Introduction

Tree growth depends on a multitude of influencing abiotic and biotic factors (Matyssek et al. 2013a). Crucial abiotic resources are CO2, water, minerals and light along with temperature (Kozlowski and Pallardy 1996). Conversely, adverse stress, such as that caused by air pollutants or drought, can lower tree growth. Outcomes from multi-factorial impacts on forest tree growth are subject to ongoing debate (Matyssek et al. 2013b).

Past and projected future increase in air temperature is the key undisputed feature of global climate change poised to most impact tree growth. Since the end of the 19th century, the decadal average temperature over land has increased by 1.3 ± 0.11 °C (Kovats et al. 2014). Further increase of average temperatures relative to the 2005 temperature by 0.3–4.8 °C until 2100 is currently under discussion, with projections of larger increases over land compared to oceans (IPCC 2013). Another important driver of tree growth, affected by global climate change, is the water supply of forests (Thomas 2000). Since 1950, precipitation has increased in northern Europe, but decreased in southern Europe. Concurrently, drought episodes have increased across Europe both in frequency (Kovats et al. 2014) and duration (Ciais et al. 2005). In addition, tropospheric background ozone levels have increased during recent decades due to human activity related to car traffic, industrial combustion processes and forest burning (Kovats et al. 2014). As ozone formation is favored under high irradiance and temperature (AQEG 2007), empirical evidence and modeling also predict tropospheric ozone peak levels to further increase in the future (IPCC 2013). Projections of future ozone regimes are linked to different scenarios. If available emission control technologies become uniformly and globally effective immediately, decreases in annual mean surface ozone levels by 3–14 nl l−1 could occur by 2100. However, under the assumption of low efficacy of emission control in combination with rising methane abundance, increases by 5–14 nl l−1 across trans-continental scales could result (Kirtman and Coauthors 2013). Even assuming unchanged emissions, climate change per se is likely to increase summer tropospheric ozone levels by 1–10 nl l−1 until the 2050s, thereby increasing the risk of productivity losses in forest trees (Wittig et al. 2009).

European beech (Fagus sylvatica L.) is the potentially dominating tree species at sub-mountainous elevation across central Europe (Ellenberg 1996). After centuries of extensive cultivation of lucrative coniferous species, beech has advanced recently to a major component in forest policy and management for restructuring coniferous monocultures into mixed-species forests to reduce ecological and economical risks in view of climate change (Baumgarten and von Teuffel 2005; Ammer et al. 2008; Knoke et al. 2008). Despite high competitiveness (Ellenberg 1996) and acclimation capacity (Kramer et al. 2010), European beech is known to be sensitive to drought (Ellenberg 1996), as well as to ozone impact (Smidt et al. 1991; Matyssek and Sandermann 2003; Geßler et al. 2004; Matyssek et al. 2010a).

Most studies to date have investigated the effects of drought and ozone on the basis of year-ring analyses due to the high availability and efficacy of data collection (Dittmar et al. 2003; Nunn et al. 2007; van der Maaten 2012). Drought episodes, as predominantly occurring in Central Europe, are mostly restricted in intensity and duration, so that usually growth reductions are not found in year-ring analysis. This does not exclude, however, minor intra-annual short-term drought effects on the radial stem growth of trees. Alternatively, short-term drought effects might incur slight but reversible reductions of radial growth, which would only be detectable by sensors with fine time resolution, such as automatic dendrometers (Herzog et al. 1995; Bouriaud et al. 2005; Deslauriers et al. 2007).

Causal effect-based evidence concerning ozone uptake and tree performance at forest sites is largely lacking (Matyssek et al. 2012a, 2013a; de Vries et al. 2014). Long-term studies have typically correlated ozone doses accumulated annually with the annual stem increment (Nunn et al. 2007; Matyssek et al. 2010a). In the present study, the immediate effects of peak level ozone uptake (Panek and Goldstein 2001; Gerosa et al. 2003) on intra-annual tree diameter growth were of primary interest. Hence, the present association analysis of both risk factors, drought and ozone, with particular focus on interaction effects (Löw et al. 2006), on the diameter growth of beech was conducted at daily time resolutions. In particular, analyses were conducted to address the following three hypotheses concerning beech in central Europe.

  1. 1.

    Short-term drought episodes that do not affect the yearly diameter increment have minor inhibiting impact on radial growth at the daily scale.

  2. 2.

    Maximum daily ozone flux has the capacity for limiting the relative radial growth of adult beech on the daily basis.

  3. 3.

    Drought and ozone interact in their influence on relative radial growth, with drought stress dominating by decreasing the ozone uptake.

Materials and methods

The posed hypotheses required measurements at high time resolution. Radial stem growth, meteorological data and xylem sap flow for the estimation of ozone flux were collected hourly throughout the three study years.

Study sites

The study was performed at seven beech forest sites belonging to the Bavarian forest ecosystem monitoring network, Level II (Bad Brückenau (BBR), Freising (FRE), Kreuth (KRE), Mitterfels (MIT), Riedenburg (RIE), Riedenburg 1 (RIE1), Würzburg (WÜR)), and at the research sites Kranzberg Forest (KRB, Pretzsch et al. 1998; Matyssek et al. 2010b) and Forellenbach (FOR, Baumgarten et al. 2000; Beudert and Breit 2004; Beudert et al. 2007; Tables 1, 2). Sites were selected along a transect, following gradients of altitude (330–1100 m above sea level, Table 1), air temperature (yearly mean 6–9 °C), precipitation (yearly sum 650–1500 mm, Table 3) and ozone exposure (SUM0: 103–193 nl l−1, Table 4).

Table 1 Site characteristics, including soil texture, available water-holding capacity, base saturation, leaf nitrogen concentration, potential natural vegetation and site coordinates
Full size table
Table 2 Tree species, percentage, age, rooting depth and basal area of F. sylvatica for the total stand, diameter at breast height (dbh), year-ring widths of study trees (with standard deviation), and years with available sap flow measurements at the respective sites
Full size table
Table 3 Total precipitation (mm) and means of temperature (daylight-air temperature, °C), humidity (daylight-relative air humidity, %) and global radiation (daylight-global radiation, W m−2) during growing seasons (May–October) of 2003–2012 and for specific years studied at individual sites
Full size table
Table 4 O3 exposure SUM0 (nl [O3] l−1·h) during the growing season (May–October), portion of days with peak O3 concentrations above 40, 60 and 80 nl [O3] l−1·h, respectively, within the growing season (%) and stomatal O3 fluxes POD0 (mmol m−2·h) and POD1 (mmol m−2·h) for the entire tree foliage (sun- and shaded leaves) at canopy height, whole growing season, for respective study years and study sites; the critical level of 4 mmol m−2·h to protect deciduous and evergreen trees, as given by LRTAP Convention (2010), is related to sun-exposed foliage and therefore cannot be applied to the data
Full size table

Making use of comparable soil types (brown earth/parabrown earth) along the transect, the available water-holding capacity in the rooting horizon (AWC) varied distinctly between the sites (106–539 mm). Generally, the sites exhibited sufficient nutrient supply (P, K, Mg, Table 1), except for Mitterfels and Forellenbach, which had rather low Mg levels (BMELF 1997). The total foliar N concentration in beech leaves ranged between 21.9 and 27.2 mg g−1 (Table 1). Four sites were stocked with pure beech, another five sites carried mixtures with other tree species (Table 2). Beech populations at the sites generally consisted of adult, even-aged trees of 86–136 years. Rooting depths varied from 0.5 m (RIE1) to 1.5 m (FRE). Nevertheless, the clayey substrates in BBR, WÜR and RIE were susceptible to dehydration. Study trees were standardized by selecting mature dominant individuals within the canopy with breast height diameters (dbh = 1.3 m above ground) above the 50 % stand-level percentile to minimize competition effects (Dittmar et al. 2003). Measurements were conducted on five beech trees per site during the growing seasons of 2010, 2011 and 2012 (Table 2).

Database

For all sites, databases containing hourly measurements of the relevant meteorological parameters (air temperature, relative air humidity, global radiation, precipitation), phenological monitoring, and characterizations of soil and nutrient supply were provided (by the Bavarian State Institute of Forestry for Level II sites, ICP Forest, 2006, and UN/ECE Monitoring in the Bavarian Forest National Park for Forellenbach, Beudert and Breit 2004). At KRB, air temperature and relative air humidity were measured with Sensor RFT-2 by UMS, München, Germany, and global radiation was measured with Pyranometer GSM 10.7 by Thies CLIMA, Göttingen, Germany. Hourly ozone concentrations above canopy and air pressure were continuously measured at FOR (Beudert and Breit 2004) and KRB (with 202 Ozone Monitor by 2B Technologies, Boulder, USA; air pressure with BARATRON type 222 B by MKS Instruments, Burlington, USA). According to Baumgarten et al. (2009), ozone data from FOR were used for MIT, and those from KRB for FRE due to proximity and comparable altitude. For the Level II sites, ozone concentrations and air pressure were adopted from nearby open-field meteorological stations operated by the Bavarian Environmental Agency (LfU, Neustadt a.D. for RIE and RIE1, Würzburg for WÜR), Deutscher Wetterdienst (DWD, Hohenpeißenberg for KRE) or the Hessian Agency for the Environment and Geology (HLUG, Wasserkuppe for BBR). Data were corrected for crown height. For use in the association models, daylight means of air temperature, relative air humidity, global radiation and air pressure were calculated.

Relative growth

Tree diameter changes at dbh were continuously measured with automatic band dendrometers (type DC2 by Ecomatik, Germany) and logged every 10 min (DT80, Datataker). Water saturation of the trees and therefore daily maximum diameters were typically reached during nights. For the calculation of daily growth, maximum diameters of preceding days were subtracted from those of current days. Days with water-related physical shrinking and swelling of stem diameters were excluded, because of imprecision of daily growth calculation. Ensuring comparability across trees, relative daily growth rates (RG) were calculated as the percent increase of the cross-sectional stem area at breast height during current days versus the total cross-sectional stem area at the beginning of the growing season.

Ozone flux

The starting point for deriving ozone fluxes was whole-tree transpiration as obtained through xylem sap flow measurements using the thermal dissipation probe method according to Granier (1985, 1987). Sensors (type SFS2 M by UP, Germany) were installed at dbh. To avoid interference by irradiance, sap flow sensors were mounted at north-exposed trunk sides and covered with light-reflecting aluminum foil. Sensors were installed at 0–2 cm of xylem depth. To assess the radial sap flow profile in the trunk, additional sensors were installed at two trunk depths (2–4, 4–6 cm) in two of the study trees per site. Sap flow density at 0–2 cm was set as 100 %, so that proportional flow density was continuously calculated at 2–4 and 4–6 cm of radial stem depth and adjusted to the stem allometries of the other study trees at the respective site. Sap flow of xylem parts deeper than 6 cm in the trunk was assumed to linearly approach zero toward the center of the tree. Calculations yielded flow rates approximately equal to zero beyond 10 cm of the radial stem depth. Sap flow was set to zero every night to avoid sensor-related artifacts in sap flow assessment. During rainy nights with substantial offsets between previous and current-day sap flow rate, zero correction was omitted to allow recording of potential nighttime sap flow.

For calculating ozone uptake, the sap flow-derived crown transpiration rate of trees was related to the respective projected foliage area (Matyssek et al. 2009), yielding unit foliage area basis (Ec). Canopy conductance (Gc) was calculated from Ec versus leaf–air mole fraction difference of water vapor (Δw) as

$${\text{Gc}} = \frac{\text{Ec}}{\Delta w}$$
(1)
$$\Delta w = \frac{\text{VPD}}{\text{p}},$$
(2)

where Δw represents vapor pressure deficit (VPD) as based on air pressure (p), presuming leaf temperature to equal air temperature (Tang et al. 2006) as an approximation at the crown level (in beech approximately 88 % of foliage area is formed by shade- or sunshade intermediate leaves, Hagemeier 2002).

Time lags between the onset of transpiration in the canopy and xylem sap flow at breast height were compensated for by mathematically synchronizing both time courses through matching at the instant of sunrise, when canopy transpiration became incited (Matyssek et al. 2004, 2008). The onset of canopy transpiration was deduced from the onset of global radiation in combination with Δw > 0. The synchronization, enforced by tree-internal water storage capacitances (Schulze et al. 1985), warranted relating diurnal ozone levels to corresponding Gc:

$${\text{FO}}_{ 3} = {\text{Gc}} \times {\text{O}}_{3} \times 0.663.$$
(3)

Ozone flux through the stomata (FO3, Eq. 3) was calculated from the ambient ozone concentration above canopy (O3), multiplied by Gc and 0.663 (representing the ratio in the molecular diffusivities of ozone versus water vapor, Grünhage et al. (2012), LRTAP Convention (2010)).

Daily ozone impact was assessed using three approaches

  • POD0 max, the hourly phytotoxic ozone dose without considering a threshold (=maximum hourly FO3 per day), and by two forms of the phytotoxic ozone dose including a threshold

  • POD1 max (=maximum hourly FO3 per day above 1 nmol m−2 pla s−1) and

  • POD1 day (daily sum of hourly FO3 above 1 nmol m−2 pla s−1).

Plant-available water (PAW) and soil water potential (SWP)

Parameters were calculated with the water budget model BROOK90 LWF, an extension of the model BROOK90 hydrologic model (Hammel and Kennel 2001; Federer 1995; Federer et al. 1996), which yields averages per site and respective rooting zones. The extended model was validated against measured soil water contents in different soil horizons (Trase System, SOILMOISTURE EQUIPMENT CORP) at the sites WÜR, RIE and MIT (Wagner et al. 2010; Weis et al. 2011). The parameterization according to Raspe (2012) was modified for beech stands across all study sites (W. Weis, K. May, personal communication; for more details see Baumgarten et al. 2014). To ensure comparability across sites, PAW was expressed as proportion of AWC (100 % level, see Table 1).

Drought intensity was assessed using two approaches

Drought 1 [%thr] = deviation of actual PAW only below threshold PAW at daily basis (percent difference), and drought 2 [MPathr] = deviation of actual SWP only below threshold SWP at daily basis (pressure difference). For the calculation of drought 1, at PAW above the threshold of 60 % of AWC, drought intensity was regarded as 0 %thr (percent below threshold). At lower PAW, drought 1 was calculated once a day as the threshold value of 60 % diminished by the actual PAW level. This approach accounted for findings of Klap et al. (2000) and Mund et al. (2010), who reported on growth reductions of beech at PAW <60 %. Drought 2 was calculated on a daily basis from SWP. As on most of the study sites, SWP of about −0.02 MPa was equivalent to PAW of 60 %, this kind of congruence was set as a provisional threshold of drought stress in beech forests. At SWP < −0.02 MPa, drought 2 was calculated as the actual SWP level diminished by −0.02 MPa, whereas larger SWPs were not accounted for in calculating drought intensities. As soil drying mostly proceeds heterogeneously throughout the different layers of the rooting zone (Bréda et al. 2006), the 10 cm layer within the rooting zone that had the most moisture was used for drought stress calculation.

Statistical analysis

The relationship between RG and the parameters drought 1, drought 2, POD0 max, POD1 max, POD1 day, air temperature, relative air humidity, ∆w, global radiation and base saturation was investigated first in an exploratory fashion, employing nonparametric smoothing to examine (i) linearity of relationships, (ii) distinction from the null hypothesis of no relationship, and (iii) whether transformations would enhance model fit. Spearman’s rank-based correlation was computed to assess the association between the multiple parameters.

A linear mixed effects (LME) model was used to estimate the association between individual-tree RG and climate parameters, while accounting for various sources of heterogeneity (function “lme” from R package “nlme”, Pinheiro et al. 2014). Next to the climate parameters above as fixed effects, a fixed slope over time (day of growing season) was considered. In addition, independent normally distributed random effects were included to model variation among different trees, and global changes over the three study years. A random effect to account for variation among the nine study sites was additionally considered but not included since it provided no statistically significant improvement to the model when viewed in addition to the other random effects. A normal autoregressive model of order 1 was used for the within-tree error component of the LME model to account for serial correlation among daily fluctuations in individual-tree RG; lags beyond one day were not statistically significant and hence not included.

Assessed parameters along with their interactions were considered as potential predictors (fixed effects) in the LME model. For enhancing convergence, all potential predictors were scaled by subtracting their mean and dividing by their standard deviation. In addition, ozone fluxes and drought intensities were log-base-2 transformed before scaling to improve model fit. A stepwise model selection procedure using the Bayesian Information Criterion (BIC) was used to find the optimal composition of fixed effects and interactions to include in the model to be reported. All statistical tests were performed at the two sided .05 alpha level of statistical significance and all analyses were conducted in the R freeware statistical package (R Core Team 2013).

Results

Meteorology and O3 exposure during growing seasons

Meteorological parameters indicated increased precipitation in the growing season of 2010 relative to the 10-year mean in all study regions (Table 3). Therefore, typically water-limited sites were only moderately affected by drought in 2010, except for a six-week drought period in June and July. In 2010, annual precipitation was increased by 25 and 11 % at the three most water-limited sites WÜR and RIE/RIE1, respectively (data not shown). During the corresponding growing seasons (May–October) increases amounted to 32 and 15 %, respectively. In 2011, the precipitation at most sites was similar to the 10-year mean, except for WÜR, which experienced reduction in precipitation by 21 % during the growing season. In 2012, precipitation was reduced compared to the 10-year mean across all study sites by up to 21 % (RIE and RIE1, Table 3). Yearly diameter growth, however, was consistent with the long-term mean, showing no reductions during the less humid years 2010 and 2012, compared to 2011 (Table 2).

Sums of hourly O3 concentration throughout growing seasons (SUM0) positively correlated with altitude while complying with 10-year means, except for varying reductions at FRE and KRB in 2010 (by 17 % each) and BBR in 2011 (by 47 %; Table 4). Hourly O3 concentrations exceeded 40 nl l−1 for 43–78 % of the days during the growing seasons across the different sites. Correspondingly, exceedances of 60 nl l−1 occurred during 13–30 % of the days, whereas 80 nl l−1 was reached only during 1–4 % of days during the growing season (Table 4). Assessments of stomatal O3 flux were available only during the three-year study period. Although SUM0 was high at mountainous sites, POD0 and POD1 did not necessarily correlate with altitude (Table 4). At BBR and FOR, these latter O3 doses ranged above the median across study sites as a consequence of high SUM0. At the sites FRE and WÜR, the phytotoxic ozone doses POD0 and POD1 ranged above the median, calculated over all nine sites, as given in Table 4).

Stem growth, meteorology, soil water and stomatal O3 influx at the daily scale

The calculated drought stress indices drought 1 and drought 2, giving respective drought intensities per day and site, are exemplified at WÜR during the growing season of 2010 in Fig. 1. As seen in the figure, both drought intensities increased proportionally as PAW and SWP dipped below their thresholds of drought stress. In parallel, RG decreased as shown in Fig. 2. As Δw increased during that period due to high air temperature (data not shown), Gc decreased, indicating stomatal closure (Fig. 2). Although O3 concentration was high, stomatal flux, therefore, mostly stayed below the threshold of 1 nmol m−2 s−1 (see POD1 max; Fig. 2). Drought events relevant for calculating drought I and II also occurred at BBR, FOR, RIE, RIE1 and WÜR, with maxima at RIE and RIE1. At other sites, PAW and SWP never dropped below stress thresholds. If drought occurred, Gc was reduced with temporary reductions in RG. One dry six-week period in June and July 2010 resulted in temporary RG reductions across the study sites, before thresholds of PAW and SWP were reached. Short-term reductions in Gc under drought resulted in temporarily decreased O3 influx at WÜR, FOR and BBR. Nevertheless, whole-seasonal POD0 and POD1 reached high levels each at the respective sites. The reason was that periods of non-limiting water supply were also accompanied by high O3 exposure. Conversely, whole-seasonal POD0 and POD1 at RIE and RIE1 remained moderate during all years due to low O3 levels.

Fig. 1
figure 1

Drought intensities, exemplarily shown for Würzburg for the growing season 2010; graph above: solid line plant-available soil water (PAW) as percent of the available water-holding capacity for total rooting area, broken line threshold for drought stress (PAW <60 %), bars intensity of drought stress (drought 1) as included in the LME model; graph below: solid line soil water potential in the wettest soil horizon (SWP), broken line threshold of drought stress (SWP < −0.02 MPa), bars intensity of drought stress (drought 2) as included in the LME model; PAW and SWP modeled with BROOK90 LWF

Full size image
Fig. 2
figure 2

The relationship between relative stem growth (RG), canopy conductance (Gc) and ∆w, as well as the impact of Gc and O3 concentration on POD1 exemplarily for Würzburg 2010; graph above: black line relative daily stem growth, bar drought period based on drought 1; graph middle: black line maximum hourly O3 uptake per day above the threshold of 1 nmol m−2 s−1 (POD1), gray line O3 concentration in the air, bar drought period; graph below: black line canopy conductance (Gc), gray line Δw, bar drought period

Full size image

The summary statistics (Table 5) including the available parameters involved during the model selection process give orientation on all potential predictor variables for RG over the years and all sites. RG was highest in 2010, but lowest in 2011 in consistency with precipitation across study sites (Table 3). The present analysis focused on the time period during the growing season, when RG was nearly linear correlated with time, to exclude phenological influences outside the main growing period. Typically such linear RG prevailed from May through the end of July at most sites, with restricted duration at high altitude. Drought occurred at five sites during the whole study period. As the number of days with drought stress was low, compared to the whole number of days within the study, the median of drought 1 and drought 2 is nil. For RIE1 with the shallowest soil (Table 2), exceptionally high drought intensities were found during each study year, representing the maxima of drought 1 and drought 2 in Table 5. The maximum hourly O3 fluxes per day (POD0 max and POD1 max) resulted in homogeneous median doses throughout the study years. Median O3 doses for 2012 were lower than those during the other years, as calculations of O3 doses in 2012 were only available for the four low-altitude sites. The median of air temperature across sites was similar during 2010 and 2012, but higher than in 2011. Global radiation reflected trends similar to those of air temperature, with maxima in 2010 and 2012. Accordingly, median relative air humidity was highest in 2011 and lowest in 2010 across sites (Table 5).

Table 5 Summary statistics for the predictors included in the association analyses with relative growth
Full size table

The foliar N concentration (Table 1) was ranked as “high” (22–25 mg g−1) or “very high” (>25 mg g−1) according to Wolff and Riek (1997) showing an absence of N limitation at our study sites. Hence, base saturation was considered as the only growth-relevant variable for nutrient supply among the predictor variables (Table 5).

All predictors tested in the model were correlated with each other, particularly the different drought intensities and O3 doses (Table 6). Global radiation, relative air humidity and Δw were also highly positively associated.

Table 6 Associations between all predictors determined by Spearman’s rank-based correlation coefficients
Full size table

Linear mixed model analysis

The optimal LME model contained the following predictors of RG: air temperature, relative air humidity, drought 1, and growth conditions of the preceding study day, as well as interactions between air temperature and relative air humidity, between air temperature and the preceding day, and relative air humidity and the preceding day (Table 7). Δw, global radiation and base saturation were not included due to co-linearity with other stronger predictors and/or lack of statistical significance. Importantly, O3 dose was not included due to lack of statistical significance.

Table 7 Estimated regression coefficients and p values of the final model selected by BIC
Full size table

A plot of RG versus POD1 max gives the illusion that RG negatively correlates with POD1 max (Fig. 3 left). However, this appears as such since POD1 max was measured to be 0 on most days and very near zero on the remaining days; the 90 % quantile of POD1 max is very low so it is only a lack of data points on the right end of the x-scale that contributes to a seemingly drop in RG for high P0D1 max values. The log-base-2 transformation monotonically expands the lower end of the POD1 max scale by stretching out closely spaced low observations while reeling in large values (Fig. 3 right). Even among the highest 10 % of POD1 max levels no negative relation to RG was indicated (Table 7; Fig. 3 right).

Fig. 3
figure 3

Scatterplot of relative growth versus POD1 max on the original scale (left) and on a log-base-2 scale (right). The vertical line indicates the 90 % quantile of POD1 max and the smooth line on the right indicates a nonparametric loess curve fitted to the transformed data. The plot shows that due to low POD1 max values and the small number of large values, the trend with relative growth appears to be decreasing. This trend does not appear on an expansion of the POD1 max values. The right graph also illustrates the large numbers of days with O3 fluxes <1 nmol m−2 pla s−1 (POD1 = 0)

Full size image

In contrast, the negative relationship between RG and drought 1 was demonstrated both with and without log-base-2 transformation (Fig. 4). A doubling of drought 1 resulted in a decrease of RG by 0.000243 % (Table 7), assuming all other parameters to remain fixed. Exemplifying the median RG (Table 5), a doubling of drought 1 (e.g., from 1 to 2 %thr) resulted in a difference of RG, decreasing from 0.012000 to 0.011757 % per day. In proportion to the median, the growth reduction amounted to 2 % at the respective day.

Fig. 4
figure 4

Scatterplot of drought intensity ‘drought 1’ versus relative daily growth as non-transformed data and on a log-base-2 scale. The smooth line is a nonparametric loess curve fitted to the data. The graph shows that the same decreasing trend is observed both at the original and the transformed scale of drought 1. The transformed scale, however, leads to a better fit of the model and is the one chosen for analysis

Full size image

Net effects of the preceding day were positive, i.e., stimulating on growth (Table 7). Days further dating back had no influence on the current RG of beech. The LME model confirmed a significant positive correlation of air temperature and relative air humidity with RG, both individually as well as through interaction (Table 7). The combined effect as visualized in Fig. 5 shows average RG increased with higher air temperatures, but also depended on relative air humidity and the respective day of year. The latter effect resulted from an additional interaction between air temperature and serial correlation among days. High air temperatures increased the positive effect of high relative air humidity levels synergistically, whereas such an effect became reduced during the later course of the year. However, the overall effect of air temperature on RG remained positive over the whole growing season. The overall effect of relative air humidity on RG also became more positive toward late summer.

Fig. 5
figure 5

Relative daily growth depending on humidity (daylight-relative air humidity), temperature (daylight-air temperature) and their interaction; the scale on the right shows the relative growth rate (% d−1), associated with the contours inside the graph; random effects were set to zero and covariates not displayed on the axis, set to their median value

Full size image

Discussion

Hypothesis I, which assumed that short-term drought would limit RG at the daily scale, was confirmed by this study. On the other hand, Hypotheses II and III, which suggested maximum daily ozone fluxes to limit RG and drought in interaction with O3 to limit O3 uptake, were not confirmed by this study. Following, we analyze the outcome of the hypothesis evaluation in detail, but open the discussion with the most influencing factor for RG.

Most prior studies have reported positive growth effects with increasing air temperature (e.g., Fritts 1958; Esper et al. 2002; Sabaté et al. 2002; Solberg et al. 2009), as deciduous trees at high latitudes profit more than warm-adapted evergreen species in tropical and subtropical regions (Way and Oren 2010). Reports have varied concerning the optimum temperature for F. sylvatica. Emberson et al. (2000a) report on 24 °C in relation to stomatal conductance for adult beech. Alternatively, Dreyer et al. (2001) referred to 34.8 and 38.1 °C regarding maximum photosynthetic electron transport rate and maximum carboxylation rate of beech seedlings, respectively, if exposed to 45 % of the ambient global radiation. They reported that a critical temperature at which the basal fluorescence began to increase was 46.3 °C. Fritts (1958) also suggested maximum growth rates of Fagus grandifolia at high daily temperatures of 37.8 °C. Biochemical processes, such as photosynthesis and respiration, speed up with increasing temperature (Way and Oren 2010), although under concurrent drought, stomatal conductance may limit CO2 uptake. This trade-off complicates the assessment of optimum air temperature for stem growth. In the present study, maximum temperatures of 31.3 °C appeared to remain too low for assessing the optimum (cf. Table 4). Mean daytime-air temperatures during the growing seasons ranged between 12.0 and 18.6 °C throughout the three study years in the region. Concordantly, the LME model indicated a statistically significant positive air temperature effect on RG. Consequently, positive effects on the daily growth of European beech may be expected under projected future increases of air temperatures in central Europe (Kovats et al. 2014). A secondary effect of global warming is an extension of growing seasons, which compounds annual growth enhancement of trees (Sabaté et al. 2002; Pretzsch et al. 2014). Also Pretzsch and Ďurský (2002) predict growth stimulation with rising temperatures for European beech in Germany. Their conclusions are in line with Boisvenue and Running (2006), who report positive growth trends for central European forests under climate change, perhaps explaining the current 77 % faster tree volume increment of beech relative to that in 1960 (Pretzsch et al. 2014).

Decreases in growth rates due to high air temperatures have been in particular identified at the southern distribution boundary of Fagus sylvatica (Ciais et al. 2005; Jump et al. 2006) and on shallow soils (van der Maaten 2012). High air temperatures are likely related to increasing frequency (Kovats et al. 2014) and severity of summer droughts (Sabaté et al. 2002; Ciais et al. 2005). High temperatures are also expected to increase both microbial and plant respiration and therefore enhance evaporative demand, leading to water deficiencies (Ciais et al. 2005; Dittmar et al. 2003). Reversely, microbial respiration and hence soil CO2 efflux are dependent on soil water availability, declining with proceeding soil drought. Jump et al. (2006) reported on increasing temperatures in parallel with declining precipitation at high altitudes in Catalonia/Spain, also concluding such a coincidence as a reason for beech decline in the Central Apennines of Italy (Piovesan et al. 2005). Dittmar et al. (2003) found different temperature–growth relations at low and high altitudes across Europe, with positive temperature effects on tree ring growth at high-altitudinal, but negative ones at low-altitudinal sites. Our analysis did not yield statistically significant differences across study sites, but temperature stimulation of beech growth in general. Dittmar et al. (2003) implied water limitation at low elevation, which was not necessarily the case in our study.

This study investigated in particular whether short-term drought at moist sites affected RG, which would provide support to findings of forest decline at sites with low drought risk (Choat et al. 2012). In our study, drought intensities were significantly correlated with RG, irrespective of being expressed on a PAW or SWP basis. Reduction in soil water content reduces stomatal conductance (Granier et al. 2000; Geßler et al. 2004; Choat et al. 2012), and hence tree growth (Thomas 2000). Respective relationships with radial tree growth were reported by Mund et al. (2010) and Rodríguez-Calcerrada et al. (2014) for PAW and by Saxton et al. (1986), Löf (2000) and Geßler et al. (2004) for SWP. We found a somewhat stronger relationship to RG limitation with PAW rather than SWP.

Different levels of drought intensity have been proposed as probable causes for growth reductions in beech. The drought threshold of PAW <60 % AWC, employed in this study, is in accordance with Klap et al. (2000) and Mund et al. (2010). However, the different findings of Granier et al. (1999) and Bouriaud et al. (2005), who predicted growth limitations at PAW <40 %, were covered as well, as calculated drought intensities increased with declining PAW. Viewing SWP of beech, Geßler et al. (2004) argued that even small reductions, such as levels below −0.05 MPa (Fotelli et al. 2003), sufficed to cause response to water depletion. A SWP of −0.02 MPa indicated incipient drought in this study, but was less negative than expected from the literature. Nevertheless, the xylem embolism threshold in beech of −1.9 MPa (Hacke and Sauter 1995) was never reached in the present study, perhaps underlining the sensitivity of Fagus sylvatica to drought (Ellenberg 1996; Geßler et al. 2004). The amount of RG reduction due to drought shown in this study was small, as data availability was focused on incipient and moderate soil drying. We assume different biological and physical processes in trees under moderate drought stress than under severe stress with ongoing embolism. We refrain, therefore, from extrapolating the presented drought–growth interaction to severe drought situations. As the threshold of drought injury was undercut only at a limited number of days during the study period, no reductions of growth were found at the yearly basis. Complete recovery of growing tissues after water limitation therefore is indicated and in accordance with van der Werf et al. (2007) and van der Maaten (2012), who reported on quick growth recovery in beech after moderate drought stress. We expect negative effects of moderate drought to be likely compensated by the currently experienced positive effects of temperature increase. Given further increase in severity and frequency of drought periods, limitations of tree growth will presumably become grave, as already reported for particular sites in the Swabian Alb region of Germany (Geßler et al. 2004; van der Maaten 2012). Hypothesis I, suggesting short-term drought to have slight negative impacts on RG and being detectable at the daily scale, has therefore been confirmed by this study.

The contribution of relative air humidity to RG has been controversially discussed. Fritts (1958) did not find consistent relationships between relative air humidity and daily stem growth. However, the fact, that high VPD can force F. sylvatica to reduce Gc and therefore radial growth under sufficient water supply was specified by Schulze and Hall (1982) and Granier et al. (2000; both cited by Geßler et al. 2004) and Lendzion and Leuschner (2008). The outcome from our model was confirmative, as RG did not only depend on soil water supply, but also on relative air humidity and interaction between air temperature and relative air humidity as drivers of VPD. The question as to why the interaction between air temperature and relative air humidity and not ∆w was chosen in the stepwise variable selection for the final model might be answered by the co-linearity of Δw with air temperature (Table 6). Relative air humidity in contrast was less correlated with air temperature.

O3 is the air pollutant presumed to be the most toxic to plants (Ashmore 2005) by generating oxidative stress decreasing photosynthesis, plant growth and biomass accumulation (Matyssek and Sandermann 2003; Wittmann et al. 2007; Ainsworth et al. 2012). Negative effects on stem diameter growth appear to be greater than on height growth (Matyssek et al. 1992; Martin et al. 2001; Pretzsch et al. 2010). Hence, ozone is discussed to limit the carbon sink strength of forests (Ollinger et al. 1997; Sitch et al. 2007; Wittig et al. 2009; Matyssek et al. 2012b, 2013a). Projections of Wittig et al. (2009) suggest decreases in tree biomass due to ambient O3 by 2050 and 2100 of 11 % and 17 %, respectively. Therefore, positive effects of increasing temperatures in central Europe on tree growth and prolonged growing periods as discussed above might be counteracted by O3. Moreover, increasing temperatures induce rising O3 concentrations (Ainsworth et al. 2012) and prolonged growing periods will likely increase O3 doses taken up by the trees. As we already know from experimental long-term free-air fumigation with twice-ambient O3 levels (Matyssek et al. 2007), diameter stem growth of adult beech was significantly reduced by 43.5 % (Pretzsch et al. 2010).

For present O3 concentrations between 40 and 60 nl l−1, Wittig et al. (2009) suggested decreases in angiosperm biomass production by around 20 %, and for concentrations between 60 and 80 nl l−1 by around 40 % relative to pre-industrial scenarios. In our study, concentrations above 60 nl l−1 were reached during 13–30 % of the days at the different sites (Table 4), indicating substantial O3 impact on the trees. However, cause–effect-related growth dependence needs to be based on stomatal O3 influx (e.g., Matyssek et al. 2007). A simulation study on O3 effects by Ollinger et al. (1997) predicted decreases of net primary production in North American forests by 3–16 % for the years 1987–1992. A study with a high-time-resolution dataset, investigating impacts of high ambient O3 concentrations from 2001 through 2003, showed negative growth effects of 30–50 % on average across 10 North American tree species (Mc Laughlin et al. 2007). Unfortunately, although sap flow data (Granier 1985, 1987) were available, authors did not derive the corresponding O3 fluxes into the trees. To our knowledge, no growth study so far exists, investigating drought and ambient O3 impacts on European beech on the base of continuous diameter and sap flow measurements. Braun et al. (2014) concluded in a POD1-simulation study, based on diameter increment, that all deciduous forests in Switzerland were impaired by 19.5 % growth losses due to O3. However, one has to keep in mind that in this study, tree diameters were only measured once every 4 years (1991–2011), the range of tree ages was quite high (76–186 years) and European beech was the only deciduous tree species investigated.

In the present study, even though negative correlation was indicated between growth rate and O3 flux, no statistically significant relation between RG of beech and maximum daily ambient O3 fluxes was observed, thereby not confirming Hypothesis II. A limitation of this study is that it is observational, with conclusions based on a mixed statistical models with correlated fixed and random effects that provide a measure of association rather than causation. Alternative cause-and-effect boundary analyses as more commonly used in experimental biology could be performed to replicate these findings (Webb 1972; Emberson et al. 2000b), but these hinge on the assumption of a true causality relationship as well as the existence of a boundary line.

O3 dose–response relationships, O3 flux models and critical levels for O3 are typically only related to sun-exposed leaves (LRTAP Convention 2010). O3 fluxes derived from the sap flux approach (Granier 1985, 1987), however, are related to the whole crown, which, for comparison, involves a problem of scaling. We want to emphasize that available critical levels for O3 (LRTAP Convention 2010) are not appropriate for sap-flow-derived O3 fluxes and hence, cannot be related. Critical levels for whole crown fluxes are expected to be smaller by a factor of 3 or 4 compared to the levels for sun-exposed leaves (M. Baumgarten, L. Grünhage, personal communication).

Interaction effects between O3 deposition and drought have been discussed controversially among different studies, though mostly expecting reduced O3 uptake under dry conditions due to stomatal closure and therefore reduced Gc (Löw et al. 2006; Wittig et al. 2007; Ainsworth et al. 2012). Closure typically results from limited soil water availability (Granier et al. 2000), and/or high VPD (Lendzion and Leuschner 2008). Corresponding relationships between POD1, Gc, ∆w, PAW and SWP were found during several drought periods in this study, as shown in Figs. 1 and 2. Furthermore, we observed reductions in yearly POD1 (Table 4) at the typically most water-limited sites RIE and RIE_1 in 2012. Such reductions co-occurred with severely decreased precipitation, reduced relative air humidity (Table 3) and prevailing SUM0 compared to other study years. Wilkinson and Davies (2009) described for a herbaceous plant species that stomatal closure under drought was suppressed due to O3. Also Matyssek and Sandermann (2003) and Wieser et al. (2012) suggested stomatal opening, among other factors as being imposed by soil water availability and O3. However, in our study, no interaction effect between drought and O3 on RG—either positive or negative—was confirmed by the optimal model, so that Hypothesis III was not confirmed. Wittig et al. (2009), who presented a metastudy on 263 single studies, similarly found no conclusive information concerning the interaction between drought and O3.

Conclusion

  • The present study demonstrated that common beech sites in central Europe with average water supply and prevailingly distinctive radial stem increment rates can currently suffer from intermittent growth reductions, driven by intra-seasonal short-term drought periods. Such growth reductions were neither restricted to edaphically water-limited sites, nor occurring during extraordinarily dry years and might be of minor significance within the whole-year stem increment. However, in view of the predicted increase in severity and frequency of drought in central Europe during progressive climate change, the demonstrated effects may gain in importance. As also shown, effects can be counteracted to some extent by temperature increase as may be occurring during warming. We conclude that high air temperature within climatic gradients across the study region at present has the capacity for compensating moderate drought impact on tree growth.

  • Ambient O3 peak fluxes in the present study remained ineffective with respect to RG. We want to point out that this study concentrated on effects, triggered by peak ozone uptake to investigate the risk of rising peak ozone concentrations. We give no statements on cumulative ozone effects over the growing season, which was already shown in an experimental free-air O3 fumigation of adult beech trees in a different study within the same region to cause significant reduction of RG. In view of still increasing O3 concentrations, it cannot be excluded that beech forests might be affected by their increasing impact.

Author contribution statement

Angelika R Kühn: measured data, performed data analyses, literature research, Figs. 1 and 2, and wrote the manuscript. Sonja Grill: performed the linear mixed model analysis, Figs. 3–5, provided statistical support to the first author. Manuela Baumgarten: supervised and contributed to the data analysis, provided advice and background knowledge. Donna P. Ankerst: supervised the linear mixed model analysis, provided comprehensive statistical knowledge, reviewed English language. Rainer Matyssek: designed the study design, acquired funding, had the overall supervision and guidance of the PhD thesis and was involved in writing the manuscript.