Quercus oleoides is an interesting study organism for several reasons. First, is widely distributed across a wide range of precipitation regimes (Cavender-Bares et al., 2015). In fact, it spans the largest gradient of dry-season aridity and wet-season rainfall within live oaks (Quercus section Virentes, Cavender-Bares et al., 2015) and encompasses one of the highest precipitation ranges within the distribution of the American oaks (Hipp et al, personal communication). Second, differences among populations have been previously reported for growth, photoprotective pigments and leaf traits in seedlings grown under controlled greenhouse and growth chamber conditions in response to cold, light and water stress (Cavender-Bares et al., 2011, 2015; Koehler et al., 2012; Ramírez-Valiente et al., 2015b; Ramírez-Valiente and Cavender-Bares, 2017). Intraspecific differences in germination (Center et al., 2016) and survival (Deacon and Cavender-Bares, 2015) have also been observed in common gardens in the field.

In 2010, we collected acorns from 19 populations belonging to five climatically contrasting regions where Quercus oleoides is present (Belize, Honduras, Mexico, dry Costa Rica and mesic Costa Rica). Seeds were stored at 4°C until sown in shadehouses. Eighty one maternal families from eight populations (between 2 and 35 maternal families per population) were finally available for the experiment due to low germination rates (Table 1). After 6 months of growth, seedlings were transferred to a common garden trial in the field in January 2011 within Zamorano University campus (14° 00′ N, 87° 00′ W, 797m, Francisco Morazán, Honduras). This area is characterized by a seasonally-dry tropical climate with mostly constant temperatures throughout the year but a marked seasonality in precipitation (Figure 1). It is located in the driest region of the entire Q. oleoides range with up to 6 months of length for the dry season (Figure 1, see Cavender-Bares et al., 2015 for more information). The climate of this area is also characterized by a little dry season in the middle of the wet season locally known as “veranillo.” This is a 2-month period (July and August) during which precipitation may be below potential evapotranspiration although its duration and intensity is highly variable across years (Cavender-Bares et al., 2015).

The common garden experiment followed a randomized block design with 24 blocks. Because of low germination rates, maternal families were not evenly represented across blocks and the resulting design was unbalanced. Seedlings were planted 50 cm apart, which minimized competition and shading. Four watering treatments (applied to six blocks each) were implemented 2 weeks after transplanting: (1) Ambient rainfall (no water added, hereafter, AMB); (2) Water added in the amount of 18 mm/week during the dry season (hereafter referred to as WDS), (3) Water added up to 25 mm/week during the wet season if it rained less than 25 mm/week (hereafter referred to as WWS), (4) Water added during both the dry and wet seasons following WDS and WWS protocols (hereafter BOTH). These treatments are representative of the natural precipitation gradient where Q. oleoides can be found. Blocks were weeded biweekly to reduce competition effects.

In May 2013, during the onset of the wet season (Figure 1), initial growth measurements were taken on all plants in the common garden (N = 634 plants). Stem height, basal diameter and number of leaves were recorded. The same process was repeated in November 2013 and March–April 2014 (Figure 1). Growth that occurred between May 2013 and November 2013 represents wet season growth, and growth between November 2013 and April 2014 represents growth during the subsequent dry season (Figure 1). Wet and dry season growth was estimated using three separate metrics: height, basal diameter and number of leaves. Metrics of growth were transformed into total dry biomass values using an allometric equation: Dry mass (g) = 0.5018x + 0.1540y + 0.3209z, R² = 0.93, P < 0.0001, where x is the height (cm), y is the diameter (mm) and z is the number of leaves. The allometric equation was constructed using 120 harvested plants up to 120 cm tall established in a greenhouse. A multiple regression was performed including total biomass as the dependent variable and height, diameter and number of leaves as dependent variables (See Cavender-Bares et al., 2004 for a similar procedure). Absolute growth rate (g day⁻¹) was calculated in each season as AGR = (M_final − M_initial)/(T_final − T_initial), where M_final is the estimated biomass at the end of the season, M_initial is the estimated biomass at the beginning of the season, T_final is the date of the measurement at the end of the season and T_initial is the date at the beginning of the season. Relative growth rate (g g⁻¹ day⁻¹) was calculated as RGR = [log_e(M_final) − log_e(M_initial)]/(T_final − T_initial).

In November 2013 (wet season) and April 2014 (dry season), one mature but recently expanded sun-exposed leaf was selected from each plant for gas exchange measurements. Gas exchange measurements were taken with LI-COR 6400 (LI-COR, Lincoln, NE, USA) from 8:30 a.m. to 12:00 p.m. in ten and nine consecutive sunny days in the wet and dry seasons, respectively. Photosynthetic rate (A) and stomatal conductance (g_s) were measured at ambient atmospheric [CO₂] ~400 μL L⁻¹ and relative humidity (~50%) with saturating light controlled at 1,500 μmol m⁻² s⁻¹ (10% blue light), block temperature set at 25°C, with vapor pressure deficit (VPD) varying between 1.51 and 3.39 in the chamber. Intrinsic water use efficiency (WUEi) was calculated as A/g_s. Leaves were collected and pressed in labeled coin envelopes. The petiole was removed and fresh leaves were scanned to obtain lamina area and then dried at 65°C. Specific leaf area was calculated as leaf area/dry weight and used to estimate photosynthetic rate and stomatal conductance on a mass basis. In a sub-sample of 7–9 randomly selected plants per block, we measured water potential at predawn (ψ_pd) using a Scholander pressure chamber (Soil Moisture Equipment Corp., Santa Barbara, CA, USA) in November 2013 (wet season) and April 2014 (dry season) at the same dates as gas exchange measurements.

In April 2014 (dry season), we measured three additional traits in the selected leaves: leaf lamina thickness, Nitrogen concentration (N_mass) and stable carbon isotope composition (δ¹³C). δ¹³C is a surrogate of integrated water use efficiency during the time for which the plant has fixed carbon (Farquhar et al., 1982; Farquhar and Richards, 1984). Leaf lamina thickness was measured in fresh leaves using a micrometer. Once leaves were dried, nitrogen concentration (N_mass) and stable carbon isotope composition (δ¹³C) were estimated using a mass spectrometer (Cornell Stable Isotope Laboratory, NY, USA).

Mixed models included population, season, watering treatment, and their interactions as fixed factors. Maternal family nested within population, and block nested within watering treatment were included as random factors. Initial plant biomass was included as a covariate. Three-way interactions were not significant for any trait and removed from final models (data not shown). Since A and g_s are non-linearly related, WUEi (A/g_s) might vary intrinsically with g_s. Prior to performing mixed models for WUEi, we tested the effect of g_s on WUEi and the interaction between fixed factors and g_s on WUEi; we plotted g_s vs. A for each combination of region/treatment/season and fitted the data to hyperboles (Figure S1). Then, we calculated WUEi at three fixed g_s values (0.1, 0.2, 0.3) that covered most of the range of variation of g_s in our dataset. We used region rather than population for this test because the small number of plants within some populations did not allow us to fit hyperbolic functions accurately. Our results did not reveal any interaction between g_s and region, season, or treatment on WUEi (Table S1), indicating that the effect of g_s on WUEi did not differ among levels of the three fixed factors. Consequently, individual values of WUEi were used for subsequent analyses.

In order to test whether growth and functional traits were associated to climate of origin; when differences among populations were found, we performed linear regressions between an index of moisture and traits. The index of moisture (I_m) was calculated as I_m = ∑(P_i − PET_i), where P_i is the monthly precipitation and PET_i monthly potential evapotranspiration (see Ramírez-Valiente and Cavender-Bares, 2017 for more details).

We also tested associations between plant performance and functional traits by analyzing trait-growth relationships using semi-partial correlations. This was done by performing Pearson correlations between the residuals of AGR and RGR obtained from mixed models and individual trait values. Relationships between traits and survival were not explored because only 14 plants died during the studied period (all of them in the wet season). Traits measured in both seasons that showed significant relationships with growth rates were used for further analyses. We then conducted two mixed-models including AGR or RGR as dependent variables and individual traits as covariates. Population, season, and treatment were included as fixed factors. Maternal family nested within population, and block nested within watering treatment were included as random factors. Initial plant biomass was included as a covariate. In order to avoid colinearity, we excluded mass-based photosynthesis and stomatal conductance because of their high correlation with area-based measurements. Results were very similar when mass-based rather than area-based measurements were included (data not shown). We also included the interactions of season by “trait” that would indicate a significant difference in the relationship between the trait and growth across seasons. Interactions of treatment by “trait” were not included in these analyses due to the absence of treatment effect for leaf traits. Interactions of populations by “trait” were not significant for any trait and were removed from final models. We also explored relationships between traits by performing Pearson correlations. Statistical analyses were performed using packages lme4 and lmerTest in R 3.2.2 (R Core Team, 2015) and STATISTICA 10.0 (StatSoft, Inc., Tulsa, OK, USA). Sigmaplot 12.5 (Systat Software Inc., San Jose, CA, USA) was used to plot some figures.

All traits measured in both dry and wet seasons exhibited high phenotypic plasticity to seasonal precipitation (question i, Figures 2–4). As observed in other oak species from seasonally-dry regions, growth rates, gas exchange and SLA showed highly plastic responses to temporal variation in precipitation (Ramírez-Valiente et al., 2010, 2014, 2015a). These traits also showed plastic responses to watering treatment (measured mainly by treatment-by-season effect) but to a lesser extent than season itself (Figure S2). This pattern indicates that differences in water input to soils during the dry season elicits plastic trait and growth responses, despite high evaporative demand and high VPD values.

Most importantly, our study revealed remarkable differences among populations of Quercus oleoides in growth rates, leaf morphology, N_mass and g_s (question ii). Wet season growth rates were negatively associated with I_m in the source populations (Figure 3). In other words, populations from the more xeric sites had higher growth rates under favorable conditions in the wet season than the more mesic populations. These results are in agreement with comparative species-level studies in seasonally-dry tropical ecosystems, which tend to show that species from xeric regions (long and severe dry seasons) evolve toward a resource-acquisition strategy with increased growth potential during the favorable season (Cornelissen et al., 1996; Givnish, 2002; Bowman and Prior, 2005; Markesteijn and Poorter, 2009; Tomlinson et al., 2013). Interestingly, these results contrast those found for evergreen oak species from seasonally-dry temperate zones, which show that populations from xeric climates with long dry seasons have reduced growth potential (Gratani et al., 2003; Ramírez-Valiente et al., 2010, 2014; Niinemets, 2015). These differences are probably explained by the fact that photosynthetic activity is limited by both water deficit in summer and cold temperatures in winter in Mediterranean-type ecosystems (Larcher, 2000; Nardini et al., 2000; Cavender-Bares et al., 2005; Flexas et al., 2014; Granda et al., 2014; Niinemets, 2016). In contrast to Mediterranean ecosystems, which have cold winters, in seasonally-dry tropical ecosystems, temperature is not a limiting factor for photosynthesis in the wet season. In Mediterranean areas where the dry season is long and winters are cold, species tend to have conservative resource-use strategies (long leaf life spans, thick leaves, high density tissues and high water use efficiency) allowing them to reach a positive carbon balance over the course of the year. In contrast, in seasonally-dry tropical ecosystems where the dry season is long, a resource-acquisition strategy that maximizes growth in the warm wet season is expected to be beneficial (Reich et al., 1997; Givnish, 2002; Villar et al., 2005; Markesteijn and Poorter, 2009; Vico et al., 2015).

Leaf economics spectrum traits (SLA, leaf thickness, N_mass, A_mass, g_{s, mass}) also showed significant differences among populations (Figure 4). Consistent with a “fast” resource-acquisition strategy, which we hypothesized would evolve in more xeric populations with longer dry seasons, N_mass and SLA were negatively related to I_m while leaf thickness was positively related to I_m. Similar patterns have been observed for growth rates in species from SDTE (Markesteijn and Poorter, 2009; Tomlinson et al., 2013). However, the patterns of trait variation in relation to I_m were less clear than in our previous greenhouse experiment with five Q. oleoides populations from the southern distribution of the species range (Figure 4, Ramírez-Valiente and Cavender-Bares, 2017). One possible explanation for the weaker associations between trait variation and climate found in the present study may result from interaction with biotic factors, such as herbivory or pathogens in the field experiment. Alternatively, or in addition, contrasting abiotic environmental variables in the field compared to the greenhouse, such as higher light levels and less variable daylengths, more variable vapor pressure deficits, more extreme soil drought and moisture conditions, or the absence of pot restrictions could have triggered differences in the expression of leaf traits between environments. For example, Conner et al. (2003) observed an increased expression of the genetic variance (i.e., greater differences among genotypes) for floral traits of the wild radish Raphanus raphanistrum in a greenhouse compared to the field. They suggested that this observation could be caused by a few specific genotypes with different trait expression across environments (Roles and Conner, 2008). In our previous experiment, a mesic population from Costa Rica had the lowest values of SLA, the smallest leaves and the highest leaf thickness, contrary to the findings obtained in this study. In fact, after removing that population from analyses, the relationships between SLA, leaf thickness and N_mass became significant or marginally significant and similar to the patterns observed in the greenhouse (Figure 4). These findings suggest a steep reaction norm between the greenhouse and the field for this population (i.e., genotype by environment interaction) compared to the other studied populations. Interestingly, plants from the mesic higher elevation site in Costa Rica, exhibit neutral genetic differentiation relative to lowland (dry) populations of Q. oleoides from Costa Rica (Deacon and Cavender-Bares, 2015). Further research is needed to determine the mechanisms causing this contrasting expression of leaf traits between experiments, particularly for the Costa Rican populations, which have been genetically isolated from the rest of the species (Cavender-Bares et al., 2011, 2015).

Populations did not differ in either instantaneous (measured by gas exchange) or integrated (measured by δ¹³C) water use efficiency. These results are contrary to our expectations for a higher WUE as part of a resource-conservative strategy in populations from short dry seasons. One possibility to explain this unexpected result could be that Q. oleoides responds to water stress by using other traits rather than increasing WUE. In fact, populations with short dry seasons did have other traits associated with a resource-conservative strategy such as low SLA and high leaf thickness, which should reduce water loss and conserve water for the plant over longer periods (Dudley, 1996; Lamont et al., 2002; Etterson, 2004). Similarly, in our previous study, we observed that populations of Q. oleoides did not vary in WUE but did differ in water potential at the turgor loss point (π_tlp) under dry conditions (Ramírez-Valiente and Cavender-Bares, 2017). In other words, all populations have similar WUE but populations from short dry seasons had lower π_tlp (i.e., increased drought tolerance) in response to drought. On the other hand, it is worth mentioning that when a given species is the subject of study, genetically-based population differences are not always consistent across studies. For example, in Mediterranean evergreen oaks, Gimeno et al. (2009) and Ramírez-Valiente et al. (2011, 2014) did not find population-level variation in (instantaneous or integrated) WUE for Q. ilex or Q. suber. However, Gratani et al. (2003) and Ramírez-Valiente et al. (2010) did find differences among populations in this trait in the same species. These findings suggest that intraspecific variation in this trait might depend on the environmental conditions or how the populations were subsampled. Consequently, the possibility of intraspecific variation in WUE for Q. oleoides should not be ruled out.

The present study showed strong relationships between photosynthetic rate per unit leaf area (A_area), specific leaf area (SLA), and growth rates (AGR or RGR) (question iii). Photosynthetic rate per unit leaf area (A_area) was positively associated with RGR in both wet and dry seasons (Table 4). A number of studies have shown that the maximum photosynthetic rate is positively associated with net assimilation rate (NAR), which represents the whole-plant carbon increase per unit leaf area and time (Poorter and Van der Werf, 1998). Although NAR is a balance between carbon gain in photosynthesis and carbon loss in respiration throughout a given period of time, light-saturated photosynthetic rate is a good predictor of NAR (Li et al., 2016). Traditionally, NAR has been considered to be only weakly related to RGR compared to other traits such as leaf mass ratio and specific leaf area or their product Leaf Area Ratio (Poorter and Remkes, 1990). Nevertheless, more recent studies have reported that net assimilation rates may be a primary determinant of RGR for specific functional groups, species or environments (Shipley, 2002, 2006; Villar et al., 2005). In oaks, A_area has been previously observed to be positively related to relative growth rate in temperate, Mediterranean and seasonally-dry tropical ecosystems (Cavender-Bares and Bazzaz, 2004; Castro-Díez et al., 2006; Ramírez-Valiente and Cavender-Bares, 2017). These observations suggest that on average, Q. oleoides seedlings with higher instantaneous carbon assimilation rates can sustain higher growth rates over longer periods of time particularly under favorable conditions of water.

Regardless of the season, specific leaf area (SLA) was negatively associated with RGR (Table 4, Table S2). These results are contrary to our hypothesis that LES traits would be positively associated with growth rates under favorable conditions. The growth benefits of low SLA have been usually explained in terms of reductions of water loss or higher water use efficiency in more sclerophyllous leaves (i.e., low SLA, high thickness) that allow higher potential to grow over longer periods of time when water is limiting during the growing season (Dudley, 1996; Lamont et al., 2002). Studies conducted with Mediterranean evergreen oaks agree with this previous evidence and have extensively reported negative associations between SLA and growth, particularly under severe drought conditions (Ramírez-Valiente et al., 2010, 2011, 2014, 2015a). Under well-watered conditions, positive relationships between SLA and growth rates are generally expected (Reich et al., 1997; Poorter and Bongers, 2006; Wright et al., 2010) although, studies with evergreen and deciduous oaks have also previously observed negative SLA-growth associations under well-watered conditions (e.g., Quero et al., 2008).

The unexpected negative SLA-growth relationship in Quercus oleoides may be mediated, at least in part, by the negative association observed between SLA and A_area (Table 5). In both seasons, SLA and A_area were negatively correlated, and A_area was positively correlated with RGR. The negative association between SLA and A_area contrasts expectations based on the worldwide leaf economic spectrum (LES) across species at the global scale (Reich et al., 1997; Wright et al., 2004), which shows that SLA is positively associated with mass-based photosynthesic rates (A_mass) but not to area-based values (A_area). In our study, SLA was positively associated with A_mass but more weakly than at the global scale (r = 0.44 in the wet season, r = 0.31 in the dry season, vs. r = 0.71 for Wright et al., 2004), whereas SLA was negatively related to A_area (r = −0.23 in the wet season and r = −0.14 in the dry season, vs. r = −0.05 for Wright et al., 2004). Recent studies have shown that global relationships between traits might not be applicable for specific group of species, biomes or environmental conditions (e.g., Edwards et al., 2014; Mason and Donovan, 2015; Grubb, 2016).

Interestingly, similar results have recently been found in other evergreen oaks. In particular, Niinemets (2015) observed a negative SLA-A_area relationship for Quercus ilex (r ≈ −0.40). He suggested that the negative SLA-A_area relationship could occur at the lower return end of the global LES spectrum due to increased mesophyll thickness. We measured leaf lamina thickness in the dry season and observed a strong negative relationship with SLA and positive relationships with A_area, RGR, and AGR (Table 4). Similar results for SLA-thickness-A_area-RGR relationships were found in our previous (greenhouse) study under both well-watered and dry treatments (Ramírez-Valiente and Cavender-Bares, 2017). Since leaf lamina thickness and mesophyll thickness are highly associated in oaks (e.g., Peguero-Pina et al., 2017), our results would support Niinemets (2015) hypothesis. “Multiple-trait” mixed models showed a significant effect of SLA on RGR even after taking A_area into account. The relationship between SLA and AGR, however, was weaker under dry conditions (significant season by SLA interaction in the AGR model), which agrees with the population-level trend showing a negative association between SLA and the index of moisture. However, the low R-coefficients observed in semi-partial correlations and the relatively weak population variation in SLA suggest that other unmeasured traits such as osmotic adjustment or biomass allocation patterns (Ramírez-Valiente and Cavender-Bares, 2017) might play a more important ecological role in Q. oleoides. In addition, survival during the earlier stages of the life cycle is known to be an important component of fitness in this species (Deacon and Cavender-Bares, 2015; Center et al., 2016), and future research needs to consider integrated measures of fitness (e.g., Shaw et al., 2008) to better understand trait-fitness relationships in Quercus oleoides seedlings.

In summary, our study revealed that xeric populations of tropical live oak exhibited higher growth rates (RGR and AGR) in the wet season but similar growth rates in the dry season. These findings are consistent with multi-species studies in seasonally-dry tropical ecosystems, which show that when dry seasons are long and wet seasons are warm species evolve toward resource-acquisitive strategies. Such species tend to have high carbon assimilation rates and fast growth rates to take advantage of favorable water availability conditions during the wet season. Our results also showed that Quercus oleoides individuals with leaves that have lower SLA and higher area-based photosynthetic rates had higher growth rates, particularly under well-watered conditions. This observation was partially explained by the negative association between specific leaf area (SLA) and area-based photosynthetic rates (A_area), which is contrary LES expectations but similar to other recent intraspecific studies on evergreen oaks. Overall, our study shows the existence of important intraspecific genetic variation in leaf economics spectrum traits and growth rates in a long-lived tropical tree. These findings are critical to understanding how this keystone species will respond to climate change in the near future (Hällfors et al., 2016).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.