Tropical dry forests (TDFs) are among the most threatened ecosystems of the world, due in large part to human settlement and transformation of these areas into agricultural lands (Murphy and Lugo, 1986; Janzen, 1988; Mooney et al., 1995; Fajardo et al., 2005, 2013). Although TDFs may grow slower than wet forests, after disturbance, they can recover their mature structure more rapidly than the latter with a more complex structure (Vieira and Scariot, 2006). For example, Lebrija-Trejos et al. (2008) found that canopy height, plant diversity and crown cover stabilized in less than 15 years in a very dry forest in Mexico. Similarly, 75% of tree species richness, canopy cover and basal area in mature forests were reached by year 5 after abandonment of slash and burn agriculture in a TDF in Bolivia (Kennard, 2002). However, recent work has shown that old growth forests, with grazing, and a 20–30 year secondary forest, with occasional grazing, differed a lot in terms of species density and accumulation of aboveground biomass in TDFs in northwestern Mexico (Álvarez-Yépiz et al., 2008), and that continued disturbance (grazing) can slow down recovery (Quisehuatl-Medina et al., 2020). Comparing tropical forests with different precipitation regimes, forests take longer to recover in dry than in wet areas, thus water seem to be one of the main abiotic factor affecting recovery of TDFs (Fajardo et al., 2013). How fast TDFs recover from disturbance will depend on the initial species pool and their functional traits. Therefore, studying functional traits of species in different successional stages along a successional gradient can be useful to understand how TDFs recover after disturbance.

Plant species from different successional groups are usually different in traits related to the acquisition, use and conservation of resources (Bazzaz and Pickett, 1980; Chazdon et al., 1996). For example, in a seasonally dry moist forest, functional group (pioneer, intermediate and shade-tolerant species) explained in average 41% of the variation in photosynthetic traits of 21 tree species, indicating that these traits can be used to differentiate plant functional groups (Ellis et al., 2000). Similarly, stomatal conductance (g_s) is lower and water-use efficiency (WUE) higher in mature TDF species than in pioneer (Fetene and Feleke, 2001; Sobrado, 2003) and invasive (Cordell et al., 2001) species. Furthermore, acquisitive species are known to dominate young plots in a succession of TDF in the Yucatan Peninsula (Sanaphre-Villanueva et al., 2022). However, recent work shows that variation within successional groups can be larger than among groups, creating a continuum of strategies (van Kuijk and Anten, 2009; Pineda-García et al., 2015; Goldstein et al., 2016; Vieira et al., 2021) and highlighting the need to study functional traits of individual species rather than generalize from so-called functional groups.

Resource-use efficiency is key for plants in disturbed habitats, as these habitats are usually limited in one or more resources (Funk and Vitousek, 2007; Funk, 2013). The stable carbon isotope composition of leaves (δ¹³C) can be used to quantify certain aspects of functional diversity in plant species (Santiago et al., 2005). In a TDF in Venezuela, deciduous species have higher δ¹³C and lower intercellular to ambient CO₂ concentration ratio (C_i/C_a), therefore predicting a higher WUE than in evergreen species (Sobrado and Ehleringer, 1997). Although δ¹³C has been studied in pioneer and late successional species of tropical rainforests (Huc et al., 1994; Bonal et al., 2007), less information is available for succession in TDFs. Previous work studying 11 evergreen and deciduous species in a TDF in Venezuela found that evergreen and deciduous plants have leaf traits that are not consistent with their expected leaf life span, with higher nitrogen (N) investment in evergreen leaves despite their lower net photosynthetic rate (A) and lower specific leaf area (SLA), and greater water-conservative traits (high δ¹³C) in deciduous plants, suggesting that this ecosystem may be more tolerant to drought than previously thought (Ávila-Lovera et al., 2019).

Many TDFs in Venezuela have one continuous arboreal stratum, but below 700 mm of precipitation, as in the dry woodland of Macanao Peninsula, Margarita Island, Venezuela, the canopy becomes discontinuous and members of arid habitats such as cacti become important members of the community defined as “dry deciduous woodland” (Fajardo et al., 2005). Previous work has shown how supplying water to seedlings in this disturbed TDF greatly increases the seedlings’ growth and survival, supporting the hypothesis that water limitation is a key factor for TDF regeneration (Fajardo et al., 2013). In this region, opencast sand mining has eliminated the topsoil of an extensive area and has generated patches of natural succession according to time since abandonment of the activity. In this study, we set out to answer the question: how do functional traits change during secondary succession in a tropical dry woodland? We hypothesize that shrub and tree species from early successional stages have trait values on the fast end of the leaf economic spectrum, e.g., high A_mass, A, SLA, N and low WUE (acquisitive traits), whereas species from later successional stages have trait values of the slow end of the leaf economic spectrum, e.g., low A_mass, A, SLA, N and high WUE (conservative traits). This hypothesis is based on the knowledge that the physiognomy of the successional stages in this TDF changes from herbaceous to shrubland, scrub and/or forest as the succession advances (Fajardo, 2007), suggesting that traits would also change along the succession, from more acquisitive traits in herbaceous plants and shrubs, to more conservative traits in trees. We also tested whether there is a seasonal response of functional traits regardless of the successional stage, especially plant-water relations and resource-use efficiency traits, as water availability decreases from the wet to the dry season and probably decreasing the nutrient availability as well. Furthermore, we tested whether there is coordination among traits of the leaf economic spectrum that differ across successional stages. Testing these hypotheses will provide valuable information for species selection in ecological restoration, especially in the understudied tropical dry forests of Venezuela (Rodríguez et al., 2010).

Results from t-tests showed that water potential differed between seasons for all but one species in the three successional stages (Figure 2A), with higher values usually found in the wet than in the dry season, with the exception of M. arenosa in the 20 years stage that did not change its Ψ, and an unusual increase in Ψ in P. carthagenensis from the wet to dry season in the DW. ANOVA results showed a significant effect of successional stage on Ψ (Table 2), with wet season Ψ being similar across sites, but dry season Ψ differing across stages, being higher in the 20 years than in the 6 years and DW stages (Figure 2A and Table 2). There was also an interaction effect of successional stage × season in Ψ (Table 2). Interestingly, RWC was not affected by successional stage, but there was an effect of season (Table 2) and an interaction effect of successional stage × season on RWC, with differences between seasons in only a few species (Figure 2B and Table 2).

T-tests showed that A declined from wet to dry seasons in seven out of 10 species (Figure 3A). The ANOVA also showed a significant effect of season and successional stage on A (Figure 2A and Table 2), with no interaction between stage and season. During the wet season, species of the 6 years stage showed the highest values of A, whilst the lowest occurred in species of the DW; furthermore, the reduction in A from wet to dry season was c. 50% in both the 6 and 20 years stages, but it was only 25% in DW (Figure 3A). Values of A in the dry season were similar across stages (Figure 3A). For g_s, we found that only five species changed their g_s between seasons, with four decreasing g_s during the dry season, and one increasing it (t-test results, Figure 3B). The ANOVA revealed an interaction effect of stage and season, with g_s decreasing from wet to dry season only in the 6 years stage (Figure 3B and Table 2). Wet season g_s decreased from the 6 years to the DW, with no difference in dry season g_s among stages (Figure 3B).

T-tests also showed that WUE significantly decreased from wet to dry season in eight out of 10 species (Figure 4A). This decrease was also observed in the ANOVA, with a significant effect of season on WUE (Table 2). There was also a significant difference in WUE among successional stages (Table 2), with the 6 years stage and DW being similar among themselves, but different from the 20 years stage. Values of IWUE changed between seasons in four species (t-test results, Figure 4B). However, the ANOVA showed no significant overall effect of season (Figure 4B and Table 2), but differences among stages (Table 2), with the greatest IWUE in the DW.

Results from t-tests showed that SLA changed little between seasons, except for three species that decreased their SLA from wet to dry season (Table 3). However, the ANOVA showed a significant effect of season and no effect of successional stage (Table 2) on SLA. Values of δ¹³C changed between seasons in five out of 10 species, with values becoming higher (less negative) in the dry season (t-tests, Table 3). ANOVA showed a significant effect of season on δ¹³C (Table 2), but there was no effect of successional stage on δ¹³C (Table 2). Leaf N concentration differed between seasons in almost all species (t-tests, Table 3), with lower values in the dry than in the wet season. The ANOVA also showed an effect of season and stage on N, with greater N values in the 6 and 20 years stages (Table 2). In contrast, PNUE only differed between seasons in a few species (t-tests, Table 3). The ANOVA also showed no significant effect of season, but successional stage had a significant effect on PNUE (Table 2).

When looking at coordination among traits, we found a positive relationship between N concentration and A_mass, but only in the DW (r² = 0.86, p = 0.008, Figure 5). Values of A_mass were similar in the 6 years and DW stage in the wet season and slightly higher than those of the 20 years stage (p = 0.420, data not shown), and they decreased in all stages during the dry season (p = 2 × 10^–5, data not shown). Values of δ¹³C were positively related to WUE in the dry season only (r² = 0.60, p = 0.008, Figure 6), with species from the 6 years stage having the greater values in both traits, indicating that these species are more water-use efficient.

Results from the first PCA on the species mean trait values using both wet and dry season data, show that the main axis of variation explained 41.8% of the data, while the second axis explained 14.3% (Figure 7A). The first axis (PC1) was positively associated with δ¹³C and negatively associated with all other traits except for IWUE. The second axis (PC2) was positively associated with IWUE. This PCA does not show a clear functional separation of the species in the three successional stages (Figure 7A). To test for different trait coordination in the two seasons, we performed two separate PCAs. The PCA using the wet season data had a PC1 that explained 26.8% of the wet season variation and showed coordination between RWC, WUE and IWUE, and these traits traded-off with A_mass, g_s and SLA (Figure 7B). The PCA using the dry season data had a PC1 that explained 31.8% of the dry season variation and showed coordination between A, A_mass, δ¹³C, WUE, and PNUE, and these traits traded-off with Ψ (Figure 7C).

We studied the water relations, gas exchange, resource-use efficiency, specific leaf area, carbon isotope composition and nitrogen concentration of seven shrub and tree species from three sites at different stages of succession in a tropical dry forest to understand how functional traits change as succession advances. We found differences among stages in traits, such as Ψ, A, g_s, WUE, IWUE, N and PNUE, and a rather strong effect of seasonality in all traits studied except for IWUE and PNUE.

We found partial support for our hypothesis that as succession advances, leaf traits become more conservative, as only A, g_s, and N decreased from the 6 years to the DW stage, with no clear trend in A_mass, WUE, and PNUE. Photosynthetic rate has been found to decrease from early to late successional stages and this is usually associated with a change in the growth form of plants as succession advances (Zhu et al., 2013; Han et al., 2020). This can partially explain our results, as shrubs are only present in the 6 and 20 years stage and not in the DW. However, the decrease in A along the succession was not accompanied by a decrease in SLA; indeed, SLA increased as succession advances. A similar increased in SLA has been found in another TDF in Guanacaste, Costa Rica (Buzzard et al., 2016). This indicates that leaves become thinner and/or less dense as succession advances, but these changes do not translate into greater photosynthetic capacity as expected from the leaf economic spectrum (Reich et al., 1997; Wright et al., 2004). However, the decrease in A as succession advances can be explained by both reduced g_s and N concentration. Species in the 6 years stage significantly decreased g_s during the dry season but with a smaller reduction in N concentration, which indicates a higher stomatal control of A, also evidenced by the higher δ¹³C values measured in these species.

Furthermore, following the expectations of the leaf economic spectrum, A_mass was positively related to N concentration, although only in the DW species. Values of A_mass represent the rate of carbon capture per unit of biomass invested in leaves, thus providing an index of photosynthetic efficiency (Ávila-Lovera et al., 2019). In our study, species from the 6 years and DW seem to be equally efficient in the C return per unit of C invested during both wet and dry season, as they have similar A_mass, while species from the 20 years stage have lower A_mass, especially in the wet season, compared to the other two stages.

Despite a decrease in N concentration as succession advances, PNUE decreased from the 6 years to the 20 years stages and then increased from the 20 years to the DW, which contrasts with the expectation that as N decreases, PNUE should increase. A study from a subtropical moist forest found that PNUE decreased as succession advances (Han et al., 2020), whereas another study from a tropical wet forest suggested that PNUE increases as succession advances (van Kuijk and Anten, 2009). In a study in Central Amazonia, PNUE was higher in pioneer species than in early and late secondary species, which was related to the pioneers’ greater photosynthetic capacity, rather than lower N availability (dos Santos et al., 2006). The values of N concentration in our study are relatively high, and can explain the relatively low PNUE we found overall compared to those found in other TDFs with lower leaf N concentration (Chaturvedi et al., 2021).

We found strong support for our expectation that functional traits, especially water relation traits and resource-use efficiency traits, change seasonally. The maintenance of RWC in the dry season in both the 6 and 20 years stages despite of a strong reduction in Ψ in the 6 years stage and a moderate reduction in the 20 years stage indicates that the change in Ψ may be due to osmotic adjustment, as turgor could have been kept at similar values between seasons and accumulation of solutes inside the cells would lead to a decreased osmotic potential and Ψ (Sanders and Arndt, 2012). Maintenance of turgor and hence RWC in the dry season was likely a result of stomatal closure, as g_s decreased in the 6 years stage. Early stomatal closure as water availability decreases is an important strategy in plant species from seasonally dry habitats (Tezara et al., 2010) and tropical forests in general (Santiago et al., 2016). At the same time, possible osmotic adjustment leading to decreased Ψ in the dry season allows for a water potential gradient between the soil and the leaf to be maintained in this period of water deficit. Given the similar Ψ values among species in the three stages along with similar responses to water deficit, species from the three successional stages seem to be functionally equivalent in this dry deciduous woodland.

Although no clear trend in WUE was observed among stages, species from the DW were more efficient in the use of water in both the wet and dry season according to values of IWUE. An increase in canopy cover during succession could have played a role in this response, as decreased air evaporative demand (low air vapor pressure deficit, VPD) would affect g_s and IWUE. Other studies have found a decreased in wet season VPD as succession advances in a TDF in Oaxaca, Mexico, and this has been associated with larger basal area (Lebrija-Trejos et al., 2010, 2011), and likely large canopy cover. The effect of canopy characteristics on gas exchange of species in different successional stages has also been highlighted in a seasonal evergreen forest in Nakhon Ratchasima, Thailand (Tor-ngern et al., 2021). High IWUE is a desirable trait for maintaining good carbon and water balances in water-limited environments and differences in this ratio suggest variation in stomatal limitation to carbon assimilation (Sobrado, 1991) in the species from the different successional stages.

We found significant relationships among some important traits supported by both the regression analyses and PCA, supporting our hypothesis of a coordination of traits across successional stages. Results from the PCA show strong coordination among most traits studied, the ones aligning with PC1, except for IWUE, which seem to be independent from the other traits (it strongly loaded with PC2). However, species mean trait values did not allow for a functional separation of the species based on the successional stages. Indeed, we found that species from the three successional stages share a big part of their functional trait space, with species from 6 and 20 years being the most similar. These results further support the idea that greater functional variation within than among successional groups is common in tropical forests (van Kuijk and Anten, 2009; Goldstein et al., 2016; Vieira et al., 2021). The separate PCAs for each of the seasons showed how traits and trait coordination changed from wet to dry season. For example, in the wet season species with high values of A_mass, g_s and SLA also had lower values of WUE and IWUE, indicating a trade-off between carbon assimilation and water-use efficiency. In the dry season, however, species with high values of A_mass also had high WUE and PNUE, indicating no trade-off between carbon assimilation and resource-use efficiency. Furthermore, species with high A_mass, WUE, and PNUE also had high values of δ¹³C and low values of Ψ indicating that the species with greater drought tolerance could maintain photosynthetic efficiency and resource-use efficiency during the period of water limitation.

Taken together our results indicate that, contrary to expectation, species along a successional gradient in a tropical dry forest in Venezuela do not necessarily become more resource-conservative as succession advances. This could reflect the low soil nutrient concentrations early in the succession (Fajardo, 2007), as opencast mining completely removes the topsoil and new plant species colonize a bare substrate. Nutrient and organic matter concentration are low in the three stages compared to the nearby forest (Fajardo, 2007), but they are lower in both the 6 years stage and DW than in the 20 years stage. Importantly, similar seasonal behavior in species from the different successional stages indicate that they may be functionally similar, as further supported by the PCA. Our results indicate that in the absence of specific trends (traits becoming more or less conservative as succession advances) in many of the traits evaluated in the different successional stages, the planting of native species from the late successional stage in the 6 and 20 years successional areas could help accelerate the succession and allow species to have a good physiological performance, and therefore help in the overall restoration of these degraded areas. We recognize that we only sampled one stand in each successional stage, and hence our conclusions may not represent what happens at the successional stage level, so we recommend further studies with a community ecology approach using a larger number of stands at each successional stage to determine whether the functional traits (mainly ecophysiological) approach used in the present study are conclusive.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

RU secured the funding. EÁ-L, RU, IC, and WT designed and conducted research. EÁ-L and WT performed the statistical analyses, made the plots, and wrote the manuscript with significant input from all other authors. All authors contributed to the article and approved the submitted version.