Introduction

Several global climate change models (GCMs) project significant changes in temperature and rainfall for the next 85 years in many tropical regions (Intergovernmental Panel on Climate Change IPCC 2015). The seasonally dry tropical forest (SDTF) that occurs in the semi-arid region of northeastern of Brazil is especially vulnerable to climate change. Specifically, future climate projections predict a substantial increase in air temperature and a decrease in annual rainfall due to a higher frequency of consecutive dry days (Marengo et al. 2009). The combined effects of higher temperatures and lower rainfall in this region will lead to an ‘aridification’, causing an expansion of semi-deserts in the driest areas (Salazar et al. 2007; Marengo and Bernasconi 2015). However, how species of this region will respond to these climate changes and which species are sensitive or resistant to drier and hotter climatic conditions is still unknown.

The SDTF are mostly deciduous ecosystems with global relevance, because they occupy a large area in the tropics and play an important role on Earth’s carbon cycle (Alström et al. 2015). Additionally, they are much more threatened by deforestation than rainforests (Olivares and Medina 1992). Global models of biogeography, vegetation, and carbon balance still consider deciduous and evergreen species as the two main functional categories for this vegetation (see Bonan et al. 2003; Reich et al. 2007). This has major implications for modeling ecosystem carbon and water balance because deciduous species have much shorter windows of opportunity for carbon gain and water loss than evergreen species. Deciduous species in some STDF exhibit higher gas exchange capacity, assimilation rate, and water use compared to evergreen species (Goldstein et al. 1989; Mediavilla and Escudero 2003; Miyazawa and Kikuzawa 2005; Fu et al. 2012; Worbes et al. 2013), but this pattern has not been consistently observed in all tropical deciduous ecosystems (Damesin et al. 1998; Bucci et al. 2005; Hasselquist et al. 2010). Additionally, the simple categorization in evergreen versus deciduous habit gives little insight about the whole range of existing ecological strategies in seasonally dry tropical ecosystems (Holbrook et al. 1995). Therefore, a more comprehensive dataset linking leaf phenology to ecophysiology and hydraulic strategies for these systems is needed to improve classification of plant functional types (PFTs) in land surface models (Bonan et al. 2003; Atkin et al. 2015).

The functional differences between deciduous and evergreen species are usually related to the balance between the costs and benefits of leaf construction (Chabot and Hicks 1982; Sobrado 1991; Wright et al. 2004) and growth strategies (Tomlinson et al. 2014). However, phenological groups might also be associated with contrasting stem traits such as wood density and capacitance, and these traits can be good predictors of the hydraulic functioning of plants in seasonal climates (Borchert 1994; Sobrado 1997; Brodribb et al. 2002, 2003; Brodribb and Holbrook 2005; Choat et al. 2006; Worbes et al. 2013; Gleason et al. 2015). For example, deciduous species exhibit phenological variations associated with xylem water potential (Ψxylem) and wood density (ρwood). Species with low ρwood show leaf sprouting and reproductive phases in the dry period, whereas leaf shedding and resprouting in species with high ρwood is usually dependent on soil water availability (Borchert 1994; Borchert and Rivera 2001; Lima and Rodal 2010; Lima et al. 2012; Oliveira et al. 2015). Furthermore, the leaf life span (LL) can be associated with other two important mechanisms: leaf fall functioning like a “hydraulic fuse” avoiding damage in hydraulic vulnerable xylem during drought (Wolfe et al. 2016; Lima et al. 2018), and deep roots allowing access to deeper water sources, which extend the duration of foliage in the tree canopy during the dry season (Hasselquist et al. 2010).

In a study carried out in a Brazilian tropical savanna experiencing a 5-month dry season, woody species with different phenological behaviors (deciduous, brevideciduous, and evergreen) showed convergence in their hydraulic strategies (Bucci et al. 2005). However, evergreen species occurring in Mediterranean climate sites (i.e., dry summers and wet winters) showed a more conservative hydraulic strategy in terms of water use (isohydric), as they had low maximum stomatal conductance (gs), high stomatal sensitivity to atmospheric drought, and high predawn leaf (Ψleaf) even during the dry season (Mediavilla and Escudero 2003). The deciduous species occurring in the same region were less conservative in water use (anisohydric) and showed high maximum gs and low stomatal sensitivity to atmospheric drought (Mediavilla and Escudero 2003). Presumably, differences in stomatal regulation strategies between deciduous and evergreen species can influence the loss in hydraulic conductance during drought, given that deciduous plants tend to have higher xylem water potential at 50% loss of hydraulic conductivity (P50) than evergreen species (Fu et al. 2012; Kröber et al. 2015; Lima et al. 2018).

Empirical data point to the existence of subgroups of deciduous species (see Lima and Rodal 2010; Lima et al. 2012; Oliveira et al. 2015), and suggest the occurrence of different patterns in the annual variation of Ψwood, and distinct drought strategies. However, these studies used commonly measured soft traits to evaluate functional differences between deciduous and evergreen species (Felsenstein 1985; Fu et al. 2012). More detailed physiological measurements and the evaluation of the effects of phylogeny with phylogenetic independent contrasts (PICs), can provide novel mechanistic insights about functional independence of groups, correlated evolution between traits, and whether the possible trait differences observed between deciduous and evergreen species are the result of evolutionary divergences (Felsenstein 1985; Webb et al. 2008).

In this study, our aim was to investigate and contrast the diversity of hydraulic strategies in DC and EV species from SDTF of the Brazilian semi-arid region. We hypothesize that (1) Leaf hydraulic strategies differ between DC and EV species and result from evolutionary divergences; and (2) DC species compensate for their short LL by being less conservative in water use; in contrast, EV should exhibit highly conservative water use behavior which reduces negative effects of desiccation throughout the dry period. Therefore, we expect that (1) DC species will show high variation in their seasonal water potential (see Borchert 1994; Lima and Rodal 2010; Oliveira et al. 2015), reducing the drop in Ψxylem when they lose the leaves (see Wolfe et al. 2016) and stomatal behavior is less sensitive to drought and will exhibit lower regulation (anisohydric behavior); (2) EV species will show lower variations in seasonal xylem water potential (Ψxylem) and higher stomatal regulation (isohydric behavior), being more sensitive to drought; (3) DC species will be more sensitive to drought than EV, decreasing crown cover in response to decreased water availability in the dry season; (4) EV species will be more conservative in the use of water, with less daily variation in gas exchange, leaf water potential and efficiency in water use compared to DC species; and (5) leaf and stem functional traits will determine distinct functional groups and will explain the variation in LL between DC and EV species from seasonally dry tropical forest.

We studied 22 woody species commonly found in the Brazilian semi-arid region within the Caatinga domain. Our study provides evidence that species with different leaf phenologies have similar hydraulic strategies and sensitivities to drought. In addition, we demonstrate that leaf and stem traits are phylogenetically independent, and that LL is correlated with hydraulic traits such as the maximum transpiration rate per unit of water transport capacity (Λ) and stomatal behaviour.

Material and methods

Study area

This study was carried out in seasonally dry tropical forests (SDTF), which occur in northeastern Brazil, from sea level up to 500 m (Pennington et al. 2000, 2009; Moro et al. 2014, 2016; Silva et al. 2017). This vegetation is locally known as caatinga (Sampaio 1995). The fragment selected (protected for approximately 40 years, with negligible human activity) is located in the Curu Valley Experimental Farm (FEVC, 3°47′S, 39°16′W), 12 km to the west of the municipality of Pentecoste, state of Ceará, Brazil. Soils in this region are shallow with low water storage capacity (Sampaio 1995). Our study area is characterized as SDTF, with plant species with different phenological patterns (according to Borchert 1994; Eamus 1999). The deciduous trees are sparse and deciduous shrubs are dominant, whereas evergreen species are less abundant. Arboreal species ranged from 4 to 15 m in height, whereas shrub species ranged from 2 to 6 m (Table 1).

Table 1 Leaf phenology, growth form, wood density, plant height, and number of individuals of woody plants in a half-hectare plot in Pentecoste, Ceará, northeastern Brazil
Full size table

Extreme droughts, of variable intensity and duration, are common in the Brazilian semi-arid region (Sampaio 1995; Marengo et al. 2013). The temperature range varies little between seasons (averages of 23 ºC in the rainy season and 27 ºC in the dry season). However, the high annual solar radiation of 2800 h and potential evapotranspiration of 2000 mm year−1 lead to water scarcity (Araújo et al. 2004). The regional climate is classified as BSh—semi-arid, dry, with summer rains and a dry winter in the Köppen–Geiger system (Peel et al. 2007). The climatic water balance estimated with data from the past 16 years (FEVC meteorological station 2000–2016) shows a marked difference between the rainy (water replacement) and dry seasons (water loss and deficit; Fig. 1). The average annual rainfall is 708.8 mm, with the rains concentrated in January to April (rainy season), followed by a long dry season (May to December). According to Thornthwaite (1948), the aridity index (Ia) and annual potential evapotranspiration (PET) are 30.2 and 2342.8 mm, respectively (Fig. 1). The average annual temperature is 29 ºC (max. 34.8 ºC and min. 22.8 ºC) with relative humidity around 60%. The average annual rainfall in the 3 years preceding this study was 519.1 mm (~ 73% lower than the average).

Fig. 1
figure 1

Climate water balance for the Curu Valley Experimental Farm, Pentecoste, Ceará, northeastern Brazil. We present data for the past 16 years (meteorological station 2000–2016; annual rainfall: 708.8 mm, annual evapotranspiration: 2342.8 mm, annual water deficit: 1634 mm and index aridity:30.2). Thornthwaite Water Balance (BHídrico GD 4.0 – 2004)

Full size image

Sampling design and model species

We set up a grid of 50 continuous plots 10 × 10 m in an area of half-hectare to sample the dominant species in the community. In total, we identified 30 woody species of shrubs and trees. However, leaf traits and gas exchange were measured in 22 species because these had more than five individuals in the area (Table 1; supporting information). The sampled species represent approximately 85% of all individuals in the area. Leaf phenology, growth form (tree, shrub, or climbing shrub), average height, and the total number of individuals per species (N) in the study area are presented in Table 1.

Soil water content and vapor pressure deficit (VPD)

To determine soil water content (SWC), we collected monthly 50 soil samples in 25 10 × 10-m plots 20 m apart from one another and evenly distributed along a half-hectare area. Data collection was carried out from December 2012 to November 2013 from 07:00 to 10:00 at two depths (0–10 cm and 11–20 cm). We determined SWC (%) using the standard gravimetric method given by the difference between fresh and dry weight, which was obtained after the samples were oven dried at 105 ºC. Data from relative air humidity and temperature sensors (Vaisala, model HMP35C) were used to calculate monthly vapor pressure deficit (VPD), according to the formula: \(VPD=es-ea\), where es is the saturation vapor pressure of air and ea is water vapor pressure. We used the averages of daily values of relative humidity and average maximum temperature in the interval of maximum radiation 9 am to 3 pm. We used these daily averages to estimate monthly VPD. All climate data were obtained from the FEVC meteorological station.

Field measurement of leaf traits

We measured 17 leaf traits to characterize leaf functional differences of deciduous and evergreen species, following the protocol proposed by Pérez-Harguindeguy et al. (2013). All leaf and stem traits measured are summarized in Table 2. Due to the logistical difficulties of tagging and monitoring branches with leaves in adult trees, we assumed leaf life span (LL; in days) to be the average number of days from the beginning of leaf sprouting to leaf fall, which we monitored in ten individuals per species at 15-day intervals. We also estimated the monthly crown cover dynamics by determining the canopy coverage from direct monthly observations of the crown leaf cover using the index of intensity of phenology activity proposed by Fournier (1974). The index consisted in estimating the leaf coverage for all species attributing a continuous percentage value of crown cover range 0–100%. To measure the other leaf traits, we used ten expanded leaves from five individuals per species (50 leaves per species). All leaves were sun-exposed and damage free. In April 2013, we collected leaves from 08:00 to 10:00, rehydrated them for 8 h before measuring leaf area (LA; cm2), petiole length (PL; cm), leaf dry mass per area (LMA; dry mass/leaf area; mg−1mm2), leaf thickness (LTh; mm), leaf density (LD; dry mass /(leaf area x leaf thickness; g cm−3), leaf dry matter content (LDMC; dry mass /water-saturated fresh mass; mg g−1), leaf saturated water content (SWCleaf; water-saturated fresh mass–dry mass / dry mass; g g−1), leaf succulence (LS; water-saturated fresh mass–dry mass/leaf area; g/m2), chlorophyll per unit leaf area (Chlarea; chlorophyll content/leaf area; µmol cm−2), and chlorophyll per unit mass (Chlmass; chlorophyll content/mass; µmol g−1).

Table 2 Comparison of 17 leaf traits among 17 deciduous and five evergreen species and phylogenetically independent contrasts (values represent average ± standard error)
Full size table

We estimated chlorophyll content using a SPAD chlorophyll meter (Minolta SPAD 502 Chlorophyll Meter, Spectrum Technologies Inc., Plainfield, IL, USA), and, then, we used the average value to calculate the polynomial model for tropical trees (chlorophyll content = 0.664 × SPAD + 0.012 × SPAD2; Coste et al. 2010). We measured the physiological leaf traits one time during the rainy season. To measure the maximum photosynthesis rate per unit area (Amax; µmol m−2 s−1), the maximum photosynthesis rate per unit mass (Am; Am = Amax/LMA; nmol g−1 s−1), and the maximum stomatal conductance (gsmax; mol m−2 s−1), we used a handheld infrared gas analyzer (Model LI-6400XT; LI-COR Inc., Lincoln, NE, USA). For measurements of gas exchange, we kept the leaves in the chamber of the handheld meter for approximately 120 s; the CO2 concentration was fixed at 400 ppm, the flow rate 400 µmol s−1, and the photosynthetically active radiation inside the chamber was 1800 µmol (LI-6400XT LED light source). We took the measurements from 08:00 to 10:00 on clear days. The relative humidity at natural conditions varied from 50 to 70% and the temperature from 25 to 32 ºC. We also calculated the intrinsic water use efficiency (WUE; Amax/gsmax; µmol CO2 mol−1 H2O). Using the same leaves from which we measured all traits, we determined the nitrogen content per unit mass (Nmass; % dry mass) and the phosphorus content per unit mass (Pmass; % dry mass) per individual. We determined leaf total nitrogen content using the micro-Kjeldahl method (Bremner 1965). To determine leaf total phosphorus content, after distillation, we used a mixture of nitric, sulfuric, and perchloric acids, at the proportion 10:1:2. The total phosphorus in the extract was determined by colorimetry, at 410 nm, using ammonia vanadomolybdate (Allen et al. 1974).

Field measurements of stem water potential, wood density, and gas exchange

To investigate the water relations of each species in the rainy and dry seasons, we measured each month the predawn xylem water potential (ΨxylemPD; MPa) using a Scholander pressure chamber with a maximum capacity of 10 MPa (Model P3005F01, Soil Moisture Corp., Santa Barbara, California, USA). We carried out ΨxylemPD measurements on three terminal branches of three individuals per species. We decided to use terminal branches, because during the dry season deciduous species lose their leaves completely. Measuring water potential in leafless terminal branches with a pressure chamber can accurately represent Ψxylem (Wolfe and Kursar 2015). We opted to measure predawn water potential (ΨPD) on branches that were not isolated in black plastic bags after we discarded the hypothesis of nocturnal transpiration by testing for differences in Ψleaf between bagged and unbagged branches. We assumed that the soil water potential at the root level equals the ΨxylemPD on terminal branches (Brodribb et al. 2002), except for species with water storage capacity (stem succulents) that maintained high water potential regardless of the soil water potential (C. leptophloeos, A. cearensis, C. vitifolium and M. carthaginensis; see Lima and Rodal 2010; Oliveira et al. 2015). We measured the ΨxylemPD from December 2012 to November 2013, from 0:00 to 4:00 on the 15th day of each month. We obtained the minimum seasonal water potential (ΨxylemPD minimum; MPa) by considering the minimum value of water potential observed during the year. To investigate the relationship between the ΨxylemPD and wood density (ρwood; g cm3), we measured ρwood in five branch samples (diameter ≥ 3 cm) of five individuals per species. Each sample was peeled, saturated in distilled water for 120 h to determine its volume and then oven dried for 72 h at 105 ºC (Pérez-Harguindeguy et al. 2013). We determined the ρwood by calculating the ratio between the dry mass and the saturated volume.

To assess the stomatal behavior, photosynthesis, and leaf water status of the species throughout the day, we measured the daily variation in stomatal conductance (gs), photosynthesis rate (A), the leaf water potential (Ψleaf) and WUE in five individuals per species. We carried out the measurements of gas exchange in rainy season (April 2014) and Ψleaf in the rainy and beginning dry season (April 2014, May 2015 and June 2016). All measurements were made on clear days from 4:00 to 18:00 at 2 h intervals, using five leaves per individual. We used ΨPD and the water potential at midday (ΨMD) to assess the stomatal regulation strategies in response to drought (Martínez-Vilalta et al. 2014). In addition, we tested for differences among species in terms of the water potential responsible for 50% of stomatal closure (ΨSC; MPa; Brodribb et al. 2003). To calculate the ΨSC, we used the measurements of Ψleaf and gs only in times of the day when light was saturating (10:00 and 12:00 pm). All measurements were taken from damage-free leaves that were exposed to full sun. We divided the 22 species in pairs and measured them for 11 consecutive days, to minimize the effects of climate variations.

We kept the leaves in the chamber of the infrared analyzer for 120 s; the CO2 concentration was fixed at 400 ppm, the flow rate at 400 µmol s−1, and the humidity in the chamber was similar to that of the ambient conditions. We calculated the photosynthetically active radiation (PAR) inside the chamber (LI-6400XT LED light source) as the averages per time (4:00–0 µmol; 6:00–300 µmol; 8:00–1200 µmol; 10:00–1800 µmol; 12:00–2000 µmol; 14:00–1200 µmol; 16:00–400 µmol; 18:00–0 µmol) obtained in clear days with the quantum sensors LI-191 connected to a data logger LI-1400 (LI-COR Inc., Lincoln, NE, USA). Immediately after the measurement of gs, we placed the leaves in a Scholander pressure chamber to measure Ψleaf.

Statistical analysis

We tested for differences in leaf functional traits between DC and EV species with a multivariate analysis of variance (one-way MANOVA) for independent samples using R (R Development Core Team 2015). The data were tested for normality (Shapiro–Wilk test), and, if necessary, log-transformed (base 10) before undergoing parametric tests. To identify the specific dependent variables that contributed to the significant global effect, we used the univariate one-way ANOVA. Additionally, we used the Bonferroni correction to adjust the alpha significance (α < 0.05), each individual test because the groups have different sample sizes.

To evaluate whether the observed differences in leaf and stem traits between DC and EV species are the result of evolutionary divergences, we calculated phylogenetically independent contrasts (PIC’s) as described in Webb et al. (2008). We constructed a working phylogenetic tree with the 22 species analyzed from the supertree R20160415 using the software ‘Phylocom’ version 4.2 (Webb et al. 2008). We compared the PIC’s values observed with the PIC’s randomly swapping each trait values across the tree, with 999 randomizations. We considered there was a phylogenetic signal when observed PIC’s were statistically different from randomized mean PIC’s (P value < 0.025; Münkemüller et al. 2012). We used the package “Picante” in R (Kembel et al. 2010) for this analysis. For each trait, we measure the signal intensity using the Blomberg's K coefficient (Blomberg et al. 2003).

To assess the stomatal regulation strategies of the species in response to the loss of conductivity during drought we used the approach proposed by Martínez-Vilalta et al. (2014). We plotted a linear regression with ΨPD as independent variable (x) and ΨMD as dependent variable (y) and used the slope value (σ) to categorize the species according to stomatal behaviour (isohydry to anisohydry) and used the intercept value (Λ) to estimate the maximum transpiration rate per unit of water transport capacity (Martínez-Vilalta et al. 2014). We tested for differences in σ and Λ between the groups with a t test for independent samples.

To investigate which groups are more sensitive to drought, we used simple and multiple regression models to test which environmental variables, VPD (kPa) and SWC (%), significantly affect species crown cover (%). We used a regression analysis to test the relationship between wood density (ρwood) with minimum predawn (ΨxylemPD minimum) and LL with hydraulic strategies descriptors (ΨSC, σ, and Λ). We also used a one-way ANOVA followed by a Bonferroni post hoc test (α < 0.05) to compare phenological groups in relation to daily variation in gs, A, Ψleaf, and WUE. All these analyzes were carried out in R (R Development Core Team 2015).

Finally, to examine combined patterns of functional traits and species ordination in relation to phenological behavior and hydraulic strategies, we performed a principal component analysis (PCA) using the software PC-ORD 6 (Multivariate Analysis of Ecological Data, Oregon, U.S.A). In this analysis, we used the minimum number of functional traits after observing a high correlation between many traits. We used multi-response permutation procedures (MRPP) based on the Euclidian distance to compare functional similarity within the groups formed after the PCA. The MRPP does not use assumptions such as multivariate normality and homogeneity of variances. This analysis provides an ‘A-value’ that varies between 0–1 and can be used as a proxy for within-group homogeneity (McCune and Mefford 2011). We considered the functional groups consistent, if the ‘A-value’ was larger than 0.3 (McCune and Grace 2002).

Results

Differences in leaf traits between evergreen and deciduous species

We observed marked differences between phenological groups in traits related to carbon assimilation and investment in leaf construction (one-way MANOVA F = 122.3, P < 0.0001). Leaf life span (LL) differed between EV (326.8 days) and DC species (159.0 days; Table 2). Leaf area (LA), SWCleaf, and Am were significantly higher for DC than EV species (Table 2). Nevertheless, LMA, LD, LDMC, Chlarea, and Chlmass were higher in EV than DC species (Table 2). There were no differences in PL, LTh, LS, Amax, gsmax, WUE, Nmass, and Pmass between DC and EV species (Table 2). All leaf traits demonstrated weak phylogenetic signal intensity indicated by Blomberg K coefficient values < 1, except SWCleaf that presented significant signal and K = 0.76 (supplementary data Table S1). The signal was not significant for these traits when compared with randomized PIC, indicating phylogenetical independency between groups (Table 2).

Seasonal maintenance of water status

All species reached their maximum ΨxylemPD after the month with the highest rainfall, then gradually decreased until reaching the ΨxylemPD minimum in the end of the dry season (Fig. 2). The ΨxylemPD varied between  – 0.1 and  – 8.3 MPa in DC species, and between  – 0.6 and  – 7.0 MPa in EV species across seasons (Fig. 2). Only one evergreen species (X. americana) showed high variation in ΨxylemPD between seasons and ΨxylemPD minimum of  – 7.0 MPa, similar behavior to most deciduous species (Fig. 2a). Only four DC species (A. subicanum, C. leprosum, P. stipulacea and P. bracteosa) reduced the drop in ΨxylemPD after leaf shedding, while other four species (M. carthaginensis, C. vitifolium, A. cearensis, and C. leptophloeos) exhibited low variation in ΨxylemPD between the dry and rainy seasons (Fig. 2b, c).

Fig. 2
figure 2

Annual patterns in predawn stem water potential (ΨPD) in 17 deciduous and 5 evergreen species. In a Deciduous and only evergreen species with high variation; b only deciduous species reduce the variation after leaf shedding; and c deciduous and evergreen species with low variation. Open square—deciduous species; black square—evergreen species. Dot line—average of leaf shedding in three individuals for species

Full size image

We found three patterns of seasonal maintenance in water status related with ρwood and leaf shedding that were independent of the two phenological categories: (1) species with high ρwood that depended on soil water availability to increase their water status, but could not maintain their status after leaf shedding in the end of rainy season; (2) species that shed leaves early in the dry season showed less variation in their water status; and (3) species with a regular maintenance of their water status regardless of the rainfall regime and duration of leaves (Fig. 2a–c). Deciduous species with high ρwood and the evergreen species X. americana depended on the rainfall regime to rehydrate or lost the leaves to avoid greater drops in ΨxylemPD, whereas the other evergreen with high ρwood and deciduous species with low ρwood either had deep roots and/or relied on stored water to maintain a stable ΨxylemPD.

We found a strong negative relation between ρwood and ΨxylemPD minimum in all species (polynomial model, r2 = 0.73, P < 0.001; Fig. 3). Deciduous species with low regulation of water status between seasons showed high ρwood and more negative ΨxylemPD minimum (Fig. 3). In contrast, DC species with low ρwood showed the least negative ΨxylemPD minimum among all species studied (Fig. 3). All EV species showed high ρwood related to less negative ΨxylemPD minimum, but X. americana showed a ΨxylemPD minimum similar to those of DC species with high ρwood (Fig. 3).

Fig. 3
figure 3

Relation between minimum seasonal xylem water potential (ΨxylemPD) and wood density (ρwood) for 17 deciduous and 5 evergreen species. Symbols represent ρwood—average of five individuals and ΨxylemPD—average of three individuals for species. Open square—deciduous species; black square—evergreen species

Full size image

Stomatal regulation strategies and gas exchange

Unlike our predictions, we observed that DC and EV species showed similar behavior in terms of stomatal control and plant conductivity loss in response to drought (Fig. 4a–v). The average values of slope (σ) of linear regressions between ΨPD and ΨMD for all species was similar for two groups, deciduous (0.77 ± 0.07 MPa MPa−1) and evergreen species (0.65 ± 0.05 MPa MPa−1; t =  – 0.84 P > 0.05; Fig. 4a–v). Most species, regardless of their phenological group, showed partially isohydric behavior, indicating more control in stomatal aperture to water loss (Fig. 4a–v). In DC species, only C. blanchetianus exhibited extreme anisohydric behavior (Fig. 4j).

Fig. 4
figure 4

Linear regression between predawn leaf water potential (ΨPD) and midday leaf water potential (ΨMD) in five evergreen species: a Cynophalla flexuosa, b Libidibia ferrea, c Piptadenia viridiflora, d Ximenia americana, e Zizyphus joazeiro; and 17 deciduous: f Amburana cearensis, g Anadenanthera colubrina, h Aspidosperma subincanum, i Bauhinia cheilantha, j Croton blanchetianus, k Combretum leprosum, l Commiphora leptophloeos, m Cordia oncocalyx, n Cordia trichotoma, o Cochlospermum vitifolium, p Helicteres heptandra, q Lafoensia pacari, r Mimosa caesalpiniifolia, s Manihot carthaginensis, t Piptadenia stipulacea, u Poincianella bracteosa, v Sebastiania macrocarpa. Horizontal line represents strict isohydric (σ = 0) behavior and inclined line strict anisohydric (σ = 1) behavior (Martinez-Vilalta et al. 2014). Points are average values of five individuals per species at different times of year (wet and dry season). Open circles—deciduous species; closed circles—evergreen species

Full size image

Deciduous species showed lower transpiration rate per unit of hydraulic conductivity (Λ) than evergreen species (DC – . 0.35 ± 0.12; EV =  – 2.41 ± 0.16; t =  – 4.19 P < 0.0001; Fig. 4a–v). The similarities in hydraulic strategies between DC and EV species are also confirmed by daily variation in gas exchange and Ψleaf (Fig. 5a–d). During the day, DC species exhibited higher gs than EV species (F = 10.12 P < 0.05), but in the hours of higher photosynthetic activity (10 am to 2 pm), the groups showed no differences (t = 1.05; 2.34; 2.67 P > 0.05; Fig. 5a). In relation to photosynthetic rates (A), the groups were similar (F = 1.97 P > 0.05; Fig. 5b). Although DC exhibited higher gs than EV species, the Ψleaf was less negative in DC species in the morning (8 and 10 am; t = 4.32; 4.83 P < 0.001) but in the rest of the day, it was not different relative to EV species (F = 1.63 P > 0.05; Fig. 5c). The phenological groups had similar WUE at all times (F = 2.02 P > 0.05; Fig. 5d).

Fig. 5
figure 5

Behavior of stomatal conductance (gs), photosynthesis rate (A), leaf water potential (Ψleaf) and water use efficiency (WUE) daily in 17 deciduous and 5 evergreen species. Average gs (a), average A (b), average Ψleaf (c) and average WUE (d) in five individuals for species by daytime. Comparison between deciduous and evergreen species for variance analysis (ANOVA and Tukey test α < 0.05). Open square—deciduous species; black square—evergreen species

Full size image

Effects of VPD and SWC on crown-cover dynamics

The VPD and SWC were strongly correlated with seasonal dynamics in crown cover (%) of DC and EV species (Fig. 6a, b). We observed different critical points in VPD affecting crown cover (%) in DC and EV species (Fig. 6b). Deciduous species had synchrony between maximum crown cover (%) and VPD values below 1.0 kPa (Fig. 6b). The crown cover (%) in EV species was not affected by VPD, even at the peak of the dry season, when the monthly VPD average was above 2.0 kPa (Fig. 6b; Table 3).

Fig. 6
figure 6

Monthly variation in crown cover (%) of 17 deciduous and 5 evergreen species in relation to soil water content (SWC %; a and vapor deficit pressure (VPD kPa; b Open square—deciduous species; closed square—evergreen species; grey bars—SWC; open circle—VPD

Full size image
Table 3 Models and R2 values from the simple regression and multiple regression analysis of crown cover (%) in relation to soil water content (%) and vapor pressure deficit (kPa)
Full size table

In DC species, crown cover (%) was positively related to SWC (simple regression, r2 = 0.67, P < 0.05) and VPD (simple regression, r2= 0.69, P < 0.01; Table 3). By analyzing the variation in SWC and VPD in relation to crown cover (%) in DC species (multiple regression, r2 = 0.86, P < 0.001; Table 3), we observed high sensitivity to the pooled effects of the atmospheric and edaphic drought. In the case of EV species, we observed a significant linear correlation only with SWC (simple regression, r2 = 0.58; P < 0.05) and no correlation with VPD (simple regression, r2 = 0.25; P = 0.77; Table 3). Crown cover (%) in EV species showed sensitivity only to edaphic drought. Even when performing a pooled analysis of SWC and VPD, there were no significant differences when we compared these multiple models with the simple model only with SWC (multiple regression, r2 = 0.58; P < 0.05; Table 3).

Interplay between leaf phenology and hydraulic traits

The variation in LL was weakly explained by the slope value (σ) that represents the isohydric and anisohydric behavior of the species (Fig. 7a). In contrast, the transpiration rate per unit of hydraulic conductivity (Λ) was the hydraulic trait more strongly correlated with LL (Fig. 7b). The negative linear relationship between LL and Λ indicates that species with high LL are less vulnerable to loss in hydraulic conductivity than species that lose their leaves (Fig. 7b). The LL was also negatively correlated with the water potential responsible for 50% of stomatal closure (ΨSC; Fig. 7c). This relationship indicates that species with high LL tolerate the increase in negative pressure (low Ψleaf) before stomata closes and consequently, are hydraulically safer than early-deciduous species.

Fig. 7
figure 7

Linear regression between hydraulic traits: a slope (σ), b intercept (Λ) and c water potential responsible for 50% stomatal closure (ΨSC) and leaf life span (LL) for 17 deciduous and five evergreen species. Value of each species of stomatal behavior iso/anisohydric in response to drought (σ) and maximum transpiration rate per unit of water transport capacity (Λ). Average of LL and ΨSC in five individuals for specie. Open square—deciduous species; black square—evergreen species

Full size image

Functional leaf and stem traits predict phenology groups and hydraulic strategies

In the principal component analysis (PCA), the first principal component represented in the PC1 axis explained 48.2% of the variation in leaf and hydraulic traits (Fig. 8). As expected, leaf traits, such as LL and LMA and hydraulic traits Λ and ΨSC, reflect a contrast between leaf construction costs and low hydraulic safety. Therefore, they were grouped oppositely and strongly loaded in the principal component 1 (PC1; Fig. 8a). Although the species were found in many directions, only those two morphological traits (LL and LMA) were strong enough to separate species according to phenological types (Fig. 8b). The second principal component, PC2, explained 18.6% of the variation in trait values (Fig. 8). Only Amax had a strong positive effect in relation to PC2 and was not crucial to separate the species according to assimilation rate (Fig. 8b). The result of the MRPP test showed a strong consistency in the groups formed based on the ordination of leaf and hydraulic traits (A = 0.46; P < 0.001). The PC1 axis was independent of the PC2 axis indicating that the cost for leaf longevity and hydraulic safety are strongly dissociated from their assimilation of CO2. The magnitude of eigenvalues and eigenvectors are shown in the supplementary data Table S2.

Fig. 8
figure 8

Ordination of seven leaf and hydraulic traits on two principal component axes constructed based on average values of species traits. The factor loadings of deciduous and evergreen species on PC1 and PC2 axis are shown in (a) ordination of species shown in (b). Open circle—deciduous species; closed circle—evergreen species. The original data were standardized. LL—Leaf lifespan; LMA—leaf dry mass per area; gs—conductance stomatal; Am—maximum photosynthesis rate per unit mass; ΨSC—water potential responsible for 50% of stomatal closure; Λ—maximum transpiration rate per unit of water transport capacity and σ—indicator of classical stomatal behaviour iso/anisohydry in response to drought

Full size image

Discussion

Our findings demonstrate that deciduous and evergreen species have slight differences in hydraulic strategies but also highlight greater divergence in hydraulic behavior within deciduous species. Differences in functional leaf traits support the well-known tradeoff between maximization of carbon gain and leaf longevity. However, we demonstrate that both groups exhibit similar degrees of isohydric behavior and significant stomatal control before loss in hydraulic conductivity. Additionally, we provide evidence that leaf shedding (deciduousness) does not necessarily stabilize stem water potential during the dry season in some deciduous species.

The functional leaf traits that differ between DC and EV species are related to leaf structure. Deciduous species invest less mass per unit leaf area while evergreen species have higher leaf mass per area. These results are consistent with the studies conducted in other seasonally dry ecosystems (Sobrado 1991; Eamus 1999; Ishida et al. 2006; Powers and Tiffin 2010; Fu et al. 2012); however, we did not find differences in physiological traits (Amax and gsmax) in contrast with some studies (Mediavilla and Escudero 2003; Ishida et al. 2006; Chen et al. 2009). These results suggest that both EV and DC species diverge in leaf functional traits but may converge in other functions.

In general, the PIC’s indicated that the leaf and stem traits from DC and EV species analyzed have a weak phylogenetic signal and that the phylogenetically close species tend to be functionally distinct. This result suggests a weak divergent selection in leaf and stem traits that differentiate DC from EV species, and which does not depend on phylogenetic proximity. The weak conservatism of these traits may be due to environmental filtering, which should select taxa that are functionally more similar, but phylogenetically more distant (Webb et al. 2002).

We found that DC and EV species exhibit similar degrees of isohydric behavior, which contradicts our hypothesis that only EV species had more conservative behavior in water use. The isohydric behavior means that early stomatal closure during the dry period can maintain high leaf water potential (Ψleaf) values and the plants avoid water stress at the expense of lower carbon uptake (Martínez-Vilalta et al. 2014; Vico et al. 2014). Commonly, stomatal closure strongly depends on atmospheric water demand and soil water available (Buckley 2005) and trees from sem-iarid regions exposed to higher vapour pressure deficits (VPD) tend to experience lower daily Ψleaf regardless of leaf habit (Klein 2014). The convergence of stomatal behavior observed in DC and EV species confirms that under certain environments and conditions, leaf phenology cannot predict hydraulic traits and water use strategies in response to drought (Borchert 1994; Brodribb et al. 2002; Bucci et al. 2004; Ishida et al. 2010; Markesteijn et al. 2011).

The leaf life span in DC species responded both to atmospheric (VPD) and edaphic drought (SWC), whereas EV species responded only to edaphic drought. Two main factors are important to determine leaf fall in response to drought: (1) duration and intensity of the dry season and (2) the capacity of each species to control water loss (Borchert 1994). Commonly, leaf fall in DC species is considered as a mechanism to avoid transpiration and favor the maintenance of high tissue water potential (Levitt 1980; Reich and Borchert 1982; Borchert et al. 2002; Markesteijn and Poorter 2009). In contrast, maintenance of leaves during the entire year and sprouting of new leaves in the dry season in evergreen species are attributed to the capacity of their root systems to obtain deep water from the soil (Borchert 1994; Ackerly 2004) or to stem capacitance (Jackson et al. 1997). Strong stomatal control throughout the year can also contribute to the maintenance of a favorable water status for a longer time (Mediavilla and Escudero 2003).

The central prediction of the hydraulic fuse hypothesis is that leaf shedding minimizes hydraulic stress in the xylem (Tyree et al. 1993). We found that after leaf shedding for several deciduous speciesm, the xylem water potential (Ψxylem) continued to drop. Similar results were described by Wolfe et al. (2016) when testing the hydraulic fuse hypothesis for species of seasonally dry forest of Panama, the Ψxylem of Genipa americana declined after leaf shedding, putting the hydraulic system at risk of failure. These authors argue that plants like G. americana should have mechanisms to recover hydraulic conductivity (i.e. embolism refilling or xylem production) and/or are programmed to senesce before hydraulic failure occurs (Wolfe et al. 2016).

Some deciduous species in our study with high ρwood and declining Ψxylem after leaf shedding (e.g. the fast-growing M. caesalpiniifolia) showed die-off of twigs, branches, and even trunks during the dry season. These parts were rapidly replaced after resprouting during a new growth phase in the following rainy season (B.C. Souza pers. observ). Even though we have not investigated thoroughly, the mechanisms of water acquisition and storage in the tissues, we suggest that further studies should assess the hydraulic mechanisms associated with tolerance to highly negative ΨxylemPD as well as the mechanisms involved in tissue production and fast water flow recovery observed in some deciduous species. We also hypothesize that fast-growing trees in this ecosystem are more prone to hydraulic failure during drought and can recover from embolism through new xylem tissue production (Eller et al. 2018).

Another finding is the low variation in ΨxylemPD throughout the seasons observed in some deciduous and evergreen species (except X. americana). For some deciduous species, low ρwood allows the maintenance of Ψxylem due to their high-water storage capacity in the stem (Borchert 1994; Borchert and Rivera 2001; Choat et al. 2005; Chen et al. 2009; Lima and Rodal 2010; Lima et al. 2012; Oliveira et al. 2015). Evergreen species have high ρwood with low capacity to store water (see Borchert 1994; Borchert and Rivera 2001; Lima and Rodal 2010; Lima et al. 2012; Oliveira et al. 2015) and yet maintain the Ψxylem slightly variable. Although wood density (ρwood) is a functional trait that predicts the capacity to regulate the seasonal water status for deciduous species, we hypothesize that rooting depth may be an important trait influencing the water status in evergreen species (Brum et al. 2017; Oliveira et al. 2014).

Finally, we showed that key hydraulic traits (Λ and ΨSC) are strongly associated to leaf phenology. The intercept (Λ) of the relation between ΨPD and ΨMD represents the maximum transpiration rate per unit of water transport capacity or, equivalently, the capacity of the plant (leaf pressure drops Ψleaf) to ‘pull’ water when there is plenty available in the soil (Martínez-Vilalta et al. 2014). The ΨSC reflects the species ability to control evaporative demand and to adjust to changes in water availability during the rainy season (Brodribb et al. 2003). Therefore, the relationship between Λ and ΨSC and leaf phenology suggests that deciduous are less tolerant to drought than evergreen species, which have more negative stem water potentials that allow higher water absorption by the roots and stomata closure at more negative thresholds (Klein 2014; Martínez-Vilalta et al. 2014). Future research should explore how differences in stomatal sensitivity affect carbon gain in species with contrasting leaf phenology.