Materials and methods

Study sites. The study sites are located in the tropical lower montane forest region of Veracruz, Mexico (19° 30' N; 96° 57' W) between 1,280 and 1,450 m asl. The climate is mild and humid throughout the year, with three distinct seasons: a relatively dry-cool season from November to March, a dry-warm season in April-May and a wet-warm season from June to October. Annual precipitation is 1,600-1,800 mm and the mean temperature is 17-18 °C. The soil has been classified as Andosol. The dominant tree species are Carpinus tropicalis (Betulaceae), Clethra macrophylla (Clethraceae), Liquidambar styraciflua (Altingiaceae), Quercus lancifolia (Fagaceae), Q. sartorii (Fagaceae), Q. xalapensis (Fagaceae) and Turpinia insignis (Staphyleaceae) (^{Williams-Linera et al. 2013}). In this region, five restoration plantations with native tree species were chosen (^{Pedraza & Williams-Linera 2003}, ^{Muñiz-Castro et al. 2015}, ^{Williams-Linera et al. 2010}, ²⁰¹⁵, ²⁰¹⁶). The sites included plantations that were categorized as recently established or young (average age = 2.5 yr) and middle-aged (average age = 14.2 yr) (Table 1). The average distance among sites was 4.6 km. All sites were situated on slopes of 11 to 36°. Soils are volcanic in origin and > 2.5 m depth, mildly acidic (pH 5.1 to 5.6), low in extractable P (4.5 mg/kg), high in organic matter (10-20 %), with a bulk density of 0.9 to 1.1 g/cm³, and texture from clay to sandy-clay-loam. Because of the sites’ proximity to each other, there are clear similarities in climate, physical and chemical soil characteristics and geomorphic environmental conditions. Thus, we were confident that the trait values of saplings and trees can be reliably compared.

Eight planted native tree species were selected from the five restoration sites (Table 2). All saplings were collected in young sites, and all adults were collected in middle-age sites. The sapling stage was < 10 cm diameter and < 10 m tall, and the adult stage included trees > ca. 10 cm diameter and > 10 m tall (Table 2).

Functional traits. Mature leaves with no herbivore damage that receive direct sunlight were collected from two branches on opposite sides of the stem or trunk with an oversized slingshot. The leaves of three saplings or three to six adults of each species were collected per site during the wet-warm season following standardized protocols (^{Cornelissen et al. 2003}). The leaves were stored in black plastic bags and transported to the Laboratory of Functional Ecology at the Institute of Ecology, Xalapa, Mexico for subsequent analysis.

We selected 10 to 20 leaves from each individual. The area of each leaf (LA, cm²), excluding the petiole, was measured in a WinFOLIA Leaf Area Meter (Software program LA2400). The leaves were oven-dried at 60 °C for at least 72 hr and dry mass determined. Specific leaf area (SLA) was calculated as leaf area divided by leaf dry mass. The leaves were then ground for analysis to determine foliar carbon, nitrogen and phosphorus. Leaf C and N were determined using the TruSpec Micro, and leaf P by digestion with HNO₃/ HClO₄. The samples were analyzed using standard techniques (^{SEMARNAT 2002}).

Stomatal density (SD) was determined in two fresh leaves per species and site. At the top, center and bottom of the abaxial surface of each leaf, we applied an imprinting paste followed by clear nail polish to produce stomatal imprints. Stomatal densities were determined using a compound microscope with 20x objective from photographs and expressed as stomata/mm².

Wood density (WD) was determined from a core extracted from each adult tree at a height of 1.3 m above ground level with a Pressler increment borer. In the laboratory, these cores were submerged in distilled water for three days until fully hydrated. The volume of the cores was determined by the water displacement method, and dry weight was measured after the cores had been oven-dried at 60 °C for 72 hr.

Relative growth rates (RGR) in height and diameter were estimated using published and unpublished databases from previous works conducted in different years in the same sites, species and permanently tagged individuals used in this study (Table 1, 2). We used the equation RGR = ln H₂ - ln H₁/(t₂ - t₁), where H₂ and H₁ are height/diameter, and t₂ and t₁ are time in years (^{Hunt 1990}). RGR was calculated for one period between two censuses and expressed over an interval of a year. The time span used to calculate RGR was 3-4 years for saplings and 8-10 years for adults.

Statistical analysis. Differences in leaf traits between stages were analyzed using a linear mixed model with stage as fixed effect, and sites as a random effect. To attain normality and homoscedasticity in the residuals of both models, we used a log₁₀ transformation. Inspection of residuals was used to verify whether the model’s assumptions had been met. WD was compared with respect to species using generalized linear model. Post hoc tests were conducted using Tukey's HSD. Analyses were conducted using the statistical platform of R version 3.4.2 (https://www.R-project.org/ 2017). A principal component analysis (PCA) was run to summarize and to visualize the main trends of sapling and adult tree species, and the relationships with foliar traits and RGR. The PCA was run in the PC-ORD software (^{McCune & Grace 2002}).

Results

Leaf area showed significant stage × species interaction, and differences among species, but LA was similar between saplings and adults (Figure 1A, Table 3). LA of H. donnellsmithii was higher for saplings than for adults, but LA of Q. xalapensis and T. micrantha was higher for adult trees. Also, LA was higher for Heliocarpus donnellsmithii, intermediate for four species (e.g., Q. xalapensis), and smaller for Trema micrantha, Carpinus tropicalis and Myrsine coriaceae (Table 4).

Overall, SLA was higher in saplings than in adults (Figure 1B, Table 3), and there were differences among species (Table 3). SLA was higher in H. donnellsmithii, and lower in Q. xalapensis, Q. germana and J. pyriformis (Table 4).

Stomatal density displayed significant stage × species interaction and differed among species, but SD was similar in saplings and adults (Figure 1C, Table 3). L. styraciflua and Q. germana had higher SD in saplings, whereas T. micrantha had higher SD in adults (Table 4). Foliar C, N and P content were similar in both stages, but species differed in their foliar nutrient content (Figure 1D - F, Table 3, 4).

Wood density differed among tree species (F = 6.39, p = 0.019), with Q. germana and Q. xalapensis having the highest WD. The other species had intermediate values (L. styraciflua, C. tropicalis, J. pyriformis, M. coriacea, T. micrantha), while H. donnellsmithii had the lowest WD (Table 2).

Relative growth rate was higher for H. donnellsmithii, M. coriacea and T. micrantha than for the other species (Figure 2). Also, RGR was higher in saplings than in adults of those three species. RGR in height and diameter were correlated (r = 0.94, p < 0.0001), so we used RGR in height for further analysis.

The PCA based on leaf traits and RGR of tree species in the sapling and adult stages is shown in Figure 3. The first three components of PCA explained 73.3 % of the total variation. LA, SLA and leaf N content had positive loading on axis 1; leaf P content and stomata were positively related to axis 2, and RGR had positive loading on axis 3 (Table 5). Q. xalapensis had the highest SD and foliar P content. H. donnellsmithii had high LA, SLA and foliar N. The clearest pattern was more separation of saplings and adults of species (H. donnellsmithii, M. coriacea and T. micrantha) in the right side of axis 1 (Euclidean distance, 141, 87, 136, respectively, mean = 121.8), and less separation between stages in species towards the central part (C. tropicalis; J. pyriformis, L. styraciflua) and left side of axis 1 (Q. germana and Q. xalapensis) (Euclidean distance, 27, 38, 40, 46, 42, respectively; mean = 38.6) within the ordination space.

Discussion

Traits are expected to influence growth rates, depending on plant size from seedling to sapling to adult (^{Martínez-Garza & Howe 2005}, ^{Martínez-Garza et al. 2005}, ^{Gibert et al. 2016}). Overall, we found that one foliar trait (SLA) differed across plant development stages while others, such as LA, SD and foliar nutrient content, were similar in saplings and adults. In addition, we found that LA, SLA, SD and foliar nutrient content differed among species, coinciding with observations about the broad spread of trait values across species within a site (^{Westoby & Wright 2006}). Relationships among LA, SLA and N content have been previously reported in tropical montane and lowland forests (^{Wright et al. 2004}, ²⁰⁰⁷, ^{Lohbeck et al. 2013}, ^{Flores et al. 2014}). However, the use of at least one trait (SLA or its inverse, specific leaf mass, SLM), characterized by intraspecific variability and correlated with other traits, has proved to be relevant for restoration (^{Martínez-Garza et al. 2005}).

As expected, SLA was higher in the saplings since it is a trait related to ontogeny (^{Martínez-Garza et al. 2005}, ^{Janse-ten Klooster et al. 2007}, ^{Spasojevic et al. 2014}), and SLA is also the most responsive attribute to different light environments in cloud forest seedlings and saplings (^{Toledo-Aceves et al. 2017}). Even though stomatal density can be affected by the availability of water and light, as well as by temperature (^{Loranger & Shipley 2010}), and several studies have reported higher SD in leaves exposed to the sun than in shade leaves (^{Popma et al. 1992}, ^{Loranger & Shipley 2010}) we found that, overall SD was similar between saplings and adults. In this study, L. styraciflua and Q. germana had a higher SD in saplings, but T. micrantha showed the opposite trend. Although there were no differences in LA, SD and foliar nutrient content between stages, we found a stage × trait effect in some species. This variation suggests that species respond individually displaying contrary trends and then the overall difference is not detected when all species were considered together.

While plant development stage was clearly related to SLA only, there is a consistency in the trait trends for groups of species. Considering the leaf economic spectrum (^{Wright et al. 2004}), the wood economic spectrum (^{Chave et al. 2009}) and RGR (^{Rüger et al. 2012}, ^{Gibert et al. 2016}), species can be categorized into different plant strategies. The SLA and WD are expected to have a negative relationship, reflecting the continuum from fast growing, light-demanding (high SLA, low WD, higher respiration rates and higher rate of nutrient uptake) to slow growing, shade-tolerant (high WD, low SLA) species (^{Wright et al. 2007}, ^{Chave et al. 2009}, ^{Rüger et al. 2012}). This trend has been observed in tropical forests, where species with high SLA have a high N content per unit leaf area, high assimilation and high RGR (^{Bongers & Popma 1990}, ^{Poorter et al. 2008}, ^{Poorter & Bongers 2006}, ^{Gustafsson et al. 2016}). Another general relationship is that SD decreases with increasing SLA, and SLA was higher and SD lower for shade leaves than sun leaves (^{Bongers & Popma 1990}, ^{Loranger & Shipley 2010}). The PCA showed that the intraspecific difference due to stage was small in comparison to the difference across species, since saplings and adults remained close within the ordination space.

We found that species with the lowest WD had the highest LA (H. donnellsmithii), and intermediate WD species included species with the smallest as well as intermediate LA. The group of species with relatively high WD included the oaks (Q. germana, Q. xalapensis). As expected, the non-pioneer forest tree species with high WD tend to have low RGR (^{Jense-ten Klooster et al. 2007}, ^{Chave et al. 2009}, ^{Muñiz-Castro et al. 2015}). Wood density has been reported as being related to leaf size, which decreases with increasing WD (^{Wright et al. 2007}). WD tends to be lower in shade intolerant than in shade tolerant species from the same habitat (^{Lawton 1984}, ^{Poorter et al. 2008}, ^{Chave et al. 2009}). In temperate forest, both SLA and WD correlated strongly and positively with shade tolerance, and WD negatively with extension growth (^{Jense-ten Klooster et al. 2007}). The pioneer species (H. donnellsmithii, T. micrantha, M. coriacea) with high RGR and low-intermediate WD had high SLA and leaf N content. The RGR of these species also changed with age. Adult trees were growing in middle-age plantations where the microenvironmental conditions more closely resemble those of the forest whereas saplings were growing in a more open environment. It has been reported that species with low WD grew fast, were able to respond to periods of higher light availability such as recently established plantations and their growth rates declined as they got bigger (^{Rüger et al. 2012}). WD has been negatively correlated with RGR across all plant sizes, as well as during a forest recovery trajectory, and changes in the light and water microenvironmental conditions may explain the switch in growth rates (^{Gibert et al. 2016}).

Finally, future research should determine how widespread intraspecific ontogenetic trait variation might be using more focal species across several restoration sites of different age. More important to restoration projects may be deciding which functional traits are relevant to relate interspecific and intraspecific variation to environmental changes during a forest restoration trajectory. One caveat that should be stated here is that this study is based on a limited number of tree species. Clearly, further work is needed to extrapolate any conclusion to other species in other areas. Our results partially supported the initial prediction about differences in traits between saplings and adults of the same tree species because sometimes the ontogenetic difference is fulfilled (e.g., SLA), although not for other traits. Overall, trait variability was higher among species than intraspecifically. Our study suggests that the variation between saplings and adults for most traits is so small that species mean trait values measured in individuals of any age, could be a useful tool to characterize group of species during the forest restoration trajectory, regardless of the ontogenetic variation.