A basic understanding of nutrition effects on the mechanisms involved in tree response to drought is essential under a future drier climate.
A large-scale throughfall exclusion experiment was set up in Brazil to gain an insight into the effects of potassium (K) and sodium (Na) nutrition on tree structural and physiological adjustments to water deficit.
Regardless of the water supply, K and Na supply greatly increased growth and leaf area index (LAI) of Eucalyptus grandis trees over the first 3 yr after planting. Excluding 37% of throughfall reduced above-ground biomass accumulation in the third year after planting for K- supplied trees only. E. grandis trees were scarcely sensitive to drought as a result of the utilization of water stored in deep soil layers after clear-cutting the previous plantation. Trees coped with water restriction through stomatal closure (isohydrodynamic behavior), osmotic adjustment and decrease in LAI. Additionally, droughted trees showed higher phloem sap sugar concentrations.
K and Na supply increased maximum stomatal conductance, and the high water requirements of fertilized trees increased water stress during dry periods. Fertilization regimes should be revisited in a future drier climate in order to find the right balance between improving tree growth and limiting water shortage.
Altered precipitation and temperature regimes as a result of climate change are expected to exacerbate the intensity and frequency of future droughts and, consequently, the mortality of forests worldwide (Wu et al., 2011). Improving our understanding of the structural and physiological mechanisms by which trees maintain tissue hydration and photosynthesis in response to water deficit is therefore of paramount importance for predicting the effects of climate change on tree survival, carbon (C) sequestration and water use in forest ecosystems (Hartmann et al., 2013; McDowell et al., 2013; Mitchell et al., 2013; Klein et al., 2014). Comprehensive studies assessing the impact of drought on water use at the stand level are still scarce in tropical and subtropical forests (Brando et al., 2008; McDowell et al., 2008). Although a strong effect of nutrition on plant acclimation to drought has been pointed out (White et al., 2009), large-scale throughfall exclusion experiments have never been designed in tropical planted forests to examine tree physiological adjustment mechanisms in response to the interaction between nutrient and water stress (Wu et al., 2011). Tropical planted forests should play a growing role in the future to supply the population's need for wood products (Paquette & Messier, 2010). Their resilience to global changes deserves further attention, as biomass production is highly dependent on fertilizer addition and the cost of fertilizers is likely to rise in a context of growing demand and limited resources (Cordell et al., 2009; Manning, 2010; Booth, 2013).
Potassium (K) fertilization could be of great interest in mitigating the adverse consequences of drought in planted forests, as foliar K concentrations are likely to influence leaf osmotic adjustment (Mengel & Kirkby, 2001; Cakmak, 2005), stomatal regulation (Fischer & Hsiao, 1968; Cochrane & Cochrane, 2009), protection against oxidative damage (Cakmak, 2005; Wang et al., 2013), and photosynthate loading into the phloem sap (Cakmak et al., 1994; Gajdanowicz et al., 2011). The positive effects of sodium (Na) supply shown recently for osmoregulatory and photosynthetic functions in Eucalyptus grandis W. Hill ex Maiden trees growing in K-depleted soils suggest that Na might also be involved in the structural and physiological adjustments to drought in this species (Battie-Laclau et al., 2013, 2014). Studies dealing with the role of K nutrition in tree response to drought have a broad significance, as primary production is strongly K-limited over large areas of the world, especially in the tropics (Römheld & Kirkby, 2010; Darunsontaya et al., 2012; Santiago et al., 2012).
Subtropical and tropical hardwood plantations are dominated by the Eucalyptus genus, which provides the raw material for wood, paper, and biofuel products as well as large amounts of firewood (FAO, 2010). A wide diversity of mechanisms to cope with drought has been reported in this genus (White et al., 2000; Pita & Pardos, 2001; Whitehead & Beadle, 2004; Callister et al., 2008). Depending on the Eucalyptus species, structural adaptations in response to water deficit can involve primarily a limitation of water loss through a reduction in leaf area (Pita & Pardos, 2001; Le Maire et al., 2011;) and/or an increase in root area in deep soil layers to maximize water uptake (Dye, 1996). At short timescales, stomatal closure is the main physiological process by which Eucalyptus trees reduce transpirational water losses in response to drought (White et al., 2000; Pita & Pardos, 2001; Warren et al., 2007). In some Eucalyptus species, osmotic and/or elastic adjustment may also contribute to maintaining cell turgor under water deficit (White et al., 1996, 2000; Callister et al., 2008). These drought responses can be greatly modified by nutrient supply through an increase in leaf area (Clearwater & Meinzer, 2001; White et al., 2010) and foliar nutrient concentrations, which are likely to change leaf osmotic adjustment (Mengel & Kirkby, 2001), source–sink relationships within trees and C allocations (Fatichi et al., 2014) as well as tree water demand (White et al., 2009). Large increases in phloem sugar concentrations during drought have been reported in Eucalyptus species (Cernusak et al., 2003; Merchant et al., 2010), which may reflect a drop in sink strength during drought. An improvement of tree water status resulting from an adequate nutrient supply might increase the C sink for tree growth during dry periods and therefore influence phloem sugar concentrations.
Within the Eucalyptus genus, the highly productive E. grandis species is the most planted worldwide in moist and warm subtropical regions (Harwood, 2011; Gonçalves et al., 2013). Despite the economic importance of this species in regions that are expected to undergo severe droughts in the future (IPCC, 2013), the mechanisms developed by E. grandis trees to cope with water deficit throughout their growth are still poorly documented. Our study aimed to understand the effects of the K and Na nutrition on the structural and physiological adjustments of E. grandis trees to water deficit. We hypothesized that K and Na supply influence tree drought-adaptive mechanisms through a faster stomatal response to water deficit; osmotic adjustment; and an increase in water extraction from deep soil layers.
Materials and Methods
The experiment (Fig. 1) was conducted at the Itatinga Experimental Station of the University of São Paulo in Brazil (23°02′S; 48°38′W). Over the last 15 yr, the mean annual rainfall has been 1360 mm and the mean annual temperature has been 20°C, with a mean monthly temperature ranging from c. 15°C (between June and September) to 25°C (between October and March). Two distinctive seasons can be distinguished: the dry season from June to September, and the rainy season characterized by high precipitation, temperature, and global radiation.
The experiment was located on a hilltop (slope < 3%) at an altitude of 850 m. The area was afforested in 1940 with Eucalyptus saligna Sm. trees. The last plantation at the study area was a highly productive Eucalyptus grandis W. Hill ex Maiden stand clear-cut in April 2010, 6 yr after planting. The soils are very deep Ferralsols (> 15 m) developed on Cretaceous sandstone, Marília formation, Bauru group, with a clay content ranging from 14% in the A1 horizon to 23% in deep soil layers. The mineralogy was dominated by quartz, kaolonite and oxyhydroxides, with acidic soil layers (pH between 4.5 and 5). Exchangeable K and Na concentrations were, on average, 0.02 cmolc kg−1 in the upper soil layer and < 0.01 cmolc kg−1 between the depths of 5 and 1500 cm (Laclau et al., 2010).
A split-plot experimental design was set up in June 2010 with a highly productive E. grandis clone used in commercial plantations by the Suzano Company (São Paulo, Brazil). Six treatments (three fertilization regimes × two water regimes) were applied in three blocks. The whole-plot factor was the water supply regime (‘exclusion’ plots (−W) vs ‘exclusion-free’ plots (+W)) and the split-plot factor (nested within each plot) was the fertilization regime (C, +Na, +K). The six treatments were as follows:
C−W, control nutrition, without K and Na application, and c. 37% of throughfall excluded;
+Na−W, 0.45 mol Na m−2 applied as NaCl, and c. 37% of throughfall excluded;
+K−W, 0.45 mol K m−2 applied as KCl, nonlimiting in terms of the availability of K for tree growth (Almeida et al., 2010), and c. 37% of throughfall excluded;
C+W, control nutrition, without K and Na application, and no throughfall exclusion;
+Na+W, 0.45 mol Na m−2 applied as NaCl and no throughfall exclusion;
+K+W, 0.45 mol K m−2 applied as KCl and no throughfall exclusion;
The area of individual subplots was 864 m2, with 144 trees at a spacing of 2 × 3 m for a specific water and fertilization regime in one block (total of 432 trees per treatment). One subplot in each treatment was 50% bigger to allow destructive sampling. The total amounts of KCl and NaCl were applied 3 months after planting. All the trees in the experiment were fertilized for the other nutrients at planting (3.3 g P m−2, 200 g m−2 of dolomitic lime and trace elements) and at 3 months of age (12 g N m−2), which was nonlimiting for tree growth at our study site (Laclau et al., 2009). The Na content in the KCl fertilizer was 1.0% and the K content in the NaCl fertilizer was 0.05% (< 15 mg K m−2 contained in the NaCl fertilizer applied).
Starting in September 2010, throughfall was partially excluded in the −W plots using panels made of clear, photosynthetically active radiation (PAR)-transmitting glasshouse plastic sheets mounted on wooden frames at a height varying between 1.6 and 0.5 m (Fig. 1). Plastic sheets covered 37% of the area in the −W plots. A 0.5-m-deep trench was dug around the perimeter of the −W plots to reduce the lateral development of roots between the +W and −W plots. Water flowed by gravity from the plastic sheets into a collecting channel that carried it away from the experiment. Litter that fell in the plastic sheets was collected weekly and spread onto the ground beneath the plastic sheets. Samples of the excluded water were collected weekly and composite samples were analyzed every 4 wk to estimate the amounts of nutrients excluded over the study period. The amounts of nutrients dissolved in the water excluded were replaced once a year in each −W plot (ammonium sulfate, KCl, NaCl, super triple phosphate fertilizers) to avoid any confounding effect between water exclusion and nutrient depletion.
Tree height and circumference at breast height were measured every 6 months on 36 central trees per subplot (excluding a minimum of three buffer rows). A scaffold tower was used to access the crown of four trees in each inner subplot (for a total of 18 towers in the three blocks).
Soil volumetric water content
Soil water content was measured with three TDR probes per layer (Trase Soilmoisture, Santa Barbara, CA, USA) installed at depths of 0.15, 0.50, 1.00, 1.50, 3.00, 4.50 and 6.00 m in each subplot in block 1. Probes were calibrated by gravimetric soil water content and bulk density measurements. Volumetric soil water contents were measured weekly from June 2010 to June 2013. Three piezometers at our study site showed that the depth of the water table fluctuated between the depths of 14.9 and 17.7 m over the study period. Soil coring down to the water table 3 yr after planting showed that the deepest roots reached a depth of 12.5 m, which was consistent with time series of soil water contents down to a depth of 17 m in +K+W and +K−W (J-P. Laclau, unpublished data).
Leaf area index and above-ground biomass
Total leaf area was determined destructively on eight trees (six trees at 8 months of age) covering the range of basal areas in each treatment at ages 8 and 11 months (rainy season 2011), 16 months (dry season 2011), 23 months (rainy season 2012), 27 months (dry season 2012) and 35 months (rainy season 2013). Foliage biomass was determined for each sampled tree by weighing all the leaves in the field and randomly subsampling 30 leaves at 8 months of age and 90 leaves (30 leaves in the upper, middle and lower thirds of the crown) at 11, 16, 23, 27 and 35 months of age. Leaf samples were immediately scanned at 300 dpi and the fresh mass measured, and they were then weighed after oven drying at 65°C for 48 h. The DWs of these subsamples were used in conjunction with their measured area to calculate specific leaf area for each crown section. The foliage DW of each crown section was calculated from the foliage FW and the DW : FW ratio of the subsamples. Tree leaf area was obtained by summing the leaf areas of the three crown sections. Living branches, dead branches, stem wood, and stem bark were separated in the field for all the trees destructively sampled and weighed at 11, 23 and 35 months of age. Subsamples were taken from all the compartments, dried until they were at constant weight, and the dry biomass of the components in each tree was calculated proportionally. Treatment-specific allometric relationships were established at each age and applied to the inventory made on the same date to estimate the total leaf area and the above-ground biomass from tree attributes (diameter at breast height and height). Above-ground biomass was estimated at 16 and 27 months after planting from the allometric equations established at the same age for leaves, and at 11 and 23 months after planting for the other above-ground tree components. Leaf area index (LAI) and above-ground biomass m–2 were calculated in each subplot by dividing the total leaf area and the total above-ground biomass estimated from the allometric relationships by the subplot area.
Tree water status and leaf water relations
Predawn leaf water potential (Ψpdwn) was used as a proxy for tree water status (Limousin et al., 2012). Ψpdwn of the four trees accessed by scaffold tower in each subplot (a total of 72 trees in the three blocks) was measured monthly on sunny days (from 05:00 to 07:00 h) using a nitrogen gas-supplied pressure chamber (PMS Instrument Company, Albany, OR, USA). Measurements were made on one fully expanded leaf per tree (c. 2 months old) in the upper third of the canopy, at the northern side (full-sun) of each tree (12 leaves per treatment). Midday leaf water potential (Ψmid) was measured the same day on another leaf of the same trees (same age and position in the crown) between 12:00 and 14:00 h. The hydrodynamic water potential gradient from roots to shoots (ΔΨplant) was calculated as the difference between Ψpdw and Ψmid Franks et al. (2007).
Leaf water relations (leaf osmotic potential, turgor pressure, bulk elastic modulus) were determined at 10, 16, 22 and 27 months after planting for two leaves sampled from two trees accessed by scaffold tower in each subplot (a total of six leaves per treatment) between 17:00 and 18:00 h. The sampled leaves were cut and immediately put into plastic bags in a saturated and cooled atmosphere and transported to the laboratory (1 km away). Leaf osmotic potential at full turgor (Ψπ100) and at turgor loss point (Ψπ0), bulk elastic modulus at saturation (εmax, inverse of cell wall elasticity), and maximum turgor pressure (P100) were estimated from pressure–volume curves (pV) constructed with a repeated pressurization method for each leaf (Scholander et al., 1965; for more details, see Battie-Laclau et al., 2013). The osmotic adjustment was calculated as the difference in Ψπ0 between the end of the rainy season (mean value at 10 and 22 months of age) and the end of the dry season (mean at 16 and 27 months of age).
Leaf gas exchange measurements
Net CO2 assimilation (Asat) and stomatal conductance (gs) were measured monthly on fully expanded leaves, sampled in the upper third of the crown of the four trees per subplot accessed by scaffold tower in block 1. Leaf gas exchanges were measured for one leaf per tree between 12:00 and 14:00 h at a PAR flux density > 1300 μmol m−2 s−1, at the same time as the measurement of Ψmid on adjacent leaves. Leaf gas exchanges were measured at near constant ambient CO2 concentration (Ca ≈ 380 μmol mol−1) with a portable gas exchange system (Li-Cor 6200; Li-Cor Inc., Lincoln, NE, USA) and a 1149 cm3 chamber equipped with a Li-Cor quantum sensor.
Total sugar and mineral element concentrations in phloem sap
Phloem sap collections were routinely made from a series of horizontal incisions as described by Pate et al. (1998) in six trees randomly selected in each subplot. Exudate was assumed to originate by pressure-induced mass outflow from cut sieve elements. Bulk samples of 20–40 μl were collected for the six trees in each subplot between 16:00 and 18:00 h, using a series of two or three cuts around the trunk at c. 1.3 m above ground level. Total sugar concentrations in phloem sap were measured immediately upon collection using a temperature-compensated hand refractometer calibrated in the range 0 ± 32% (w/v) sucrose. The rest of the sap sample was then filtered and stored at −20°C to await analysis of K and Na by inductively coupled plasma emission spectroscopy (Jobin Yvon 50, Longjumeau, France).
Mixed-effect models were used to test the effects of water supply, fertilization, season, stand age, and interaction between water supply and fertilization, water supply and stand age, fertilization and stage age as well as between water supply and fertilization and stand age (as fixed effects) on above-ground biomass (without the season factor), leaf area index, predawn and midday water potentials, leaf water relations (Ψπ0, Ψπ100, εmax, P100), and concentrations of total sugars, K and Na in phloem sap. Blocks and water supply × block were considered as random effects. Residues were modeled by a first-order autoregressive correlation model to account for the correlations between sampling dates. The ASReml statistical package v.3.0 was used (ASReml, Hemel Hempstead, UK). The level of significance applied in all testing was 0.05.
In order to analyze the response of stomatal conductance to changes in liquid- and gas-phase driving forces (Ψpdwn and vapor pressure deficit (VPD), respectively), an exponential model was used to fit the relationships between gs and Ψpdwn, as well as between gs and VPD (Oren et al., 1999). Thus, gs data from each treatment were analyzed based on:
where b1 and b2 are gs at Ψpdwn = −0.25MPa (no soil water limitation) and at VPD = 1 kPa (no atmospheric limitation), respectively (hereafter designated as the reference stomatal conductance, gsref−Ψpdwn and gsref−VPD), and m1and m2 are the sensitivities of gs to Ψpdwn and VPD, respectively (−dgs/d logeΨpdwn and −dgs/d logeVPD).
Local exponential models of midday gs for each treatment and global exponential models for the whole data set were fitted using Proc Nlin (SAS, Cary, NC, USA). Differences between local and global models were evaluated using F-tests based on the error sum of squares (SSE) and the total number of parameters involved in the models. It compares Fobs and Ftab, calculated as:
where p1 is the number of parameters for the local model (one model per treatment), p2 is the number of parameters for the global model for the whole dataset (p2 < p1), SSE1 is the error sum of squares for the local model, SSE2 is the error sum of squares for the global model and n is the number of measurements. F(p1–p2,n−p1) is the theoretical value given in Fischer's table (Ftab). If Fobs > Ftab then the local model describes the data set better than the global model and the factor studied has a significant effect (Brown & Rothery, 1993).
Above-ground growth and seasonal variations of LAI
Potassium fertilization increased above-ground biomass by 52% relative to Na supply and by 146% relative to the control nutrition at 35 months after planting across the two water supply regimes (Fig. 2a). Significant interaction between stand age, water supply, and fertilization regime showed that the effects of treatments on above-ground biomass accumulation changed over tree growth (Table 1). While throughfall exclusion reduced above-ground biomass accumulation by 14% in +K from 23 to 35 months after planting, tree growth was not influenced by the water supply regime in +Na and C.
Table 1. P-values for the effects of season (S, rainy vs dry season, except for above-ground biomass, BABG), water supply regime (W, undisturbed rainfall vs 37% of throughfall exclusion), fertilization regime (F, control, supply of Na, supply of K), stand age (age), interaction between water supply and stand age (W × age), water supply and fertilization (W × F), fertilization regime and stand age (F × age) and among water supply, fertilization and stand age (W × F × age) on BABG, leaf area index (LAI), predawn and midday leaf water potentials (Ψpdwn and Ψmid, respectively), osmotic potential at the turgor loss point (Ψπ0), osmotic potential when leaves are fully saturated (Ψπ100), turgor pressure at full turgor (P100), maximum values of bulk modulus of elasticity (εmax), and concentrations of total sugars, potassium, and sodium in phloem sap (Sphloem, Kphloem, and Naphloem, respectively) for Eucalyptus grandis trees in a split-plot design
W × F
W × age
F × age
W × F × age
BABG and LAI are shown in Fig. 2, Ψpdw and Ψmid in Fig. 4, Ψπ0, Ψπ100, P100, and εmax in Fig. 7, and Sphloem, Kphloem and Naphloem in Fig. 8.
The time course of LAI over the first 3 yr after planting reflected the seasonality of rainfall and VPD (Figs 2b, 3a). LAI at the end of the dry season (at 16 and 27 months of age) decreased by c. 20% compared with the end of the rainy season (at 11 and 22 months of age), across the fertilization and water supply regimes. This pattern was more pronounced from 11 to 16 months after planting than from 22 to 27 months after planting. Mean LAI over the study period was 31% higher in +Na and 70% higher in +K compared with C, across the two water supply regimes. Throughfall exclusion did not significantly influence LAI during the first 3 yr after planting in the experiment (Table 1). However, throughfall exclusion led to a decrease in LAI in K-fertilized trees at the end of the study period (4.3 and 5.2 m2 m−2 at age 35 months in +K−W and +K+W, respectively), whereas it was not affected in +Na and C (2.8 and 1.8 m2 m−2, respectively, irrespective of the water supply regime).
Soil water storage
Climatic conditions as well as water and nutrient manipulations strongly influenced soil water storage down to a depth of 6 m (Fig. 3). Time series of soil water stocks reflected high rainfall concurrent with low VPD over the hot periods, from October to May, and low rainfall concurrent with high VPD over the cold periods, from June to September. Water storage down to a depth of 6 m was at a maximum in C+W and at a minimum in +K−W over the study period, and the differences between treatments increased with stand age. This pattern reflected the higher water demand of K-supplied trees and, to a lesser extent, Na-supplied trees than that of K/Na-deficient trees, as well as large differences in water input between the +W and −W plots. Throughfall exclusion prevented the percolation of gravitational waters at a depth of 4.5 m under K- and Na-supplied trees from c. 1.5 yr after planting onwards. Time series of soil water stocks in the 4.5–6.0 m soil layer suggested that the stocks of water stored in deep soil layers were withdrawn earlier during the second rainy season after planting by trees in +K and +Na than in C, and that the trees with the highest water deficit (high LAI despite throughfall exclusion) had the capacity to take up more water in deep soil layers than the trees with a lower water deficit (low LAI or undisturbed rainfall).
Seasonal variations of leaf water potential
Ψpdwn exhibited strong seasonal variations during the second and third years after planting, which were more pronounced in −W than in +W (Fig. 4). Ψmid time series were 1.0–1.5 MPa lower than Ψpdwn time series in all treatments, except during the driest periods. The lowest Ψpdwn values reached −0.9 MPa in C−W, −1.4 MPa in +Na−W, and −1.9 MPa in +K−W at the end of the last dry season (at 821 d after planting). Throughfall exclusion lowered Ψpdwn by −0.22 MPa and Ψmid by −0.15 MPa, on average, from 1.5 yr after planting onwards. Fertilization significantly reduced both Ψpdwn and Ψmid, and the effect of fertilization increased with stand age (Table 1). The hydrodynamic water potential gradient from roots to shoots (ΔΨplant) remained relatively constant throughout the seasons, whatever the treatment and despite large fluctuations in Ψpdw (Fig. 4c). The mean ΔΨplant across seasons and treatments was 1.15 MPa.
Seasonal variations of midday leaf gas exchanges
Midday stomatal conductance (gs) and net CO2 assimilation rate (A) reflected the time course of leaf water potentials (Fig. 5). The lowest gs and A values coincided with the dry periods, when soil water storage down to a depth of 6 m was at its lowest. Mean gs and A values were c. 2.5 times higher in the K-supplied trees than in the C trees over the study period, across the two water supply regimes. The Na supply led to gs and A values intermediate between those measured in +K and C. While the effects of throughfall exclusion on gs and A remained low throughout the first 3 yr after planting in C, K supply led to a decrease by c. 30% of gs and A in +K−W relative to +K+W from 1 to 3 yr after planting. The drop in A was larger than the decrease in biomass accumulated above ground, which amounted to only 18% over the same period. Leaf gas exchanges dropped during the first rainy season after planting in C and were thereafter little influenced by climatic conditions. By contrast, large temporal variations of gs and A were observed in +Na and +K over the first 3 yr after planting.
Leaf stomatal conductance vs Ψpdwn and VPD
Midday gs decreased exponentially with decreasing Ψpdwn and with increasing VPD (Fig. 6a,b). K (and to some extent Na) supply sustained much higher gs for most of the range of potential. The relationships between gs and Ψpdwn as well as between gs and VPD were not affected by throughfall exclusion for each nutrient supply regime (Fig. 6). Treatment-specific exponential models for gs vs Ψpdw and gs vs VPD were not significantly different from a global model for C+W and C−W (F2,36 = 0.12 for gs vs Ψpdw and 0.58 for gs vs VPD), +Na−W and +Na+W (F2,36 = 0.06 for gs vs Ψpdw and 0.59 for gs vs VPD), as well as +K−W and +K+W (F2,43 = 0.64 for gs vs Ψpdw and 2.15 for gs vs VPD). By contrast, Na and K supply strongly influenced the relationships between gs and Ψpdwn as well as between gs and VPD: a single exponential model for C+W, +Na+W, and +K+W was significantly different from individual models for each treatment (F4,58 = 54.80 for gs vs Ψpdwn and 15.41 for gs vs VPD), and a single exponential model for C−W, +Na−W, and +K−W was also significantly different from individual models for each treatment (F4,57 = 30.87 for gs vs Ψpdwn and 5.37 for gs vs VPD). In order to be able to compare gs between treatments without having the confounding effect of the seasonal variation in Ψpdw and the diurnal variation in VPD, we determined two gs references. Both gsref−Ψpdw and gsref−VPD increased from 200 mmol m−2 s−1 in C to 500–1100 mmol m−2 s−1 in the fertilized treatments (inserts of Fig. 6). In addition, there was a fourfold increase in gs sensitivity to Ψpdw (−dgs/d logeΨpdwn) and to VPD (−dgs/d logeVPD) in the fertilized treatments. These results also showed that there was a common and positive relationship among treatments between the responses of gs to Ψpdw and VPD and the variations in gsref−Ψpdw and gsref−VPD, indicating that as gsref−Ψpdw and gsref−VPD increased, so did stomatal sensitivity.
Osmotic potential, bulk modulus of elasticity, and maximum turgor pressure
Osmotic potential at the turgor loss point (Ψπ0), osmotic potential when leaves are fully saturated (Ψπ100), maximum values of bulk modulus of elasticity (εmax) and turgor pressure at full turgor (P100) were strongly influenced by the season and the fertilization regime (Fig. 7, Table 1). The lowest osmotic potentials and the highest modulus of elasticity and full turgor pressure were observed during the dry seasons (at 16 and 27 months after planting) in +Na and +K. Throughfall exclusion led to a significant decrease in Ψπ100 and a significant increase in P100. The minimum Ψπ0 values (at 27 months of age) reached −2.1, −2.6, and −2.7 MPa in C−W, +Na−W, and +K−W, respectively, and −2.0, −2.4, and −2.5 in C+W, +Na+W, and +K+W, respectively. The mean osmotic adjustment between the end of the rainy and dry seasons ranged from 0.3 MPa in C+W to 0.6 MPa in +K−W.
Phloem sap composition
Concentrations of total sugars in phloem sap shifted from 17 to 21% in all the treatments during the rainy seasons to 25–28% during the driest periods in −W (Fig. 8). The sugar concentrations in phloem sap were significantly influenced by the water supply regime, the season, and the stand age (Table 1). While K fertilization led to a significant increase in K concentration in phloem sap sampled in +K, Na supply significantly increased the Na concentration in +Na. Throughfall exclusion did not significantly influence the concentrations of K and Na in phloem sap.
Effects of throughfall exclusion on tree growth
The artificial exclusion of 37% of throughfall triggered tree water deficit in +K and +Na but not in C, and the drought response remained small relative to the growth stimulation of the nutrients. Excluding c. 500 mm yr−1 did not reduce above-ground biomass accumulation in the first 2 yr after planting, although high growth rates, LAI, and leaf gas exchanges indicated high water requirements relative to controls. This pattern highlights the major role of the large amounts of water stored in these deep soils after clear-cutting, which supported fast growth rates in the first years after planting. Soil water contents in the 4.5–6.0 m soil layer decreased earlier and more sharply in −W than in +W, which might result from a faster depletion of the upper soil layers in –W. E. grandis roots can reach a depth of 6 m at 1 yr after planting in deep tropical soils (Christina et al., 2011; Laclau et al., 2013). We speculate that the increase in withdrawal of water from deep soil layers during dry seasons may be favored by an increase in C partitioning to roots. The increase in sugar concentration observed in phloem sap during drought probably resulted from a reduction of the activity (i.e. growth) of major above-ground sinks such as leaves and wood (Muller et al., 2011; Sala et al., 2012; Lemoine et al., 2013), thus making more assimilates available for root growth (Costa e Silva et al., 2004). A rise in Na concentration within the phloem sap in +Na, while K concentrations did not increase relative to C, suggested that Na addition did not modify K nutrition and that tree growth enhancement in response to Na supply was the result of a functional substitution of K by Na in our K-deficient soil (Subbarao et al., 2003; Wakeel et al., 2011; Battie-Laclau et al., 2014). Total sugar, K and Na concentrations in our study were within the range of values observed for Eucalyptus globulus (Pate et al., 1998).
Contrasting LAI values between treatments influenced the amounts of water stored in the soil, which led to more severe water deficit conditions in +K and +Na than in C. The large tree response to K supply in our study suggests that applying fertilizers to compensate for exporting dissolved nutrients from the −W plots (0.03–0.05 mol K m−2 over the first 3 yr after planting) was essential to disentangle the effects of water exclusion and tree nutrition on the physiological mechanisms examined. Previous studies showed that long-term throughfall exclusion is likely to reduce K availability in forest soils (Johnson et al., 2008) as a result of the export of nutrients with excluded water. A network of experiments in European coniferous forests showed that throughfall exclusion dramatically reduced K cycling and demonstrated the necessity to replace the nutrients exported through foliar leaching (Gundersen et al., 1995). Our results suggest that special care must be taken to compensate for potassium exports from the exclusion plots in experiments manipulating throughfall over long periods in nutrient-poor soils.
Stomatal and osmotic behaviors of E. grandis trees
A relatively constant hydrodynamic water potential gradient from roots to shoots (ΔΨplant) over large fluctuations in soil and atmospheric moisture conditions showed that E. grandis trees exhibited a stomatal response intermediate between isohydric and anisohydric behaviors. This intermediate stomatal functioning was also observed in Eucalyptus gomphocephala DC. trees and has been referred to as isohydrodynamic (Franks et al., 2007). Stomatal closure occurred over large ranges of Ψpdw and VPD (−0.1 to −1 MPa and 1–4 kPa, respectively). Stomata of eucalyptus trees responded to Ψpdw and to VPD in a manner consistent with protection of the xylem integrity for water transport (Oren et al., 1999; Domec et al., 2009). The emergent behavior was a decreasing gs with decreasing Ψpdw and increasing VPD at rates that were predictable and proportional to gs at high Ψpdw and low VPD. Based on a hydraulic model, −dgs/d logeVPD was shown to be proportional to gsref−VPD, with the proportionality averaging c. 0.60 for isohydric plants and varying predictably depending on the range of VPD used in the analysis. The slope of the relationships between gs and logeVPD in response to contrasting water and nutrient supply was < 0.6 (Fig. 6), indicating that the stomatal response to VPD was intermediate between isohydric and anisohydric species (Oren et al., 1999). Mirroring the stomatal behavior in response to VPD, the response of gs to Ψpdw was related to gsref−Ψpdw, which indicated a strong coordination between soil water content and stomatal regulation. Because of this coordinated stomatal response to the liquid and gas phase fluxes, treatment-induced differences in gsref−VPD and gsref-Ψpdw have implications for gas exchange on both short (diurnal through VPD) and long (soil drying cycle through Ψpdw) timescales.
Osmotic adjustment contributed to the maintenance of leaf turgor. The stomatal regulation of transpiration rates in response to increasing soil water deficit and VPD did not prevent seasonal variations of Ψmid in +K and +Na over the first 3 yr after planting. The lowest values of Ψmid reached −2.4 MPa in +K−W in the third year after planting, which was close to the turgor loss point (Ψπ0 estimated at −2.7 MPa at age 27 months in this treatment). Decreasing leaf osmotic potential (osmotic adjustment) and increasing cell wall elasticity (elastic adjustment) are alternative mechanisms for maintaining leaf turgor at low water contents (Tyree & Jarvis, 1982; White et al., 2000). As cell wall elasticity decreased while water deficit increased in our study, the elastic adjustment did not contribute to lowering the threshold of leaf turgor loss during dry seasons. By contrast, osmotic adjustment was involved in the maintenance of leaf turgor under low Ψmid values. The treatment that experienced the highest amount of water stress (+K−W) exhibited the greatest osmotic adjustment (0.6 MPa) and the highest P100 and εmax values. Changes in osmotic potential and/or tissue elasticity have been shown in various Eucalyptus species submitted to drought (White et al., 1996, 2000; Pita & Pardos, 2001; Lemcoff et al., 2002), including E. grandis (Callister et al., 2008). High osmotic adjustment capacity associated with rigid cell walls in our E. grandis clone could facilitate the maintenance of the cell volume and the integrity of the photosynthetic apparatus during the water restriction periods, and could therefore promote a fast recovery in leaf gas exchanges after soil rewetting (Clifford et al., 1998; White et al., 2000; Lemcoff et al., 2002).
Effects of K and Na supply on drought-adaptive mechanisms of E. grandis trees
In agreement with our hypothesis, K and Na supply increased maximum gs (as seen with the increase in gs−ref) as well as osmotic adjustment. Stomatal response to soil water availability and VPD was much higher in +K and +Na than in C. As a consequence, seasonal fluctuations in stomatal conductance were greatly amplified by K supply and, to a lesser extent, Na supply. This short-term regulation allowed the trees to react rapidly to environmental changes. While K and Na supply increased stomatal conductance over most of the study period, the reverse occurred at the end of the last dry season. The stomatal response to increasing VPD suggests that K and Na deficiency impair stomatal opening under low to moderate VPD (< 5 kPa) and impair stomatal closure under high VPD (> 5 kPa). However, the weak stomatal response to high VPD in C may also result from higher amounts of water stored in the soil than in +K and +Na. Depending on plant species, K supply can favor either stomatal closure (Arquero et al., 2006; Benlloch-Gonzalez et al., 2008) or stomatal opening (Tomemori et al., 2002; Jin et al., 2011). Potassium and Na supply also enhanced leaf turgor during dry periods by increasing osmotic adjustment. Battie-Laclau et al. (2013) showed that the accumulation of K and Na ions in E. grandis leaves in response to K and Na supply contributes to increasing leaf turgor through a reduction in osmotic potential.
Benefits of nutrition offset any detriment of declining soil water over the first 3 yr after planting our E. grandis trees. However, this tentative conclusion would be countered if mortality increased for fertilized trees in the future. The beneficial effect of K and Na supply on the physiological and structural adjustments to drought may not be sufficient to counterbalance high values of LAI and gs, and consequently high tree water demand, during extreme drought. The significant effect of fertilization on Ψpdw shows that tree water deficit was exacerbated by K and Na supply. Ψpdw dropped down to −1.9 MPa in +K−W and −1.4 MPa in +Na−W, whereas it remained above −0.9 MPa in C−W throughout the study period. Higher LAI and gs in +K and +Na than in C led a strong reduction of soil water storage down to a depth of 6 m, from 1 yr after planting onwards. Although our E. grandis clone reduced the total transpiring area during each dry season, the drop in LAI relative to the last rainy season was relatively low (c. 20%) and scarcely affected by the treatments. Stomatal closure was almost complete when Ψpdw dropped down to −1 MPa. Thus, the safety margin to confront more severe drought seemed to be strongly reduced in +K compared with C. Indeed, the difference between Ψpdw and Ψmid (soil-to-leaf water potential gradient, which drives water absorption) at the end of the second dry season was much lower in +K−W (c. 0.4 MPa) than in +Na−W and C−W (c. 0.7 and 1.0 MPa, respectively). Consequently, even though there was no tree mortality in our study, the time series of Ψpdw and Ψmid suggest that K and Na supplies increase the risk of hydraulic dysfunction following extreme water deficit. This risk might increase throughout the rotation as a result of the progressive water depletion of the deepest soil layers.
The isohydrodynamic behavior concurrent with the maintenance of high LAI values during short dry periods displayed by our E. grandis clone led to large withdrawals of water in the soil in response to K and Na supply, which may make this species vulnerable to prolonged droughts. Moreover, the lowest A values were observed in +K−W at the end of the dry seasons, which suggests that K supply might also be detrimental under severe drought through C starvation. There is still some controversy as to whether drought-induced death in forests is primarily related to hydraulic failure or to C starvation (McDowell et al., 2008; McDowell, 2011; Sala et al., 2012). Mitchell et al. (2013) showed that fast-growing species with low control on stomatal conductance (E. globulus and E. smithii in their study) maintained photosynthesis rates under prolonged drought that were sufficient to avoid C starvation, but their high water-use resulted in rapid soil water depletion, declines in water status, and death through hydraulic dysfunction. By contrast, species with tighter control of gas exchanges avoided rapid depletion of soil water, but at the expense of lower C uptake, negative C balance, depletion of stored carbohydrates, and death through C starvation. The isohydrodynamic stomatal behavior of our E. grandis clone might be intermediate between the two contrasting behaviors reported by Mitchell et al. (2013). However, the rapid soil water depletion associated with increases in phloem sap sugar concentrations under drought, as observed for other fast-growing Eucalyptus species (Pate et al., 1998; Cernusak et al., 2003; Merchant et al., 2010), suggests that death under severe drought is more likely to result from hydraulic dysfunction than from C starvation. Our results suggest that deep roots may play a major role in the strategy to cope with drought, a factor that was not considered in Mitchell et al. (2013) as trees were grown in pots. Increased water stress and tree mortality have been shown for Eucalyptus plantations in response to nitrogen fertilization (Stoneman et al., 1996; White et al., 2009), and our study suggests that K fertilization can have similar consequences, despite the beneficial effects of K supply on drought-adaptive mechanisms. Fertilization regimes should thus be adapted to reduce tree water demand in regions where droughts are predicted to increase.
We acknowledge the staff at the Itatinga Experimental Station (ESALQ-USP) as well as Eder Araújo da Silva and Floragro (www.floragroapoio.com.br) for their technical support. The study was funded by FAPESP (www.fapesp.br, 2010/50663-8), CIRAD, USP-COFECUB (Project 2011-25), AGREENIUM (Plantotrem project) and SOERE F-ORE-T. Additional support was provided by the US Department of Energy Terrestrial Ecosystem Sciences program (11-DE-SC-0006700), the National Science Foundation (NSF IOS-1146746 and NSF EAR-1344703) and USDA-NIFA.