δ¹³C signature of tree rings and radial increment of Fagus sylvatica trees as dependent on tree neighborhood and climate

Abstract

We conducted dendroecological analyses in 80-year-long tree ring chronologies to detect neighborhood effects (competition intensity, species identity) on the δ¹³C signature of tree rings and radial stem increment of Fagus sylvatica trees growing either in monospecific or mixed patches of a temperate forest. We hypothesized that tree ring δ¹³C is a more sensitive indicator of neighborhood effects and the impact of climate variability on growth than is ring width. We found a closer correlation of summer precipitation to δ¹³C than to ring width. While the ring width showed a decline over the test period (1926–2005), the mean curve of δ¹³C increased until the mid of the 1970s, remained high until about 1990, and markedly decreased thereafter. Possible explanations related to ontogeny and environmental change (‘age effect’ due to canopy closure; elevated atmospheric SO₂ concentrations in the 1960s–1980s) are discussed. Beech target trees surrounded by many allospecific trees had a significantly lower mean δ¹³C in the period 1926–1975 than beech with predominantly or exclusively conspecific neighborhood, possibly indicating a more favorable water supply of beech in diverse stands. Contrary to expectation, trees subject to more intense competition by neighboring trees (measured by Hegyi’s competition index) had lower δ¹³C values in their tree rings, which is thought to reflect denser canopies being linked to increased shading. We conclude that tree ring δ¹³C time series represent combined archives of climate variability, stand history and neighborhood effects on tree physiology and growth that may add valuable information to that obtained from conventional tree ring analysis.

Introduction

One important issue in the biodiversity–ecosystem functioning debate is the dependence of ecosystem stability on diversity (Odum 1953; Loreau et al. 2002; DeClerck et al. 2006). Frequently discussed stability parameters of ecosystems are the resistance to, and the resilience after, disturbances such as drought events or herbivore attack. Most of the relevant research on the relationship between diversity and stability has been conducted in herbaceous plant communities while woody associations have been studied only exceptionally. It is generally accepted that mixed forests show greater resilience with regard to herbivore attack than monospecific stands (Jactel et al. 2005; Pretzsch 2005). However, the relationship between tree species diversity and the resistance to, or the resilience after, drought events in forests is not clear yet (Larsen 1995; DeClerck et al. 2006). This question is of high relevance to forestry because natural forests are widely being replaced by monospecific plantations in temperate and in tropical regions while the consequences for ecosystem functioning and stability are poorly known.

Common reactions of trees to water limitation are reductions in height and diameter growth, which can last for several years or even decades (Peterken and Mountford 1996; Archaux and Wolters 2006; Bréda et al. 2006). Drought effects in forests can be enhanced by intraspecific or interspecific competition for water (Gouveia and Freitas 2008) which may be reflected in the chronology of annual tree rings (Saurer et al. 2008). Another archive of environmental changes is the tree-ring δ¹³C signature. It can be used as a proxy for stomatal conductance and thus as a tool for obtaining a long-term record of changes in soil moisture and/or the evaporative demand of trees. δ¹³C values of tree rings have been reported to show drought signals more precisely than tree-ring width does (Andreu et al. 2008). In a similar manner as tree rings, δ¹³C time series do not represent pure physical archives but may also reflect biological processes such as competition for light or water in the forest stand.

The intensity of interspecific or intraspecific competition in forests is often approximated by indices of stand density such has Hegyi’s competition index which is based on stem distance and diameter (Orwig and Abrams 1997; Piutti and Cescatti 1997; Gouveia and Freitas 2008). Like other measures of competition intensity, this index does, however, not take into account that species often differ in their competitive abilities. For most of the investigated mixed forest stands, interspecific competition between different tree species has been reported to be asymmetric (Yoshida and Kamitani 2000; Canham et al. 2004, 2006). Niche complementarity can reduce the intensity of interspecific competition in comparison with intraspecific competition (Kelty 2006). As a consequence, interspecific competition can also lead to positive effects on the growth and water status of one or more partners of the interaction.

Aboveground competition may also result in changes of the canopy structure and the light regime, thereby affecting the δ¹³C signature of leaf mass (Medina et al. 1991; Buchmann et al. 1997; Hanba et al. 1997; West et al. 2001). Further, competition could affect the water availability for the competing species in a mixed stand. Both mechanisms can have consequences for tree growth and the δ¹³C signature in the annual rings. Thus, long-term records of these growth and water status proxies can provide valuable insight into a tree’s long-term water regime and possibly also into competition-induced changes of the water balance (McNulty and Swank 1995; Buchmann et al. 1997; Skomarkova et al. 2006; Grams et al. 2007; Saurer et al. 2008).

For a time period of 80 years, we analyzed the radial increment and the δ¹³C signature of tree rings of selected Fagus sylvatica trees. These trees were carefully selected for their specific neighborhood constellations and competition intensity in monospecific and mixed patches of a species-rich temperate deciduous forest. Our study tests two hypotheses, (1) the δ¹³C signature in tree rings is influenced by the competition intensity and the species identity of a tree’s neighborhood, and (2) tree ring δ¹³C signatures are more sensitive indicators of neighborhood effects and climate variation than tree ring series are.

Methods

Study site

Dendrochronological and dendrochemical investigations were conducted in 16 mature Fagus sylvatica L. (European beech) trees in the temperate broad-leaved forests of Hainich National Park (western Thuringia, Central Germany) close to the village of Weberstedt (51°05′28″N, 10°31′24″E) at about 350 m elevation. Besides the Galio–Fagetum and the Hordelymo–Fagetum associations, i.e., beech forests on slightly acidic to basic soils, the Stellario–Carpinetum community, a broad-leaved mixed forest rich in hornbeam, linden and ash (Mölder et al. 2008, 2009), is abundant in the study region. The most common tree species are F. sylvatica, Fraxinus excelsior L. (European ash) and Tilia cordata Mill. (little-leaved linden), whereas T. platyphyllos Scop. (large-leaved linden), Carpinus betulus L. (European hornbeam) and Acer pseudoplatanus L. (sycamore maple) are admixed at lower densities.

The trees were chosen at a maximum distance to each other of 4.9 km on eutrophic loess-derived soils with a profile depth of about 60 cm, situated in level or gently sloping terrain on limestone (Triassic Upper Muschelkalk). The soil type of the study sites is (stagnic) Luvisol according to the World Reference Base for Soil Resources (FAO/ISRIC/ISSS 1998). Since the forest exists for at least 200 years, it represents ancient woodland in the definition of Wulf (2003). During the past 40 years, only single stems have been extracted. On the study sites, the last extractions of stems were conducted between 1991 and 1998 (E. Kinne, pers. communication). All trees were selected in stand sections with a closed canopy and a more or less homogeneous stand structure. The recent investigation is part of the Hainich Tree Diversity Matrix Study, which analyzes the functional role of tree diversity in a temperate mixed forest (Leuschner et al. 2009). We conducted soil chemical and physical surveys on all prospective study sites prior to tree selection in order to guarantee sufficient site comparability with respect to edaphic conditions (see Guckland et al. 2009). The study area is characterized by an annual mean temperature of 7.5°C and about 590 mm of precipitation per year (1973–2004, Deutscher Wetterdienst Offenbach, Germany).

Tree selection and neighborhood characterization

For investigating radial increment and the δ¹³C signature in annual rings of beech in its dependence on variable stem neighborhoods, we selected 16 trees (Table 1) from a pool of 152 adult Fagus trees, which had been analyzed for tree ring chronologies in a precedent study (Mölder 2010). The tree selection was based on predefined criteria of the neighborhood constellation, suitable biometric characteristics of the tree and the quality of the extracted core. All target Fagus trees were part of the upper canopy, had a diameter at breast height (dbh) of 40–60 cm, and the crown area was at least 30 m² large. Another criterion was that the ring series could be successfully cross-dated to stand chronologies and no questionable tree rings occurred, the time period between sample extraction, ring measurement and sample drying did not last longer than one or two days, and the samples were free from signs of injury or infection. This selection procedure reduced the sample size to 16 target trees to be considered. Subsequently, we grouped the target trees according to the importance of Fagus and non-Fagus trees in the neighborhood (Fagus100 group, all neighbors being Fagus; Fagus70-99 group, 70-99% of the competition index (CI) value being contributed by Fagus–Fagus interactions; Fagus<70 group, less than 70% of the CI value being due to Fagus–Fagus interaction but more than 30% being due to allospecific interactions). Allospecific neighbors belonged to the genera Tilia, Fraxinus, Quercus and Acer. The three neighborhood groups contained four Fagus100, five Fagus70-99 and seven Fagus<70 target trees. The 4–7 trees were treated as replicates in the analysis. Even though we ended up with a rather small number of suitable trees in each group, we preferred to apply these strict selection criteria to obtain beech trees with a well-defined neighborhood and to conduct the analysis with rather homogeneous data sets in terms of neighborhood structure, instead of including further target trees with somewhat different neighborhoods which would have increased the data heterogeneity. We accepted that the smallest sample size (n = 4) was realized in the group with exclusively intraspecific neighborhood (Fagus100) because these tree clusters were more homogeneous than the Fagus70-99 and Fagus<70 groups with a variable species identity of the neighbors, and thus, a more heterogeneous structure of the neighborhood.

Table 1 Characterization of Fagus target trees and their immediate vicinity in the three neighborhood categories from purely intraspecific neighborhoods (Fagus100) to neighborhoods with a contribution of Fagus of less than 70% (Fagus<70)

In the direct neighborhood of the target trees, we recorded the species identity, dbh, height and relative position (i.e., distance and angle between neighbor and target tree) of those trees >7 cm dbh whose crowns had direct contact with the beech target tree. The 16 chosen tree groups consisted of 3–5 (in a few cases, up to 8) trees surrounding the beech target tree and covered stand areas of about 100–600 m² in size. In winter 2006/2007, dbh, tree height and species composition were recorded in the tree clusters with the aim to characterize the neighborhood of the beech target trees qualitatively and quantitatively. We also quantified the crown dimensions by 8-point crown projections using a sighting tube equipped with a 45° mirror and cross-hairs to ensure the proper view of canopy elements from the ground (Johansson 1985). For approximating the projected crown area by a polygon, eight points along the edge line of the crown were selected in a manner that approximated the estimated crown area best. In summer 2007, hemispheric photos were taken with a digital camera equipped with a fisheye lens, thus providing information on canopy dimensions, gap fraction and canopy openness in the neighborhood of the central beech tree. To calculate canopy openness, we used the software Gap Light Analyzer 2.0 (Simon Fraser University, British Columbia, Canada; Institute of Ecosystem Studies, New York, USA) and restricted the canopy perspective to an opening angle of 30° from the zenith which is in agreement with the protocol for analyzing tree competition in forests applied by Pretzsch (1995). We calculated the coefficient of variation (CV) of tree height in the tree clusters in order to provide a measure of canopy heterogeneity. To estimate the intensity of competition in the tree clusters, we calculated CI after Hegyi (1974) for all those trees in the neighborhood of the target beech tree that were present with a part of their crown in the “influence sphere” of this tree, i.e., a cone with an angle of 60° turned upside down with the apex being positioned at 60% of the target tree’s height. The more trees being present in this cone and the smaller the distance to the target tree, the higher is the competition index:

$$ {\text{CI}}_{i} = \sum\limits_{j = 1}^{n} {{\frac{{{{d_{j} } \mathord{\left/ {\vphantom {{d_{j} } {d_{i} }}} \right. \kern-\nulldelimiterspace} {d_{i} }}}}{{{\text{Dist}}_{ij} }}}} $$

(1)

where d _i is the diameter at breast height of the target tree i (cm); d _j is the diameter at breast height of the competitor j (cm); and Dist _ij is the distance between target tree and competitor (m). Trees with a competition index larger than 0.9 were classified as trees exposed to higher competition intensity (n = 8), target trees with a CI smaller than 0.9 as trees with lower competition intensity (n = 8).

We further expressed the tree diversity of the clusters with the Shannon diversity index H′ (Magurran 2004):

$$ H^{\prime} = - \sum\limits_{i - 1}^{S} {p_{i} \;{\text{In}}\,p_{i} } \quad{\text{with}}\,p_{i} = {\frac{{n_{i} }}{N}}\, $$

(2)

where S is the species richness of the target tree’s neighborhood and p _i is the fraction of trees belonging to species i. The fraction p _i is calculated from the ratio between the number of stems n _i of species i and the total number of neighbors N.

Sample preparation and analysis

In summer 2006, we cored all 16 Fagus target trees at 1.3 m height (5 mm corer) on that side of the trunk that showed lowest influence of wood tension or compression. Since we had to meet the conservation regulations of the Hainich National Park, each tree was cored only once. After recutting the surface of the cores with a razor blade, we used titanium dioxide to enhance the visibility of the tree rings before ring analysis. Annual tree-ring width was measured to the nearest 0.01 mm using a LINTAB-5 dendrochronological measuring table (Rinn Tech, Heidelberg, Germany) and TSAP-Software (TSAP-Win Version 0.59 for Microsoft Windows, Rinn Tech, Heidelberg, Germany). In a pre-analysis, we searched for unrecognizable or questionable rings in the cores in order to reconsider them during cross-dating. As quality criteria, we considered the t value (Baillie and Pilcher 1973; Hollstein 1980), the co-linearity of increment (Gleichläufigkeit, Eckstein and Bauch 1969), and the cross-dating index (Grissino-Mayer and Kaennel Dobbertin 2003). Cross-dating of a chronology is accepted as being reliable, if it reaches a minimum t value of 3.5 (Baillie and Pilcher 1973; Hollstein 1980), a minimum co-linearity of 70% for a 50-year overlap (Eckstein and Bauch 1969; Frech 2006), and a minimum cross-dating-index (CDI) >20 (Müller 2007). The tree age at coring height (1.3 m) was calculated as follows: we took pictures from the core centers and determined the distance between the innermost visible tree ring and the point of intersection of the medullary rays. The distance was then divided by the mean ring width of the ten innermost rings to estimate the number of missed tree rings, which were then added to the number of measured tree rings (Schmidt et al. 2009). After the dendrochronological analysis, the samples were dried at 65°C and cut ring by ring for the period 1926–2005. Both the latewood and earlywood of a ring were included in the samples in order to reduce the variation caused by anatomical properties (Smith and Shortle 1996). The wood of a tree ring was cut into small pieces with a razor blade and 1 mg of a ring was weighed out in tin capsules for determination of the δ¹³C signature. We used samples of 0.4–1 mg of acetanilide as internal standard. The analyses were carried out with a Delta V Advantage isotope ratio mass spectrometer (Thermo Fisher Scientific, Waltham, Massachusetts, USA), which was combined with a Conflo III interface (Thermo Fisher Scientific) and a NA 1500 C/N Elementar Analyzer (Carlo Erba Strumentazione, Milan, Italy). By using the internal standard acetanilide, the ¹³C/¹²C isotope ratios were related to the Peedee belemnite limestone standard using the equation δ¹³C (‰) = ([R _sample/R _standard] − 1) × 1,000, with R = ¹³C/¹²C. Partial stomatal closure may be indicated by an enrichment of ¹³C, i.e., higher (less negative) values of δ¹³C.

Statistical methods

Individual ring-width (w) series were standardized following mainly Andreu et al. (2008). After a Box–Cox transformation of the raw width values (in mm) to stabilize the variance, we detrended the series by fitting a linear regression line. Subsequently, standard chronologies were built with robust means. Furthermore, we removed autocorrelation from the single detrended ring series by using an autoregressive model (w _ac). δ¹³C values were first corrected for long-term changes in the atmospheric ¹³CO₂ signal by addition of the difference between modeled atmospheric δ¹³C and a standard value (δ¹³C_cor). As standard we used the “pre-industrial” atmospheric δ¹³C of −6.4‰ as suggested by McCarroll and Loader (2004). Subsequently, we applied an autoregressive model in order to remove autocorrelation in the δ¹³C_cor time series as was done in the ring-width series (w _ac). In the following, the δ¹³C_cor time series corrected for autocorrelation will be referred to as δ¹³C_ac.

Descriptive statistics on ring-width series were calculated with the package dplR, yielding mean sensitivity according to Eq. 3 (MS_I) and Eq. 4 (MS_II) after Biondi and Qeadan (2008). MS_II takes present trends into account and gives with its absolute value, in a similar way as MS_I, a measure for temporal dissimilarity:

$$ {\text{MS}}_{\text{I}} = {\frac{2}{n - 1}}\sum\limits_{t = 2}^{n} {{\frac{{\left| {w_{t} - w_{t - 1} } \right|}}{{w_{t} + w_{t - 1} }}}} $$

(3)

$$ {\text{MS}}_{\text{II}} = {\frac{n}{n - 1}}{\frac{{\sum\nolimits_{t = 2}^{n} {w_{t} - w_{t - 1} } }}{{\sum\nolimits_{t = 1}^{n} {w_{t} } }}} $$

(4)

with w = width of the tree ring, n = length of the tree-ring series, t = 1, 2,…, n = years in the tree-ring series.

Tests for differences in absolute stem increment and δ¹³C signatures among beech trees of different neighborhood categories were conducted with a non-parametric multiple comparison procedure after Hothorn et al. (2008), implemented for Dunnett contrasts or, for two groups, with a two-sample test for the non-parametric Behrens–Fisher problem (Brunner and Munzel 2000). Differences between individual tree-ring series were tested for significance with Friedman’s non-parametric test.

For these statistical analyses, we used the software R (version 2.10.1, R Development Core Team, 2009) with the following packages and scripts: sarima, dplR, nparcomp and zoo.

Climate (monthly precipitation and temperature) data were derived from the data set CRU TS 2.1 (Mitchell and Jones 2005) for the coordinates 51.25°N and 10.25°E. The sum of monthly totals of precipitation and averages of temperature for the period between January and December were used to build chronologies of whole-year climate data (hereafter referred to as annual values). We calculated a climate index as the quotient of the precipitation total and the mean temperature of the months April–September (Frech 2006). Bootstrapped Pearson correlations (number of bootstrapped iterations = 1,000) of monthly precipitation and temperature were calculated with the program DendroClim2002 (Biondi and Waikul 2004) for the year of tree-ring formation (current year) and the year prior to ring formation (preceding year). In order to avoid the problem of multi-colinearity, which would occur in data sets on meteorological parameters, we also calculated response functions (Fritts 1976). Correlation coefficients and response function coefficients are only indicated if they were significant at p < 0.05.

Results

Beech stem increment and δ¹³C signatures as dependent on climatic parameters

For our study site, neither annual values (precipitation, temperature and climate index) nor values for the growing season (April–September) revealed significant linear trends with the year as independent variable over the 80-year study period. However, we detected a significant linear increase of temperature in the growing season for the period after 1976 (R ²_adj  = 0.48, p < 0.001, y = 0.087x − 160.183; x = Gregorian year). In contrast, precipitation and climate index showed no trend for the period 1976–2005.

We detected both negative (July–September, bootstrapped correlation coefficients r between −0.42 and −0.19) and positive (June and October, r values between 0.21 and 0.40) correlations between ring-width chronologies (w _ac) and monthly mean temperatures during the growing season of the year prior to the reference year (Fig. 1c). Temperature values of the current year showed exclusively a negative correlation with ring width (June and July, r values between −0.33 and −0.21). Response function coefficients for temperatures of the current year were only significant for June–July and were always negative. In contrast to the correlation analysis, the more rigorous response function coefficients were never significant for a given time period for the sampled trees in all three neighborhood groups (Fagus100, Fagus70-99 and Fagus<70).

Fig. 1

Bootstrapped correlations (bars, r) of δ¹³C_ac and radial increment data (w _ac) with monthly precipitation and temperature values of the preceding and current year. In addition, the coefficients of response functions are given as horizontal lines (rf). Only significant correlations and response functions are shown (p < 0.05)

Precipitation in June of the preceding year and ring width (w _ac) were negatively correlated (correlation coefficients between −0.30 and −0.24), while precipitation in the growing season of the current year was positively related to w _ac (r values ranging from 0.20 to 0.33). The response function coefficient was only significant for the conspecific group Fagus100 in June (r = 0.17). Precipitation was a relevant factor for all neighborhoods only in June of the preceding year.

Monthly temperature values in the growing season (June–August) of the year prior to ring formation showed a significant negative correlation with the δ¹³C_ac chronologies (r values between −0.28 and −0.20), while growing season temperature (June–August) of the current year was positively related to the δ¹³C_ac signature (r 0.21–0.29). All three neighborhood groups were similar in showing a relationship of δ¹³C_ac to temperature in July of the current year, while the response function coefficients were not uniformly significant in all three neighborhood groups in a given month. Only July precipitation of the current year was negatively correlated with the δ¹³C_ac signature from all neighborhood groups (r −0.39 to −0.37) and the response function coefficients were significant and negative for all three neighborhoods as well.

Relationship between δ¹³C signals and annual radial increment

Whereas the δ¹³C_cor mean curve of all 16 sampled trees showed a continuous increase until the mid of the 1970s with a steep decline after about 1990, the increment curve generally declined over the 80 investigated years (Fig. 2). If we assume that increment and δ¹³C_cor signals should be negatively correlated, only 6 of the 16 sampled trees showed the assumed relationship for the period 1926–2005 with correlation coefficients between −0.42 and −0.24 (Table 2). Trees from all neighborhood groups equally showed a negative relationship between ring-width chronologies (w _ac) and δ¹³C_ac series. Neither tree height, variability in tree height in the investigated tree cluster, or crown area, nor CI showed significant effects on the direction of the correlation between w _ac and δ¹³C_ac chronology (data not shown). In contrast, trees, where δ¹³C_ac and annual increment were negatively correlated, were significantly younger (p = 0.037, two-sided) than trees not showing this negative relationship (Table 2, trees with a significant negative relationship in any of the analyzed periods versus trees with no significant or a positive relationship). Moreover, younger trees generally revealed a closer and older trees a less tight δ¹³C_ac–w _ac relation. Further, the δ¹³C_ac-ring width relationship was different between the 1926–1975 and 1975–2005 periods for most of the sampled trees with only four trees showing a significant relationship between the two variables in the 1975–2005 period.

Fig. 2

Chronologies of δ¹³C_cor and radial increment w for all 16 Fagus trees (smoothed with a 5-year running mean). In addition, annual precipitation totals (bars) and mean annual temperatures are given

Table 2 Pearson correlation coefficients (r) and p values for relationships between radial increment and δ¹³C_ac

Beech stem increment and δ¹³C signals influenced by competition intensity and neighborhood diversity

Comparison of w and δ¹³C_cor in the two allospecific neighborhood categories (Fagus70-99 and Fagus<70) with the purely conspecific group (Fagus100) as control (Table 3) revealed no significant differences for the 80-year period from 1926 to 2005 (w, p = 0.394; δ¹³C_cor, p = 0.138; one-sided). Further, radial increment w and δ¹³C_cor were not significantly different between the three categories in the study periods before and after 1975 (w: 1926–1975, p = 0.408; 1976–2005, p = 0.170; δ¹³C_cor: 1926–1975, p = 0.113; 1976–2005, p = 0.478), even though a tendency for a difference in particular between the Fagus100 and the Fagus<70 group was visible (Figs. 3, 4).

Table 3 Descriptive statistics for the 1926–2005 chronologies of raw ring-width series (w) and δ¹³C_cor tree-ring chronologies of Fagus trees in three different neighborhoods (only conspecific neighbors: Fagus100, few allospecific neighbors, many Fagus neighbors: Fagus70-99, many allospecific and also Fagus neighbors: Fagus<70)

Fig. 3

δ¹³C_cor chronologies of beech (period 1926–2005) corrected for changes in atmospheric δ¹³C (lines) in three different neighborhood categories (only conspecific neighbors: Fagus100, few allospecific neighbor and many Fagus neighbors: Fagus70-99, many allospecific and also Fagus neighbors: Fagus<70) and corresponding precipitation in July (bars)

Fig. 4

Ring-width chronologies (w) in three different neighborhood categories (only conspecific neighbors: Fagus100, few allospecific neighbors, many Fagus neighbors: Fagus70-99, many allospecific and also Fagus neighbors: Fagus<70) and corresponding precipitation in July (bars)

When pooling the data of the purely conspecific group (Fagus100) with the Fagus70-99 category and comparing the distribution with that of the allospecific group (Fagus<70), a significantly lower δ¹³C_cor value is found for the Fagus<70 than for the Fagus70-100 category in the study period from 1926 to 1975 (p = 0.045, Fig. 5). This difference was not significant for the period 1976–2005 (p = 0.475). The higher radial increment (w) in the Fagus<70 group in 1976–2005 was only significant on a significance level of 10% (p = 0.076), and there was no significant difference for the earlier period 1926–1975 (p = 0.183). Even though the differences in mean annual growth and δ¹³C_cor signature between Fagus trees with contrasting neighborhood diversity were not significant for the period 1976–2005 (Figs.  3, 4, 5), the plotted mean curves of the ring-width chronologies (w) showed different levels of annual increment indicating a smaller increment since approximately 1976 for beech trees with a predominantly conspecific neighborhood (Fig. 4).

Fig. 5

Mean annual radial increment (w) and δ¹³C_cor in annual rings of Fagus trees in the periods 1926–1975 and 1976–2005 in aggregated neighborhood categories (relative contribution of Fagus to competition index (CI) 70-100%: Fagus70-100, n = 9; relative contribution of Fagus to competition index CI < 70%: Fagus<70, n = 7). Different letters indicate significant differences at p < 0.05 (one-sided, non-parametric two-sample test after Hothorn et al. 2008)

Analogous to comparisons between mainly conspecific neighborhoods and allospecific neighborhoods (Fig. 5), we grouped beech trees with regard to the competition intensity in the neighborhood according to Hegyi’s index CI (Figs. 6, 7), irrespective of con- or allo-specific interactions, and investigated the mean δ¹³C_cor values and mean radial increment for the two periods 1926–1975 and 1976–2005. Differences between δ¹³C_cor values of the two competition intensity groups were significant for the period 1976–2005 (p = 0.014, one-sided) and existed as a tendency for the earlier period 1926–1975 as well (p = 0.09). Radial increment (w) in the period 1926–1975 tended to be higher in the beech trees being subject to a higher competition intensity, but this difference was not significant (p = 0.176; 1976–2005: p = 0.284).

Fig. 6

Mean annual radial increment (w) and δ¹³C_cor in annual rings of Fagus trees in the periods 1926–1975 and 1976–2005 in two categories of competition intensity. Different letters indicate significant differences at p < 0.05 (one-sided, non-parametric two-sample test after Hothorn et al. 2008)

Fig. 7

Chronologies of radial increment w (a) and δ¹³C (b) in two different classes of competition intensity in the neighborhood of the target trees (n = 8). ‘Lower competition intensity’ stands for Hegyi competition indices <0.9, ‘higher competition’ for CI values >0.9. Standard errors of the means associated with the period 1926–1975 and 1976–2005 are shown for the lower competition group (solid lines) and the higher competition group (dotted lines)

Effect of age on temporal changes in radial increment and δ¹³C

Cambium age was negatively correlated with radial increment w (r ² = 0.67, p < 0.001, y = −1.62x + 376.61) regardless of tree age and competition intensity in the neighborhood. The correlation of cambial age with the δ¹³C_cor signature was significant and positive when the entire period (1926–2005) was considered (r ² = 0.43, p < 0.001, y = 0.007x − 25.18).

Discussion

Correlation of climate parameters with tree ring δ¹³C signatures and radial increment

Often, δ¹³C_cor and ring-width data (w) reveal a negative correlation because wet conditions in summer lead to better growth (wider tree rings) and a stronger discrimination against ¹³C (lower δ¹³C_cor values) under the condition of higher leaf conductance. However, in our sample the two parameters frequently did not show this relationship. Fagus trees lacking a correlation between δ¹³C and ring width tended to be older than those showing a significant negative correlation and were exposed to a relatively high competition intensity (CI, 1.12–1.65). One explanation of this mismatch is that ring width and the δ¹³C signature of tree rings at least partly depend on different environmental factors. While both parameters were influenced by precipitation and temperature in our study, the months with a significant effect were different between the two signals (see Fig. 1). In addition, other factors than climate may also affect the δ¹³C signature including stand thinning (McDowell et al. 2003; Skomarkova et al. 2006) and changes in atmospheric δ¹³C (McCarroll and Loader 2004). Skomarkova et al. (2006) explain the partial mismatch between δ¹³C signal, ring width and climate with the remobilization of carbohydrates stored in the earlier growing season. In their study, only the mid-season δ¹³C value of wood tissue was related to the season’s actual climate and associated climatic constraints on the assimilation rate. Beech wood isotope ratios matched modeled isotope ratios in the assimilates only in the mid-part of the growing season while wood growth was found to be disconnected from carbon assimilation during the early and late part of the year (Skomarkova et al. 2006). Contrasting carbon allocation patterns between younger and older trees may also be an explanation for our observation that a significant ring width-δ¹³C_ac relation existed only in younger beeches. When trees reach maturity, additional C sinks (such as masting events) compete with wood growth, and existing C sinks (as the root system) may operate more independently from climate signals as the organs reach a larger size than early in life, thus decoupling δ¹³C and w to a certain degree.

Only for the δ¹³C_ac signal, all three neighborhood categories showed a uniform response to current year precipitation (in particular that of July). This influence was less clear for ring width (w _ac) which seems to support our second hypothesis that the isotope signal is more sensitive to climate variation than is ring width, at least with respect to precipitation. Similarly, Robertson et al. (1997) reported that significant components of a climate signal may be included in δ¹³C values even if ring width is only poorly correlated with climate fluctuation. Other authors also found a closer dependence of the δ¹³C signal on precipitation, compared to the precipitation dependence of ring width, in particular at drought-influenced sites, what matches with the findings in the Hainich stand (Saurer et al. 1995; Gagen et al. 2004; Andreu et al. 2008).

For temperature variation, the situation was different with a closer relation to ring width than to δ¹³C_ac in annual rings. Particularly influential was the temperature of the previous year. For this variable, we found a larger number of significant correlation and response function coefficients than for other temperature parameters. The significance of the temperature of the previous year for current wood growth has been linked to carbohydrates stored in the previous year that support radial growth in the following year (Hoshino et al. 2008; Lo et al. 2010). Carbohydrate storage and associated carry-over effects are likely causes not only for the frequently observed correlation of δ¹³C_ac signals with climate parameters of the previous year, but also for the autocorrelation in tree ring width (Skomarkova et al. 2006; Vaganov et al. 2009).

Since our study sites are characterized by pronounced summer droughts (Frech 2006; Guckland et al. 2009), the strong relationship between summer (July) precipitation and δ¹³C_ac in tree rings makes it likely that the variation in δ¹³C_ac values in the Hainich forest is a reflection of interannual variation in mean stomatal conductance in the growing season. Drought-induced stomatal closure has been found to significantly reduce the canopy carbon gain of temperate broad-leaved trees in drier summers even in the mostly humid climates of Central and Northern Europe (Gagen et al. 2004; Granier et al. 2007).

δ¹³C in tree rings is dependent on competition intensity

Competition can change the intensity of mutual shading among neighboring crowns (Canham et al. 2004) which could influence the δ¹³C signature of tree rings. Indeed, various studies revealed an increase of δ¹³C in tree biomass with increasing irradiance (e.g., Hanba et al. 1997). The structure of tree canopies is influenced by the stem density in the neighborhood and may also depend on the functional traits of neighboring trees (Jack and Long 1991). Therefore, the neighborhood could leave traces in the δ¹³C signature of tree rings. In our study, trees subject to more intense competition by neighboring trees (measured by Hegyi’s CI) had lower δ¹³C_cor values in their rings, on average by about 0.8‰ for CI values between 0.94 and 1.72 versus CI values between 0.58 and 0.90 (Fig. 7b, p = 0.057 for the comparison of intercepts). If competition for water were a key factor, one would expect the opposite, i.e., reduced discrimination of ¹³C due to lowered stomatal conductance. In a water-limited oak forest, Gouveia and Freitas (2008) found stand density-dependent differences in leaf carbon isotope discrimination and were able to define an optimal stand density from comparisons of δ¹³C signatures in stands differing in stem density. They argued that higher tree densities would lead to increased competition for water resources while lower densities were associated with lower water retention in the ecosystem, since trees in this forest type improve the water storage capacity, resulting in the lowest δ¹³C values at moderate tree densities. Thus, a simple positive relationship between competition intensity and δ¹³C is unlikely. Water limitation, as reported for the forests studied by Gouveia and Freitas (2008), is also characteristic of our study site, which is located close to the eastern drought-induced range limit of Fagus in Central Europe (Mölder et al. 2009). We hypothesize that mechanisms leading to reduced drought stress of beech such as self-shading of leaves and shading by other trees (West et al. 2001) may facilitate the existence of Fagus at this site. This effect may mask the expected positive effect of competition intensity on the δ¹³C signature of beech leaf and wood tissue and may result in lowered stomatal limitation. However, it is important to mention that trees exposed to lower competition intensity were on average older than trees with a higher competition index (p = 0.01). Competition intensity and tree age were in fact negatively correlated in our data set, which may have influenced any relationship between CI and δ¹³C. Studies with a more complete control of influencing factors are needed to disentangle this relationship.

Another mechanism, which could contribute to the observed pattern in δ¹³C among trees differing in their exposure to competition intensity, is that leaves within the canopy may assimilate CO₂ released from respiration of lower canopy strata. Respiration leads to discrimination of ¹³C (see Berry et al. 1997), thus reducing the abundance of ¹³CO₂ in the air and resulting in more negative δ¹³C values in the lower canopy where heterotrophic processes dominate. The lower and upper canopies of forests have been found to differ in atmospheric δ¹³C by 1.7–5.5‰ (Sternberg et al. 1989; Knohl et al. 2005). Thus, in denser stands with higher CI, the closer packing of crown elements could lead to an intensified assimilation of ¹³C-depleted CO₂. In spite of carry-over effects due to carbon storage, the signal is likely to be manifested in the corresponding tree rings.

Competition with conspecific versus competition with allospecific neighbor trees

In predominantly allospecific neighborhoods competition intensity appeared to be higher than in conspecific ones (Table 1, differences not significant). In order to separate diversity effects from competition intensity effects, we pooled all trees that were exposed to mainly intraspecific competition (Fagus70-100). This led to a harmonization of competition intensity in the neighborhood categories to be compared (average CI, Fagus 100 = 0.72; Fagus70-99 = 1.23; Fagus<70 = 0.93; Fagus70-100 = 1.00). Since competition intensity was not significantly different between the Fagus<70 and Fagus70-100 categories, the specific properties of the neighbors (tree identity) or tree diversity must have been influential and not competition intensity itself.

For the period 1926–1975, the δ¹³C values of beeches from primarily conspecific neighborhoods were found to be higher than corresponding values of trees in the neighborhood of allospecific competitors. Since intraspecific competition for water between beech trees is likely to be an important factor, allospecific neighbors may facilitate better growth of beech indirectly through reduced water consumption if the neighbors use water more conservatively than beech (Köcher et al. 2009). In fact, xylem flux measurements in the stem and measurement of leaf conductance in the dominant tree species of the Hainich mixed forest revealed that beech coexists with species that generally use less water than Fagus when soil moisture content is moderate to high (Hölscher et al. 2005). A higher water availability in more diverse stands should be associated with higher stem increment rates of beech. This, however, was only observed as a tendency in the more recent period 1976–2005 but not in the 1926–1976 period. Moreover, the δ¹³C values of the recent period do not show a significant effect of neighborhood diversity. Thus, another factor than neighborhood patterns must have influenced δ¹³C additionally that was active only for the past 30 years.

Long-term trends in δ¹³C and stem increment

A conspicuous result is the characteristic optimum curve of δ¹³C_cor values since 1926. The 80-year δ¹³C_cor record revealed a period of elevated values between 1965 and 1990, when the signature was about 1‰ higher than before and after this period. The ascending curve might be explained by the ‘age effect’ when δ¹³C_cor increases in ageing trees with increasing irradiance due to crowns reaching higher canopy strata (Francey and Farquhar 1982; Saurer et al. 1997). Another cause underlying the curve pattern could be changes in atmospheric chemistry over the past decades. We speculate that elevated SO₂ concentrations in the atmosphere of Central Europe from the 1960s to the late 1980s may have resulted in partial stomatal closure of sensitive trees, thus decreasing ¹³C discrimination during photosynthesis (Savard et al. 2004; McNulty and Swank 1995; Sakata et al. 2001). The drop in δ¹³C from 1990 to 1995 coincides with a sharp decrease in SO₂ concentrations due to strict emission reduction measures in the European Community and especially in the former German Democratic Republic (Thüringer Landesanstalt für Umwelt und Geologie 2002) in that time. Another reason could be the development of denser canopies in the past two decades after selective cutting ceased. A denser canopy is typically linked to reduced δ¹³C values due to lower radiation penetration (Francey and Farquhar 1982) and less stomatal limitation (Farquhar et al. 1989).

In a period when many Central European forests on poorly buffered soils suffered from acid rain, annual stem increment in the beeches of our stand showed a lasting reduction from about 1985 onwards, and δ¹³C_cor decreased as well. For the Hainich forests on limestone, however, anthropogenic acidification is not a likely cause of this change in growth patterns. We rather suggest that forest management practices may have had a substantial influence on both radial growth patterns and the δ¹³C signal. With the cessation of selective cutting in the 1990s (E. Kinne, pers. communication), competition between beech and the other species has most likely increased. Beech trees growing in a species-rich neighborhood (Fagus<70 group) had been favored in the past by an extensive selection cutting regime. Now, these trees face a more closed and darker stand which simultaneously reduces radial growth and leads to lower δ¹³C values (see also Duquesnay et al. 1998). In turn, selective cutting in earlier decades may have increased soil moisture availability due to reduced stand water use (Sucoff and Hong 1974; McDowell et al. 2003) which might have caused reduced δ¹³C signatures in the period before 1975.

Another reason for the simultaneous decrease in ring width and δ¹³C since 1990 could have been elevated summer temperatures since the 1990s accompanied by an increasing number of masting events in beech (Schmidt 2006). Frequent masting depletes the carbohydrate resources and leads to reduced radial growth as does the more frequent occurrence of extreme summer heat waves as they happened in 2003 (Granier et al. 2007).

Finally, a possible explanation is also offered by the age effect with a decelerating height and diameter growth in maturing forests. Since competition intensity and tree age were negatively correlated in our data set, age could have confounded an assumed effect of CI on ring width and δ¹³C.

We conclude that chronologies of width and δ¹³C in annual rings of beech can be significantly influenced by the structure of the tree’s neighborhood. However, this signal is typically weaker than the influence of climate variation, tree age and effects of forest management. Current neighborhood constellations are only a snapshot of the community structure, which change substantially over the lifetime of a tree. Nevertheless, for parts of the chronology we obtained evidence of lowered δ¹³C signatures (significant) or elevated radial growth (non-significant tendency) in beeches growing in mainly allospecific neighborhoods as compared to trees in a predominantly conspecific neighborhood. This indicates, at least for certain periods, that beech grew better in a mixed than in a monospecific neighborhood in this broad-leaved forest. Our study is probably the first to investigate dendrochronological climate archives in their dependence on the simultaneous action of climate, tree age, neighborhood structure, and management effects in a species-rich mixed forest. Further studies in other mixed forest types are needed to substantiate the effects of neighborhood identity and diversity on tree growth using larger sample sizes and by proving the effect to be independent from species and forest management. We suggest that the synchronous analysis of ring width and δ¹³C chronologies for target trees with contrasting neighborhoods may be a promising tool for improving our understanding of the mechanisms of negative and positive interactions between trees in mixed stands.