Trees

, Volume 20, Issue 5, pp 571–586

Inter-annual and seasonal variability of radial growth, wood density and carbon isotope ratios in tree rings of beech (Fagus sylvatica) growing in Germany and Italy

Authors

  • M. V. Skomarkova
    • Institute of Forest SB RASAkademgorodok
  • E. A. Vaganov
    • Institute of Forest SB RASAkademgorodok
  • M. Mund
    • Max-Planck Institute for Biogeochemistry
  • A. Knohl
    • Max-Planck Institute for Biogeochemistry
    • ESPM DepartmentUniversity of California
  • P. Linke
    • Max-Planck Institute for Biogeochemistry
  • A. Boerner
    • Max-Planck Institute for Biogeochemistry
    • Max-Planck Institute for Biogeochemistry
Original Article

DOI: 10.1007/s00468-006-0072-4

Cite this article as:
Skomarkova, M.V., Vaganov, E.A., Mund, M. et al. Trees (2006) 20: 571. doi:10.1007/s00468-006-0072-4

Abstract

We investigated the variability of tree-ring width, wood density and 13C/12C in beech tree rings (Fagus sylvatica L.), and analyzed the influence of climatic variables and carbohydrate storage on these parameters. Wood cores were taken from dominant beech trees in three stands in Germany and Italy. We used densitometry to obtain density profiles of tree rings and laser-ablation-combustion-GC-IRMS to estimate carbon isotope composition (δ13C) of wood. The sensitivity of ring width, wood density and δ13C to climatic variables differed; with tree-ring width responding to environmental conditions (temperature or precipitation) during the first half of a growing season and maximum density correlated with temperatures in the second part of a growing season (July–September). δ13C variations indicate re-allocation and storage processes and effects of drought during the main growing season. About 20% of inter-annual variation of tree-ring width was explained by the tree-ring width of the previous year. This was confirmed by δ13C of wood which showed a contribution of stored carbohydrates to growth in spring and a storage effect that competes with growth in autumn. Only mid-season δ13C of wood was related to concurrent assimilation and climate. The comparison of seasonal changes in tree-ring maximum wood density and isotope composition revealed that an increasing seasonal water deficit changes the relationship between density and 13C composition from a negative relation in years with optimal moisture to a positive relationship in years with strong water deficit. The climate signal, however, is over-ridden by effects of stand density and crown structure (e.g., by forest management). There was an unexpected high variability in mid season δ13C values of wood between individual trees (−31 to −24‰) which was attributed to competition between dominant trees as indicated by crown area, and microclimatological variations within the canopy. Maximum wood density showed less variation (930–990 g cm−3). The relationship between seasonal changes in tree-ring structure and 13C composition can be used to study carbon storage and re-allocation, which is important for improving models of tree-ring growth and carbon isotope fractionation. About 20–30% of the tree-ring is affected by storage processes. The effects of storage on tree-ring width and the effects of forest structure put an additional uncertainty on using tree rings of broad leaved trees for climate reconstruction.

Keywords

Carbohydrate storageClimateDendrochonologyDroughtStable carbon isotopes

Introduction

Tree-ring growth and wood density have been used extensively as indicators of climate change where growth is generally related to rainfall in the growing season and wood density is a predictor for summer temperatures. While these relations have been established for conifers (Schweingruber and Briffa 1996; Briffa et al. 1998, 2004), deciduous species have not been studied as detailed, especially with respect to wood density because of the more complicated structure and variability in their annual tree-ring growth (Bouriaud et al. 2004). Tree-ring carbon isotope composition could potentially help to expand the dendroclimatic analysis of deciduous trees because the 13C/12C ratios of wood and of cellulose are expected to also reflect climatic conditions (Wilson and Grinsted 1977; Leavitt 1993; McNulty and Swank 1995; Duquesnay et al. 1998; McCarrol and Loader 2004). Assuming that wood is formed from current assimilated carbohydrates, the fractionation of CO2 during diffusion through the stomata and during photosynthesis, which is strongly related to drought, would be “frozen” in wood cellulose (Brugnoli and Farquhar 2000). Long-term chronologies of δ13C in tree rings already detect changes in carbon isotope ratios in the atmosphere due to fossil fuel combustion (de Silva 1979; Freyer and Belacy 1993; Stuiver et al. 1984; Leavitt and Long 1985; Tang et al. 1999). However, recent studies of intra-annual variability of δ13C (Helle and Schleser 2004; Scartazza et al. 2004; Schulze et al. 2004) have questioned the simple correlation between δ13C in tree rings and climate due to the effects of carbohydrate storage (Gäumann 1935), especially early and late in the growing season.
https://static-content.springer.com/image/art%3A10.1007%2Fs00468-006-0072-4/MediaObjects/468_2006_72_Fig1_HTML.gif
Fig. 1

Left: Location of the studied sites in Germany (Hainich, Leinefelde) and Italy (Collelongo). Right: Climate diagrams of for studied sites. Hatched areas indicate periods of a positive hydrological balance, while dotted areas indicate periods of drought

Given the uncertain relation between wood density and wood isotope ratios and their responses to climate, we studied the climate induced variability of radial growth and tree-ring structure using beech (Fagus sylvatica L.) from Central Germany (Hainich and Leinefelde site) and Italy (Collelongo), regions which are quite different with respect to the seasonal course of climatic conditions. Tree-ring width and intra-annual density and δ13C changes were analyzed in order (i) to identify which climatic factors influence tree radial growth and to what extent, (ii) to examine the relationship between intra-annual wood density, isotope composition and climate. The hypothesis was that tree-ring growth at the northern latitude forest of Hainich and Leinefelde would be limited by temperature, while growth of the Italian Collelongo would affected by summer drought despite of high annual rainfall. And (iii) to detect a potential management effect between the managed age class forest of Leinefelde and the protected national park of Hainich.

Materials and methods

Experimental sites

This study was conducted in three beech (Fagus sylvatica L.) forests located in Germany (Hainich and Leinefelde) and Italy (Collelongo) (Fig. 1).

The Italian site, Collelongo, is located in the Abruzzo region (41°50′58″N, 13°35′17″E, 1560 m a.s.l.). Environmental conditions and stand parameters are representative of beech forests in central Apennine. Stand density and basal area were very high (775 trees ha−1, 39.4 m2 ha−1) (Table 1) with a mean diameter at breast height of 24.6 cm and a mean height of 21.2 m (determined in 2002). In 2001, the mean age of trees was 109 years. The soil, developed on calcareous bedrock, has a variable depth (40–100 cm) and can be classified as a humic alisol. Site topography is gently sloping (Scartazza et al. 2004). Mean annual temperature and precipitation between 1951 and 1998 were 6.3°C and 1180 mm year−1, respectively (http://www.knmi.nl/samenw/eca).
Table 1

Basal area and stand density of the three study sites in Hainich, Leinefelde and Collelongo

 

Hainich

Leinefelde

Collelongo

Basal area (m2 ha−1)

34.2

35.2

39.4

Tree density (>7 cm) ha−1

334

224

775

Canopy height (m)

34.0

36.9

21.2

BHD class (trees ha−1)

   

 0–5

141

0

75

 6–10

99

8

80

 11–15

49

4

165

 16–20

30

0

140

 21–25

34

0

105

 26–30

19

0

115

 31–35

22

8

65

 36–40

33

28

50

 41–45

19

56

25

 46–50

25

80

10

 51–55

16

28

0

 56–60

15

12

0

 61–65

14

0

0

 66–70

7

0

0

 71–75

4

0

0

 76–80

3

0

0

 81–85

2

0

0

 86–90

1

0

0

BHD: Breast height diameter (cm)

The German sites, Leinefelde (51°19′42″N, 10°22′04″E, altitude 420–450 m a.s.l.) and Hainich (51°04′45″N, 10°27′07″E, altitude 445 m a.s.l.) are approximately 30 km apart and located in western Thuringia. Mean annual precipitation is 750–800 mm, and mean annual temperature ranges from 6.5 to 7.0°C (http://carbodat.ei.jrc.it). The Leinefelde forest soil is a uniform, fertile, silty clay loam Luvisol (FAO 1998). Soil depth is 40–90 cm (Mund 2004). The forest is an intensively managed even-aged beech stand (mean tree age in 2003 was 114 years) with sparse single oak trees. Thinning removed dead, damaged, co-dominant and suppressed trees to favor the best dominant stems (Bascietto et al. 2004). Stand density was three times lower compared to Collelongo, and as a result, basal area in Leinefelde was lower despite higher maximum BHD and canopy height (Table 1). The Hainich site is a national park that has not been managed regularly for about 40 years and only single trees were removed. The stand is uneven-aged with old dominant trees originating from a “coppice with standards” system about 150 years ago. Stand density was higher than in the managed Leinefelde forest due to the uneven-age structure, but maximum breast height diameters (BHD) were much larger at Hainich than in both other stands and reached 0.9 m (Table 1). The soil of the Hainich site is a silty clay Cambisol with a depth of 50–70 cm soil depth (Mund 2004).

Long-term daily meteorological data (temperature and precipitation) were taking from the following weather stations: Roma-Campino (Italy) for the period 1951–1998 and Leinefelde (Germany) for the period 1957–2003. We are aware that Roma-Campino is at a lower elevation (105 m a.s.l.) than Collelongo (1560 m a.s.l.), but data for Collelongo are only available since 1996. In Fig. 1 we plotted the climate diagram for Collelongo and added the long-term temperature and precipitation data of Roma. It becomes clear, that the montane climate of the Apeninns is dominated by the regional Mediterranean climate with a distinct dry period in the main part of the growing season (June–July). The precipitation is higher than at the Roma station but most of this precipitation falls as snow and runs off during snow melt. Thus, the greatest difference between the Italian and the German site is in the seasonal distribution of precipitation. In contrast to Italy summer drought periods are generally not observed at the German sites.

Tree sampling and dendrochronological analysis

Twelve beech trees were sampled and analyzed at each site. Two cores were collected from each tree at a height of 1.3 m with a 5 mm-diameter increment borer. One core was used for measurement of wood density and tree-ring width, the other core for measurement of δ13C.

Measurements of tree-ring width were carried out using a semi-automatic device LINTAB-III (Cook et al. 1990; Rinn 1996; Vaganov et al. 1996). Density profiles of tree rings were measured in the densitometric laboratory at Krasnoyarsk, Russia (Walesch Electronics, Switzerland, Schweingruber 1988; Vaganov and Shashkin 2000). Longitudinal strips were sawn from cores at right angles to the fiber direction. The strips have a constant thickness close to 1.2 mm. They were exposed for 1 h to X-ray radiation, the source being 3.5 m from samples, using Kodak TL, and standard electrical conditions (accelerating tension = 8.5 kV; flux intensity = 15.0 mA; Polge 1966; Schweingruber 1988). The detailed density profiles were obtained using a densitometer DENDRO-2003 (Schweingruber 1988; Kirdyanov 1999). Since wood of beech is diffuse-porous and in a number of cases the resolution of the light beam during scanning of the X-ray photography is the same as vessels size (5 μm), measurements of wood density for each tree-ring was repeated in triplicate with some displacement of the scanning line. The average curve was used to determine the maximum density during the course of growth of the tree-ring (Vaganov 1990; Vaganov and Shashkin 2000). To define the average density profiles, density data from the three measurements of the same tree-ring were averaged for segments which represent 1/10th of the tree-ring (Vaganov 1990; Sass and Eckstein 1995).

Cross-dating of individual cores was carried out based on tree-ring width (Schweingruber 1988) using the program COFECHA (Cook and Peters 1981; Holmes 1992). Age-specific changes of tree-ring width and maximum density were eliminated, in accordance to the standards in dendrochronology. Absolute values of each tree-ring width were standardized by calculating the ratio of the tree-ring deviation in a specific year and the long term average of tree-ring width to remove the effect of age and diameter on tree-ring width (Shiyatov 1986; Cook et al. 1990; Vaganov et al. 1996). This process filters the low-frequency component from the short-term fluctuation of tree-ring width and other characteristics. A spline function was used to determine the long-term trend for each tree core using the program ARSTAN (Cook et al. 1990). The following parameters were calculated:

The “interserial correlation” describes the variation of a certain tree-ring parameter in a certain year between all sample trees. This parameter gives information about the heterogeneity of the stand and the characteristics of the site. The parameter also reveals the similarity in tree growth to inter-annual climatic variability.

The “mean sensitivity” is a measure of the variability of the tree rings in consecutive years. Thus, the within-series statistics characterizes the common signal in a standardized chronology. The mean sensitivity is related to the variance of the standardized chronology, but uses the absolute change of consecutive years to express variation.

The variance of the First Eigenvector explains the fraction of the variability of tree-ring parameters that is explained by environmental factors common for all trees at each site (Cook et al. 1990).

The First order autocorrelation explains the correlation between the tree-ring width in the past year (t−1) and the ring width in the year under investigation (t).

The effect of climatic factors on the inter-annual variability of tree-ring growth and wood density was estimated through correlations between the chronologies and monthly climatic data from the above mentioned meteorological stations. These calculations were performed each month for 12 months (from October of the previous year to September of the current year) using the program RESPONSE (Holmes 1983).

Carbon isotope analysis

The carbon isotope analysis in tree rings was determined using a laser ablation-combustion line coupled to an isotope-ratio-mass-spectrometer (Nd:YAG 266 nm UV Laser, Merchantek New Wave, Fremont, Calf. Coilped to Finnigan Delta+XL) as described in detail by Schulze et al. (2004). The exact location of each ablation spot was visualized with a camera mounted directly on the laser ablation station. The wood samples were enclosed in a quartz glass chamber which was flushed by helium as carrier gas. For quantitative combustion of the ablation particles to CO2 and H2O the sample was passed at 700°C through a Al2O3 tube containing CuO wire as source of oxygen. The reaction gas was passed through a GC column (HayeSep D, Bandera, Texas) to separate CO2 from other gases. Water was removed with a Nafion water trap. Isotope ratios were expressed in the ‰ notation based on Pee Dee Belemnite (VPDB) as standard. The spatial resolution of a laser shot is 70 μm. A series of 4–5 shots was used to characterize the wood at one location. The series of shots was repeated every 120 μm.

Profiles of δ13C were measured on five sample trees per site for the 6-year-period, 1998–2003. The previous 13 years (1991–2003) were measured for two sample trees per site. The distance between laser ablation spots was 0.2 mm. Since width of tree rings formed in different years is different, the number of measurements of density and isotopes within these tree rings varied, accordingly. Data were interpolated when comparisons were made between wood density and δ13C on an absolute time scale.

Fluxes measurements and modeling of δ13C in assimilates

At the Hainich and the Leinefelde sites in Thuringia, net ecosystem exchange was continuously measured using the eddy covariance technique at a height of 43.5 m since September 1999 and since April 2002, respectively. The flux system consisted of a triaxial sonic anemometer (Gill Solent R3, Gill Instruments, Lymington, UK) and a fast response closed-path CO2/H2O infrared gas analyzer (IRGA) in absolute mode (LiCor 6262-3, LiCor Inc. Lincoln, NE, USA). Fluxes were 2-D coordinate rotated, corrected for fluctuation dampening in the tube and quality controlled (Knohl et al. 2003). Nighttime flux measurements during turbulent conditions were used to derive ecosystem respiration (RE) based on an exponential statistical model (RE=keaTe), where T denotes soil temperature at 2 cm (°C), θ denotes soil moisture content (vol.%) and k, a, and b are fitting parameters. Gross primary productivity (GPP) was calculated from net ecosystem exchange plus ecosystem respiration (RE).

Carbon isotope ratios of assimilated carbon were calculated as δ13C of assimilates = δairΔcanopy, where δair denotes the carbon isotope ratio of CO2 in tropospheric air (approximately −7.9‰), as measured at the site within the EU-project AEROCARB, and Δcanopy denotes canopy discrimination. Δcanopy was derived from the ratio of CO2 concentration at the site of carboxilation (cc) to CO2 concentration in the atmosphere (ca), according to Farquhar et al. (1989):
$$\Delta _{{\rm canopy}} = \bar a + (b - \bar a)\frac{{c_{\rm c} }}{{c_{\rm a} }},$$
where \(\bar a\) summarizes several fractionations from the diffusion of CO2 from the canopy air space to the site of carboxylation and b denotes the fractionation of the enzyme-catalyzed fixation of CO2.
The CO2 concentration at the site of carboxylation (cc) was estimated from the ratio of GPP to total canopy conductance (gc) assuming a big-leaf approach
$$c_{\rm c} = \left( {c_{\rm a} - \frac{{G\!P\!P}}{{g_{\rm c} }}} \right)$$
Total canopy conductance summarizes aerodynamic, stomata and mesophyll conductance. Aerodynamic conductance was calculated based on continuous turbulence measurements. Mesophyll conductance was estimated from literature (0.6 mol m−2 s−1, based on Epron et al. (1995) and scaled with leaf area index). Stomata conductance was calculated from continuous eddy covariance measurement with an inverted Penman–Monteith equation during dry conditions to minimize evaporation effects from the soil and then interpolated based on a simple regression model as developed by Collatz et al. (1991). Details of the procedure are presented in Knohl and Buchmann (2005). It is important to note, that this approach to model δ13C of assimilates reflects an ecosystem perspective since ecosystem scale fluxes of carbon, water and energy are used. Therefore, the δ13C of assimilates is flux-weighted and thus biased towards large, dominant trees that contribute the most to the ecosystem flux. Dominant trees were also used for the analysis of δ13C in wood to allow for a meaningful comparison. Since microclimatic conditions inside the canopy varie strongly from those outside, resulting in vertical differences in carbon discrimination (Ehleringer et al. 1986; Knohl et al. 2005), we compared modeled δ13C of assimilates with δ13C of bulk material and sugars from leaves (four sampling dates with six replicates each) in the top canopy, where the top contribution to the flux is expected. Sampling approach and sugar extraction and analysis are explained in Knohl et al. (2005) and Götlicher et al. (2005). Understory at the Hainich site consists of herbal geophytes that have a relevant contribution to the overall carbon uptake only in spring before bud break of the trees and should not influence the discrimination calculation afterwards. Modeled and measured values of assimilates and leaf material agreed well within an uncertainty of 1‰  (see Fig. 7).
https://static-content.springer.com/image/art%3A10.1007%2Fs00468-006-0072-4/MediaObjects/468_2006_72_Fig2_HTML.gif
Fig. 2

Schematic showing the transformation of within tree-ring density measurements and δ13C to absolute time scale using the seasonal cumulative growth curve as measured by girth band for the years 2002 and 2003. Depending on growth, the transformation is different for each year

Table 2

Statistical parameters of tree-ring width and maximum tree-ring density for the sites Hainich, Leinefelde and Collelongo based on 12 trees and for the period 1951–1998 (Italy) and 1957–2003 (Germany)

 

Tree-ring width

Maximum wood density

 

Hainich

Leinefelde

Collelongo

Hainich

Leinefelde

Collelongo

Ring width (mm), maximum density (mg cm−3)

1.68

1.60

1.49

930

990

930

Standard deviation

0. 26

0.29

0.19

0.03

0.02

0.02

Number of samples

12

12

12

12

12

12

Average interserial correlation

0.49

0.46

0.50

0.20

0.21

0.22

Sensitivity coefficient

0.23

0.15

0.18

0.02

0.02

0.02

Variance in first eigenvector

37.96

31.93

43.74

21.03

19.61

27.95

First order auto-correlation

0.45

0.42

0.30

0.15

0.05

0.30

Radial growth monitoring

Intra-annual radial growth was measured using automatic girth-bands. At the Hainich site 17 beech trees representing all diameter classes have been measured since 2002. For comparison with the tree cores in this study, weekly mean growing season increment (2002 and 2003) of the five dominant trees were measured. The results were expressed as a percent of total annual girth increment (Bouriaud et al. 2004). To compare the seasonal growth dynamics of wood density and δ13C with tree-ring growth (Fig. 2), data were converted into absolute time scale by aligning the change of wood parameters (density, δ13C; Fig. 2 middle) along fractions of the whole tree-ring with the cumulative growth curve (scaled from 0 to 100%; Fig. 2 left). This results in changes of δ13C and density on an absolute time scale. Depending on growth the synchronization is specific for each year (Fig. 2 right).

Results

Long-term climatically induced tree-ring width and maximum density variations

Mean radial growth rates (tree-ring width) of the unmanaged Hainich and the managed Leinefelde site were very similar and both tended to be higher than at the Collelongo site (Table 2). In contrast, maximum wood density was higher at Leinefelde than at Hainich and Collelongo. Since the highest wood density is always reached at the end of the growing season, differences in wood density may be an effect of small changes in the duration of cambial activity which could result from differences in management or from complex interactions of temperature and rainfall. The inter-annual variability was higher for radial growth (+50%) than for maximum wood density (+5%; Fig. 3). Climate induced variability of beech tree-ring width (percentage of the explained variability in the first Eigen-vector) was 32–43% across all sites. This variability is lower than in conifers growing at sites where tree radial growth is limited by a single dominant climatic factor, such as temperature at the northern and upper timberline. In such cases the variance of first Eigenvector may reach 60–70% (Cook and Peters 1981; Shiyatov 1986; Vaganov and Shashkin 2000). Climate induced variability of the maximum density was even less and did not exceed 28%. A relatively small effect of climate on tree-ring width was also supported by a low coefficient of interserial correlation (correlation between individual series) (see Table 2). The first order autocorrelation was 30–40% for tree-ring width, meaning that 10–15% of the tree-ring width in a given year is explained by the growth conditions in the previous year. This is an indication that carbohydrate storage from the past year is affecting growth in the present year, which appears to be characteristic of deciduous trees (Kozlowski and Pallardy 1997).
https://static-content.springer.com/image/art%3A10.1007%2Fs00468-006-0072-4/MediaObjects/468_2006_72_Fig3_HTML.gif
Fig. 3

Standardized site tree-ring width (A) and maximum density (B) chronologies for the three studied sites. The standardized chronologies are based on 12 trees per site

The first order autocorrelation of wood density was not significant for all sites. Despite a large variation, there is a trend of increased density with increasing tree-ring width for both climatic regions (Fig. 3).

Figure 3 indicates a periodicity in growth at the German sites with a frequency of about 20 years. Some of this variation could be due to masting years (Hilton and Packam 2003), although this periodicity has disappeared since 1970. The pattern is also not apparent at Collelongo. The maximum density data do not show such periodicity at any of the sites. Even in the closely neighbored sites of Leinefelde and Hainich the maximum density data do not always go in parallel. The curves show a decrease in ring width and density for the dry year of 2003.

Correlation coefficients of tree-ring width and maximum density with monthly temperature and precipitation (1900–2005) show that both temperature and precipitation influence annual variability of tree-ring width and wood density (Fig. 4). May temperature and June precipitation in Collelongo are significantly correlated with radial growth, while maximum wood density was correlated with warm temperatures during the second half of a growing season (July–September) and decreases in years with a moist September.
https://static-content.springer.com/image/art%3A10.1007%2Fs00468-006-0072-4/MediaObjects/468_2006_72_Fig4_HTML.gif
Fig. 4

Correlation coefficients of tree-ring and maximum density chronologies with climatic data: lines indicate correlations with temperature, columns indicate correlations with monthly precipitation, (*) denotes a significant correlation (p<0.01)

In contrast, growth at the Hainich site is mainly correlated with late winter and spring precipitation. While this effect is not visible at Leinefelde, maximum density correlates with July temperatures at both sites. A closer inspection reveals a very weak correlation between early summer precipitation and ring width. It was surprising to see that the large variation in precipitation between February and June had no effect on growth at Collelongo (Fig. 5). The large variability in ring width at Hainich and Leinefelde indicates that precipitation between February and June had only a minor effect (Fig. 5). Maximum wood density was positively correlated with temperature across all sites. The Italian and the German sites show similar trends, but at different ranges of temperature. The Italian site had a 7°C higher July temperature, but a similar slope of the density–temperature relation as the German site. Thus, a correlation between wood density and July temperature is distinct for the climatic region and only valid within the range of observations. This indicates that there is either a major acclimation or the correlation with maximum temperature is a measure for the length of the growing season rather than a physiological response to temperature.
https://static-content.springer.com/image/art%3A10.1007%2Fs00468-006-0072-4/MediaObjects/468_2006_72_Fig5_HTML.gif
Fig. 5

Regressions between standardized tree-ring width and precipitation (left), and between standardized maximum wood density and air temperature (right)

https://static-content.springer.com/image/art%3A10.1007%2Fs00468-006-0072-4/MediaObjects/468_2006_72_Fig6_HTML.gif
Fig. 6

Example for intra-annual variations of wood anatomy (A: the laser shots for isotope determination can be seen as vertical bars), and of wood density and δ13C for beech tree number 7 at Hainich, Germany. Density resolution is 10 μm, isotope determinations were spaced by 120 μm

Within tree-ring variation of wood density and isotopic composition

Figure 6 shows the temporal changes in wood density and δ13C during two consecutive years (2002 and 2003) using one predominant tree of Hainich out of five trees per site as example (sample H7, see Table 3). Lower values of wood density were observed at the beginning of the growing season. This was followed by a period with more or less constant wood density, which then increased later in the season to reach a maximum at the end of the growing season. Smaller changes in the density profiles as well as the seasonal trend were mainly caused by seasonal changes in the frequency of vessels in the wood matrix.
Table 3

Breast height diameter, tree height, crown height, tree-ring width and δ13C value of wood in June for the sample trees at Hainich (H), Leinefelde (L) and Collelongo (C) used for isotope analysis

Tree

D1.3 (cm)

Height (m)

Crown area (m2 tree−1)

Ring width (mm)

δ13C of wood

    

2002

2003

2002

2003

Hainich

       

 H2

39

31.8

42.4

0.63

0.79

−29.2

−27.4

 H6

57

30.4

44.8

1.39

1.35

−30.6

−28.5

 H7

52

34.3

107.5

4.60

2.51

−27.1

−25.7

 H10

65

35.7

115.0

1.01

0.89

−25.3

−24.7

 H12

52

30.0

76.2

1.32

1.83

−25.9

−24.7

Average of five δ13C-trees

   

1.79

1.47

−27.6

−26.2

Standard deviation of five δ13C-trees

   

1.43

0.63

−2.0

−1.5

Average of 12 density trees

   

1.33

1.03

  

Standard deviation of 12 density trees

   

1.14

0.61

  

Leinefelde

       

 L3

40

34.8

20.4

2.51

1.32

−27.8

−26.6

 L7

46

37.6

35.8

2.17

2.03

−24.9

−23.4

 L8

52

36.4

47.8

1.41

1.67

−26.0

−25.5

 L9

44

36.9

26.0

0.93

0.77

−25.5

−24.5

 L11

37

33.6

27.4

2.29

1.54

−26.3

−25.0

Average of five δ13C-trees

   

1.46

1.47

−26.2

−25.0

Standard deviation of five δ13C-trees

   

0.60

0.42

−1.0

−1.2

Average of 12 density trees

   

1.46

1.25

  

Standard deviation of 12 density trees

   

0.71

0.46

  

Collelongo

       

 C1

25

19.3

11.0

1.11

1.12

−26.9

−25.4

 C2

31

21.2

24.0

1.37

1.79

−25.4

−24.2

 C6

32

21.4

26.9

0.66

0.39

−26.0

−25.6

 C9

35

21.7

30.0

1.47

1.95

−26.0

−25.0

 C11

31

22.2

37.1

1.27

1.37

−25.7

−24.6

Average of five δ13C-trees

   

1.18

1.32

−26.0

−25.0

Standard deviation of five δ13C-trees

   

0.28

0.55

−0.5

−0.5

Average of 12 density trees

   

1.01

0.94

  

Standard deviation of 12 density trees

   

0.35

0.56

  

All trees were dominant in the canopy. Differences between trees used for δ13C (n=5) and for density (n=12), and differences between year 2002 und 2003 and sites were not significant

In contrast to the seasonal trend in wood density, δ13C increased at the beginning of growth, and then continually decreased in 2002 until the end of tree-ring growth. Between tree rings there is a sharp rise in δ13C values of wood. In the dry year of 2003 the δ13C values initially decreased to reach an almost constant value until about 50% of the tree-ring was formed. This was followed by a distinct increase in δ13C. The tree in this example had a smaller tree-ring in the dry year of 2003 than in the wetter year 2002.

Seasonal trends of canopy assimilation, wood density and δ13C are shown for the Hainich/Leinefelde sites for the wet year of 2002 and for the dry year of 2003 (Fig. 7). Total rainfall was evenly distributed throughout the season in 2002 but decreased during March and summer 2003 (Fig. 7A). Total rainfall between May and August was 50% less in 2003 (181 mm) compared to 2002 (389 mm). Precipitation occurred again in mid August 2003. Thus, 2003 was a significantly warmer summer than in 2002, with maximum daily temperatures reaching 25°C compared to 19°C in 2002 (Ciais et al. 2005).

Canopy CO2 assimilation (GPP) and canopy conductance (Fig. 7B) differed during the two years, while GPP showed a broad summer maximum in 2002. In 2003 flux rates were very high in May and early June, followed by a sharp decline of fluxes from late June to August. Canopy conductance showed a marked decrease in the dry period of 2003. During the wet August, photosynthesis and canopy conductance recovered, and rates were similar for both years at the end of the growing season. Leaves developed in spring and were shed in autumn about at the same time in both years.

Stomata (gc) and CO2-flux (GPP) changed in parallel (Fig. 7B), and the ratio of GPP/gc remained constant. Thus, the modeled δ13C of assimilates was very similar (−26‰) in both years (Fig. 7C), except for a short increase in late May and early August of 2003 when canopy conductance decreased more that photosynthesis. In order to validate the model for δ13C of assimilates we compared model values with δ13C of bulk leaf material and sugars of leaves from an intensive campaign from 29 July to 19 August 2002 (four sampling days, one per week). Bulk material (−26.6±0.5‰) was less variable from sampling to sampling than sugars (−26.5±1.0‰), but showed an almost identical average as model output indicating the reliability of our approach within the uncertainty associated with natural variability (Fig. 7C).
https://static-content.springer.com/image/art%3A10.1007%2Fs00468-006-0072-4/MediaObjects/468_2006_72_Fig7_HTML.gif
Fig. 7

The seasonal courses of (A) bi-weekly precipitation (column) and temperatures (lines), (B) bi-weekly canopy conductance and gross primary production (GPP), (C) weekly growth rates, modeled δ13C of assimilates and measured δ13C of leaf sugar for model validation (in August 2002), (D) δ13C of wood and (E) wood density for individual sample trees during the 2002 and 2003 growing seasons. Hai: Hainich site, Lei: Leinefelde site

Tree-ring growth (Fig. 7C) started in early May during both years, and reached a maximum in late May and early June. In the wet year 2002, 45% of the tree-ring was formed by July, while growth peaked much earlier than in 2003 (end of May–beginning of June).

In 2002, the δ13C values of wood were either constant or showed a continuous decrease over time (Fig. 7D). This contrasts 2003, where δ13C generally increased starting in May in trees with large canopy area, to reach the highest level late in the growing season. The average level of δ13C increased by 0.5‰  in 2003 as compared to 2002. Growth almost ceased during the period of summer drought in 2003.

Wood density (Fig. 7E) showed less variation between trees, and increase during the growing season. The maximum density was similar between dry and wet years, but maximum density was reached earlier in the dry year.

The δ13C of wood showed large variation between trees, ranging between −30.5 and −25.5‰  in the wet year 2002 (Fig. 8). This large variation between individual trees is difficult to explain. Generally there is a trend of increasing δ13C in wood with an increasing crown area relative to stem diameter (Table 3). This means that trees with large crowns per sap wood area have greater problems stabilizing their water relations, and are more sensitive to drought than trees with smaller crowns. More importantly, the managed forests with high stand density and even-aged canopies have higher δ13C relative to their crown size than uneven-aged structured stands at same rainfall. This lead to the unexpected observation that the water balance of the managed Leinefelde stand as indicated by the δ13C value of wood (−25.27‰±1.59) was more similar to the stand at Collelongo (−25.20‰±0.54) than to the un-managed stand at Hainich (−26.14‰±1.39). The climatic differences between the two sites seem to be over-ridden by stand structure, which is dominated by current (Leinefelde) or historical (Collelongo) management. Thus, the climatic signal in δ13C can only be extracted with knowledge of stand density and crown structure.
https://static-content.springer.com/image/art%3A10.1007%2Fs00468-006-0072-4/MediaObjects/468_2006_72_Fig8_HTML.gif
Fig. 8

The relation between δ13C of wood and the crown area/breast height diameter-ratio in wet and dry years at the study sites

Table 3 shows large variability in ring width between trees and on average no significant change in ring width between the dry and wet year and the German and Italian sites. There is only a tendency for slower growth in Italy and only individual trees showed the expected drought response in 2003.

At Collelongo precipitation was below normal in 2001 and 2003, and higher-than-normal in 2002 (Fig. 9). The δ13C values showed large variation between trees in dry years, while δ13C values were similar in wet years. The general level of δ13C was similar to the German sites (−26‰). Also, there was a seasonal trend toward decreasing δ13C during the wet year but increasing δ13C in the dry years. Seasonal trends of wood density did not respond to summer drought, except that there was a larger variation between trees in dry years.

Only for the German Hainich site δ13C of assimilates were modeled, and Fig. 10 shows that the seasonal trend in modeled δ13C of assimilates is not related to the δ13C in wood (Fig. 9). In 2002, the isotope ratio of wood was initially higher than that of modeled δ13C of assimilates indicating that wood was formed from a different source than assimilation. The isotope ratio of assimilates increased and that of wood decreased to reach a level by mid season that is about 2‰  lower in wood than in assimilates. At the end of the growing season, δ13C of assimilates increased but δ13C in wood remained constant suggesting that assimilates at the end of the growing season are not deposited into wood. In the following dry year of 2003 growth starts as in the wet year of 2002 with wood that is isotopically enriched, indicating that the mobilization of carbohydrates from storage of the previous autumn seemed to cause a shift in δ13C. The isotope ratio of assimilates increased during the course of the dry year to very high values (closed stomata) in August, while wood remained at constant δ13C. In autumn, similar to the wet year, δ13C in assimilates decreased, but again these assimilates apparently were not used for wood growth which remained at a high δ13C level. Each data point in Fig. 10 represents 10% of the tree-ring width. Thus, about 30–40% of the data points are affected by mobilization and accumulation of storage products.

The isotope ratio in wood of Fagus is similar to the modeled isotope ratio in assimilates only for a very short period during the growing season. Despite the large seasonal variation in δ13C of assimilates, the isotope ratios in wood show less seasonal variation. Also, growth in the dry year did not show the reduction which one would have expected this from the trend in assimilation. Total C-assimilation was reduced by 22% (N=12 bi-weekly flux data) in 2003 compared to 2002, while ring growth was reduced by only 17.5% (N=12 biweekly increment data). The use of stored carbohydrates for growth could explain the strong correlation between tree rings of consecutive years (First order autocorrelation in Table 2).

The relationship between wood density and δ13C differed between dry and wet years across all sites (Fig. 11). In wet years, a strong negative correlation was observed, that is dense wood correlated with isotopically depleted wood. This contrasts to a dry year, where a positive correlation was observed between wood density and isotope composition. Overall, there was a large variation between trees in dry and wet years.

Table 4 summarizes the correlations between wood density and δ13C for the periods of 1999 and 2003 based on five sample trees, and for 1991 and 2003, based on one sample tree in Hainich and Collelongo. With few exceptions, wood density and δ13C at Collelongo where positively correlated during dry years. In contrast, the German sites generally exhibited a negative correlation between wood density and δ13C during wet years. Dry years in Italy as well as in Germany were characterized by low rainfall in June–July.
https://static-content.springer.com/image/art%3A10.1007%2Fs00468-006-0072-4/MediaObjects/468_2006_72_Fig9_HTML.gif
Fig. 9

(A) Bi-weekly average temperature and precipitation, (B) δ13C of wood and (C) wood density of individual sample trees on a relative time scale at the Collelongo site

https://static-content.springer.com/image/art%3A10.1007%2Fs00468-006-0072-4/MediaObjects/468_2006_72_Fig10_HTML.gif
Fig. 10

Relationship between δ13C in assimilates and average δ13C in wood at the Hainich site

https://static-content.springer.com/image/art%3A10.1007%2Fs00468-006-0072-4/MediaObjects/468_2006_72_Fig11_HTML.gif
Fig. 11

Relationship between δ13C in wood and wood density for individual sample trees at three studied sites for a wet year (2002) and a dry year (2003)

Discussion

There was not a strong correlation between the climatic response of beech tree-ring width and maximum density under the investigated conditions. One reason for this observation is that radial growth and formation of wood, as indicated by tree-ring width, wood density and δ13C, is determined by different climatic factors during the seasons and the interactions differ between years (Vaganov and Shashkin 2000). Higher precipitation and snow are typical for the autumn–winter period at the Italian site. This results in high soil water contents mainly at the very beginning of the growing season. Therefore, radial growth correlated positively with temperature in May (the beginning of the season after snowmelt) and precipitation in June when summer drought affects growth. The correlation with temperature shows that the conditions in the first half of the growing season are important for cambial activity and production of the new wood elements (Sass and Eckstein 1995; Bouriaud et al. 2004; Scartazza et al. 2004). Surprisingly the temperature effect is smaller in Germany. Also precipitation at the German sites was equally distributed throughout the year with a summer maximum, and the correlation to precipitation was less pronounced.

The climate response of maximum density showed that denser wood was formed in the years with a warmer late summer. This is the period when latewood in beech tree rings is formed (Bouriaud et al. 2004), and carbohydrates are stored for the next season. Beech trees at the Hainich and Leinefelde sites showed some differences in climatic responses despite growing in vicinity and on similar soils. Systematic thinning at Leinefelde produced a uniform dense canopy, while the Hainich site exhibited a diverse crown structure typical of unmanaged forests. The significant correlation between maximum density and temperatures for July to September indicates that radial growth and wood formation continue longer in Leinefelde than in Hainich, which may be related to stand density (Hauser 2003; Kramer and Kätsch 1982; Zahner and Oliver 1962).

If tree-ring width responds to temperature and precipitation at the beginning of the growing season and maximum density to climate at the end of the growing season, then δ13C variations should mainly reflect the current conditions of assimilation during the middle of the growing season. An influence of inter-annual changes of water availability on 13C in wood has been described before (Leavitt 2002; Warren et al. 2001; Kagawa et al. 2002). In cases where water is not a limiting factor during the season, δ13C variations should be negatively related to wood density, because the production of the larger number of vessels in wood is thought to be associated with lower δ13C. We confirm this hypothesis for wet years, where δ13C and wood density are negatively correlated. This correlation changes in dry years, where δ13C increases with increasing wood density. Apparently storage interacts with these correlations needs further attention.
Table 4

Correlation between δ13C and wood density during growth of individual tree rings for three investigated sites (for the period 1998–2003 and for the period 1991–2003)

Site and no. of sample

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

Collelongo-1

       

0.86

−0.50

0.86

0.92

−0.72

0.76

Collelongo-2

       

0.73

0.91

0.75

0.75

−0.79

0.87

Collelongo-6

−0.73

0.73

0.70

0.90

−0.95

0.40

0.88

0.93

−0.56

0.86

0.79

−0.89

0.65

Collelongo-9

       

0.42

−0.40

0.79

0.30

−0.63

0.43

Collelongo-11

       

0.86

−0.73

−0.81

0.85

−0.86

0.74

Hainich-2

0.93

−0.77

−0.64

0.80

0.74

−0.47

−0.89

−0.68

−0.76

−0.67

−0.85

−0.51

0.79

Hainich-6

       

−0.78

−0.93

−0.82

−0.90

−0.75

0.71

Hainich-7

          

−0.63

−0.88

0.81

Hainich−10

       

−0.78

−0.83

−0.92

−0.79

−0.91

0.79

Hainich-12

       

0.82

−0.91

−0.31

−0.59

−0.29

0.81

Leinefelde-3

       

0.41

0.57

−0.91

−0.76

−0.54

0.96

Leinefelde-7

       

−0.69

−0.69

−0.74

−0.96

−0.25

0.88

Leinefelde-8

       

−0.88

−0.78

−0.14

−0.94

−0.68

0.31

Leinefelde-9

       

−0.61

−0.57

−0.80

−0.68

0.66

0.38

Leinefelde-11

       

−0.71

−0.88

−0.91

−0.91

−0.86

0.07

Bold numbers indicate a significant negative correlation, thin numbers a significant positive correlation, and non-significant correlations are printed in Italics (p<0.05)

The seasonal change of δ13C in wood contains additional information when related to the δ13C in assimilates. Although there are uncertainties associated with potential fractionations during transport of assimilates and biosynthesis of secondary compounds (Gleixner et al. 1993), the differences in seasonal changes of δ13C in wood and assimilates are striking and suggest a significant interaction with carbohydrate storage. The isotope ratio of new assimilates at the beginning of the growing season are opposite to the isotopic changes in wood and we could interpret the difference between wood δ13C and a 1:1 line with δ13C of assimilates as the contribution of carbohydrate mobilization. By the end of season the growth rate and number of vessels decrease, while wood density increases. δ13C values decreased or remained constant while the isotope ratio in assimilates reached a seasonal minimum. This would indicate the time of filling the parenchyma cells with carbohydrates. Similar changes of isotope content connected to anatomy were observed for maple (Leavitt and Long 1991), beech and oak (Helle and Schleser 2004) and for pine (Schulze et al. 2004). Obviously during the early and late part of the year, wood growth seems to be disconnected from carbon assimilation and also wood density is independent of δ13C for a large part of the season. Several studies showed a strong correlation between δ13C of phloem sap and environmental drivers controlling stomatal conductance and thus carbon discrimination (Keitel et al. 2003; Barbour et al. 2005). This suggests that differences found between δ13C of wood and in assimilates occur during the formation of wood and not during transport.

Apparently, during mobilization in spring the isotopically light carbohydrates are in Fagus used by metabolism (respiration) and only the enriched carbohydrates are deposited in wood. This implies that all carbohydrates from storage are broken down to triose-phosphates and that respiration has higher priority over growth. In contrast, in autumn light carbohydrates were deposited in wood and heavy carbohydrates apparently disappeared in storage. Decreasing day-length and lower temperatures likely send the signal to change from wood growth to storage. Although the correlation between wood density and temperature suggests a control by high temperatures, maximum density is reached at the end of the season, a period when temperatures decline. Thus, if wood growth has priority in the autumn carbon cycle, then it is the length of the growing season rather than temperature per se which determines wood density. Surprisingly little is known about the metabolic chain between assimilation, respiration, storage and wood growth.

The correlation between tree rings as well as the change in wood isotopes and assimilates suggests that about 10–20% of a tree-ring in spring is build by remobilization products or additional 10–20% of the tree-ring are affected by storage processes in autumn. This is less than suggested by Gäumann (1935) who proposed that up to 75% of tree-ring growth in Fagus would originate from storage products.

The pattern of 13C content in the concurrent tree rings leads to uncertainty using isotope data averaged for tree rings or measured separately for latewood and earlywood. The average values can be very similar between years despite large differences in climate and photosynthesis (for example, in 2003 and 2002). We could not fully resolve the observed large variation over time in tree-ring width. Mast years with strong beech nut formation has been suggested as one factor causing growth depressions (Gäumann 1935; Hilton and Packam 2003), and late frost events which kill the new leaves in spring may be another. Obviously, the patterns and processes causing the intra-annual changes can differ significantly and they are not directly linked to climate. Thus, trees having abundant storage parenchyma, could be less suitable for climate re-construction.

Conclusions

The results confirm that different parameters of tree rings (width, density and δ13C) respond quite differently to climatic factors and they supplement or compensate each other with respect to the analysis of dendroclimatology. At the studied sites tree-ring width is correlated with the climatic conditions at the beginning of the growing season, maximum density correlates with temperatures (this is possibly a surrogate for growing season length) in the second part of the growing season and δ13C variations are connected to the seasonal dynamics of the water regime. However, the correlations with single climate factors are very weak. Thus, a multi-proxy approach is needed to understand the relation between tree-ring width, density, isotopic composition and climate.

The comparison of seasonal changes in wood density and isotope composition within tree rings provides the possibility to recognize periods of summer drought as indicated by a positive correlation between wood density and isotopic composition. Our dara also show the importance of high-resolution δ13C measurements to provide adequate interpretation of the climate effects.

Despite this correlative evidence, we still have little understanding of carbon discriminating processes between CO2-assimilation, storage and wood formation. Ten to 20% of the concurrent tree-ring in Fagus seems to be affected by the C-cycle of the tree in the previous year and 10–20% of the current year will affect the next year. Only an understanding of the metabolic chain and priorities between assimilation, respiration, storage and wood formation can help to improve the existing models of carbon isotopes fractionation and its relation to climate change.

Forest management may over-ride the effects of climate on isotope ratios in wood.

Acknowledgments

This work was supported by Alexander von Humbold (Research Award 2003 for E. Vaganov), Russian Ministry of Science and Education Funds (Scientific School 2108.2003.4) and Russian Foundation of Basic Research (RFBR 05-04-48069). We thank Dr Giorgio Matteuci for providing the climate data and crown dimensions for Collelongo

Copyright information

© Springer-Verlag 2006