivores. Both of these syndromes—massive water-storing photosynthetic bark and very thick dead outer bark—underscore again the exceptional functional breadth covered by TDF ecological strategies.

Our hypervolume analyses highlight how, in combination, wood and bark traits contribute to diverse ecological strategies in TDFs. Although the hypervolumes showed strong overlap across major vegetation types (Figure 6 and Figure S4), TDFs had larger hypervolumes for practically all sets of traits (Table 4). Moreover, TDFs had a higher absolute range in the hypervolumes for wood traits, highlighting the role of residual vessel diameter and wood density in the diversification of TDF ecological strategies (Figure 6a). Hypervolumes also highlighted constrasts between TDFs and other systems worldwide. With regard to wood traits, TDFs were maximally different from temperate deciduous forests (Figure 7a), vegetation subject to cold and freezing periods, conditions that seem to restrict vessel diameter [72]. These conditions and short growing seasons seem to drive convergence on similar phenologies and trait values across temperate deciduous forest species [81], e.g., spanning a narrow range of wood densities [40] and leaf lifespans [82] (Table 3). Although just as seasonal as temperate forests, seasonality in TDFs is associated with warm temperatures, high productivity, and drought, which seem to favor much wider ranges of ecological strategies [13,67].

With regard to bark traits, TDFs were most different from Mediterranean systems and most similar to temperate woodlands (Figure 6b). Differences with Mediterranean species would be expected, given that most species in these systems are shrubby, with stems that do not persist after fires [83], and thus lack thick OB; Mediterranean systems also lack species with water-storing stems, thus lacking thick IB. Surprisingly, TDFs were similar to temperate woodlands, which largely include sclerophyll woodlands, which, unlike TDFs, are made up of mostly evergreen species subjected to fire. That two systems with very different selective factors should coincide in their residual bark thickness belies the commonly-held belief that fire is the main selective agent favoring this trait [76]; it also highlights that water storage, mechanical support, and photosynthesis are all evident in many of the very thick barks of TDF species [47]. Finally, when bark and wood traits were combined in hypervolume comparisons, TDFs were again very similar to desert and xerophytic systems and most different from treeline ecotones (Figure 6c), which have very short, cold-limited growing seasons.

4.2. Integrating the Functional Diversity of Wood, Bark, and Leaves in Tropical Dry Forests

Although we did not include leaves in our study, it is possible to integrate the patterns recovered here for wood and bark with those for leaves that have been previously documented in TDF species. Species of dry forests have been shown to have very high functional diversity in their leaves [13,84,85,86]. For example, specific leaf area (SLA) can vary six-fold (from 4.2 to 26.9 mm²/mg) in TDFs, in comparison with only 2.2-fold in deserts, tropical evergreen forests, or tundras [12]. Increasing with leaf nitrogen and photosynthetic rate and decreasing with leaf lifespan, SLA is central to the economic spectra of leaves [87] and plants in general [88]. High variation in leaf traits such as SLA and phenology and wood traits such as wood density can be explained by the diverse water-use strategies in TDFs. Our study shows that high variation in poorly studied traits such as residual vessel diameter and residual bark thickness is likely explained by this same diversity in water-use strategies. As a result, drought-avoidant deciduous species will also show wider vessels and thicker bark for their size, whereas drought-resistant evergreen species will have narrower vessels and thinner bark for their stem size [35,52,89]. Unlike moist environments in which light availability is a strong selective factor [85], water availability and the resulting strategies to cope with drought seem to explain the high functional diversity of leaves, wood, and bark in tropical dry forests and other seasonal environments [33,38,74].

4.3. Implications of High Functional Diversity in Tropical Dry Forests for Restoration Efforts

Basic information on the functional diversity of TDFs is essential for maximal and minimal intervention strategies for restoration. With regard to maximal intervention, i.e., restoration plantings [90], optimizing functional diversity will maximize the delivery of multiple ecosystem services [91,92]. Given that the recovery of ecosystem function is mediated by plant functional traits, trait-based approaches in restoration are increasingly becoming more important for species selection [6,91,93,94]. Species for TDF restoration plantings thus need to be selected based on a good understanding of the functional diversity of the original vegetation, not just the original species composition [95,96]. For example, this understanding could guide species selection based on functional redundancy [97]. This underscores the need for basic trait data from TDFs and other dry habitats that are understudied and poorly represented in international functional trait datasets [98]. In the case of minimal intervention, also called unassisted forest regeneration sensu [99], i.e., exclusion of disturbance to allow natural succession, understanding the breadth of trait values in the original TDFs is essential to compare successional stages and make decisions regarding restoration intervention accordingly [94].

In trait-based restoration, wood density has been especially important because of its association with drought resistance [91]. However, vessel diameter is very likely a more important target trait for resisting drought. All else being equal, wider vessels seem to be more vulnerable to drought-induced embolism than narrow ones [46]. At the same time, taller plants have wider vessels as a result of natural selection buffering the increase in hydraulic resistance that occurs with height growth. For this reason, we used residual vessel diameter here, i.e., vessel diameter taking into account plant height. The large variation in residual vessel diameter recovered for TDFs is congruent with the wide variation in drought tolerance of TDF species discussed previously. Selecting species from this wide range of vessel size would help increase planting resilience under the increasingly erratic precipitation and more extreme droughts TDFs are experiencing [100]. The relationship between vessel diameter, plant height, and embolism vulnerability could have an additional implication for TDF restoration. If taller plants have wider conduits and wider conduits are more vulnerable to embolism, then all else being equal, taller plants should be more vulnerable to embolism [42]. Within species, this could mean that variants with vessels that are relatively narrow, given their height, will persist best because they will be the most resistant to embolism for a given height. It also suggests that restored forests are likely to be shorter than the original ones [101], causing changes to ecosystem function and services that are dependent on plant height.

Another target trait in restoration plantings has been bark thickness. It has been suggested that in sites with high fire frequency, species with thick bark could increase the resilience of plantings [102]. Although thick total bark can protect plants from fire, OB seems to be the region favored by natural selection in this context, whereas IB responds to metabolic demand and water availability [47]. Restoration plantings in seasonal areas that are expected to experience more frequent fire and drought would thus benefit from species with both thicker OB and IB. This is exactly the case with TDFs [100,103]. Our results show that both OB and IB thicknesses are highly diverse in TDFs, suggesting a wide range of species is available for restoration plantings. Although we did not examine associations between thicker bark, lower wood density, thicker stems, and larger leaf area [35], these associations would need to be taken into account to optimize target ecosystem services from restoration plantings.

5. Conclusions

Our results highlight that restoring TDFs is challenging not just because of their high phylogenetic diversity but mainly because of their high functional diversity. Our findings show that this diversity is not restricted to leaves, as has been shown in previous studies, but also includes wood and bark. Together, wood, bark, and leaf functional diversities in TDF species reflect the wide span of ecological strategies in response to drought, the most important selective factor shaping the traits of species in TDFs and other seasonally dry environments. Documenting and understanding the functional diversity of TDFs is essential to monitor the recovery of ecosystem services and also to inform species selection for restoration plantings to ensure high tree performance. The wood and bark traits studied here are particularly important in the context of climate change, especially with the more frequent fires and more extreme droughts that threaten original and restored TDFs. Given the unique combination of high human pressures and increasingly erratic climates and representing the apogee of terrestrial plant functional diversity, the challenges in restoring and sustainably managing TDFs are highly distinctive and represent some of the most complex restoration situations faced globally.