Drought-induced defoliation and long periods of near-zero gas exchange play a key role in accentuating metabolic decline of Scots pine



  • Drought-induced defoliation has recently been associated with the depletion of carbon reserves and increased mortality risk in Scots pine (Pinus sylvestris). We hypothesize that defoliated individuals are more sensitive to drought, implying that potentially higher gas exchange (per unit of leaf area) during wet periods may not compensate for their reduced photosynthetic area.
  • We measured sap flow, needle water potentials and whole-tree hydraulic conductance to analyse the drought responses of co-occurring defoliated and nondefoliated Scots pines in northeast Spain during typical (2010) and extreme (2011) drought conditions.
  • Defoliated Scots pines showed higher sap flow per unit leaf area during spring, but were more sensitive to summer drought, relative to nondefoliated pines. This pattern was associated with a steeper decline in soil-to-leaf hydraulic conductance with drought and an enhanced sensitivity of canopy conductance to soil water availability. Near-homeostasis in midday water potentials was observed across years and defoliation classes, with minimum values of −2.5 MPa. Enhanced sensitivity to drought and prolonged periods of near-zero gas exchange were consistent with low levels of carbohydrate reserves in defoliated trees.
  • Our results support the critical links between defoliation, water and carbon availability, and their key roles in determining tree survival and recovery under drought.


Forests are key regulators of land–atmosphere interactions at the global scale, including energy balance processes and carbon (C) and water cycles (Bonan, 2008). Among all the abiotic factors constraining forest functioning, water availability is often the most limiting factor in many forest ecosystems (Boisvenue & Running, 2006). Under extreme and/or chronic drought, partial canopy dieback and increased tree mortality may occur (Bréda et al., 2006). These phenomena are being increasingly reported worldwide (Allen et al., 2010; Carnicer et al., 2011) and can ultimately lead to the collapse of many forest-related ecosystem services (Breshears et al., 2010). Drought-induced dieback is a complex multifactor process (McDowell, 2011) in which climate change emerges as a key element (Van Mantgem et al., 2009), but often in combination with other interacting drivers, such as forest densification (Vilà-Cabrera et al., 2011). This complexity and our limited understanding of the physiological mechanisms involved in drought-induced dieback processes (Sala et al., 2010) hinder our capacity to predict the potential feedbacks between global change and forest ecosystems (McDowell et al., 2011).

Tree growth and survival largely depend on an adequate supply of water to transpiring leaves via root absorption and xylem transport in a tension-driven hydraulic continuum. Physiological and structural responses to fluctuations in water availability have been observed at various time scales and organizational levels, resulting in an operative range of water potentials for the plant (Maseda & Fernández, 2006). In the short term, stomatal closure regulates the decrease in leaf water potential that results from the rates of evaporative loss to the atmosphere and water supply by soils (Oren et al., 1999). The mode of stomatal regulation is tightly co-ordinated with whole-plant hydraulic architecture to prevent the impairment of xylem transport as a result of excessive cavitation, whilst maximizing water extraction from the soil (Sperry et al., 2002). Under long-term drought conditions, some tree species reduce their leaf area so that they can maintain moderate transpiration rates without increasing tension in the xylem. As a result, the capacity of their hydraulic system to transport water to the canopy (whole-plant leaf-specific hydraulic conductance, kS-L) can be maintained, or even increased (Mencuccini, 2003; Maseda & Fernández, 2006). At the whole-tree scale, experimental partial leaf removal usually leads to hydraulically mediated improvements in the water balance of the remaining leaves and subsequent increases in stomatal conductance (Pataki et al., 1998; Quentin et al., 2011).

Pines are particularly plastic in their patterns of biomass allocation to leaves (DeLucia et al., 2000; Martínez-Vilalta et al., 2004). Within pine species, the leaf-to-sapwood area ratio (AL : AS) decreases with increased climatic aridity in field-grown populations (Mencuccini & Grace, 1994; Maherali & DeLucia, 2001). These plastic hydraulic adjustments (Creese et al., 2011) enhance kS-L and allow pines in drier populations to display higher transpiration rates per unit leaf area than their mesic counterparts, without increasing the water potential gradient between soil and canopy (Maherali & DeLucia, 2001). Likewise, in long-term experimental studies, treatments with comparatively reduced water availability show lower leaf areas and higher leaf-specific transpiration rates, especially when soil water is plentiful (Cinnirella et al., 2002).

Scots pine (Pinus sylvestris) thrives under contrasting water balance conditions across its wide geographical distribution, largely because of an efficient hydraulic adjustment via reductions in AL : AS and enhanced stomatal control in response to evaporative demand (Poyatos et al., 2007; Martínez-Vilalta et al., 2009). In addition, an enhanced sensitivity to soil water availability has been shown in warm and dry sites, also related to low AL AS (Poyatos et al., 2007). Despite these adjustments, Scots pine drought-induced dieback is being increasingly reported throughout southern Europe and the Mediterranean basin (Martínez-Vilalta & Piñol, 2002; Bigler et al., 2006; Galiano et al., 2010), which points to the biological costs and limits to the plasticity of the species in response to drought.

Defoliation in drought-exposed Scots pine populations (Dobbertin et al., 2010; Galiano et al., 2010) has been related to reduced canopy development (Dobbertin et al., 2010) and low growth rates (Galiano et al., 2011). Crown defoliation and associated symptoms can thus be regarded as indicators of low vigour, although it appears to be reversible once water stress is relieved (Dobbertin et al., 2010). In chronically dry sites, however, reduced leaf area coupled with Scots pine's characteristic isohydric stomatal control (Irvine et al., 1998; Zweifel et al., 2007; Duursma et al., 2008; Poyatos et al., 2008) may lead to increased reliance on recent assimilation (Eilmann et al., 2010). The continued metabolic demand may eventually deplete carbohydrate reserves, a situation that has been associated with increased mortality risk in Scots pine (Galiano et al., 2011). This scenario is consistent with the C starvation hypothesis (McDowell et al., 2008; McDowell, 2011). However, it still remains to be tested whether whole-plant hydraulic capacity in field-grown declining trees maintains its integrity and its ability to recover in response to extreme drought. If defoliated trees display an enhanced physiological sensitivity to water deficits, extreme reductions in leaf area may be regarded more as an unavoidable symptom of drought stress rather than as a strategy for long-term tree survival.

Here, we analysed the patterns of seasonal sap flow, needle water potentials and whole-plant hydraulic conductance in Scots pines affected by drought-induced defoliation compared with that of apparently healthy trees in one of the first sites in which drought-induced dieback was reported for Scots pine (Martínez-Vilalta & Piñol, 2002). The coexistence of healthy and defoliated Scots pines in the study area provides an ideal model system to explore the physiological implications of drought-induced defoliation and the mechanisms underlying eventual tree death and possible C starvation (Galiano et al., 2011). We hypothesized that: (1) defoliated and nondefoliated pines would display similar regulation of leaf water potential, with minimum values of c. −2.5 MPa (Martínez-Vilalta et al., 2009); (2) the maximum values of sap flow per unit of leaf area and soil-to-leaf hydraulic conductance in spring would be higher in defoliated than in nondefoliated pines; but (3) both variables would show stronger declines during the summer in defoliated pines, indicating higher sensitivity to increasing soil and atmospheric drought. We finally postulated that, in terms of C gain: (4) the increased hydraulic capacity during spring in defoliated pines would not compensate for their reduced canopy leaf area and the prolonged periods with near-zero gas exchange, leading to more strongly depleted carbohydrate reserves and increased mortality risk associated with drought-induced defoliation.

Materials and Methods

Study site

The study was conducted on a northwest-facing hillside within the Tillar valley, at the Poblet nature reserve (Prades Mountains, northeast Spain). The climate is Mediterranean, with a mean annual rainfall of 664 mm, peaking in spring and autumn, and a mean annual temperature of 11.3°C (period 1951–2010), according to a spatially explicit climatic database (cf. Ninyerola et al., 2007a,b). The experimental area (41°19′58.05″N, 1°0′52.26″E; 1015 m asl) is located on a 35° hillslope, on fractured schist, which results in fairly rocky and shallow (c. 40 cm deep) xerochrept soils with a loamy texture and a high gravel content of c. 46% (J. Barba, CREAF, Barcelona, unpublished). Additional information about the study site can be found in Hereter & Sánchez (1999). In this valley, holm oak (Quercus ilex L.) coppiced woodland occupies lower elevations (below 800 m asl), whereas Scots pine stands dominate the upper areas (Gutiérrez, 1989).

Our study focuses on a Scots pine population which is at least 150 yr old and has remained unmanaged for the past 30 yr (Hereş et al., 2012). The studied area has been affected by severe drought episodes since the 1990s (Martínez-Vilalta & Piñol, 2002). Average standing mortality and crown defoliation in the Tillar valley are 12% and 52%, respectively, whereas Scots pine regeneration is very low (Vilà-Cabrera et al., 2013). In some parts of the forest, standing mortality is > 20% and cumulative mortality is as high as 50% in the last 20 yr (J. Martínez-Vilalta, unpublished). Our study plot is located in one of these areas (Table 2), in which defoliated Scots pines survive side by side with nondefoliated individuals and holm oak growing in the understorey.

Meteorological and soil moisture measurements

A data acquisition system (CR1000 datalogger and AM16/32 multiplexers, Campbell Scientific Inc., Logan, UT, USA) was used to store 15-min means of meteorological variables, soil moisture and sap flow sampled every 30 s. Sensors for measuring air temperature and air relative humidity (CS215, Campbell Scientific Inc.), precipitation (52203, R.M. Young Company, Traverse City, MI, USA), total solar radiation (SP1110, Skye Instruments Ltd, Llandrindod Wells, Powys, UK) and wind speed (05103-5, R.M. Young Company) were installed at the top of a 16-m-tall tower within 20 m of the plot centre. Average volumetric soil water content (θ) in the upper 30 cm of soil was monitored using six frequency domain reflectometers (CS616, Campbell Scientific Inc.) randomly distributed within the plot. Volumetric soil water contents measured in soil samples within the plot were regressed against θ to correct values from automatic sensors, which were strongly affected by soil stoniness. The measurement period lasted from the end of April 2010 until December 2011.

Sap flow of defoliated and nondefoliated pines

In November 2009, we selected 10 nondefoliated and 11 defoliated Scots pines of similar size for sap flow measurements (Supporting Information Table S1), within a maximum distance of 70 m from the centre of the plot. Defoliation was expressed as the percentage of green leaves and was visually estimated relative to a healthy canopy of a similar sized tree in the same population (Galiano et al., 2010). Those Scots pines which had 50% or less leaves were considered to be defoliated (Table S1). The foliage of two of the defoliated pines (35% green leaves) had turned completely brown by March 2010, and hence these trees were not included in our analyses.

Sap flow density was measured with constant heat dissipation sensors (Granier, 1985) manufactured in our laboratory. All sensors (= 38) were installed in April 2010. Probe pairs (length, 2 cm) were inserted into the xylem at breast height, with a vertical separation of 12 cm, and were covered with reflective bubble wrap to minimize natural temperature gradients. Two sensors were installed in all pines, on the north- and south-facing sides of the main stem.

Periods during which the sensors were not powered (on average 46 d per tree) were used to estimate natural temperature gradients in the stem. The upper (95%) and lower (5%) quantiles of natural gradients were 0.24 ± 0.02 and −0.20 ± 0.02°C (across trees mean ± SE), respectively; therefore, their effect on sap flow measurements could be considered to be very low (Do & Rocheteau, 2002). Nevertheless, a multiple regression model of natural temperature gradients was fitted using environmental variables (including lags) as predictors, for each individual probe. On average, these models explained 30% of the variance in natural temperature gradients and were used to correct the signal from Granier sensors. Sap flow density in the outer xylem was then calculated according to the original calibration (Granier, 1985), estimating the maximum temperature difference under zero flow conditions (ΔTmax) only for those nights on which the temperature difference between the probes was stable (2-h running coefficient of variation of < 0.5%) and the vapour pressure deficit (D, kPa) was low (2-h running minimum < 0.2 kPa). When these conditions were not met, ΔTmax was linearly interpolated from neighbouring days (Oishi et al., 2008).

Sap flow measurements made using single-point sensors installed in the outer sapwood were integrated to the entire xylem depth using measured radial profiles of sap flow (Nadezhdina et al., 2002). We measured sap flow at six depths in the xylem using the heat field deformation (HFD) method (Nadezhdina et al., 2006), in three defoliated and three nondefoliated Scots pines, during at least 7 d per tree. Two HFD sensors with 8- and 10-mm spacing between measuring points were used (RP20/8-10, Dendronet, Brno, Czech Republic). The shallowest measuring point was always located at 5 mm beneath the cambium, and therefore sap flow could be measured up to a depth of 45–55 mm into the sapwood. Sap flow was expressed as a percentage of maximum sap flow, which usually occurred in the shallowest measuring point. Radial profiles of relative sap flow (% of maximum) were fitted to Gaussian functions (Microsoft Excel 2007 Solver and Microsoft VBA) using the relative xylem radius as the explanatory variable (Ford et al., 2004). Radial correction coefficients (Cr) were obtained as the ratio between the integrated radial profile over the shallowest 2 cm of sapwood and that integrated over the entire sapwood. For each tree in which radial patterns of sap flow had been measured, we applied the corresponding Cr to obtain whole-tree sap flow on a sapwood area basis (JS). For the rest, we used an average value for each crown defoliation class (Cr, defoliated = 0.58 ± 0.05; Cr, nondefoliated = 0.41 ± 0.04).

Sapwood and needle analyses

In November 2009, the main stems of 15 defoliated and 15 nondefoliated Scots pines in the studied population were cored with a Pressler increment borer (5 mm in diameter; Suunto, Vantaa, Finland). The sapwood depth was visually delimited in the field and then used to calculate the tree sapwood area (AS, cm2). The relationship between the tree basal area (AB, cm2) and tree sapwood area was AS = 0.406AB – 83.145 (adjusted R2 = 0.76, n = 30) and did not differ between crown defoliation classes (= 0.91; not shown). In the same cores, sapwood nonstructural carbohydrates (NSCs) were analysed according to Galiano et al. (2011) and expressed as a percentage of dry matter. In addition, for the same sampled pines, δ13C and nitrogen (N) content (% DW) were measured in 20–30 mid-canopy current-year needles using standard methods (Galiano et al., 2011).

Leaf area and shoot parameters

In order to obtain allometric equations to estimate the total leaf area of the measured trees, we followed the approach used by Martínez-Vilalta et al. (2007). In September 2010, a total of 13 (2.9–15 cm in diameter) and 12 (3.2–14.3 cm in diameter) primary branches were harvested from defoliated and nondefoliated trees, respectively. Once in the laboratory, we selected three representative twigs per branch and measured annual shoot growth. We also measured the number of needles, their projected area (Licor 3100 Leaf Area Meter, Licor Inc., Lincoln, NE, USA) and dry weight (oven dried for 48 h at 60°C) separately for cohorts of years 2008 (and older), 2009 and 2010. The rest of the needles for each branch were dried and weighed, converting whole-branch needle dry biomass to projected, one-sided needle area using branch-specific values of specific leaf area (SLA).

A predictive model of branch-level needle area (AL,br, m2) was built using branch diameter (dbr, cm) and defoliation class as explanatory variables (adjusted R2 = 0.69, n = 25). Defoliation class was represented by parameter C in the following equation (= 5.56 for defoliated trees and = 6.07 for nondefoliated trees):

display math(Eqn 1)

This relationship was used to estimate the leaf area of all the primary branches of pines monitored with sap flow sensors. The basal diameter of primary branches of these trees was measured from the ground using large callipers with laser pointers (Mantax Black and Gator Eyes, Haglöf, Sweden). Laser-measured branch diameters showed a tight 1 : 1 relationship with tape-measured values (R2 = 0.95), with the slope and intercept not significantly different from one and zero, respectively.

Leaf water potential

Leaf water potentials were measured monthly from June to October (except for September 2011) using a pressure chamber (PMS Instruments, Corvallis, OR, USA). Measurements were taken at predawn (Ψpd; just before sunrise at 03:00–05:00 h, solar time) and at midday (Ψmd; 11:00–13:00 h, solar time). On each sampling date, one exposed shoot tip was excised with a pruning pole from four to seven trees of each crown defoliation category, from those monitored with sap flow sensors. Once sampled, shoot tips were stored in a plastic bag with moist paper towels to avoid leaf water loss during the time lag between shoot excision and measurement in the field, which was typically < 2 h.

Whole-tree sap flow, canopy conductance and hydraulic conductance

Whole-tree sap flow rates were obtained by multiplying JS (average of radially corrected sap flow density measured by north- and south-facing sensors) by AS, and expressed on a leaf area basis (JL) by dividing JT by AL. According to on-site phenological observations, Scots pine AL was assumed to increase linearly from mid-May to mid-July and to decline linearly from August to November to two-thirds of its maximum value (Beadle et al., 1982). Sap flow data were then aggregated to daytime averages (RS > 1 W m−2), yielding sap flow per tree (JT,dt, kg H2O d−1), sap flow per unit sapwood (JS,dt, g H2O cm−2 d−1) and sap flow per unit leaf area (JL,dt, kg H2O m−2 d−1). Daytime averages were used to avoid problems related to lags between transpiration and water uptake.

Midday canopy stomatal conductance (Gs,md) was calculated with the simplified Penman–Monteith equation for aerodynamically rough canopies (Whitehead & Jarvis, 1981):

display math(Eqn 2)

(γ, psychrometric constant (kPa K−1); λ, latent heat of vaporization of water (J kg−1); ρ, air density (kg m−3); cp, specific heat of air at constant pressure (J kg−1 K−1); Dmd, midday vapour pressure deficit (averaged between 11:00 and 14:00 h) (kPa); JL,md, midday leaf area-based sap flow converted to molar units (mmol m−2 s−1)). We used midday values so that canopy stomatal conductance was more representative of instantaneous stomatal regulation; and lag effects were minimized (Irvine et al., 2004).

We also used JL,md to calculate whole-tree hydraulic conductance (kS-L, mmol m−2 MPa−1 s−1), assuming that trees had reached equilibrium with the soil during the night, and that Ψpd represents an estimate of soil water potential (Irvine et al., 2004):

display math(Eqn 3)

Data analysis

All variables, with their symbols and units, are described in Table 1. All analyses, unless otherwise stated, were carried out with R Statistical Software version 2.12.0 (R Development Core Team, 2009). Seasonal sap flow data, leaf water potential and hydraulic conductance data were analysed using linear mixed models (lme function, package nlme) with tree as a random factor (Pinheiro & Bates, 2000). Fixed factors included date (or seasonal period) and defoliation class. Shoot and needle parameters were similarly analysed, introducing cohort and defoliation class as fixed factors and branch as a random factor. We thus tested for differences in JL,dt, JS,dt, ψpd, ψmd and kS-L across defoliation classes and periods (hypotheses 1 and 2). Differences in relationships between continuous variables in defoliated vs nondefoliated pines were analysed using generalized least squares (gls function, package nlme), in order to account for variance heterogeneity and temporal autocorrelation, when present. Here, we tested how defoliation influenced the effects of θd on ψpd and kS-L (hypothesis 3), the regulation of ψmd with respect to ψpd (hypothesis 1) and controls on Gs,md by ψpd (hypothesis 3). Differences in NSC, leaf δ13C and N content between defoliation classes were assessed using t-tests or Mann–Whitney–Wilcoxon tests (hypothesis 4).

Table 1. List of abbreviations used for the measured variables and parameters in this study (fitted parameters are shown in bold)
a Asymptote of the relationship between JL,dt and Ddt kg H 2 O m −2  d −1
A B Tree basal areacm2
A L Tree leaf aream2
A L  : A S Leaf-to-sapwood area ratiom2 cm−2
A L,br Leaf area of primary branchm2
A S Tree sapwood areacm2
B Initial increase in JL,dt with Ddt Adimensional
C Defoliation class coefficient in the AL,br model Adimensional
c p Specific heat of air at constant pressureJ kg−1 K−1
C r Coefficient for radial integration of sap flow density Adimensional
d br Branch diametercm
D Instantaneous (15-min) vapour pressure deficit of the airkPa
D dt Daytime-averaged vapour pressure deficit of the airkPa
D md Midday averaged vapour pressure deficit of the airkPa
G s,md Canopy stomatal conductance at middaymmol m−2 s−1
G s,ref Reference Gs,md at Dd = 1 kPa mmol m −2  s −1
J L Instantaneous (15-min) sap flow per unit leaf areakg H2O m−2 s−1
J L,dt Daytime-averaged sap flow per unit leaf areakg H2O m−2 d−1
J L,asym Asymptote of response of JL,dt to θd kg H 2 O m −2  d −1
J L,md Leaf area-based sap flow during middaymmol m−2 s−1
JSInstantaneous (15-min) sap flow per unit sapwoodkg H2O m−2 s−1
J S,dt Daytime-averaged sap flow per unit sapwoodg H2O cm−2 d−1
J T Instantaneous (15-min) sap flow per treekg H2O s−1
J T,dt Daytime-averaged sap flow per treekg H2O d−1
k S-L Whole-tree hydraulic conductancemmol m−2 MPa−1 s−1
m Sensitivity of Gs,md to Ddmmol m−2 MPa−1 s −1 (loge kPa)−1
NSCNonstructural carbohydrates% DW
R S Total solar radiationW m−2
R S,dt Daytime average of total solar radiationW m−2
SLASpecific leaf areacm2 g−1
ΔTmaxTemperature difference measured by sap flow probes under zero-flow conditions°C
Γ Psychrometric constantkPa K−1
δ13CCarbon stable isotopic composition
θInstantaneous (15-min) soil water contentcm3 cm−3
θdDaily average of soil water contentcm3 cm−3
θ mid Inflection point in the response of JL,dt to θd cm 3  cm −3
θ scal Scaling parameter in the response of JL,dt to θd cm 3  cm −3
λ Latent heat of vaporization of waterJ kg−1
ρ Air densitykg m−3
Ψ md Midday water potentialMPa
Ψ pd Predawn water potentialMPa
Ψ 50 Water potential causing 50% loss of hydraulic conductivityMPa

Differences in the environmental controls on sap flow of defoliated and nondefoliated pines were assessed by examination of the functional response of JL,dt to evaporative demand and soil water availability (hypotheses 2 and 3). The asymptotic relationship between JL,dt and average daytime D (Ddt) was analysed using an exponential saturation function (Ewers et al., 2002):

display math(Eqn 4)

where a is related to maximum sap flow and b is the initial increase in sap flow with respect to Ddt. Only data with daytime-averaged radiation (RS,dt) > 150 W m−2 and daily soil moisture (θd) > 0.20 cm3 cm−3 were used to fit the function. The nonlinear reduction in sap flow with decreasing θd was described by a three-parameter sigmoid (based on results by Duursma et al., 2008):

display math(Eqn 5)

fitted to data when RS,dt > 150 W m−2 and Ddt > 0.8 kPa, with JL,asym being the asymptote, θmid the inflection point (for which JL,dt = JL,asym/2) and θscal a scaling parameter. Both functions were fitted to individual tree JL,dt and environmental data using nonlinear mixed models (nlme, package nlme), with tree as random factor and defoliation class as fixed factor.

In all cases, model selection was based on the Akaike information criterion (AIC), including the selection of appropriate random effects, variance structures and autocorrelation parameters. Model assessment was carried out by visual inspection of residual plots. The significance of fixed effects was assessed using F tests. Post-hoc tests were carried out with the general linear hypothesis function (glht) in the package multcomp (Hothorn et al., 2008).

The relationship between Gs,md and Dmd (hypothesis 3) was analysed using boundary-line analysis, fitting the following model using quantile regression (Cade & Noon, 2003):

display math(Eqn 6)

We assumed that the linear regression results for the 95th quantile were representative of the relationship between Gs,md and loge Dmd when other factors were not limiting (Cade & Noon, 2003) and identified the resulting intercept and slope with the model parameters Gs,ref (reference Gs,md at Dmd = 1 kPa) and m (sensitivity to vapour pressure deficit, −dGs,md/dloge D), respectively. Only data with Dmd > 0.6 kPa were analysed to avoid potential errors in Gs,md estimation under low evaporative demand (Ewers & Oren, 2000).


Seasonal course of environmental variables and sap flow

Over the study period, meteorology was typical of a low-elevation Mediterranean mountain climate, with occasional cold spells in winter (daytime minimum of −3.8°C), moderately high temperature and evaporative demand during summer (maximum of 26.6°C and 3.0 kPa, respectively) and irregular precipitation (Fig. 1). The two growing seasons studied were not exceptional in terms of temperature, with May–October average values of 15.9 and 17.0°C for 2010 and 2011, respectively, compared with a climatic average (1951–2010) value of 16.6°C. Likewise, a normal drought period occurred in 2010, with 333 mm of total precipitation recorded from May to October (92% of the climatic average). However, the drought in 2011 was especially acute, with just 113 mm of rain between May and October, representing only 31% of the climatic average. Accordingly, soil water content remained below 0.10 cm3 cm−3 between July and mid-October in 2011 (Fig. 1c).

Figure 1.

Seasonal course of environmental variables and Scots pine (Pinus sylvestris) sap flow over the study period: (a) daily precipitation; (b) daytime-averaged solar radiation (RS,dt, grey bars) and vapour pressure deficit (Ddt, closed dots); (c) mean volumetric soil water content (θd, cm3 cm−3, closed circles) and standard errors (grey bars); (d) across-trees mean of daytime sap flow per unit sapwood area (JS,dt); (e) per unit leaf area (JL,dt) in defoliated (closed circles and continuous line) and nondefo-liated (open circles and dashed line) pines. Coloured bars depict standard errors. Asterisks in (a) show the sampling dates for water potential measurements. Gaps in the time series are a result of power failures. Note that labels on the x axis represent year in decimal format (i.e. January 2011 is 2011.0).

Sap flow was higher during spring and decreased strongly after June for both years. Rainy periods in late summer and autumn raised soil water availability, leading to a recovery in sap flow rates in 2010, but not in 2011 (Fig. 1d,e). In general, sap flow during peak summer and autumn drought was lower for 2011 relative to 2010 (Fig. 1d,e; Table S2). Nondefoliated pines showed similar sap flow per unit sapwood area (JS,dt) to defoliated pines (Table S2; Fig. 1d). This pattern changed when sap flow was expressed per unit leaf area (JL,dt) and, especially during late spring and early summer, JL,dt was considerably higher for defoliated pines relative to nondefoliated pines (hypothesis 2). However, in response to summer drought, JL,dt was more strongly reduced in defoliated pines, eventually reaching similar near-zero values to those found for nondefoliated pines (Fig. 1e; Table S2), consistent with hypothesis 3. Tree-level water use was consistently lower in defoliated pines, but the difference was not statistically significant when averaged over seasonal periods, except for the summer of 2010 (Table S2).

Environmental responses of sap flow in defoliated and nondefoliated pines

The increase in JL,dt with Ddt was well described by the exponential saturation function employed (Fig. 2). Defoliated and nondefoliated pines differed in their maximum JL,dt, with higher a values for defoliated individuals, but not in the initial increase in JL,dt with Ddt (parameter b) (Fig. 2; Table 3). Sap flow started to decline appreciably at values of θd ≈ 0.25 cm3 cm−3 and was almost zero below θd ≈ 0.10 cm3 cm−3 (Fig. 2). The analysis of the response of JL,dt to θd showed that defoliated pines displayed a higher JL,asym and also a slightly higher θmid (Table 3), implying a steeper decline of JL,dt with θd.

Figure 2.

Responses of daytime averages of leaf area-related sap flow (JL,dt) to vapour pressure deficit (Ddt) (a) and soil water content (θd) (b) in Scots pine (Pinus sylvestris). Average values for defoliated (closed circles) and nondefoliated (open circles) pines are shown. Standard errors are not displayed to improve clarity. Different fits for defoliated (solid line) and nondefoliated (dashed line) pines reflect the significance of the defoliation effect in the nonlinear mixed models of individual tree JL,dt data (Table 3).

Water potentials and hydraulic conductance

Leaf water potentials gradually decreased from late spring/early summer to early autumn for both years (Fig. 3). Across dates, nondefoliated pines showed slightly higher water potentials (c. 0.1 MPa higher), both when measured before dawn and at midday. The water potential difference varied with date (< 0.001; not shown), but not with defoliation class (= 0.41; not shown), although a significant interaction of date and defoliation class was found (= 0.038; not shown).

Figure 3.

Seasonal course of predawn water potential (a, b), midday water potential (c, d) and soil-to-leaf hydraulic conductance (kS-L) in defoliated (closed circles) and nondefo-liated (open circles) Scots pines (Pinus sylvestris) for 2010 and 2011. Different line patterns depict significant differences between defoliated and nondefoliated pines for the entire study period according to linear mixed-effects models (cf. 'Materials and Methods' section). Where a significant interaction between time and defoliation class was present, asterisks (< 0.05) or dots (0.05 < < 0.1) identify significant differences between defoliation classes.

The lowest ψpd values were measured in September and October for 2010 and 2011, respectively. In 2011, they reached a significantly lower seasonal minimum value (–2.2 MPa) relative to 2010 (−1.6 MPa), for both defoliated and nondefoliated pines (both < 0.001; Fig. 3). Likewise, seasonal ψpd values were lower in 2011 (October, −2.5 MPa) relative to 2010 (August, −2.3 MPa) (both < 0.001; Fig. 3). ψpd declined with decreasing θd following a nonlinear relationship (Fig. 4a), with similar slopes but different intercepts for defoliated and nondefoliated trees (Table 4), implying that, at a given θd, ψpd was lower in defoliated trees.

Figure 4.

Relationships between daily average soil water content (θd) and (a) predawn water potential and (b) soil-to-leaf hydraulic conductance (kS-L) in defoliated (closed circles) and nondefoliated (open circles) Scots pine (Pinus sylvestris). Different line patterns depict significant differences (< 0.05; Table 4) between defoliated (solid) and nondefoliated (dashed) pines. The two low values of kS-L at high soil moisture values correspond to a lack of recovery in kS-L after autumn rains in 2010, and were not used to fit the regression lines (cf. 'Results' section).

Whole-plant hydraulic conductance (kS-L) decreased as the drought progressed (Fig. 3e,f) and did not show a short-term recovery in response to precipitation. In 2010, kS-L still showed very low values after a rainy period in early autumn which raised θd above 0.20 cm3 cm−3 (Fig. 3e). Nevertheless, early summer values of kS-L in 2011 did recover, although, apparently, not at the same level as the maximum values in 2010. This was probably caused by a delayed first measurement of kS-L in 2011. Defoliated pines displayed higher values of kS-L early in the growing season, but, in response to drought, kS-L dropped to values similar to those found in nondefoliated pines (Fig. 3e,f; significant interaction between date and defoliation class, = 0.016).

Despite the decoupling observed between kS-L and θd in autumn 2010, kS-L was positively related to θd for the rest of the dates (Fig. 4b). This relationship differed for defoliated and nondefoliated pines (Table 4), with the rate of decline in kS-L with decreasing θd being higher for defoliated pines (hypothesis 3). The capacity of Scots pine to supply its canopy leaves with water was greatly reduced when θd fell below c. 0.10 cm3 cm−3 (Fig. 4b).

Stomatal control in defoliated and nondefoliated pines

Defoliated and nondefoliated pines shared a common relationship between ψpd and ψmd (Fig. 5a, Table 4). However, defoliated and nondefoliated Scots pines showed differences in the response of Gs,md to ψpd (Fig. 5b). The function relating Gs,md and log(–ψpd) showed different intercepts and marginally higher slopes, in absolute value, for defoliated pines (Table 4).

Figure 5.

Relationships between predawn water potential and (a) midday water potential and (b) canopy stomatal conductance measured at midday (Gs,md) in defoliated (closed circles) and nondefoliated (open circles) Scots pine (Pinus sylvestris). Different line patterns depict significant differences (< 0.05; Table 4) between defoliated (solid) and nondefoliated (dashed) pines. The 1 : 1 line is also shown.

Reference canopy stomatal conductance was generally higher for defoliated than for nondefoliated trees, except for one individual (Fig. 6); Gs,ref was higher in defoliated trees (t-test, = 0.006), showing an average of 255 ± 27 mmol m−2 s−1 relative to 139 ± 31 mmol m−2 s−1 observed for nondefoliated pines. The sensitivity of Gs,md to the vapour pressure deficit (m) shared the same proportionality with respect to Gs,ref across defoliation classes, with a slope of 0.79 ± 0.02 mmol m−2 s−1 loge kPa−1 and a nonsignificant intercept (= 0.841).

Figure 6.

Stomatal sensitivity to vapour pressure deficit parameters for defoliated (closed circles) and nondefoliated (open circles) Scots pine (Pinus sylvestris). The solid line represents the common relationship between reference stomatal conductance (Gs,ref) and sensitivity to vapour pressure deficit (m) across defoliation classes (cf. 'Results' section). The dotted line has a slope of 0.6 and zero intercept, representing the predicted relationship for species showing an isohydric control of leaf water potential (Oren et al., 1999).The dashed line depicts the fit between Gs,ref and m observed across Scots pine populations (Poyatos et al., 2007).

Shoot and leaf parameters and sapwood NSCs

The patterns in shoot and leaf parameters (Table S3) revealed higher growth constraints for defoliated pines. Shoot growth was 42% higher for nondefoliated pines in 2009 and 2010, relative to that of defoliated pines, but no significant difference was observed in 2008. Most of the needle-level differences were observed for the 2010 cohort, when nondefoliated pines had 20% more needles and lower SLA relative to defoliated pines (Table S3). Average needle retention was c. 3 yr for both defoliation classes (not shown).

At the leaf level, no differences in either 13δC or leaf N content were found between defoliated and nondefoliated pines, and leaf N content was not correlated with either 13δC or sapwood NSC (not shown). Finally, defoliated pines displayed similar sapwood density to that of nondefoliated pines, and significantly lower (33%) sapwood NSC levels (Table 2), a result consistent with hypothesis 4.

Table 2. Characteristics of the studied stand, together with leaf and sapwood parameters of defoliated and nondefoliated Scots pines (Pinus sylvestris)
  P. sylvestris Q. ilex Other
  1. Stand-level variables were measured in two 18.2 × 18.2-m2 plots (J. Barba, unpublished). LAI was estimated from the tree-level leaf area obtained by site-specific allometric relationships (cf. 'Materials and Methods' section). Differences in tree-level traits were assessed with t-tests or Mann–Whitney–Wilcoxon tests for non-normally distributed variables. Different lowercase letters represent statistically significant differences (< 0.05) between defoliated and nondefoliated trees. See Table 1 for abbreviation and symbol descriptions. DBH, tree diameter at breast height; LAI, leaf area index.

Plot level
Density (stems ha−1)1061511736242
DBH (cm)37.90 ± 3.026.65 ± 5.38.4 ± 0.46.7 ± 1.2
Basal area (m2 ha−1)12.3311.4624.862.9
Height (m)13.4 ± 0.514.8 ± 0.8  
LAI (m2 m−2)0.160.421.39n. a.
Tree level
Sapwood density0.46 ± 0.01a0.44 ± 0.01a  
Trunk sapwood NSC% DW0.70 ± 0.09a1.05 ± 0.06b  
Leaf 13δC‰−26.7 ± 0.66a−26.4 ± 0.33a  
Leaf N1.19 ± 0.04a1.37 ± 0.06a  
% dry mass    


Seasonal patterns of sap flow in Scots pine at its dry limit

The present study portrays a detailed picture of the water relations of Scots pine at the dry edge of its distribution and under exceptionally dry conditions. The study area experienced the driest growing season on record, as the driest May–October period since 1951 had received 195 mm of rain (in 2007), compared with the value of 113 mm for the same period in 2011. Accordingly, the seasonal course of sap flow showed that Scots pine at this dry site is extremely sensitive to summer drought. Moreover, reduced precipitation during the winter and the associated lack of recharge of soil water reserves (Fig. 1) may limit the amount of water available to meet evaporative demand during the summer (Llorens et al., 2010).

We observed a steep decline in sap flow rates at soil water content values below c. 0.25 cm3 cm−3 (Figs 1, 2), consistent with similar threshold-like responses observed elsewhere (Irvine et al., 1998; Lagergren & Lindroth, 2002; Duursma et al., 2008; Poyatos et al., 2008). However, here, we found that extremely low sap flow rates continued for over 2 months in 2010 and > 5 months in 2011, a more persistent drought response than that observed in other drought-affected populations (Zweifel et al., 2009). The sensitivity analysis of Gs,md with respect to Dmd also showed a strict stomatal closure in response to increasing evaporative demand in the studied Scots pine population. Stomatal sensitivity to Dmd (m) was proportional to the gas exchange capacity (Gs,ref), but the slope of the relationship between m and Gs,ref was greater than the value of 0.6 found for mesic tree species with an isohydric regulation of leaf water potential (Oren et al., 1999). These results suggest a hypersensitive stomatal regulation with respect to D in this Scots pine population, as recently found for other species (Ogle et al., 2012).

Sap flow in defoliated pines is more sensitive to summer drought

Defoliated Scots pines showed higher JL,dt and kS-L, consistent with hypothesis 2, and in agreement with the patterns observed along climatic gradients in Scots pine (Poyatos et al., 2007) and other pine species (Maherali & DeLucia, 2001). This increased transport capacity determined higher water use per unit leaf area in defoliated than in nondefoliated pines, especially under well-watered conditions (Fig. 1, Table S2). Accordingly, the hydraulic capacity per unit sapwood area was similar across defoliation classes (Fig. 1, Table S2), as also reported for drought-adapted Pinus ponderosa relative to mesic populations (Maherali & DeLucia, 2000). Likewise, artificial defoliation resulted in a perfect compensatory increase in transpiration of the remaining leaves in Pinus taeda (Pataki et al., 1998), although this response may not occur in other conifers (Brooks et al., 2003).

Here, we also showed that defoliated pines were more sensitive to summer drought (hypothesis 3), as shown by the steeper decline in JL,dt with decreasing θd (Fig. 2, Table 3), and the enhanced sensitivity of Gs,md to ψpd (Fig. 5b, Table 4). Defoliated pines were also more sensitive to Ddt in absolute, but not relative, terms (Table 4, Fig. 2) (cf. Poyatos et al., 2007). This pattern was confirmed by the analysis of Gs,md, whereby defoliated and nondefoliated pines shared the same proportionality between the maximum gas exchange and its response to Ddt (Fig. 6). The slope of the m vs Gs,ref relationship was similar to the population-averaged value of m/Gs,ref (0.81) found for another Mediterranean Scots pine stand (Vallcebre, eastern Pyrenees), in which no drought-driven mortality has been reported so far (Poyatos et al., 2007). In that study, however, m tended to plateau with increasing Gs,ref across populations along a climatic gradient spanning from Scandinavia to southern Spain. Here, we found a linear relationship between both parameters for trees with contrasting AL : AS within a site. This discrepancy and the higher Gs,ref values observed for defoliated pines relative to those found across a climatic gradient suggest that the mechanisms behind climate-driven hydraulic adjustments across populations differ from those leading to spatially diffuse defoliation patterns in a declining stand.

Table 3. Summary of the nonlinear mixed models relating daytime values of Scots pine (Pinus sylvestris) sap flow per unit leaf area (JL,dt) with vapour pressure deficit (Ddt) and soil water content (θd)
ModelParameterEstimateSE t P
  1. Defoliation class (defoliated, ≤ 50% green leaves; nondefoliated, > 50% green leaves) was introduced as a factor in the models; nondefoliated is used as the reference level. Parameters with a significant effect of defoliation class are shown in bold. See Table 1 for abbreviation and symbol descriptions.

display math
a: nondefoliated 1.548 0.331 4.677 < 0.0001
a: defoliated 1.539 0.482 3.196 0.0010
b −1.4380.09714.896< 0.0001
display math
JL,asym: nondefoliated 1.267 0.257 4.934 < 0.0001
J L,asym : defoliated 1.634 0.363 4.501 < 0.0001
θmid: nondefoliated 0.161 0.003 51.491 < 0.0001
θ mid : defoliated 0.011 0.003 3.098 0.0020
θscal0.0280.00216.977< 0.0001
Table 4. Coefficients and summary statistics for some relevant relationships among environmental and ecophysiological variables related to whole-plant water status in Scots pines (Pinus sylvestris)
Response variableModel termsEstimateSE t P
  1. Defoliation class (defoliated, ≤ 50% green leaves; nondefoliated, > 50% green leaves) was also introduced as a factor in the linear models; nondefoliated is used as the reference level. All models were fitted using generalized least squares to account for variance heterogeneity and temporal autocorrelation, where necessary. Parameters with a significant effect of defoliation class are shown in bold. See Table 1 for abbreviation and symbol descriptions.

k S-L Intercept−0.9620.366−2.6280.0144
θ d : nondefoliated 9.048 2.690 3.364 0.0024
Defoliated 2.086 0.860 2.426 0.0225
θ d : defoliated 16.557 6.514 2.542 0.0173
Defoliated 0.128 0.056 2.27 0.0380
Ψ md Intercept−1.1410.087−13.14<0.0001
Ψ pd 0.5310.0598.992<0.0001
log(– Ψ l , pd ): nondefoliated 1.260 0.534 2.360 0.0212
Defoliated 1.103 0.223 4.952 <0.0001
log(–Ψpd): defoliated−0.8790.460−1.9100.0605

Near-homeostasis in needle water potential across years and crown condition

Minimum needle water potentials were at the lower limit of the range reported for European Scots pine populations, including some drought-exposed sites (Martínez-Vilalta et al., 2009). However, a perfect homeostasis in midday leaf water potentials could not be strictly maintained under the extremely dry conditions in 2011 (Fig. 3c,d). Hypothesis 1 was only partly confirmed, as a perfect homeostasis in Ψmd was not observed either across defoliation classes, as defoliated pines displayed (slightly but) consistently lower Ψmd in comparison with nondefoliated pines (Fig. 3c,d).

Despite this lack of perfect regulation, Ψmd values were still far from site-specific estimates of branch-level Ψ50 (water potential causing 50% loss of xylem conductivity), which did not vary with crown condition (−3.01 MPa; Gómez, 2012). However, for both defoliation classes, root xylem was comparatively more vulnerable to embolism (Ψ50 = −1.87 MPa), with roots of defoliated trees showing a steeper vulnerability curve (Gómez, 2012). Hence, assuming that Ψpd is representative of root water potential (Fig. 3a,b), defoliated pines would have lost 61% of root xylem conductivity during the peak drought in 2011, compared with 53% loss in nondefoliated pines. This result suggests that small hydraulic differences between defoliation classes could contribute to their differential water use and, ultimately, to a greater risk of mortality in defoliated individuals (cf. Galiano et al., 2011).

Soil-to-leaf hydraulic conductance declines more rapidly with drought in defoliated pines

Leaf-specific hydraulic capacity in both defoliated and nondefoliated Scots pines declined with drought and failed to recover quickly after the autumn rains in 2010 (Fig. 3e,f). This seasonal pattern was also observed in Pinus palustris (Addington et al., 2004) and is consistent with the lack of an immediate recovery in transpiration after drought in Pinus pinaster (Duursma et al., 2008). The faster decline of kS-L in response to drought in defoliated pines (hypothesis 3) may reflect differences in local water availability, as suggested by the lower Ψpd value in defoliated than in nondefoliated pines (Fig. 4b). A combination of slightly higher θmd in the vicinity of the roots of nondefoliated pines, together with their slightly higher resistance to xylem embolism, probably explains the differences in kS-L dynamics observed in this study (Fig. 3). Overall, these results are consistent with the negative association between defoliation and soil parameters related to increased water-holding capacity observed in another declining Scots pine population (Galiano et al., 2010).

Soil heterogeneity may also mediate root acclimation to local soil hydraulic properties (Addington et al., 2006) differently across crown condition classes, which would explain the higher Ψpd at a given soil moisture observed for nondefoliated pines (Fig. 4a). However, a recent study on a similarly drought-exposed Scots pine population, showing that crown transparency did not affect either root standing crop or root morphology (Brunner et al., 2009), does not support this explanation. Alternatively, fungal infection-driven impairment of stems and/or roots may underlie their steeper seasonal decline in kS-L (Croise et al., 2001; Heiniger et al., 2011), a possibility that remains to be explored in our study population.

Defoliation and physiological mechanisms underlying drought-induced mortality

Defoliated trees showed reduced needle number, shoot length and branching, and increased SLA (Table S2), in accordance with drought-induced constraints on canopy development reported elsewhere for Scots pine (Irvine et al., 1998; Thabeet et al., 2009; Dobbertin et al., 2010). For these trees, reduced leaf area, enhanced sensitivity to drought and a lack of compensatory increase in needle photosynthetic capacity (as inferred from leaf N content and gas exchange data; Table 2; Y. Salmon, University of Edinburgh, unpublished) all probably contribute to reduced NSCs relative to nondefoliated pines (hypothesis 4). Moreover, the long periods (up to 5 months) of severely restricted gas exchange do not appear to be compensated at the tree level by more favourable conditions for C assimilation because of the increased hydraulic capacity during spring. In any case, regardless of crown condition, NSCs were very low relative to those measured in more mesic Scots pine stands (Hoch et al., 2003; Gruber et al., 2011). Although more detailed observations on the dynamics of NSCs across multiple plant tissues and over longer time periods are required, our results suggest that persistent constraints on C gain during long droughts may be critical to determine the survival of pines. Pinus edulis has been reported to recover after < 4 months of restricted gas exchange (Breshears et al., 2009), but longer droughts may act as contributing factors preceding tree death (Breshears et al., 2009; Plaut et al., 2012). A recent study in another Scots pine population affected by drought-induced decline in the central Pyrenees found a close association between defoliation, similarly low stem NSC values to those reported here and increased mortality risk in the event of a new drought (Galiano et al., 2011).

Our results are consistent with the C starvation hypothesis (McDowell et al., 2008; Galiano et al., 2011), although the numerous feedbacks between plant hydraulics and C availability and translocation within the plant (Sala et al., 2010) preclude the identification of the ultimate factor(s) triggering tree death (McDowell, 2011). For the studied drought-exposed Scots pine population, reduced NSCs mediated by poor crown condition and prolonged stomatal closure may feed back on tree survival and resilience under drought by limiting leaf production through bud impairment (Bréda et al., 2006), increasing vulnerability to pathogen attacks (Wermelinger et al., 2008) and preventing the recovery of hydraulic function after drought. Radial growth is severely restricted in our study area (Hereş et al., 2012), especially in defoliated pines (A. Vilà-Cabrera et al., CREAF, Barcelona, unpublished). As post-drought recovery of kS-L in conifers is highly dependent on new xylem growth (Brodribb et al., 2010), reduced C allocation to xylem growth might also constrain the hydraulic capacity of drought-exposed trees. Moreover, enhanced C limitations in defoliated pines may render them more vulnerable to fungal infections (Heiniger et al., 2011), and associated hydraulic dysfunction (Croise et al., 2001). Finally, other mesic-adapted tree species have shown continuous C investment to fine roots to compensate for increased fine root mortality under drought (Meier & Leuschner, 2008), possibly enhancing the depletion of local carbohydrate reserves.

Concluding remarks

In conclusion, this study shows that drought-induced defoliation may be more adequately seen as an inevitable consequence of drought in inherently vulnerable trees than as a strategy to cope with water stress. This distinction is important in the context of climate change and the widespread increase in defoliation recently observed in many southern European forests, including Scots pine populations (Carnicer et al., 2011). Extreme hydraulic adjustments may allow short-term survival, but jeopardize future resilience in response to drought. In this regard, Mediterranean populations of Scots pine may be especially vulnerable to the longer droughts projected for the region (Bates et al., 2008). The replacement of Scots pine by more drought-resistant vegetation will eventually depend on other demographic processes, past management history, site factors (Vilà-Cabrera et al., 2013) and species interactions (Lloret et al., 2012).


Thanks to J. Llagostera, X. Buqueras, A.Vallvey and all the staff at PNIN de Poblet for allowing us to carry out research at the ‘Barranc del Tillar’ nature reserve and for their logistic support in the field. We also acknowledge the field and laboratory support by J. Barba, J. Curiel, M. Mejía-Chang, P. García, P. Llorens, A. Vilà, H. Romanos, B. Ros and E. Sànchez. M. Ninyerola and M. Batalla (Unitat de Botànica, UAB) provided the climatic database (CGL2006-01293, Spanish Ministry of Science and Innovation, MICINN) and E. Marimon some meteorological data (www.meteoprades.net). The comments by David Ackerly and two anonymous reviewers greatly improved the manuscript. This research was funded by MICINN via competitive grants CGL2007-60120, CGL2010-16373, CSD2008-0004, a Juan de la Cierva postdoctoral fellowship awarded to R.P., Natural Environment Research Council (NERC) grant NE/I011749/1 to M.M. and an FPI doctoral fellowship awarded to L.G. D.A. was supported by an FPU doctoral fellowship from the Spanish Ministry of Education.