Radiation modifies the effect of water availability on the carbon isotope composition of beech (Fagus sylvatica)


Author for correspondence: Mark A. Adams Tel: +61 89380 3445 Fax: +61 89380 1001 Email:adamsma@cyllene.uwa.edu.au


  •  Water availability and flux and carbon isotope and other chemical compositions of wood, foliage and phloem sap were quantified in a replicated field test examining the effects of exposure (north-east, south-west) and stem density on beech ecophysiology (Fagus sylvatica).
  •  Standard techniques for quantifying water potential, water flux in stems and carbon isotope composition of wood and foliage were complemented with chemical analysis of phloem sap collected using the ‘phloem bleeding’ technique.
  •  Phloem sap was similar in composition to other hardwoods. The δ13C signatures of sap reflected short-term fluctuations in water availability and intercepted radiation, whereas foliar δ13C was a poor reflection of current environmental conditions.
  •  δ13C of tree rings suggested that the south-west-facing site experienced a greater shortage of water during summer than the paired north-east-facing site. However, phloem δ13C analysis showed that interpreting the effects of water availability using δ13C might be confounded by radiation. Better integration of hydrological models with reliable indices of water stress in beech will be required for reliable predictions of growth, especially if (when) climates change.


Owing to their effects on internal concentrations of carbon dioxide (CO2) (Ci), intercepted radiation and atmospheric and soil water deficits modify the ratio of 13C : 12C in plant carbon (δ13C, e.g. Leavitt & Long, 1986; Livingston & Spittlehouse, 1996; Korol et al., 1999). For example, rates of photosynthesis are generally greater in sunlit foliage than in shaded foliage, leading to a greater reduction in Ci and a greater depletion in 13C (Leavitt & Long, 1986). Even more well described is the reduction in Ci caused by stomatal closure in response to water deficits (e.g. Guehl et al., 1995; Stewart et al., 1995; Lauteri et al., 1997; Panek & Waring, 1997).

The relatively large body of literature dealing with δ13C in conifers shows that a number of other factors (e.g. altitude, nutrition, hydraulic architecture) also play roles in determining δ13C (e.g. Walcroft et al., 1996; Korol et al., 1999; Hultine & Marshall, 2000; Warren & Adams, 2000) in wood and foliage. For example, δ13C varied with shoot water potential at a rate of −0.18‰ MPa−1 in Quercus spp. (Damesin et al., 1998) and with altitude at a rate of approx. 2‰ km−1 in numerous species and studies (e.g. Körner et al., 1991). Most of the δ13C studies have been ‘observational’ and there are few field-based, controlled tests of the effects of environmental conditions (e.g. availability of radiation, water, nutrients) on δ13C in plant tissues (e.g. Morecroft & Woodward, 1990). From one controlled field experiment, Högberg et al. (1993) provided evidence for the effects of nitrogen fertilization on δ13C signatures in foliage of Picea abies and Pinussylvestris.

As for conifers, both simple assessments of water availability (e.g. Macfarlane & Adams, 1998) and modelled water balances (Dupouey et al., 1993) show moderately strong relationships with δ13C signatures of the wood in recent (e.g. < 10 yr) annual growth rings of hardwoods. Over the longer term, increases in atmospheric concentrations of CO2, and variations in the leaf-to-air humidity gradient and in leaf temperature will all influence δ13C (Comstock & Ehleringer, 1993). For both conifers and hardwoods, modelled water balances for the growing-season, and sometimes for periods as short as 1 month, provide the strongest relationships (up to 93% of variance in δ13C explained) between δ13C of wood and indices of environmental conditions (Dupouey et al., 1993; Livingston & Spittlehouse, 1996). Like most, these models were developed using data from more or less ‘ideal’ sites with homogeneous vegetation and soil.

Recently, Yoneyama et al. (1997) provided a first report of the isotope ratios in phloem sap for wheat that was followed by Pate & Arthur (1998) and Pate et al. (1998) for Eucalyptus globulus. Part of the attention on phloem sap originates from a desire to obtain a better estimate of the δ13C signatures of leaf sugars – some of the first products of carbon fixation – and thus an estimate of the effects on δ13C of current environmental conditions. Hence the determination of δ13C in phloem sap, containing carbon assimilated in a period of hours to days before sampling, is useful to assess short-term variations in Ci/Ca (Brugnoli et al., 1998; Pate & Arthur, 1998). Such data are required to help explain why the relationship between foliar δ13C signatures and current environmental conditions varies so greatly among species, genera and climates. In the case of beech, new season’s foliage, like that of many other species, is formed largely from stored carbon and nutrients (e.g. Millard & Proe, 1993; Kozlowski & Pallardy, 1997) and thus more likely to carry a δ13C signature that reflects the environmental conditions for carbon fixation in previous seasons than the current one. Clearly, the determination of δ13C in phloem sap provides a strong guide to Ci/Ca during the present growing-season. Yoneyama et al. (1997) used aphids to sample phloem sap, whereas Pate et al. (1998) relied on phloem sap that bled from shallow incisions in the bark. Irrespective of the method, the ability to obtain phloem sap, will ‘enable more rigorous tests of prediction and reality for δ13C and δ15N, and determination of the uses of natural abundance measurements in improving our understanding of C and N metabolism and transport’ (Yoneyama et al., 1997).

Water availability is a limiting factor for growth of beech (Fagus sylvativa) in central Europe (Ellenberg, 1995). Climate models for central Europe predict an increase in temperature and longer periods of drought during the growing-season (Enquete Kommission, 1994). Reductions in soil water availability will, arguably, strongly influence patterns of competition between drought-sensitive beech and less sensitive species. Both mature and regenerating stands of beech grow on the shallow, rendzina soils derived from limestone that are widespread in southern Germany (e.g. Schwäbische Alb). These soils often have poor water-holding capacity. Equally importantly, there is strong political and social support for the transformation of spruce monocultures into site-adapted deciduous forests with beech as the dominant species. At the present time, we lack the necessary physiological basis for reliable prediction of either the likely success of re-establishment of beech on shallow, rendzina soils or effective management strategies, especially the management of stand density.

Advantage was taken of the opportunity presented by a replicated field trial in southern Germany to characterize the interacting effects of radiation and water availability on δ13C signatures in phloem sap, foliage and wood, in order to obtain information about temporal variation in Ci. By combining these data with the determination of water potential of twigs and soil, tree water use and stand transpiration, it was hoped to validate δ13C signatures of different tissues as indices of water availability and incident radiation in beech forests that were sensitive to typical management treatments. The trial included paired sites that differed mainly in aspect on either side of a narrow valley and, within each site, replicated plots of differing stand density. The broad hypotheses, based on past records of growth and vegetation analysis, were that the more south-facing site would have less available water and greater radiation than the more north-facing site, and that thinning would increase the availability of water. Additionally, an aim was to characterize ‘phloem bleeding’ as a means of sampling phloem sap from beech. This sampling technique is a prerequisite to characterizing δ13C signatures in the phloem sap of adult beech trees, since other techniques are difficult to use in the field (aphid technique) or introduce C compounds into the sampling solution (Schneider et al., 1996), thereby confounding interpretation of δ13C signatures.

Materials and Methods

Study site, climate and experimental design

The experimental sites are located in the south of Germany, about 100 km south-south-west from Stuttgart (longitude: 8°40′ E; latitude: 48°00′ N) in a low mountain range (Schwäbische Alb, 740–760 m above sea level). The mean annual air temperature at the site is about 6.6°C, and the mean temperature during the growing-season (April to October) is about 11.5°C. The average annual precipitation is 810 mm, with monthly maxima in June and July. In 1999 both mean air temperature and total precipitation were greater than long-time averages (Table 1). In July and August, potential evaporation (Vp) exceeded precipitation.

Table 1.  Precipitation (P), potential evapotranspiration (Vp) and mean air temperature (T) for the experimental sites in 1999. Precipitation and mean air temperature were recorded at a climate station of the Deutscher Wetterdienst in Tuttlingen, approx. 5-km distance from the experimental plots. Vp was modelled according to Matzarakis et al. (1999)
MonthP (mm)Vp (mm)T (°C)
  • *

    Sums for P and Vp; average for T.

January 55.2 19.2 0.1
February 97.5 16.7−2.2
March 48.0 32.4 4.0
April 77.4 44.5 7.2
May130.2 72.713.0
June110.7 76.113.7
July 87.9 94.817.1
August 71.4 81.315.8
September 77.4 63.714.4
October 50.6 36.3 7.3
November 42.9 20.2 0.6
December134.1 18.5 0.1
Sum/average*983.3576.3 7.6

The experimental sites are located on the two opposing sides (not more than 500 m apart) of a single, narrow valley. On one site, a set of experimental plots face the north-east (NE), while on the other site the plots face the south-west (SW). Rainfall does not vary significantly across the valley. The slope of both sites is moderately steep (NE, 58–100%; SW, 36–58%). Soil profiles are characterized as terra fusca–rendzina derived from limestone (Weißjura beta and gamma series) and on both sites (SW and NE) are shallow, averaging less than 20-cm depth of topsoil before becoming dominated by parent rock interspersed with pockets of organic matter and mineral soil. The soil profile at the SW site is especially rocky, containing more than 40% (volumetric basis) rocks and stones (> 63-mm diameter) in the top 20 cm of the soil, rising to 80% below 0.5 m. The soil of the NE site contains 15% rocks and stones in the uppermost 20 cm of the soil and approx. 30% below 0.5 m. The soil pH (water) is 5.7 in the surface organic layer and 7.5 at 0.6-m depth. On both sites beech (Fagus sylvatica L.) is the dominant species, making up > 90% of the total basal area of trees.

In 1999, trees on both sites had reached an average age of 70–80 yr. Also in 1999, and before experimental treatments (see below) were applied, the trees on the NE site had a mean height of 27.3 m. The mean diameter at breast height was 23.9 cm and the stocking amounted to approx. 530 stems ha−1. The trees on the SW site had a mean height of 25.4 m, a mean diameter at breast height of 21.5 cm and a stocking of approx. 530 stems ha−1. The understorey vegetation of the two sites differs according to the radiation budget (see below) and leads to a classification of the stand on the NE site as Hordelymo-Fagetum, with that on the (putatively drier) SW site classified as Carici-Fagetum (Oberdorfer, 1992).

On each site, two silvicultural (thinning) treatments plus controls (unthinned) were established in 1999. The experimental design consisted of two blocks, each containing a single plot (each approx. 0.53 ha in size) of the two silvicultural treatments plus a control plot. The total basal area (BA) of trees on the untreated control plots varied between sites – on the NE site the mean BA in control plots was close to 27 m2 ha−1, while on the SW site the control plot BA was about 20 m2 ha−1. The two thinning treatments on both sites were BA = 15 and 10 m2 ha−1. Thinning decreased the leaf area index (LAI) from 5.16 (control) to 3.15 (BA = 15 m2 ha−1) and 1.68 (BA = 10 m2 ha−1) on the NE site, and from 5.12 to 3.24 and 2.12 on the SW site.

The difference in aspect (NE or SW) produces a clear difference in radiation interception. Calculation of global radiation according to Keding (1984) and Fritsch (1998) shows that the maximum daily radiation for the NE site amounts to 79% of radiation available on the SW site in July and only 47% of that available in October (Fig. 1). This difference between sites is reflected in rates of evaporation and hence available soil water. Fig. 2 illustrates the variation in soil water potential for a 10-d period (20 September to 1 October 1999). At the beginning of the period, the lack of recent rainfall greatly reduced water potential in soils on both sites. The reduction was more severe for the SW site, and here soil water potential reached minimum values of −0.08 MPa. After rains starting in the middle of September, soil water potential increased at both sites, and more quickly at the SW site. Shoot pressure potential (ψ, MPa) and water use (ml cm−2 s−1) by trees were strongly influenced by water availability. For example, in 1999 shoot pressure potentials (measured using standard techniques with a Scholander-type pressure bomb (Rennenberg et al., 1996)) were similar at both sites in mid-July (when there had been good recent rains). However, shoot pressure potentials declined, and more strongly at the SW site (Table 2), during August when almost no rain fell in the week preceding the sampling (Table 2) and soil water potentials were strongly negative (Fig. 2). These patterns were reflected in those of tree water use.

Figure 1.

Diurnal course of global radiation on a horizontal area (G, thick solid line) and on the north-east-facing (GNE, dashed line) and the south-west-facing (GSW, thin solid line) experimental sites near Tuttlingen, southern Germany, for cloudless days in July (top) and October (bottom). GSW > G since the surface of the south-west slope is irradiated by the incident parallel sunlight with a less acute angle than the horizontal area.

Figure 2.

Water potential of soils in control plots of the south-west- and north-east-facing sites in the autumn of 1999, and daily sums of precipitation. Soil water potential was measured using tensiometers placed at depths of 40 cm (dashed line, SW; dotted line, NE) and 60 cm (thin solid line, SW; bold solid line, NE).

Table 2.  Xylem water potential in beech for consecutive months in summer 1999. Rainfall in the month (week in parentheses) preceding the July sampling was 108 (19.3) mm, and rainfall in the month preceding the August sampling was 71 (0.5) mm. Similarly, there were approx. 58 h of full sunlight in the week preceding the July sampling, and approx. 69 h preceding the August sampling. The modelled, plant-available soil water content (water balance model WBS3: Schmidt, 1990; Fritsch, 1998) was some 55% of field capacity at the north-east site in July and 45% in August
AspectTreatment (basal area, m2 ha−1)Monthψ (Mpa)SE
NorthControl (approx. 25)July−1.24−0.14
 15 −1.23−0.12
 10 −1.38−0.16
SouthControl (approx. 25)July−1.34−0.41
 15 −1.18−0.41
 10 −1.20−0.11
NorthControl (approx. 25)August−0.54−0.22
 15 −0.89−0.32
 10 −0.71−0.22
SouthControl (approx. 25)August−2.45−0.11
 15 −2.35−0.17
 10 −2.40−0.12

The prevailing conditions of water availability and climate were established by measuring xylem (water) flow densities using Granier-style probes. Briefly, flow densities in the water-conducting sap wood of beech were determined using the constant-heating method according to Granier (1985) and Köstner et al. (1996). Flux densities (ml cm−2 sapwood area s−1) were determined every 5 min in seven to nine beech trees per plot and calculated as means of every 30 min. Thus in July 1999 xylem flow densities in trees on the SW site (control and 10 m2 ha−1 treatments) were greater than those in trees on the NE site (Fig. 3a). However, 1 month later when the soil water potential became more negative (Fig. 2), this pattern was reversed and trees on the NE site showed higher xylem flow densities than those on the SW site (Fig. 3b).

Figure 3.

Effects of site (south-west aspect, SW; north-east aspect, NE) and treatment (control and 10 m2 ha−1) on sap flow densities of beech growing at the experimental sites near Tuttlingen, southern Germany, in July (a) and August/September (b). Sap flow densities were determined in the sapwood of seven to nine beech trees per site and per treatment, and displayed in the figure as mean (solid line) ± SD (dotted line). The trees chosen were representative for the forest stand and covered the range of diameters (at breast height, over bark) at the experimental sites.

Flux densities can be converted to estimates of water use using the stand sapwood area according to the following equation:

image( Eqn 1)

(ST, stand transpiration (mm); SA, sapwood area (cm2 m−2); FD, mean flow density in the xylem (ml cm−2 sapwood area s−1)). Half-hourly rates of stand transpiration were summed for 24 h. Sapwood area was determined according to the technique of Glavac et al. (1989). A solution of berberine chloride was used to displace the xylem sap from sections (cut from the height at which the sensors had been installed) of representative beech trees. The infiltration of berberine chloride also stained the sapwood area and allowed subsequent estimation of the depth of xylem tissue that was active in water transport. About 80% of the total cross-sectional area of the trunks of adult trees was active in water transport.

In July 1999 the mean transpiration of the stand on the NE control site (within the period shown in Fig. 3a) amounted to 1.3 mm day−1 and was comparable to the transpiration of the SW control site (1.4 mm day−1). Thinning (to 10 m2 ha−1: Fig. 3a) decreased stand transpiration to about 60% of the control on the SW site and to about 45% on the NE site. At the NE site, mean daily transpiration in August/September (during the period shown in Fig. 3b) was comparable to that in July for both control (1.4 mm day−1) and thinned (10 m2 ha−1, 0.6 mm day−1) stands. However, during this period both control and thinned stands on the SW sites transpired between 55% and 65% of the water transpired in July, and significantly less water than transpired by the stands on the NE site. Hereafter, the data are reported as xylem flux densities rather than rates of transpiration, since relative rates of water use rather than absolute rates are of more concern.

Sampling and analysis of wood, foliage and phloem

The δ13C of wood was characterized from both the NE and SW sites, as described in previous studies of F. sylvatica (Dupouey et al., 1993) or of other hardwoods such as Eucalyptus (Macfarlane & Adams, 1998) or softwoods (Saurer et al., 1995; Panek & Waring, 1997). Trees for these analyses were felled in March 1999, before the growing-season had started. In brief, wood samples were collected from discs cut at breast height from felled trees (four co-dominant trees per site). For determination of mean δ13C in growth rings, a radial–tangential section was taken from each disc and a sample of wood from each whole growth ring was oven-dried at 65°C and then ground into a fine, homogeneous powder of which 2.0–2.5 mg per sample were analysed. As noted by Dupouey et al. (1993), earlywood–latewood transitions ‘cannot be unambiguously located’ in beech and, moreover, are ‘unlikely to represent the same date for each year’. Whole rings were thus analysed, excluding material obviously strongly contaminated with lignin or resin, such as that in resin ‘pockets’ or at tree ring boundaries. Furthermore, holo-cellulose was not separated from the whole wood samples. This introduces a slight error but one that is generally strongly systematic (e.g. Livingston & Spittlehouse, 1996; Macfarlane et al., 1999; Warren & Adams, 2000) over the measurement range – the measured δ13C for wood in tree rings is generally less than 1‰ depleted in 13C vis-à-vis holocellulose or cellulose.

Phloem sap was collected, as described by Pate et al. (1998) for E. globulus, at approximately monthly intervals from at least three trees in each plot on each site. The bark of each sample tree was cut at heights between 1.3 and 1.8 m to the depth of the wood, at about 15° to the horizontal, using a single-sided razor blade or scalpel. The ‘bleeding’ phloem sap was always exuded either immediately after the incision was made or not at all. Trees bled poorly or not at all in late spring/early summer, and increasingly readily through the summer into autumn (Table 3). Different trees were used at each sampling date in order to avoid cumulative effects of bark damage.

Table 3.  Qualitative analysis of the rate of phloem exudation from incisions (approx.10-cm length) in bark at a height of 1.3 m above ground level
MonthRate of phloem bleeding*
  • *

    Slight, < 5 µl per bleeding incision; moderate, 5–10 µl; heavy, > 20 µl.


On selected sampling dates, four trees were climbed in each plot and foliage was sampled from one to two branches excised from the sunlit part of the tree crowns. Different trees were used for phloem sap and foliage sample in order to avoid artefacts in the composition of phloem sap as a consequence of the harvest of twigs for foliage samples. However, material from neighbouring trees of the same social class (co-dominant or dominant) was collected.

Finely ground samples of foliage and wood were combusted to CO2 under a surplus of oxygen (O2) using an elemental analyser (NC 2500, Carlo Erba Instruments, Milan, Italy) before passing via a Conflo II interface (ThemoQuest Finnigan Instruments, Bremen, Germany) into an isotope ratio mass spectrometer (Deltaplus ThermoQuest, Finnigan Instruments, Bremen, Germany). An aliquot of undiluted phloem sap (10 µl) was pipetted into tin capsules and oven-dried for 15 min at 70°C, and then analysed for 13C : 12C. δ13C (in ‰ units) was calculated with respect to the Pee Dee Belemnite (PDB) standard, [(13C : 12C)sample/(13C : 12C)standard− 1] × 1000.

Phloem sap was further analysed for sugar content as follows. Diluted (1 : 250) aliquots of 100 µl of sap were injected into a HPLC system (Dionex DX 500: Dionex, Idstein, Germany). Separation was achieved on a CarboPac 1 separation column (250 × 4.1 mm: Dionex, Idstein, Germany) with 36-mM sodium hydroxide as an eluent at a flow rate of 1 ml min−1. Carbohydrates were measured by means of a pulsed amperometric detector equipped with a gold working electrode (Dionex DX 500: Dionex). Individual carbohydrates were eluted 8–16 min after injection and were identified and quantified by internal and external standards.

Statistical analysis

All data were analysed using the routines contained within the commercial packages StatView, SuperAnova (Abacus Concepts, Berkeley, CA, USA) and SAS (SAS Institute Inc., Cary, NC, USA). The effect of site (NE, SW) and thinning treatment on δ13C signature and sugar content of phloem sap and on δ13C in foliage was assessed using a two-way ANOVA procedure. The same test was applied to differentiate the effects of site and sampling location (top or bottom of treatment plots) on δ13C signature of the phloem sap.


Retrospective analysis of climate and δ13C in wood

Patterns of rainfall and radiation during the growing-season over the 20-yr period 1978–98 were more or less in keeping with longer-term trends. Rainfall in the growing-season averaged more than 600 mm or 75% of annual rainfall and there were typically more than 1100 h of full sunlight for the same period (Fig. 4a). Growing-season temperatures averaged 10.8°C (the long-term mean being 11.5°C). The years 1985, 1989 and especially 1991 were relatively dry and rainfall was less than 500 mm in the growing-season; only 425 mm of rain fell in this period in 1991.

Figure 4.

Historical data for rainfall and sunshine, δ13C and yearly radial growth of beech. (a) Sum of rainfall (solid diamonds) and sunshine hours (shaded bars) during the growing-season measured at the permanent weather stations of the Deutscher Wetterdienst in Tuttlingen and Möhringen between 1982 and 1998. (b) Weighted mean δ13C of wood in tree rings of beech growing on north-east-facing (NE, open circles) and south-west-facing (SW, filled circles) sites near Tuttlingen, southern Germany, for the period from 1979 to 1998. (c) Yearly radial growth of beech from the north-east-facing (NE, open triangles) and south-west-facing (SW, filled triangles) experimental sites. δ13C and the radial growth data are means ± SE (for δ13C) from measurements of four and 16 trees per aspect, respectively.

Wood from the SW site was significantly less depleted in 13C (δ13C was less negative) than the wood from the NE site (Fig. 4b) from 1979 to 1991. None the less, and after an abrupt change in δ13C at the SW site in 1991–92, this trend was reversed for the years 1992–95, probably associated with both the below-average rainfall in 1991, and a minor thinning at the SW site in the subsequent year. Rates of radial growth of trees at the two sites were generally closely matched over the 20-yr period, despite generally greater year-to-year variation than exhibited by δ13C (Fig. 4b,c). As for δ13C, there was a marked change in radial growth at the SW site in the latter part of the study period, albeit beginning 2 yr later (1993–94) than the change in δ13C. Hence the decrease in δ13C in the wood of the trees from the SW site in 1991–92 was followed by an increase in radial growth in 1993–94. None of the directly measured climatic variables, nor any of a number of simple indices of water balance (e.g. ratio of precipitation to evapotranspiration), was strongly related to growth or to δ13C (results not shown). On the other hand, radiation (hours of full sunlight during the growing-season) was significantly related (P < 0.05, r = 0.62) to δ13C, but only at the SW site.

Phloem and foliage analysis

Phloem sap collected via the ‘bleeding’ technique was dominated by sucrose (generally > 80% of total sugar concentration) with smaller concentrations of glucose and fructose (Fig. 5). Several other sugars were present at minor concentrations but these were not identified or quantified. Raffinose is often found in phloem sap and during the analytical technique employed (see Materials and Methods) may have co-eluted with fructose and been quantified thereas.

Figure 5.

Effects of site (north-east, NE; south-west, SW) and treatment (control, 15 m2 ha−1 and 10 m2 ha−1) on sugar composition of phloem sap collected from beech growing near Tuttlingen, southern Germany, in summer/autumn 1999. The significance of the main effects from analysis of variance is shown (n.s., not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001), as are standard errors for each treatment. In the histograms, solid black denotes glucose, no shading denotes fructose, and stippled shading denotes sucrose. See the text for details of experimental design and replication.

Owing to the relatively large size of each sample plot and the steep slope, an objective was to characterize the effect of sampling location (top or bottom of plot) on the δ13C signature in the phloem sap. This comparison showed that, across all thinning treatments, there was a strongly significant effect of site (NE or SW) on δ13C, whereas sampling location (top or bottom of plot) had no effect. The mean values of δ13C for phloem sap collected at the bottom and the top of the NE plots amounted to −26.43 ± 0.46‰ and −26.95 ± 0.43‰, respectively. For the SW plots the same quantities were −20.61 ± 0.2‰ (top) and −20.80 ± 0.31‰ (bottom). Consequently, on all sampling dates, samples were collected from trees randomly selected within, alternately, either the top or bottom third of each plot. Sampling was alternated so as to avoid repetitive sampling and possible injury (and thus artefacts generated by wound responses) of individual trees.

Over the 3-month period in the peak of the growing-season (July–September), the δ13C of phloem sap varied by about 13‰ (Fig. 6) across all sampling dates, sites and treatments; a much greater variation than seen between years in the δ13C of wood. In July, there was no effect of thinning treatment, but phloem sap from the SW site was consistently and significantly (P < 0.05) more depleted in 13C than sap from the NW site. In August, the driest month of the 1999 growing-season, there were both highly significant (P < 0.001) site effects and treatment (P < 0.01) effects. In contrast to July, phloem sap from the NW site was now always more depleted in 13C than sap from the SE site, by approx. 7‰ in control plots, 6.3‰ in plots with a BA of 15 m2 ha−1, and 5.4‰ in plots with a BA of 10 m2 ha−1. By the end of September after intensive rainfall, phloem sap from both sites and all treatments had a δ13C signature < −32‰ but only the effect of site was significant (P < 0.05), the SW site again producing sap that was more depleted in 13C than sap from the NW site.

Figure 6.

Effects of site (north-east, NE; south-west, SW) and treatment (control, 15 m2 ha−1) on carbon isotope composition δ(13C) of foliage and phloem sap. The significance of the main effects from analysis of variance is shown (n.s., not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001), as are standard errors for each treatment. In the histograms, no shading denotes foliage, and solid black denotes phloem sap.

Foliage of beech carried a δ13C signature that varied little over the growing-season (Fig. 6) and was not significantly different either between sites or among treatments. The mean δ13C signature of foliage from both sites was −28.1 ± 0.1‰, whereas that of wood of the most recent (1996–98) three annual rings was −25.7‰ for the NE site and −25.9‰ for the SW site.

The sugar content of phloem sap and its δ13C signature are compared in Fig. 7. Only August and September samples were analysed for both sugar and carbon isotope composition, owing to the small samples of sap that could be collected in July. For both sites in September and for the SW site in August, sugar concentrations were unrelated to δ13C. For the NE site in August, sugar concentrations were weakly and inversely related to δ13C (P < 0.01, n = 35, r2 = 0.29). Shoot pressure potential (SWP, see Materials and Methods) was also closely related to phloem δ13C signatures: δ13C(‰) = −3.93 (‰ MPa−1) × SWP( Mpa) − 30.65(‰) (r2 = 0.83, P < 0.01), although the relationship was not robust since data were lacking for the intermediate range.

Figure 7.

The relationship between carbon isotope signatures and sugar concentrations of beech phloem sap sampled in consecutive months in 1999. Filled circles and filled upward-pointing triangles, control; open circles and open squares, 15 m2 ha−1 treatment; filled downward-pointing triangles and open upward-pointing triangles, 10 m2 ha−1 treatment; open downward-pointing triangles, all treatments.


The first hypothesis (i.e. that aspect would determine water availability) was broadly supported by the data. On the basis of previous studies of beech (Saurer & Siegenthaler, 1989; Schleser, 1990, 1992; Dupouey et al., 1993), the 20-yr record in tree rings shows that water availability was less at the SW site since wood from this site was significantly less depleted in 13C than wood from the NE site. Equally, August was the driest month during the 1999 growing-season, and in this month phloem sap from the SW site was also much less depleted in 13C than sap from the NE site. The relation between soil water availability and δ13C signature is illustrated in Fig. 2, and soil water potential was distinctly more negative at the SW site than at the NE site at the time of the August sampling. Only slight differences in δ13C signatures of the phloem were observed between sites (SW and NE) in September when soil water potential had increased because of heavy rainfall. Pate & Arthur (1998) and Pate et al. (1998) showed recently that the δ13C signature of phloem sap from E. globulus was an excellent predictor of current water availability. Here, phloem δ13C was closely (albeit not robustly) related to water potential, although the slope of the regression line (approx. −3.9‰ MPa−1) was much greater than that recorded for Quercus spp. (−0.18‰ MPa−1, Damesin et al., 1998) and still greater than that recorded for Pinus spp. in much more seasonally arid environments in Australia (Warren et al., 2001). The analysis of the relationship between δ13C and water potential remains speculative, and further work is needed to confirm the linear range for beech.

As with previous studies where growth (measured as either BA increment or radial growth) has been compared with δ13C, there was reasonable agreement between the two parameters (see Fig. 4) in the present study. However, both growth and δ13C exhibited a variable ‘time-lag’ with environmental conditions (rainfall, radiation) that precluded a significant correlation. For growth, the lag effect has been at least partly explained by Sass & Eckstein (1995). The mean area of the first xylem vessels formed in each growth ring of beech is strongly influenced by water availability in the previous summer. Sass & Eckstein (1995) concluded that vessel formation at the annual beginning of cambial activity was mainly controlled by ‘internal factors’ with little influence of rainfall, while towards the end of cambial activity vessel formation was strongly influenced by recent rainfall.

Based on Pate & Arthur (1998), it could be proposed that water availability would be the major influence (or even the sole determinant) of the δ13C signature of carbon in beech tissues and phloem sap contents. If this were the case then thinning, which almost universally increases the amount of water available to remaining trees (e.g. Breda et al., 1995), should produce tissues that are more depleted in 13C than controls (the second hypothesis). However, in August when water supply on both sites was strongly limited by climatic conditions as indicated by the increase in soil water potential (Fig. 2), plots with the least tree density and BA (10 m2 ha−1) produced phloem that was the least depleted in 13C. An alternative explanation is that thinning – which caused a drastic decrease in LAI – has substantially increased the radiation intercepted by remaining trees, thereby increasing rates of photosynthesis and reducing Ci (e.g. Ehleringer et al., 1986; Leavitt & Long, 1986; Pearcy & Pfitsch, 1994; Israeli et al., 1996; Berry et al., 1997). A significant though weak relationship was also observed between hours of sunlight (a measure for the plant-available radiation) in the growing-season and δ13C in wood for the SW site. In part, the expression of an irradiance effect in δ13C only when water was in limited supply may stem from the rather low light compensation point of beech (approx. 350 µmol m−2 s−1, Kreuzwieser et al., 1997). Only when Ci is already constrained by partly closed stomata will the response to light (reduction owing to increased rates of carbon fixation) be realized.

Apart from direct tests of the hypotheses, the data allow useful comment on several other aspects of the interactions of climate and soils/landforms. It can be argued that, for any given climatic conditions, soil/landform characteristics hold soil water more ‘tightly’ at the NE site, against both evapotranspiration and evaporation, than at the SW site. This is mostly due to the poor water storage capacity (see soil descriptions and soil water potential, Fig. 2) of the soil at the SW site relative to that on the NE site. When water was in plentiful supply (e.g. in July), trees on the SW site transpired freely (Fig. 3a), at higher rates than trees on the NE site, and phloem carbon was significantly less depleted in 13C at the SW site (Fig. 6). A similar pattern of δ13C was seen in September. According to the literature, the higher rate of depletion of available soil water at the SW site (Fig. 2) should cause at least partial closure of stomata and a lower δ13C. Indeed, this was confirmed in August when water flux densities in the xylem of trees at the SW site fell by more than half. In addition SWP doubled and phloem δ13C was increased by at least 5‰. This pattern was partially reversed for the NE site – water fluxes in the xylem of trees at the NE site were greater than at the SW site during the dry month of August (Fig. 3b), which is supposed to be a consequence of the open stomata. In addition, SWP actually increased between July and August (Table 2), and δ13C signatures of phloem sap were little changed from those in July (Fig. 6). These results are strongly suggestive that during the dry month of August, the supply of available water at the NE site had been moderately well maintained, whereas it had been substantially depleted at the SW site (see Fig. 2).

In particularly dry years and the year immediately thereafter, there were few differences in δ13C of wood between the SW and NE sites. For example, the mean δ13C of wood from the years 1985, 1986 and 1989–92 was −25.5‰ at the NE site and −25.3‰ at the SW site. It is concluded that the soil/landform characteristics set ‘lower limits’ to water availability that are not particularly different between the two sites but that, in any given period, availability is largely determined by exposure to radiation and thence by vapour pressure deficit. The results are consistent with the findings of Walcroft et al. (1997) that the observed close relationships between soil water availability (e.g. Stewart et al., 1995) or cumulative transpiration (e.g. Livingston & Spittlehouse, 1996) and δ13C owe much to the integration of a number of environmental variables (e.g. precipitation, irradiance, temperature and air saturation deficit) by Ci, and to the direct relationship of transpiration and δ13C through stomatal conductance.

It was not intended to model δ13C and water availability at the available sites and, in any event, the required long-term climatic data (radiation and evaporation) were lacking. Moreover, these sites pose considerable difficulties for water balance modelling (in contrast to most sites for which models have been generated) owing to their shallow surface soils, rocky subsurface soils that quickly grade to ‘blocky’ parent material, and steep slopes. It must also be noted that water balance–δ13C relationships have seldom been tested beyond sites at which they were generated. It remains to be seen if on-going research can develop a successful water availability model for these sites and, then, if it has any predictive capacity for δ13C. Given the observed strong influence of radiation on phloem δ13C (and on wood δ13C), inclusion of radiation in the latter will probably improve sensitivity (Walcroft et al., 1997).

The sugar and carbon isotope analysis of the phloem sap of beech proved instructive, first for the range of δ13C values observed, and secondly for the relative concentrations of carbon isotopes and sugars.

Over the growing-season in 1999, the mean δ13C of sap from the 12 experimental plots (six per site) varied by more than 13‰: from approx. –20‰, values close to those observed for foliage from desert plants (e.g. Ehleringer & Cooper, 1986), to approx. –33.5‰, values typical of plants in high-rainfall, tropical forests (e.g. Jackson et al., 1993). This range is perhaps best interpreted in light of the soil profile at the SW site which, when coupled with the steep slope, clearly holds little water for plants for any length of time. Growth of beech on this site is probably a fine balance between evapotranspiration and rainfall during the growing-season, and is not buffered by a huge soil water storage pool. Changes to this balance via global warming may influence growth and vitality of beech, and may cause changes in the vegetation composition owing to a shift in competition patterns. On the other hand, the extreme values of δ13C for phloem sap are integrated within the tree to produce wood and foliage with δ13C signatures close to those found in other studies of beech (e.g. Schleser, 1990, 1992; Dupouey et al., 1993). The ability to sample phloem sap from beech, as demonstrated here, should enable future studies to elucidate relationships among carbon transport and metabolism and account for the differences in δ13C signatures among recently fixed carbon, transported sugars, and developing wood and foliage.

Phloem sap from beech has a sugar ‘profile’ comparable to that of other hardwoods for which there are data (e.g. Zimmerman & Ziegler, 1975; Pate & Arthur, 1998; Pate et al., 1998). While the sap was notably rich in sucrose (> 200 mM), it also contained glucose and fructose at concentrations in the range 0–10 mM. Sugar concentrations varied greatly between trees within treatments and sites, and never as predictably with δ13C as shown by Pate et al. (1998) for E. globulus in late summer. There, sugar concentrations in phloem sap increased as water availability (as indicated by δ13C) decreased. While there is some suggestion (Fig. 7) that sugar concentrations in beech phloem sap may be related to δ13C, the relationship is tenuous and largely obscured by tree-to-tree variation. Even when only those data from trees under the greatest water stress are considered (SW site in August), sugar concentrations were independent of δ13C. Pate & Arthur (1998) attributed their result to ‘independent responses of the sugar-loading systems of the mesophyll : minor vein system to current water status of the leaf, xylem and parent plant’. The differences between the two studies can be attributed to plot-level climate or (given the high tree-to-tree variability) even individual tree microclimate. Severe shortages of water for beech (as indicated by δ13C of −20‰) are restricted to days, or at most weeks, whereas in western Australia the ‘summer drought’ may last for 3 months or more.

This study has shed some light on past accounts of the abundance of carbon isotopes in beech, and should stimulate further attempts to model the relationships among environmental variables and δ13C. δ13C signatures in the phloem of beech are sensitive to seasonal changes in water availability, whereas δ13C signatures in the foliage are far less sensitive. In addition, the data suggest an interactive effect of low water availability and increased radiation on δ13C in the phloem sap. The demonstrated ability to sample phloem sap of mature beech in the field, both directly and easily, opens the door to many future field studies – most obviously of carbon and nutrient transport but also of hormone transport between shoots and roots.


The authors thank the DFG (SFB 433) for financial support. MAA acknowledges the support of an Alexander von Humboldt Fellowship. The authors thank S. Hauser, H. Spiecker, S. Augustin and E. Hildebrand for providing data on radial growth, LAI and soil water potential.