Increased water-use efficiency does not lead to enhanced tree growth under xeric and mesic conditions

Authors


Summary

  • Higher atmospheric CO2 concentrations (ca) can under certain conditions increase tree growth by enhancing photosynthesis, resulting in an increase of intrinsic water-use efficiency (iWUE) in trees. However, the magnitude of these effects and their interactions with changing climatic conditions are still poorly understood under xeric and mesic conditions.
  • We combined radial growth analysis with intra- and interannual δ13C and δ18O measurements to investigate growth and physiological responses of Larix decidua, Picea abies, Pinus sylvestris, Pinus nigra and Pseudotsuga menziesii in relation to rising ca and changing climate at a xeric site in the dry inner Alps and at a mesic site in the Swiss lowlands.
  • iWUE increased significantly over the last 50 yr by 8–29% and varied depending on species, site water availability, and seasons. Regardless of species and increased iWUE, radial growth has significantly declined under xeric conditions, whereas growth has not increased as expected under mesic conditions. Overall, drought-induced stomatal closure has reduced transpiration at the cost of reduced carbon uptake and growth.
  • Our results indicate that, even under mesic conditions, the temperature-induced drought stress has overridden the potential CO2 ‘fertilization’ on tree growth, hence challenging today's predictions of improved forest productivity of temperate forests.

Introduction

Understanding the acclimation capacity of different tree species to the combined effects of increasing atmospheric CO2 concentration (ca) and temperature (along with decreasing water availability) is important for the assessment of the vulnerability of European forests to global climate change (see also Lindner et al., 2010; Sarris et al., 2013). Hotter and drier climatic conditions can lead to severe reductions of tree growth and even cause extensive tree mortality and forest decline under extreme conditions (Allen et al., 2010; Rigling et al., 2013), fundamentally affecting forest communities, functions and services (Anderegg et al., 2013). While drought stress impedes tree growth and vigor, rising ca may, in theory, promote tree growth by enhancing photosynthetic activity and increasing intrinsic water-use efficiency (iWUE) of trees (Ainsworth & Rogers, 2007; Keenan et al., 2013). Therefore, under moderate dry conditions, elevated ca may improve tree gas exchange and survival. However, if droughts become severe and long-lasting, reduction in photosynthetic activity as a result of stomatal closure and biochemical down-regulation may cancel out direct benefits of elevated ca (Franks et al., 2013). Despite these facts, the combined influence of rising ca, decreasing water availability and warming on long-term growth and gas-exchange responses of trees under natural conditions is still poorly understood.

Species-specific long-term responses of iWUE and growth to rising ca have often been analyzed based on the stable carbon isotopic composition in tree rings and radial growth of trees (Peñuelas et al., 2011; Silva & Anand, 2013 and references therein). While iWUE generally increases with rising ca, the likelihood that a tree will show enhanced growth in response to elevated ca depends strongly on site conditions. While growth rates usually increase with rising ca in moist temperate forests (Cole et al., 2010; McMahon et al., 2010), declines in growth are often observed for tropical and dry temperate forests (Nock et al., 2011; Silva & Anand, 2013). In Europe, declines in tree growth were reported for some species at dry sites in the Iberian Peninsula following warming, despite increasing iWUE (Peñuelas et al., 2008; Andreu-Hayles et al., 2011; Linares & Camarero, 2012), whereas Koutavas (2013) and Martinez-Vilalta et al. (2008) found evidence of growth enhancement for Greek fir in western Greece and Scots pine in northeastern Spain. It is therefore unclear to what extent rising ca and changing climatic conditions increase iWUE and affect growth rates of different species under xeric and mesic conditions. Still, by analyzing changes in iWUE and growth rates alone, it is impossible to determine the physiological adjustments (i.e. changes in assimilation rates and/or stomatal conductance) responsible for the changes in iWUE. Therefore, a detailed account of long-term changes in Δ13C, growth rates, and iWUE, along with the analysis of δ18O in tree rings, is essential to understand the effects of rising ca and changing climatic conditions on growth and gas exchange of trees. However, very few studies have done such a detailed assessment so far (but see Nock et al., 2011; Battipaglia et al., 2013; Sarris et al., 2013). Furthermore, little information is available about the long-term species-specific growth and gas-exchange responses to rising ca and decreasing water availability in the Alps and central Europe where unprecedented warming and drought periods have been reported (Ciccarelli et al., 2008; Ceppi et al., 2012).

The carbon (δ13C) and oxygen (δ18O) stable isotope compositions in tree rings are valuable proxies for retrospectively understanding ecophysiological responses of trees to rising ca and changing climatic conditions over seasonal to decadal timescales. δ13C in tree rings can be used to estimate the degree of drought stress experienced by trees during photosynthesis (Eilmann et al., 2010). The δ13C ratio is directly related to CO2 uptake, water balance and iWUE, that is, the ratio between photosynthetic assimilation rates (A) and stomatal conductance (gs) during photosynthesis (Farquhar et al., 1982, 1989). Still, shifts in δ13C and iWUE can be caused by variations in A or gs, or both, and both δ13C and iWUE can change because of changes in atmospheric δ13C and ca over time. By contrast, δ18O in tree rings is mainly influenced by the isotopic signature of source water; evaporative enrichment at the leaf level, as a result of stomatal transpiration (Dongmann et al., 1974; Farquhar & Lloyd, 1993); ambient air humidity and the isotopic composition of water vapor; and biochemical fractionation during oxygen incorporation (Sternberg, 2009). As δ18O in tree rings is independent of variations in A and is linked to variations in leaf water in response to transpiration, δ18O can be used to separate the effects of A and gs on iWUE because it records changes in gs (Scheidegger et al., 2000). On the other hand, δ13C and δ18O in early- and latewood carry different physiological information. Earlywood formation depends partially on stored photosynthates from the previous years, whereas latewood formation relies essentially on current-year photosynthates (Helle & Schleser, 2004) and records the current environmental impacts on trees. Further, by analyzing δ18O in early- and latewood separately, seasonal source water of trees can be inferred (An et al., 2012). Thus, important environmental and physiological information could be undetected when focusing on the whole annual ring or the latewood section only.

Here, we present for the first time a detailed account of the long-term growth and gas-exchange responses at the intra- and interannual scales of five co-occurring tree species growing under xeric and mesic conditions to rising ca and changing climate. Specifically, we analyzed the basal area increment (BAI) as well as the δ13C and δ18O compositions in early- and latewood in five conifers at two ecologically contrasting sites in the dry inner Alps and Swiss moist lowlands for the period 1960–2009. Three of the studied species are native to the sites, that is, Larix decidua Mill, Picea abies (L.) Karst. and Pinus sylvestris L., whereas Pinus nigra Arn. and Pseudotsuga menziesii var. menziesii (Mirb.) Franco are nonnative. Our specific objectives were as follows: to test if rising atmospheric CO2 concentrations and changing climatic conditions (i.e. temperature, precipitation and air humidity) in the dry inner Alps and Swiss moist lowlands have been causing systematic changes in growth rates and iWUE of these species; to determine how changes in iWUE are related to radial growth rates; and to assess species-specific physiological adjustments (changes in photosynthetic assimilation rate and/or stomatal conductance) in dry vs moist years.

Materials and Methods

Study sites and species

We selected two afforestations with contrasting climate and site conditions (xeric vs mesic). The xeric site is located in the dry inner-Alpine Aosta Valley in north-western Italy (45°46′59″N, 7°32′52″E; 1150 m above sea level (asl)) on a steep south-facing slope (75%). This afforestation was established between 1905 and 1910 at an initial density of c. 1200 trees ha−1, but drought-induced mortality in the first years after planting reduced the stand density considerably (Vescoz, 1909). The actual stand is composed of low-stature trees (Table 1) and is relatively open (c. 400 tree ha−1; C. Letey, pers. comm.), so that intertree competition is minimal. The climate is dry in summer and the mean annual temperature and precipitation sum are 7.3°C and 682 mm (norm period 1961–1990, Fig. 1). The soil is a regosol with an available water capacity of 51 mm (P. Weber, unpublished). Under these site conditions, L. decidua, P. abies and P. sylvestris grow near their dry distribution limit in the inner Alps (Ellenberg, 1988) and are thus expected to be more vulnerable to climate change. The mesic site is located 3 km from the town of Biel in Switzerland (47°09′57″N, 7°16′06″E; 750 m asl) on a south-facing slope (35%) in the foothills of the Jura Mountains. This afforestation was established in 1889 (Marcet, 1975) at a density of c. 200 trees ha−1 (Stauffer & Adams, 1993). In recent decades, occasional harvesting of single trees has occurred to maintain the stand relatively open and to minimize intertree competition (B. Hadorn, pers. comm.). This site is close to the southern and altitudinal distribution limits of L. decidua and P. abies in the Jura Mountains (Ellenberg, 1988), where it is unclear how the expected decrease in water availability will affect their physiological and growth performance in the future. The climate is temperate with a mean annual temperature of 8.0°C and mean annual precipitation of 1203 mm (norm period 1961–1990, Fig. 1). The soil is a leptosol with an available water capacity of 119 mm (P. Weber, unpublished). At both sites, a mixed stand comprising the native P. abies, P. sylvestris and L. decidua, and the nonnative P. nigra and P. menziesii was sampled. The age of the sampled trees varied between 100 and 115 yr. Thus, all trees were at least 50 yr old at the beginning of the investigation period in 1960.

Table 1. Mean characteristics of the trees studied at the xeric (Aosta) and mesic (Biel) sites
SiteSpeciesNo. of treesAgeDBH (cm)Height (m)Mean BAI 1930–2009 (cm2 yr−1)Mean BAI 1951–1980 (cm2 yr−1)ΔBAI/Δt 1951–1980 (cm2 yr−1)Mean BAI 1980–2009 (cm2 yr−1)ΔBAI/Δt 1980–2009 (cm2 yr−1)
  1. Values are means ± 1 SE. Mean basal area increment (BAI) for the periods 1930–2009, 1951–1980, and 1980–2009 and rates of change yr–1 of BAI (ΔBAI/Δt) for the periods 1951–1980 and 1980–2009 were calculated. Trends in BAI time series were tested with Mann–Kendall trend test. Rates of changes correspond to the Theil–Sen trend estimates. Significance levels: *, < 0.05; **, < 0.01; ***, < 0.001. DBH, diameter at breast height.

Aosta (xeric) Pinus nigra 1510637.2 ± 1.415.6 ± 0.99.53 ± 0.5410.86 ± 0.790.314***11.12 ± 0.70−0.154*
Pseudotsuga menziesii 1510142.4 ± 1.718.9 ± 0.313.88 ± 0.7214.58 ± 0.950.10915.81 ± 1.12−0.313*
Larix decidua 1510425.0 ± 1.213.0 ± 0.54.47 ± 0.344.16 ± 0.470.0304.71 ± 0.53−0.143*
Pinus sylvestris 1510624.7 ± 1.19.5 ± 0.74.29 ± 0.334.51 ± 0.490.198**4.72 ± 0.53−0.096
Picea abies 1210426.4 ± 2.214.1 ± 0.67.52 ± 0.537.59 ± 0.720.193*9.39 ± 0.77−0.228*
Biel (mesic) P. nigra 1510048.6 ± 1.325.0 ± 0.512.74 ± 0.5013.71 ± 0.760.211***13.30 ± 0.66−0.111
P. menziesii 1510078.8 ± 2.739.1 ± 1.046.70 ± 1.7246.44 ± 1.990.41553.61 ± 2.390.161
L. decidua 1510253.4 ± 1.631.3 ± 1.020.07 ± 1.1720.78 ± 1.420.11419.06 ± 1.99−0.563**
P. sylvestris 1511541.8 ± 1.223.1 ± 0.99.87 ± 0.5710.55 ± 0.73−0.0966.71 ± 0.51−0.063
P. abies 1410053.4 ± 2.334.3 ± 1.222.81 ± 1.2121.37 ± 1.260.17325.34 ± 1.750.116
Figure 1.

Mean monthly precipitation sum (gray bars, ± 1 SE) and mean monthly temperature (black line) for the norm period (1961–1990) at the xeric site in Aosta (a) and at the mesic site in Biel (b).

Dendrochronological methods

At each site, 12–15 codominant or dominant healthy trees per species were sampled for dendroecological analysis. Two increment cores were taken from each tree at c. 50 cm height. The wood samples were air-dried and their surfaces were prepared using a core-microtome. Earlywood, latewood and total ring widths were measured separately to the nearest 0.01 mm using a stereomicroscope linked to a LINTAB digital positioning table and the software TSAP (Rinntech, Heidelberg, Germany). Early- and latewood tree-ring sections were defined according to visual aspects (darkening and cell size). For species with gradual transition between early- and latewood (P. abies), we defined the earlywood/latewood boundaries as the middle of the transition zone. The individual tree-ring series were visually crossdated and verified with the program COFECHA (Holmes, 1983). To overcome the problem of declining tree-ring width with increases in tree diameter and to detect long-term growth changes, individual early- and latewood width series were converted to earlywood BAI and latewood BAI (Biondi & Qeadan, 2008). For each species and site, BAI for each year was averaged over all individuals and calculated for the period 1930–2009.

Stable isotope analysis

The four trees showing the highest correlation with the species-specific site tree-ring chronology were selected for isotopic analysis. The eight cores per species and site were put in a Soxhlet apparatus with 95% ethanol for 24 h to extract resin and mobile extractives. The early- and latewood of each annual ring for the period 1960–2009 were separated from each core (two cores per tree) with a scalpel under a stereomicroscope and the samples from the four different trees (eight cores) of the same year were pooled regardless of mass (Leavitt, 2008). The pooled samples were milled and homogenized using an ultracentrifugal mill (ZM 200, Retsch, Haan, Germany), weighted and put into a tin capsule for 13C and a silver capsule for 18O mass spectrometer analysis. In total, 1300 samples per stable isotope were measured. To determine the C isotope ratio, the samples were combusted to CO2 under an excess of oxygen at 1020°C in an elemental analyzer (EA-1110, Carlo Erba Thermoquest, Milan, Italy) linked to a Delta S mass spectrometer (Finnigan MAT, Bremen, Germany) operating in continuous-flow mode. To determine the oxygen isotope ratio, the samples were pyrolyzed to CO at 1080°C in an elemental analyzer (EA-1108, Carlo Erba Thermoquest, Milan, Italy), which was connected to a DELTAplus XP mass spectrometer (Finnigan MAT, Bremen, Germany) via a variable open split interface (Finnigan ConFlo III, Thermo Electron, Bremen, Germany). The isotopic ratio was expressed in delta notation in per mil (‰) as the relative deviation from the international standard (V-PDB for 13C/12C and V-SMOW for 18O/16O). The precision of the analysis was ≤ 0.10‰ for δ13C and ≤ 0.25‰ for δ18O.

Calculations of water-use efficiency of trees

Tree water-use efficiency (WUE) at any time during photosynthesis corresponds to the ratio of net CO2 assimilation to transpiration rates, meaning that the greater the WUE, the more carbon is fixed per unit water lost (Farquhar et al., 1989).

Carbon isotopic discrimination (Δ13C) occurring in C3 plants is a result of preferential use of the lighter 12C atoms over the heavier 13C atoms during photosynthesis and is defined as:

display math(Eqn 1)

where δ13Catm is the isotopic value of atmospheric CO2, and δ13Ctree is the isotopic value of tree rings. According to Farquhar et al. (1982), Δ, as a first approximation, is linearly related to the ratio of intercellular (ci) to atmospheric (ca) CO2 mole fractions as expressed by (Eqn 2):

display math(Eqn 2)

where a (4.4 ‰) is the fractionation during CO2 diffusion between the ambient atmosphere and the intercellular spaces (O'Leary, 1981), and b (27 ‰) is the fractionation during carboxylation in C3 plants. We used δ13Catm estimated values for the period 1960–2003 from McCarroll & Loader (2004), and measured values for the period 2004–2009 available online (http://www.esrl.noaa.gov/gmd/). Measured ca values for the period 1960–2009 were obtained from Keeling et al. (2009) (http://scrippsco2.ucsd.edu/). It should be noted that Δ13C is determined by the ratio of the chloroplast to ambient CO2 mole fraction (cc/ca) rather than ci/ca, making it sensitive to mesophyll conductance (Seibt et al., 2008). Further, mesophyll conductance varies according to changes in environmental conditions, such as light, temperature, CO2 and water availability (Flexas et al., 2008). Thus, using ci may be problematic if mesophyll conductance to CO2 is not constant (Seibt et al., 2008). However, information on mesophyll conductance of the investigated species is missing and using mean values of mesophyll conductance from the literature would not improve results (Cernusak et al., 2013). Therefore, the simpler linear model of Farquhar et al. (1982) was used, which also allowed comparison with other studies.

Intrinsic water-use efficiency can be estimated according to Farquhar & Richards (1984) from values of Δ13C and ca as follows:

display math(Eqn 3)

where 1.6 is the ratio of diffusivities between water vapor and CO2 in air.

Climate data and standardized water deficit index (SWDI)

Mean daily temperature and precipitation sum were obtained from the nearest weather stations to the study sites for the period 1960–2009. For the mesic site, the closest station (433 m asl) was 3 km away in Biel (MeteoSwiss). For the xeric site in Aosta Valley, the nearest station (750 m asl) was 5 km away at the hydroelectric power station in Promiod Covalou (Department of Soil Conservation and Water Resources, Autonomous Region of Valle d'Aosta, Italy). Missing data for the station in Promiod Covalou were filled in by linear regressions with the nearby stations of Saint-Christophe (13 km) and Brusson (15 km). Owing to the difference in altitude between the weather stations and the study sites, temperature data were adjusted using lapse rates (Supporting Information, Table S1).

Leaf-to-air vapor pressure deficit (VPD), that is, the difference between the vapor pressure inside the stomata (es) and actual vapor pressure (ea), was calculated at both study sites using air temperature (T) data from the meteorological stations and ea data from the gridded data set CRU TS3.21 (Harris et al., 2013). Vapor pressure inside the stomata was estimated according to Allen et al. (1998), as follows:

display math(Eqn 4)

Mean daily temperature and precipitation sum were used to compute SWDI. For the calculation of this index, monthly water budget, that is, the difference between potential and actual evapotranspiration (PET − AET), was calculated using a soil water balance model based on a modified Thornthwaite method (Willmott et al., 1985). Monthly soil water budgets (SWB) were then aggregated over 9 months for the site in Aosta and over 6 months for the site in Biel based on a modified version of the aggregation procedure by Vicente-Serrano et al. (2013), as follows:

display math(Eqn 5)

where k (in months) is the timescale of aggregation and n is the calculation number. The SWB time series were then adjusted to a log-logistic distribution with a Gaussian kernel, so that the past months have a decreasing weight, and standardized using the algorithm developed by Vicente-Serrano et al. (2010) within the R package SPEI (Beguería & Vicente-Serrano, 2012). This aggregation procedure and the number of months included were chosen to reflect the memory effect of the soil water reservoir and corresponded to the time window at which trees were most sensitive to soil water deficits at both sites (Lévesque et al., 2013). Values of SWDI usually vary between −3 (extremely dry) and +3 (extremely wet), while values close to 0 indicate normal moisture conditions.

Data analysis

Temperature, precipitation, VPD, SWDI, BAI, Δ13C, iWUE and δ18O temporal trends and their significance were tested with Mann–Kendall trend tests and Theil–Sen trend estimates after prewhitening in order to remove the first-order autocorrelation using the R package zyp (Bronaugh & Werner, 2012). Partial correlations were calculated to disentangle the relative effect of temperature and VPD vs ca on tree iWUE. Partial correlation is a statistical test used to assess the relationships between two variables while controlling for the effects of a third variable. Pearson correlations were used to test the influence of water deficit (SWDI) on early- and latewood BAI, Δ13C and δ18O. Earlywood BAI, Δ13C and δ18O were correlated with the mean SWDI values calculated from September of the year before growth to June of the year of growth. Latewood BAI, Δ13C and δ18O of the species were correlated with the mean SWDI values calculated from June to September of the year of growth. These selected months corresponded to the periods in which the early- and latewood formation depended at most on water availability (Lévesque et al., 2013). Relationships between BAI and iWUE were tested using least-square linear regressions. Correlations were used to assess the influence of δ18O in precipitation, precipitation amount and VPD on δ18O in tree rings.

To differentiate whether changes in iWUE in moist vs dry years are the result of changes in photosynthetic assimilation rates and/or stomatal conductance, we used the conceptual model of Scheidegger et al. (2000). The dual-isotope conceptual model was originally developed for leaf organic matter and not for interpreting stable isotope variations in tree rings over time, as source water and water vapor δ18O must remain constant over time (Roden & Siegwolf, 2012). To assess the magnitude of the changes in source water over the study period, monthly measurements of δ18O in precipitation were obtained from the Global Network of Isotopes in Precipitation (GNIP). We used the monthly δ18O records from the closest station (Bern) to our study sites for the period 1970–2008 available online (http://www.univie.ac.at/cartography/project/wiser/). The difference of monthly δ18O in precipitation recorded in Bern and at the study sites was assessed with the Online Isotopes in Precipitation Calculator (Bowen et al., 2005; Bowen, 2013). Variations in source water and water vapor δ18O were also reduced by sampling trees in close proximity to each other and by calculating averages of isotopic composition over all moist (SWDI > 0) or dry (SWDI < 0) years in order to reduce the year-to-year variability. Changes in isotopic values were considered to represent a canopy temporal integration of A and gs over either the early or late growing season (early- or latewood) as discussed by Barnard et al. (2012), and hereafter referred as Aint and gs int. Still, the species-specific changes cannot be easily and directly compared, as different tree species can have different rooting pattern and depth, thus having access to different source water. Here, the dual-isotope conceptual model was used as a first approximation of gas-exchange response of trees to soil water deficits.

Results

Climatic conditions and trends

At the xeric site, the total annual precipitation and mean annual temperature varied between 434 and 1183 mm and 5.4 and 9.6°C, respectively, during the period 1960–2009. In this period, annual temperature increased significantly, by 2.2°C, which was mainly caused by the rise in temperature observed since the beginning of the 1980s (Table 2, Fig. S1). For the period 1980–2009, a significant warming trend (2.1°C) was observed during the growing season (April–September) while precipitation remained unchanged. At the mesic site, the total annual precipitation and mean annual temperature varied between 807 and 1575 mm and 6.7 and 9.6°C, respectively, over the 1960–2009 period. The mean annual temperature increased significantly, by 2.3°C, during this period. For the period 1980–2009, mean growing season temperature (April–September) increased by 1.5°C. Total annual precipitation did not change significantly over the study period at either site, but tended to decrease at the mesic site after 1980 (Table 2). As rising temperatures were not followed by increases in precipitation, VPD increased significantly at both study sites, whereas soil water availability remained unchanged or tended to decrease (Table 2).

Table 2. Temperature, precipitation, vapor pressure deficit (VPD), and soil water deficit index (SWDI) trends at the xeric site in Aosta and the mesic site in Biel
Climate variableAosta (xeric)Biel (mesic)
Growing season (April–September)AnnualGrowing season (April–September)Annual
  1. Trends in time series were tested with Mann–Kendall trend test. Changes correspond to the Theil–Sen trend estimates. Significance levels of changes within the periods: *, < 0.05, **, < 0.01.

Temperature (°C)
Mean 1960–200913.537.7813.478.00
Change 1960–20091.292.20*1.84**2.32**
Mean 1980–200913.678.0913.828.33
Change 1980–20092.08**1.54**1.50**1.62**
Precipitation (mm)
Mean 1960–20093847015871184
Change 1960–200966−42218
Mean 1980–20094027005971200
Change 1980–20090−5−78−235
VPD (kPa)
Mean 1960–20090.760.540.450.30
Change 1960–20090.020.030.09*0.06*
Mean 1980–20090.770.550.460.31
Change 1980–20090.13**0.09*0.07*0.05*
SWDI
Mean 1960–20090.000.010.000.01
Change 1960–2009−0.070.29−0.46*−0.01
Mean 1980–2009−0.040.03−0.07−0.01
Change 1980–2009−0.020.35−0.090.19

Trends in BAI, Δ13C, iWUE and δ18O

Long-term changes in BAI, Δ13C, iWUE and δ18O were assessed to test if rising ca and temperature have caused significant shifts in growth rates and gas exchange for the period 1960–2009. At the xeric site, all species showed an increasing trend in BAI until c. 1980 (Fig. 2). However, since the early 1980s, all species except P. sylvestris have shown significant growth declines (Table 1), coinciding with the steady increase in temperature since 1980 (Fig. S1). In contrast to the xeric site, radial growth remained relatively constant in time at the mesic site (Fig. 2) despite the constant increase in temperature since the 1960s (Fig. S1). Significant growth declines were only found for L. decidua at the mesic site (Table 1).

Figure 2.

Mean annual basal area increment (solid line) for the period 1930–2009 for five species at the xeric site in Aosta (left column) and at the mesic site in Biel (right column). The gray shaded area represents ± 1 SE. The dashed line represents a locally weighted polynomial regression.

At the xeric and mesic site, iWUE increased significantly, by at least 8–29%, during the period 1960–2009 (Table 3, Figs 3, 4), and was significantly related to the rise in atmospheric CO2 concentrations (Table 4). The iWUE of P. nigra and P. sylvestris at the xeric site and in latewood of all species at the mesic site was positively correlated with temperature and VPD (Table 4). No significant trends were found for the Δ13C chronologies for the period 1960–2009 except for the latewood section of P. sylvestris at the xeric site and the earlywood section of P. menziesii at the mesic site, which showed a significant positive increase (Table 3). Over this period, the ratio of intercellular to ambient CO2 concentration remained constant for most species (data not shown), which resulted in improved iWUE. At the xeric site, the increase in iWUE was highest in P. abies (0.53 μmol mol−1 yr−1) and lowest in P. sylvestris (0.16 μmol mol−1 yr−1; Table 3). The increase in iWUE was similar in both early- and latewood of P. menziesii, L. decidua and P. abies, whereas it was more pronounced in the earlywood than in the latewood of P. nigra and P. sylvestris. At the mesic site, there was a significant increase in iWUE in all species, except for the earlywood of P. menziesii, during the period 1960–2009 (Table 3, Fig. 4). As for the xeric site, P. abies had the highest increase in iWUE, with an increase of 0.56 μmol mol−1 yr−1. P. menziesii showed the lowest increase with a rise of only 0.11 μmol mol−1 yr−1 (Table 3). The increase in iWUE was higher in the latewood than in the earlywood of P. menziesii, L. decidua and P. abies. The increases in iWUE of P. nigra and P. menziesii were less pronounced than at the xeric site, whereas P. abies showed similar increases in iWUE at both sites.

Table 3. Mean discrimination (Δ13C), δ18O and intrinsic water-use efficiency (iWUE) values, and their rates of change yr–1 for the period 1960–2009 for the xeric site in Aosta and the mesic site in Biel for the earlywood (EW) and latewood (LW) sections
SiteSpeciesRing sectionΔ13C (‰)Change Δ13C (‰ yr−1)δ18O (‰)Change in δ18O (‰ yr−1)iWUE (μmol mol −1)Change in iWUE (μmol mol−1 yr−1)
  1. The rates of change and their significance were calculated with the Mann–Kendall trend test and Theil–Sen trend estimates. Significance levels: *, < 0.05; **, < 0.01; ***, < 0.001.

Aosta (xeric) Pinus nigra EW17.23−0.01027.010.00494.00.491***
LW16.110.00827.040.002104.40.375***
Pseudotsuga menziesii EW16.97−0.00724.710.00996.30.468***
LW15.86−0.00526.100.006107.00.493***
Larix decidua EW16.260.00625.56−0.001102.90.365***
LW15.700.00725.760.009108.50.360***
Pinus sylvestris EW17.03−0.00226.05−0.01095.70.414***
LW15.970.031***26.01−0.019105.70.156**
Picea abies EW15.350.00026.320.018111.90.507***
LW14.350.00226.320.019121.40.530***
Biel (mesic) P. nigra EW18.550.00225.100.00481.00.327***
LW18.140.00525.800.00385.10.311***
P. menziesii EW16.150.034***23.730.001103.90.112
LW15.920.02225.31−0.005106.20.242*
L. decidua EW18.84−0.00624.330.00678.30.385***
LW18.74−0.01625.350.00379.30.458***
P. sylvestris EW18.30−0.00223.540.00683.60.370***
LW17.830.00424.540.00288.00.319***
P. abies EW17.05−0.00124.540.019*95.60.518***
LW16.72−0.01325.310.01398.90.564***
Table 4. Partial correlations between intrinsic water-use efficiency (iWUE), mean temperature (Tmean), vapor pressure deficit (VPD) and atmospheric CO2 concentrations
SiteSpeciesEarlywoodaLatewooda
iWUE vs TmeanbiWUE vs VPDbiWUE vs CO2ciWUE vs TmeanbiWUE vs VPDbiWUE vs CO2c
  1. a

    For the earlywood section, the mean temperature or VPD from April to June was used whereas for the latewood section the mean temperature from June to August was used.

  2. b

    Partial correlation controlled for changes in atmospheric CO2.

  3. c

    Partial correlation controlled for changes in temperature.

  4. Significance levels: *, < 0.05; **, < 0.01; ***, < 0.001.

Aosta (xeric) Pinus nigra 0.220.37**0.71***0.31*0.43**0.76***
Pseudotsuga menziesii 0.030.020.70***0.180.220.77***
Larix decidua −0.100.190.75***0.090.27*0.77***
Pinus sylvestris 0.31*0.27*0.66***0.150.200.47***
Picea abies 0.090.090.83***0.080.100.86***
Biel (mesic) P. nigra 0.100.140.69***0.43**0.57***0.36**
P. menziesii −0.230.050.45**0.27*0.35**0.29*
L. decidua −0.090.000.72***0.42**0.51***0.39**
P. sylvestris 0.140.42**0.83***0.150.27*0.60***
P. abies −0.27*−0.090.85***0.230.40**0.69***
Figure 3.

Tree-ring Δ13C in earlywood (a) and latewood (b), intrinsic water-use efficiency (iWUE) in earlywood (c) and latewood (d), and δ18O in earlywood (e) and latewood (f) for five tree species (Larix decidua, Picea abies, Pinus sylvestris, Pinus nigra and Pseudotsuga menziesii) at the xeric site in Aosta for the period 1960–2009.

Figure 4.

Tree-ring Δ13C in earlywood (a) and latewood (b), intrinsic water-use efficiency (iWUE) in earlywood (c) and latewood (d), and δ18O in earlywood (e) and latewood (f) for five tree species (Larix decidua, Picea abies, Pinus sylvestris, Pinus nigra and Pseudotsuga menziesii) at the mesic site in Biel for the period 1960–2009.

At the xeric site, no significant trend was detected for the δ18O chronologies (Table 3). δ18O in earlywood of all species except L. decidua was significantly correlated with δ18O in precipitation (Table 5). Latewood δ18O of all species did not correlate with δ18O in precipitation and VPD, whereas significant correlations were found between latewood δ18O of all species except L. decidua and precipitation amount (Table 5). At the mesic site, only the earlywood of P. abies showed a significant trend in δ18O, with an increase of 0.95‰ over the period 1960–2009 (Table 3). Earlywood δ18O of all species was significantly positively correlated with δ18O in precipitation and VPD, and negatively correlated with precipitation amount (Table 5). Latewood δ18O of all species was significantly positively correlated with δ18O in precipitation. The estimated values of the monthly δ18O in precipitation at the study sites were similar to the measured δ18O values in Bern (Fig. 5a), hence justifying the use of the δ18O in precipitation records of Bern. In contrast to tree-ring δ18O, δ18O in precipitation increased by 1.18‰ between 1970 and 2008, and followed the warming trend observed since 1980s (Fig. 5b). However, only δ18O in precipitation in June and October showed significant positive trends during this period (Table S2). In addition, δ18O values in precipitation in dry vs moist years was not statistically different (Fig. 5c), making the assumption of similar source water in dry vs moist years valid.

Table 5. Pearson's correlation coefficients between δ18O in tree rings (δ18Otree) and climate variables
SiteSpeciesEarlywoodLatewoodδ18OEW vs δ18OLW
δ18Otree vs δ18Oprecipitationaδ18Otree vs precipitationbδ18Otree vs VPDcδ18Otree vs δ18Oprecipitationaδ18Otree vs precipitationbδ18Otree vs VPDc
  1. a

    For the earlywood (EW) section, the mean value of the δ18O in precipitation from March to June was used, whereas for the latewood (LW) section the mean value of the δ18O in precipitation from May to August was used. Correlations between δ18O in tree ring and δ18O in precipitation were calculated for the period 1970–2008.

  2. b

    For the EW section, the precipitation sum from March to June was used, whereas for the LW section the precipitation sum from May to August was used. Correlations between δ18O in tree ring and precipitation sum were calculated for the period 1960–2009.

  3. c

    VPD, vapor pressure deficit. For the EW section, the mean VPD value from March to June was used, whereas for the LW section the mean VPD value from May to August was used. Correlations between δ18O in tree ring and VPD were calculated for the period 1960–2009.

  4. Significance levels: *, < 0.05; **, < 0.01; ***, < 0.001.

Aosta (xeric) Pinus nigra 0.56***−0.150.51***0.080.36*0.080.53***
Pseudotsuga menziesii 0.63***−0.200.50***0.030.31*0.090.58***
Larix decidua 0.120.050.21−0.070.170.020.73***
Pinus sylvestris 0.47**−0.200.36**−0.230.40**−0.220.51***
Picea abies 0.39*0.090.29*−0.040.33*−0.190.55***
Biel (mesic) P. nigra 0.60***−0.33*0.35*0.55***−0.33*0.37**0.56***
P. menziesii 0.62***−0.32*0.44**0.42**−0.130.230.49***
L. decidua 0.52***−0.36*0.33*0.51***−0.260.36**0.56***
P. sylvestris 0.55***−0.33*0.38**0.30*−0.070.030.54***
P. abies 0.71***−0.35*0.45**0.44**−0.250.27*0.51***
Figure 5.

(a) Mean monthly δ18O in precipitation for the period 1970–2008 for the GNIP isotope station in Bern (solid line) in comparison to the mean monthly δ18O estimated values at the study sites in Aosta (dashed line) and Biel (dotted line) obtained from the Online Isotopes in Precipitation Calculator (Bowen et al., 2005; Bowen, 2013). (b) Mean annual δ18O in precipitation for the period 1970–2008 for the GNIP isotope station in Bern. (c) Mean δ18O in precipitation in dry vs moist years for the period 1970–2008 for the GNIP isotope station in Bern with the t-statistic and P-value of a Welch's two-sample t-test. In panel (c), each box represents the interquartile range, the whiskers extend to the fifth and 95th percentiles, the thick black solid line inside the box shows the mean, and outliers are shown as open circles beyond the whiskers. Moist years were defined as those with soil water deficit index (SWDI) values > 0, whereas dry years were defined as those with SWDI < 0.

Responses of BAI, Δ13C and δ18O to water deficits

At the xeric site, the early- and latewood BAIs of P. nigra, P. sylvestris and P. abies were significantly positively correlated with water availability (Table 6). By contrast, water deficits did not significantly influence the radial growth of P. menziesii and L. decidua. Earlywood Δ13C of P. nigra, P.menziesii and L. decidua was positively correlated to water availability, whereas no or marginal correlations were found between earlywood δ18O and SWDI (Table 6). For all species, Δ13C and δ18O in latewood were positively correlated to water availability.

Table 6. Pearson's correlation coefficients between earlywood or latewood basal area increment (BAI), Δ13C, δ18O and the mean standardized water deficit index (SWDIa)
SiteSpeciesEarlywoodLatewood
BAI vs SWDIΔ13C vs SWDIδ18O vs SWDIBAI vs SWDIΔ13C vs SWDIδ18O vs SWDI
  1. a

    For the earlywood section, the mean SWDI from September of the previous year to June of the current year was used, whereas for the latewood section the mean SWDI from June to September of the current year was used.

  2. Significance levels: *, < 0.05; **, < 0.01; ***, < 0.001.

Aosta (xeric) Pinus nigra 0.45**0.38**−0.090.54***0.38**0.43**
Pseudotsuga menziesii 0.220.32*0.030.090.210.33*
Larix decidua 0.27*0.34*0.070.110.57***0.24
Pinus sylvestris 0.28*0.260.040.34*0.200.45**
Picea abies 0.28*0.090.000.30*0.35*0.39**
Biel (mesic) P. nigra 0.090.28*−0.030.32*0.62***−0.37**
P. menziesii 0.46***0.15−0.040.49***0.52***−0.06
L. decidua 0.030.120.040.220.67***−0.29*
P. sylvestris 0.100.48***−0.010.53***0.67***−0.05
P. abies 0.090.32*−0.080.34*0.65***−0.24

At the mesic site, only the earlywood BAI of P. menziesii was significantly positively correlated with water availability, whereas the latewood BAI of all species, except L. decidua, was significantly positively correlated to water availability (Table 6). Earlywood Δ13C values of P. nigra, P. sylvestris and P. abies were positively correlated to water availability (Table 6). No correlations were found between earlywood δ18O of any species and water availability. Latewood Δ13C of all species was significantly positively correlated to water availability, whereas latewood δ18O of P. nigra and L. decidua was significantly negatively correlated to SWDI.

Relationships between iWUE and radial growth rates

The influence of gas exchange on radial growth of trees was tested by assessing the relationships between iWUE and BAI. At the xeric and mesic sites, the early- and latewood BAIs of all species, except P. abies, were negatively correlated to iWUE (Fig. 6). As the relationships between BAI and iWUE had also a temporal aspect as iWUE increased over time, the long-term assessment of the growth–gas exchange relationships was possible. At the xeric site, the BAI of P. nigra, L. decidua and P. sylvestris significantly decreased while iWUE increased, whereas the BAI of P. menziesii and P. abies remained rather stable despite increasing iWUE. At the mesic site, all species except P. abies showed a significant reduction in latewood BAI regardless of increasing iWUE.

Figure 6.

Relationships between the earlywood basal area increment (BAI) and the intrinsic water-use efficiency (iWUE) in earlywood (empty circles; dashed regression line) and relationships between the latewood BAI and iWUE in latewood (black circles; solid regression line) for five species (Larix decidua, Picea abies, Pinus sylvestris, Pinus nigra and Pseudotsuga menziesii) at the xeric site in Aosta (left column) and mesic site in Biel (right column) for the period 1960–2009. Pearson correlation coefficients between the BAI and iWUE are indicated by the r-values. Significance levels: *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Species-specific physiological adjustments in dry vs moist years

Changes in photosynthetic assimilation rates and/or stomatal conductance in dry vs moist years were estimated by assessing the shifts in the δ13C–δ18O space (Fig. 7). At the xeric site, δ18O in earlywood of P. abies and P. sylvestris was lower in dry than in moist years, while δ13C in earlywood of all species except P. abies was greater in dry than in moist years (Fig. 7a). Similarly, δ18O in latewood was lower and δ13C in latewood was higher in dry than in moist years (Fig. 7b). By contrast, δ18O in earlywood of all species at the mesic site was higher in dry than in moist years, while δ13C was higher only for L. decidua, P. sylvestris and P. abies (Fig. 7c). For the latewood section at the mesic site, all species showed higher δ13C values in dry than in moist years, while δ18O was higher in dry than in moist years only for P. nigra, L. decidua and P. abies (Fig. 7d).

Figure 7.

Relationships between δ13C and δ18O in moist (open symbols) vs dry years (closed symbols) in early- and latewood of five species (Ld, Larix decidua; Pa, Picea abies; Ps, Pinus sylvestris; Pn, Pinus nigra and Pm, Pseudotsuga menziesii) at a xeric (Aosta) and a mesic (Biel) site. Mean values (symbols) ± 1 SE are shown. Shifts in the isotopic space are represented by arrows. Moist years were defined as those with soil water deficit index (SWDI) > 0, whereas dry years were defined as those with SWDI < 0. The insets display the corresponding Aintgs int (photosynthetic assimilation rate vs stomatal conductance) responses of the different species according to the dual-isotope conceptual model of Scheidegger et al. (2000) (Fig. S2). Please note the different x- and y-axis scales between sites.

Discussion

Trends in iWUE and BAI and their relationships

Over the period 1960–2009, ca increased from c. 317 to 387 ppm, an increase of 22% (Keeling et al., 2009), while iWUE increased significantly in both early- and latewood, by 8–29%, depending on species, sites and seasons (Table 3, Figs 3, 4). Previous studies have reported that trends in iWUE could potentially be related to changes in tree height over time as Δ13C can decline with increasing height (McDowell et al., 2011) as a result of: adjustments in hydraulic conductivity (Monserud & Marshall, 2001); assimilation of δ13C-depleted air near the forest floor (Schleser & Jayasekera, 1985); changes in irradiance and photosynthetic capacity (Francey & Farquhar, 1982); and variations in VPD with height in the canopy (McCarroll & Loader, 2004). However, we did not observe any significant changes in Δ13C for all species, except P. sylvestris at the xeric site and P. menziesii at the mesic site (both species showed a significant increase in Δ13C), during the period 1960–2009 (Table 3). Further, all sampled trees were at least 50 yr old at the beginning of the investigation period and they had nearly reached their maximum height, as it is estimated that tree height has only increased by 3–12 m in the following 50 yr depending on species and site conditions (WSL Swiss Federal Institute for Forest, unpublished). Such increases in height in time would have no or only marginal effects on Δ13C and iWUE (Table 3; Monserud & Marshall, 2001; McDowell et al., 2011). Additional evidence is provided by studies showing that before the rise in CO2 concentration, trees generally do not show any age-related δ13C trends after an initial juvenile phase of c. 50 yr (Gagen et al., 2007). Therefore, the potential age/height effects on Δ13C and iWUE in our study can be ruled out. In fact, all species showed a constant ci/ca ratio over time (data not shown), leading to a moderate increase in iWUE. The increases in iWUE observed in our study are within the range of rising iWUE values reported (between c. 5 and 30%) for various conifers at various sites during recent decades in Europe (Gagen et al., 2011; Maseyk et al., 2011; Linares & Camarero, 2012).

The partial correlations analysis allowed us to disentangle the influence of temperature and VPD vs CO2 on tree iWUE (Table 4). Higher CO2 concentrations significantly enhanced iWUE of all species at the xeric and mesic sites, while temperature and VPD had minor influences on iWUE, except for P. nigra and P. sylvestris at the xeric site and in the latewood of all species at the mesic site (Table 4). iWUE was higher in latewood than earlywood (Table 3) because of the lower soil water contents and higher VPD in summer when latewood formation occurs. This difference was even more pronounced at the xeric than at the mesic site because of high evaporative demand prevailing in summers in inner-Alpine dry valleys (Fig. 1). High air temperature and VPD can strongly reduce stomatal conductance, hence reducing Δ13C and increasing iWUE in trees (McCarroll & Loader, 2004). This can be accentuated during summer when VPD and water deficits are at their highest, making Δ13C and iWUE in latewood of most species at both study sites sensitive to VPD and soil water availability (Table 4 and 6). At the xeric site, the increases in iWUE were more pronounced from the 1980s onwards than during the period 1960–1980 (Fig. 3), coinciding with the steep increase in CO2 concentrations since the early 1980s and rising temperature and VPD (Table 2, Fig. S1). Similarly, at the mesic site, the rise in iWUE followed the increase in temperature and VPD recorded since 1960 (Table 2).

Early- and/or latewood growth of all species was negatively correlated to iWUE (Fig. 6). As iWUE increased over time, negative relationships between radial growth and iWUE (Fig. 6) strongly indicate that radial growth has been declining despite the long-term increases in iWUE, particularly at the xeric site. If radial growth is regarded as a good proxy of net carbon gain, decreasing trends in radial growth and increasing trends in iWUE suggest that a reduction in stomatal conductance has prevailed, rather than an increase in assimilation rates over time (Voltas et al., 2013). All species at the xeric site showed a consistent decrease in BAI since the 1980s, whereas at the mesic site growth remained rather stable for all species, except for L. decidua which showed a significant reduction in BAI (Table 1, Fig. 2). The declines in BAI reported here at the xeric site are in agreement with recent studies that have found warming-induced growth reductions in spite of increasing iWUE for a multitude of species at dry sites in the Iberian Peninsula (Peñuelas et al., 2008; Andreu-Hayles et al., 2011; Linares & Camarero, 2012). However, the results found at the mesic site contrast with the CO2-induced (Cole et al., 2010; McMahon et al., 2010) and warming-induced growth enhancement (Way & Oren, 2010) found in other temperate forests. In fact, our results suggest that drought-induced stomatal closure resulting from warming and increased VPD has reduced transpirational water loss at the cost of reducing CO2 uptake, hence overriding the potential CO2 ‘fertilization effect’. This could have been exacerbated at the xeric site by the low soil water availability resulting from the low soil water-holding capacity, the low soil fertility and the higher intertree competition than at the mesic site.

Physiological responses to water deficits

At the xeric site, we found systematic and consistent declines in δ18O and increases in δ13C in the early- and latewood of all species in dry years compared with wet years (Fig. 7a,b). Based on the dual-isotope conceptual model of Scheidegger et al. (2000), a decrease in δ18O would indicate an increase in stomatal conductance, and an increase in δ13C would suggest that trees would enhance carbon uptake (Fig. S2). However, under drought stress, trees face the dilemma of reducing stomatal conductance to avoid hydraulic failure and desiccation at the cost of carbohydrate depletion through decreasing photosynthesis and carbon assimilation (McDowell et al., 2008; Mitchell et al., 2013). This is still consistent with an increase of δ13C. The unexpected decrease in δ18O in dry years, however, results from water uptake from deeper soil water source, which is more depleted in H218O (Sarris et al., 2013; Treydte et al., 2014) as indicated by the positive correlations between latewood δ18O and SWDI at the xeric site (Table 6). In dry years, the signature of depleted source water also dominates because most wood is formed during spring (Eilmann et al., 2011) when precipitation is highest in the Aosta Valley (Fig. 1a) and strongly depleted in H218O (Fig. 5a) and mixed with snowmelt water that is also highly depleted in H218O (Tang & Feng, 2001). As precipitation is low in summer in the Aosta Valley (Fig. 1a) and surface water is scarce, trees must acquire water from deeper layers, which again is more depleted in H218O (Sarris et al., 2013; Treydte et al., 2014). The remaining water from the topsoil, which is enriched in H218O (Moreno-Gutierrez et al., 2013; Treydte et al., 2014) during summer, is scarce and only a limited amount of biomass can be produced from this enriched water. Consequently, the biomass with the signature of the depleted water from spring and deeper soil water source dominates in dry years (Treydte et al., 2014), overriding the signal from the summer periods, which also carries the signal of reduced stomatal conductance (see also Sarris et al., 2013). On the other hand, the reduced amount of assimilates formed during drought is used almost entirely for maintenance of metabolic processes, while the remaining C is invested in the roots (Dewar et al., 1994), as C investments into below-ground growth during drought stress are of higher priority than above-ground growth (Kozlowski & Pallardy, 2002) to ensure water acquisition (Saxe et al., 1998). This leaves little carbon for tree-ring growth and may mask the summer δ18O signal in tree rings in dry years at the xeric site (Sarris et al., 2013), and agrees with the nonexistent relationship between latewood δ18O and summer VPD (Table 5). Changes in source water acquisition by trees between spring and summer periods at the xeric site are also supported by the positive correlations between earlywood δ18O and δ18O in precipitation from March to June, and the nonexistent relationship between latewood δ18O and δ18O in precipitation from May to August (Table 5).

At the mesic site, δ13C and δ18O were greater in dry than in moist years in the early- and latewood of all species (Fig. 7c,d). These shifts in the δ13C–δ18O space indicate that trees in dry years reduce their stomatal conductance at the cost of carbon uptake. Precipitation at the mesic site is abundant and rather uniformly distributed over the year (Fig. 1b) and the soil has a relatively high water-holding capacity (119 mm). These characteristics make trees at the mesic site less dependent on deep soil water pools during dry periods, hence reducing the signal of the depleted H218O in tree rings (see earlier). This is further supported by the significant positive correlations between δ18O in early- and latewood and δ18O in precipitation (Table 5). Moreover, δ18O in early- and latewood was positively related to spring and summer VPD (Table 5). These coherent positive correlations between early-and latewood at the mesic site suggest that variations in source water between seasons and from year to year (Fig. 5c) are minimal, so that changes in δ18O can be related to leaf evaporation and not shifts in source water, as is the case at the xeric site. Further, most of the tree growth at the mesic site takes place during the summer months when topsoil is enriched in H218O (Lévesque et al., 2013). Thus, the biomass with the signature of the enriched water from summer dominates. This is consistent with the δ18O theory, which indicates that δ18O in organic matter increases when relative humidity decreases (McCarroll & Loader, 2004), and the expected reduction in stomatal conductance of trees under drought conditions (Mitchell et al., 2013). Reduced stomatal conductance during dry summers at the mesic site was also supported by the strong coupling between Δ13C in latewood of all species and water availability, as well as by the negative relationships between latewood δ18O and soil water availability (Table 6).

Conclusions

Irrespective of tree species, their origins (native vs nonnative), and their increases in iWUE, radial growth has significantly declined under xeric conditions since the beginning of the 1980s, whereas no growth enhancement was found under mesic conditions. In fact, our results contradict the CO2- and temperature-induced growth enhancement reported in other mesic forests (Cole et al., 2010; McMahon et al., 2010). Mainly, our study suggests a similar physiological response between tree species (i.e. reduced stomatal conductance and carbon uptake followed by growth reduction or stagnation) to rising ca and drought under xeric and mesic conditions. This finding agrees with the recent study of Choat et al. (2012), which indicates a global convergence in the vulnerability of xeric and mesic forest types to drought. Assuming that these results can be extrapolated to similar xeric and mesic sites in central Europe, such changes in growth rates and transpiration may challenge today's predictions of improved forest productivity, particularly under mesic conditions, and may have strong consequences on the terrestrial carbon and hydrological cycles (Bonan, 2008). Finally, our study highlights the fact that the dual-isotope conceptual model must be used with great care in the interpretation of plausible gas-exchange patterns under xeric conditions, as changes in source water violated the assumptions of the dual-isotope framework. While soil water pools used by trees at the xeric site changed over the seasons and in dry vs wet years, the water source used by trees at the mesic site remained fairly constant between seasons and years. Therefore, a solid knowledge of the precipitation patterns and soil water content, particularly under xeric environments, is essential for the correct application of the dual-isotope approach and for the understanding of tree responses to changes in ambient CO2 concentrations and water availability.

Acknowledgements

We would like to thank Bernhard Hadorn and Corrado Letey for the sampling permissions and Fabio Brunier (Department of Soil Conservation and Water Resources, Autonomous Region of Valle d'Aosta, Italy) for providing the climate data for the site in Aosta. We also thank Pascale Weber and Roger Köchli for their help with the soil analyses, and Francesca Carnesecchi for her help with sample preparation. This research was funded by the Swiss State Secretariat for Education and Research under the COST action FP0703 and by the BAFU/WSL Research Program ‘Forests and Climate Change’.

Ancillary