The dataset is based on a 37 years time series including 179176 annual observations for 9826 individual beech trees in 102 plots across Switzerland. The plots cover gradients in drought, nitrogen deposition, ozone, age, altitude, and soil chemistry (Braun et al., 2017b). Trees have been assessed annually for crown transparency and discoloration according to ICP Forests (2016). Trees with a crown transparency of >60% are hereafter called “strongly damaged” trees. The number of dead trees is related to the population of the previous year, respectively.

Shoots for nutrient analysis have been collected in the years 1984, 1987, 1991 and thereafter every 4 years by helicopter from the upper sun canopy (nine harvest events in total). Each harvest consisted of eight trees per plot. Leaves were dried and analysed for N, P, and K as described in Braun et al. (2020). Photobleaching was assessed by selecting randomly 25 leaves from each harvested branch. Leaves showing signs of photobleaching, e.g., not uniformly green, were counted. For regression analysis, the number of unbleached and the number of bleached leaves per tree was used.

During the harvest 2019 which followed the dry summer 2018 an analysis for actual cavitation was performed in 145 branches sampled in the sun canopy in 10 plots between July 22 and August 3 in 2019. The samples were wrapped into black plastic bags immediately after sampling and brought into the lab within the same day. Branch pieces of 15 cm were recut under water. Reversible embolies were removed by rinsing them under a pressure of 0.1 Pa (Hacke and Jansen, 2009) with a freshly membrane filtered (0.45 μm) solution of 0.1% safranine (Harvey and van den Driessche, 1997) in water during 30 min. The stained branches were kept frozen until image analysis of the stained areas with ImageJ¹. The stained area was related to the total possible cross sectional area without the core.

Meteorological data were interpolated for each plot from the nearest eight monitoring stations of the Federal Office of Meteorology and Climatology (MeteoSwiss) as described by Braun et al. (2017b). The resulting daily averages were used in the regression analysis or as input to the hydrological model Wasim-ETH (Schulla, 2019). With this model the following drought indicators were calculated (i) ratio between actual and potential evapotranspiration (ETa/ETp), (ii) site water balance (SWB) which is the cumulated difference between precipitation and potential evapotranspiration at a daily basis, with added water storage capacity of the soil (available water capacity, AWC). The cumulation was limited to +AWC. The cumulation started on January 1st. The lowest value reached during summer was used in the data analysis. Temperature, precipitation, and drought indicators for the harvest year were averaged over the time between start of the growing season and the harvest date. Start and end of the season were taken from phenological data of beech budbreak and discoloration observed by MeteoSwiss and adjusted for altitude as described in Braun et al. (2017b). The season length thus differs between plots.

All statistical analyses were performed using R 4.0.3 (RCore Team, 2020). Foliar concentrations were log transformed when the transformation improved the model. Staining results and discoloration were analysed using linear, mortality, numbers of strongly damaged trees per plot and photobleaching using binomial mixed regression models (Bates et al., 2015), with plots and years and – in the case of photobleaching also trees – as random factors, following the data analysis strategy by Braun et al. (2017a). In brief, backward selection of predictor variables using the Akaike Information Criterion (AIC) was applied. Estimated explained variance (Pseudo-R²) of the fixed effects only and of fixed and random effects were calculated according to Nakagawa and Schielzeth (2013). Natural cubic splines (one or two knots) were used to model non-linear functional relationships. All models including R scripts and diagnostic plots are provided in the Appendix. Residuals were checked for normality, homoscedasticity and outliers. Effect plots with estimated marginal means and the corresponding 95% confidence intervals taking into account all fixed and random effects were calculated according to Lüdecke (2018).

In 2019 and 2020, the proportion of trees with a crown transparency of >60% increased by a factor of 6.1 and 6.7 compared to the long term average before 2019 (Figure 1). This increase was mainly due to an increase in dead branches. The mortality was also increased but less strongly: by a factor of 5.0 (2019) and 3.9 (2020), respectively (Figure 2). In the years 2018, 2019, and 2020 death occurred more suddenly as shown by the crown transparency one year before dieback (Figure 3).

Both mortality and the proportion of strongly damaged trees were increased when foliar concentrations of P and K were low (Table 1 and Figures 4, 5). The relation of mortality with P was non-linear. For both parameters, drought was also an important predictor. The best drought indicator explaining mortality and strong crown transparency was the minimum site water balance of the previous year, two and three years before, for mortality also the precipitation in spring.

Discoloration was another phenomenon which clearly increased in 2019 and 2020 (Figure 6). In addition, the phenomenon rated as photobleaching at the harvested shoots was greatly increased in 2015 and 2019 (Figure 7). For both parameters foliar P concentrations played a role. The relation with P suggested that photobleaching may be regarded as P deficiency symptom. It was analysed in a separate spline regression for showing the non-linear relations (Figure 8 left). The other predictors differed for photobleaching and discoloration. For discoloration, observed in the field together with crown transparency, the two climatic variables drought and irradiation and the fructification from the previous year improved the model (Table 2). For photobleaching the maximum July temperature and the irradiation were relevant predictors (Table 3). There was also a highly significant interaction between P concentration and maximum July temperature which is illustrated in Figure 8 right: when foliar P concentrations were low (e.g., 0.6 mg g^–1), the effect of the maximum July temperature on photobleaching was stronger than when P concentrations were sufficient (e.g., 1.2 mg g^–1).

Actual cavitation quantified as vessels stained with safranine (Figure 9) was related with drought in the previous year and with foliar N:K ratio in the current year (Figure 10). Neither drought of the actual year nor other foliar nutrients were related to current year cavitation.

The observations on parasites naturally remain episodic. But there is a number of observations on known weakness parasites. Neonectria coccinea was observed frequently on dying beech trees, both the typical dark dots as well as fruit bodies. In a heavily drought affected forest district a number of uprooted beech trees were observed. Detailed examination revealed a number of primary root rot fungi which seem to be promoted by the drought: Armillaria sp., Pholiota aurivella, Meripilus giganteus.

The extent of the drought induced crown dieback observed in the long-term Intercantonal Forest Observation Program in Switzerland after the severe summer drought in 2018 was out of previous experience. There have been records on drought damages of beech after the years 1947 and 1951 (Schindler, 1951; Wagenhoff and Wagenhoff, 1975). Both studies describe, however, an increased incidence of beech bark disease, a phenomenon, which has been common also in later years. In the study sites this has been associated with increased beech mortality in the 1980s. It seems that the extensive crown dieback due to hydraulic failure is a symptom which has not been known up to now in Central Europe. The drought damages observed after 2018 were much more severe than after the summer 2003 with a similar heat wave. The main differences between 2003 and 2018 were a dry spring in 2018 and a series of dry years preceding 2018 (2015 and 2017). 2003 followed a series of rather wet years, and the drought only started in June.

The increased numbers of beech trees with heavy crown defoliation and tree mortality in 2019 and 2020 provide evidence of a breakdown of the hydraulic system for parts or the entire tree crown. Beech is not known as species where catastrophic hydraulic failure is common (Leuschner, 2020). However, the actual cavitation measurements confirm the hypothesis (i) of hydraulic failure as one of the explaining factors for the observed drought damages in beech trees. The observations are in agreement with the results of Schuldt et al. (2020) who made similar staining experiments in the drought experimental plot in Hölstein, Switzerland. Branch cavitation seems to be the main reason for the serious branch dieback. Similar results were obtained by Walthert et al. (2021) who measured actual percent loss of conductivity in branch samples. The partial hydraulic failure of the branches may later lead to complete hydraulic failure of the whole tree.

The refilling of cavitated vessels in spring is based on two processes (Cochard et al., 2001): production of new vessels, which is common in all plants with secondary cambium and active refilling, which is more species specific. This latter process has been shown to occur in Fagus sylvatica during the time of leaf expansion and sap flow (Cochard et al., 2001). Our data show that this refilling process was disturbed following the severe summer drought in 2018. This was indicated by the correlation of the proportion of active vessels with the drought of the previous year. Furthermore, the proportion of active vessels was also correlated with K nutrition. The importance of N and K nutrition for cavitation vulnerability has been shown already by Harvey and van den Driessche (1999) and more recently by Oddo et al. (2014). In the present study, low K concentrations were related to higher proportions of dead branches and higher tree mortality which may have a similar physiological background with increased cavitation.

The increased observed occurrence of photobleaching can be interpreted as a result of energy shortage, leading to strong discoloration in the presence of phosphorus deficiency and at high temperature levels. Photooxidation of chlorophyll by radiation is a normal concomitant process of photosynthesis requiring energy for reduction. When this energy is lacking the chlorophyll is oxidized (Powles, 1984). Therefore, photobleaching may be interpreted as a sign of “starvation”. However, as Körner (2015) suggested, plant growth is not carbon limited and in the current case the carbon starvation hypothesis should be replaced by a phosphorus limitation hypothesis which is also supported by the mortality increase at low foliar P concentrations. The strong increase in time is the result of high temperatures in 2015 and 2019, but also an effect of the interaction between P deficiency and high temperatures in July. The foliar P concentrations have decreased over the last decades, probably due to high N deposition (Braun et al., 2020). While in arable crops acute P deficiency symptoms are well known as yellowing, for tree species there are only a few descriptions of visible symptoms only for the seedling stage (Bergmann, 1992). We suggest that the observed photobleaching is regarded as a P deficiency symptom, as it starts to increase when foliar P concentrations fall below 40% of the threshold for normal nutrition.

It is well known that weakened trees are less resistant against parasites. Several field observations from a number of plots from the long-term Intercantonal Forest Observation Program of strongly damaged trees support this hypothesis (iii). As these events are based on a small number of trees, no modeling could be done. However, the time delay observed between drought and mortality suggests that weakness parasites play an important role in the drought response of beech.

Before the severe drought in 2018, the higher crown transparency before dieback suggests that mortality due to weakness parasites played a more important role in Swiss beech. With the persisting cavitation damages since 2018 it appears to be difficult to separate between hydraulic failure effects and parasite infestation in living beech trees. Therefore, we argue that an overall assessment of the 2018 drought response of European Beech needs to take into account the drought damages due to hydraulic failure [hypothesis (i)], but also effects such as the lack of energy [hypothesis (ii)] as well as increased occurrence of weakness parasites [hypothesis (iii)].