Introduction

The frequency and magnitude of extreme weather events such as drought, heat waves, and heavy rain are expected to increase in the future (IPCC 2012). Understanding the impact of extreme events on plant performance is an important research goal in ecology and is increasingly investigated in experimental studies (Gutschick and BassiriRad 2003; Schroter and others 2005; Jentsch and others 2007, 2011; Suttle and others 2007; Knapp and others 2008; Kreyling and others 2008a, 2013; Beierkuhnlein and others 2011; Walter and others 2011). Extreme weather events may cause stronger effects on plants, plant communities, and ecosystems than gradual warming. How powerful a single event can be was demonstrated in the heat wave year 2003 (Schär and others 2004), in which plant gross primary production was reduced by approximately 30% in Europe (Ciais and others 2005). However, so far, research has mostly focused on the impact of single extreme events (van Peer and others 2004; Gallé and others 2007; Kreyling and others 2008a; Noormets and others 2008).

We argue that climate change impact research should be extended to studying recurrent events and interacting pulse pressures, to gain insights on the resilience of plant species, communities, and ecosystems. The duration and magnitude of the first event and the point in time when the next extreme event occurs can have an influence on the respective plant individual and its ability to recover between and after the two events. Walter and others (2011) found a higher percentage of living biomass in the grass Arrhenatherum elatius after two pulsed drought events compared to individuals which were subjected to only one pulsed drought event over one growing season. An improved photoprotection in the double-stressed grasses was indicated by reduced maximum quantum efficiency, which was caused by reductions of maximum fluorescence (Maxwell and Johnson 2000). An ecological stress memory in double-stressed A. elatius seems to exist and enables the plant to acclimate within its lifespan (Walter and others 2011). Accumulating evidence suggests that plants subjected to recurrent extreme events are able to deal better with subsequent extremes by epigenetic changes or accumulation of signaling proteins or transcription factors, for instance (Bruce and others 2007; Boyko and Kovalchuk 2011). Meisner and others (2013) detected legacy effects in soil biota induced by drought in an inoculation experiment with soil previously exposed to drought and then planted with native and exotic plant species. In particular, this legacy of drought in the soil biota influenced the exotic plant species positively and the natives negatively, whereas the opposite was found for controls, that is, when the soil was not treated with drought before planting. Meisner and others (2013) suggest that this legacy effect might be associated with alterations in the soil biota and their soil processes. Soil previously exposed to drought showed a higher inorganic nitrogen availability compared to soil without drought immediately before planting.

However, positive effects are not guaranteed. The efficacy of the stress memory effect may depend on the plant species, life forms, and threshold effects due to the magnitude and frequency of extreme weather events. Zavalloni and others (2008), for instance, did not find an increased drought resistance in grasslands, which experienced mild short droughts and warming over 3 years followed by a prolonged drought. Moreover, hints exist that perennial species such as trees may not positively “remember” a previous drought period when exposed to another drought several years later. Quercus ilex, for example, showed a reduction in resprouting and survival after a second drought event in 1995 compared to the first drought in 1985 (Lloret and others 2004). For Pinus edulis, a higher mortality rate was found after a recurrent drought event in 2002 in comparison to a preceding drought in 1996 (Mueller and others 2005).

A plant’s response to extreme events can further depend on their neighboring species (van Peer and others 2004; Wang and others 2007; Saccone and others 2009; van Ruijven and Berendse 2010; Kreyling and others 2011; Otieno and others 2012; Arfin Khan and others 2014; Grant and others unpublished data). For instance, the dwarf shrub V. myrtillus revealed a stronger reduction in biomass production due to drought when growing together with another dwarf shrub and two grasses than when growing only with the other dwarf shrub (Kreyling and others 2008a). Moreover, shifts in flower phenology due to drought were found for Calluna vulgaris in a plant community with another dwarf shrub and two grasses compared to a plant community with only another dwarf shrub (Jentsch and others 2009). Novoplansky and Goldberg (2001) detected a lower survival time in Scleropogon brevifolius plants under drought conditions if Sporobolus airoides plants were located in its’ neighborhood. On the other hand, S. airoides survival time was independent from plant neighbors. These examples indicate an influence of the plant neighborhood or community composition on drought response. Therefore, an effect of community composition on plant species’ ecological stress memory seems conceivable.

The aim of our study was to examine the role of preceding drought occurrences for the response of plants in different plant community compositions (temperate grassland and heath communities of varying species and functional group number) and monocultures to a very severe drought event. All plants had previously been involved in a 6-year long-term field experiment, which included annually recurrent experimental and natural drought periods (EVENT-I in Bayreuth, Germany; Jentsch and others 2007, 2011).

We hypothesized that (i) a 6-year long pre-exposure to different drought occurrences influences the stress resistance of plant communities and species during a very severe drought event, with better performance of plants formerly subjected to drought. Furthermore, we expected that (ii) this ecological stress memory effect of single species is altered by plant community composition.

Methods

Experimental Site

The EVENT-I experiment (Jentsch and others 2007) was established in the Ecological Botanical Garden of the University of Bayreuth, Germany (49°55′19″N, 11°34′55″E, 365 m a.s.l.) in 2005. The long-term mean annual temperature at the site is 8.2°C and the long-term mean annual precipitation is 724 mm (1971–2000) with a precipitation peak in December/January and June/July (data: German Weather Service). The previously homogenized and drained soil consisted of loamy sand (82% sand, 13% silt, 5% clay) with a pH 4.5 in the upper (0–20 cm) and pH 6.2 in the lower (20–80 cm) soil layer (measured in 1 M KCl).

Experimental Design of Pre-exposure Manipulations

From 2005 to 2010, the experiment was carried out in a two-factorial design: (1) precipitation manipulations (“ambient control,” “drought,” “heavy rain,” and “regular watering”) (Table 1) and (2) plant community composition (grassland and heath in different community compositions) (Table 2).

Table 1 Overview of Pre-exposures (2005–2010)
Full size table
Table 2 Experimental Heath Plant Communities, Grassland Plant Communities, and Monocultures in the EVENT-I Experiment (Jentsch and others 2007)
Full size table

Intensity of “drought” and “heavy rain” was based on the local 100-year and 1000-year extreme event in each category. The 100-year (for 2005–2007) and 1000-year extreme events (for 2008–2010) were calculated based on the precipitation data of the growing season (April to September) from 1961 to 2000 (data: German Weather Service) using Gumbel I distributions (Gumbel 1958). Further specifics of the climatic manipulations are described in Table 1 and Jentsch and others 2011. The plots of the “drought” manipulation were covered with rain-out shelters (steel frames: Hochtunnel, E & R Stolte GmbH, Germany) that permitted nearly 90% penetration of photosynthetically active radiation (transparent plastic sheets: 0.2 mm polyethylene, SPR5, Hermann Meyer KG, Germany) during the simulated drought events. The “heavy rain” and “ambient control” plots were exposed to the ambient weather conditions of Bayreuth without rain-out shelters. In 2005, the “regular watering” plots were treated identically as the “ambient control.” In 2006, the “regular watering” plots were covered during the same time as the “drought” exposure and received the rainfall amount of the “ambient control” to reveal roof artifacts, resulting in no significant roof artifact effects (Kreyling and others 2008b). In 2007, the regularly watered plots were covered during the whole growing season with rain-out shelters like the ones used for the drought exposure. During this period, these plots received the weekly long-term precipitation sum with one part irrigated at the beginning of the week and the second 3–4 days later ensuring continuous water supply. “Regular watering” started in 2008 (without using rain-out shelters). The “regular watering” plots received at least the long-term (1971–2000) precipitation sum per week during the growing season. If natural rainfall was less than the long-term average sum for the same week, the missing amount was added by irrigation. If weekly rainfall exceeded the long-term sum, it was not subtracted from the next irrigation.

The weekly precipitation sums for the pre-exposures and the volumetric soil water content (vol%) in the years 2005–2010 are provided in Appendix Figure S1. Natural drought periods in the four different pre-exposures during the growing seasons 2008–2010 are provided in Table 3 (continuous soil moisture data for the growing seasons 2005–2007 are not available).

Table 3 Number of Days and the Maximum Number of Consecutive Drought Days with Volumetric Soil Water Content Less or Equal to the Permanent Wilting Point (7 vol%) During the Growing Seasons 2008–2010 for the Pre-exposures “Ambient Control,” “Drought,” “Heavy Rain,” and “Regular Watering”
Full size table

Plant community compositions were established in combinations with an increasing number of plant functional groups (Table 2). For heath, there were combinations of two species of one functional group (H2−) and four species of two functional groups (H4−). For grassland, four species of two (G4−) or three functional groups (G4+) were combined. Additionally, there were monocultures (M−) for “ambient control” and “drought” of key species of heath (C. vulgaris and V. myrtillus) and grassland (A. elatius and Holcus lanatus). The total setup consisted of five replicates of each factorial combination, 75 plots of two by two meters in size and 10 plots of two by two meters in size divided into four small plots for the four monoculture target species (each monoculture 1 m2). The factors were applied in a split-plot design with the different community compositions blocked and randomly assigned within each weather manipulation (Jentsch and others 2007). The originally planted species composition of the year 2005 was maintained by periodical weeding.

Very Severe Drought Manipulation

In the year 2011, we conducted a very severe drought event exceeding projected climate change scenarios, which lasted for 57% (104 days) of the growing season (April to September) for all pre-exposures. We covered the whole experimental site with a steel frame (Haygrove Tunnels Ltd., Ledbury, UK) and a transparent polyethylene sheet (0.18 mm, UV M 42, folitec Agrarfolien-Vertriebs GmbH, Westerburg, Germany; total area 50 m length × 31.5 m width; 3.75 m height). The sheet edges of the rain-out shelters were at a height of 1.05 m on the long sides and 2.55 m on the front/back side. The polyethylene sheet permitted nearly 90% penetration of photosynthetically active radiation. Prior to starting the drought manipulation, the plants received a watering treatment of 46.6 mm divided into two applications (11th/13th of May 2011) to adjust all plants to the same initial condition. The amount of the watering treatment (46.6 mm) was calculated out of the difference of long-term average precipitation sum to natural rainfall sum from April 2011 (the start of the growing season to the start of the very severe drought).

The drought manipulation took place from the 17th of May until the 28th of August 2011 (104 days). Volumetric soil water content (vol%) was measured weekly in each plot in a depth of 10 cm (Field Scout TDR Soil Moisture Meter, Spectrum Technologies Inc., Plainfield) over the course of the experiment (Figure 1). The volumetric soil water content of all pre-exposed plots dropped below the permanent wilting point (7 vol%) on the 26th of May 2011 and stayed there for the rest of the drought manipulation.

Figure 1
figure 1

Volumetric soil water content (vol%) (weekly mean) over the course of the very severe drought experiment in the year 2011. The permanent wilting point (pF = 4.2 = 7 vol%) is indicated by the dotted black line.

Full size image

Response Parameters

Aboveground Biomass, Tissue Die-Back, and the Ratio of Reproductive to Total Biomass

We harvested the aboveground biomass in the heath and grassland plant communities biweekly in different subplots for each single plot (harvest 1 and 2: 17th/30th of May, harvest 3 and 4: 14th/27th of June, harvest 5 and 6: 11th/25th of July, harvest 7 and 8: 8th/22th of August 2011) and three times in the monocultures (harvest 2: 30th of May, harvest 4: 27th of June, harvest 7: 8th of August 2011) during the 104 days of drought. During each harvest, plant material in a subplot of 20 cm × 40 cm within each plot was cut at the surface of the soil. In the heath plant communities (H4− and H2−), subplots were required to encompass one individual of C. vulgaris and V. myrtillus, respectively, in the 20 cm × 40 cm frame. Subsequently, the plant material (entire aboveground biomass) was sorted into three groups per species: vegetative, reproductive (seeds and flowers), and dead parts of the plant. The plant parts would be defined as vegetative/reproductive (alive), if the plant material or parts of the plant material were still green. Dead parts of the plant were defined as brown and wilted that lost its chlorophyll. The woody part of the dwarf shrubs was also separated into dead and alive based on visual evaluation of the transport system. All plant material was dried at 60°C for 72 h before weighing.

Tissue die-back (TD) was calculated as

$$ {\text{TD}} = \frac{{{\text{dead}}\,{\text{biomass}}\,({\text{g}})}}{{{\text{total}}\,{\text{biomass}}\,({\text{g}})}}\,*\,100. $$
(1)

The ratio of reproductive to total biomass (RR) was calculated as

$$ {\text{RR}} = \frac{{{\text{reproductive}}\,{\text{biomass}}\,({\text{g}})}}{{{\text{total}}\,{\text{biomass}}\,({\text{g}})}}. $$
(2)

Belowground Biomass and Root to Shoot Ratio

To estimate the belowground biomass, the roots were collected with a root core (4 cm diameter and 14 cm length; corresponding to 0.00126 m2 surface area). The main rooting zone was within the upper 20 cm. Hardly any roots penetrated the soil deeper than 20 cm. Root core samples were taken from harvest 2–8 in all four plant communities. Due to sampling problems, some data sets are missing, in particular the second and eighth harvest of the heath plant community H2− as well as the second harvest of the heath plant community H4−. Three root core samples (corresponding to 0.00378 m2 surface area in total) per subplot were combined to one mixed sample at each sampling date after harvesting the 20 cm × 40 cm (corresponding to 0.08 m2 surface area) aboveground biomass sample. In the heath plant communities, the root core was positioned 2 cm beside the root beginnings of a C. vulgaris individual, 2 cm beside the root beginnings of a V. myrtillus individual, and one root core in the middle between both individuals. In the grassland plant communities, three root cores were randomly positioned. After harvesting, the root samples were washed, dried at 60°C for 72 h, and the remaining stones/sand were removed before weighing.

The root to shoot ratio (RSR) was calculated as

$$ {\text{RSR}} = \frac{{{\text{belowground}}\,{\text{biomass}}\, ( {\text{g}})/0.00378\,{\text{m}}^{2} }}{{{\text{aboveground}}\,{\text{biomass}}\,({\text{g}})/0.08\,{\text{m}}^{2} }}. $$
(3)

Aboveground biomass represents the sum of the vegetative, reproductive, and dead parts of plant.

Statistical Analysis

Analysis of variance (repeated measure ANOVA) combined with linear mixed effect models were applied to test for the main effect of the factor “pre-exposure” (levels: “ambient control,” “drought,” “heavy rain,” and “regular watering”) for aboveground biomass (per plant community and per single species), tissue die-back (per plant community and per single species), reproductive biomass (seeds and flowers; only for the species analyses), ratio of reproductive to total biomass (only for the species analyses), belowground biomass (only for the plant community analyses), and root to shoot ratio (only for the plant community analyses). We analyzed the first harvest separately from the other seven harvests to test for pre-exposure effects before the beginning of the very severe drought (that is, simple lag effects, not stress memory effects). No harvest one was available for the monocultures as well as for the response parameters belowground biomass and root to shoot ratio of the plant communities. Sometimes the sample weights of the response parameters reproductive biomass and ratio of reproductive to total biomass of the plant species within the entire harvest one were zero and thus no analysis could be performed. The “harvest date,” the “plot number,” and the “repetition number” were included as random factors in the mixed effect models from harvest 2–8, thus taking the split-plot design and the repeated measurements into account. The “harvest date” was included as a random factor in the mixed models, because of no significant interaction of this factor with the factor “pre-exposure” in a pre-analysis. The mixed effect models of harvest one included the random factor “repetition number.” The pre-exposure “regular watering” does not exist in the plant community with a legume species (G4+) and in the monoculture (M−), and “heavy rain” is also missing in the monoculture (M−). Each experimental plant community, each monoculture, and each species were separately analyzed to achieve full factorial combinations for all analyzed subsets. The A. elatius monoculture had to be taken out of the species analysis because of insufficient plant replicates in the pre-exposures. If necessary, the data were square root-, log-, power-, or (asin(sqrt(y)/100))-transformed to improve the normality of residuals and the homogeneity of variances prior to analysis (Faraway 2006). In case of a significant “pre-exposure” effect of the linear mixed effect model, post hoc comparisons with the Tukey’s test were performed according to Hothorn and others (2008). The level of significance was set to P < 0.05. All statistical analyses were conducted with the software R 2.13.1 (R Development Core Team 2011) and the additional packages nlme (Pinheiro and others 2012) and multcomp (Hothorn and others 2008).

Results

Lag Effects of the Pre-exposures Before the Beginning of the Very Severe Drought

No significant pre-exposure effects were found in any of the eight different community cases (four communities, two response parameters) before the beginning of the very severe drought:

Tissue Die-Back of Plant Communities

Regarding the first harvest, no pre-exposure effect was found in tissue die-back for any plant community before the beginning of the very severe drought (first harvest; see Appendix Figure S2; Table S1).

Aboveground Biomass of Plant Communities

No pre-exposure effect could be detected in aboveground biomass for any plant community before the beginning of the very severe drought (first harvest; see Appendix Figure S3; Table S2).

Tissue Die-Back of Plant Species

None of the plant species showed a pre-exposure effect in tissue die-back before the beginning of the very severe drought event (first harvest; see Appendix Figure S4; Table S3).

Aboveground Biomass of Plant Species

P. lanceolata revealed a significant pre-exposure effect in the grassland plant community with a legume species (G4+), whereby plants pre-exposed to “heavy rain” showed a higher amount of aboveground biomass than plants previously exposed to “drought” (first harvest; see Appendix Figure S5; Table S4).

Furthermore, no pre-exposure effect could be detected for C. vulgaris, V. myrtillus, A. elatius, and H. lanatus in aboveground biomass before the beginning of the very severe drought (first harvest; see Appendix Figure S5; Table S4).

Reproductive Biomass of Plant Species

None of the five plant species, where an analysis was possible, exhibited a pre-exposure effect in reproductive biomass and in the ratio of reproductive to total biomass before the beginning of the very severe drought (first harvest; see Appendix Tables S3 and S4).

Only one significant pre-exposure effect was found for one plant species out of 28 different species cases before the beginning of the very severe drought (five species, four community combinations, four response parameters; not all combinations were available).

Pre-exposure Effects on Plant Communities in the Face of a Very Severe Drought

Tissue Die-Back of Plant Communities

Tissue die-back due to the very severe drought differed between the pre-exposures (“ambient control,” “drought,” “heavy rain,” and “regular watering”) in all plant communities within harvest 2–8 (Figure 2, top panel; Table 4). The plants subjected to the pre-exposure “regular watering” were most negatively affected by the very severe drought event (Figure 2, top panel). The plants of heath plant communities (H4− and H2−) and of the grassland community without a legume species (G4−), which had in previous years been exposed to “ambient control,” “drought,” and “heavy rain” conditions revealed no significant differences between the pre-exposures (Figure 2, top panel). In contrast, the plants of the “ambient control” pre-exposure in the grassland plant community containing a legume species (G4+) showed a significantly higher tissue die-back due to the very severe drought in comparison to the plants previously subjected to “drought” and “heavy rain.” Generally, the species of the heath plant communities (H4− and H2−) were less affected by the very severe drought and had a lower tissue die-back than the grassland plant communities (G4− and G4+).

Figure 2
figure 2

Tissue die-back (%) (top panel) and root to shoot ratio (bottom panel) of heath and grassland plant communities from harvest 2–8. The lower case letters represent significant differences as revealed by the post hoc test. Displayed are the mean percentages over all sampling dates and the standard errors. n. s. not significant.

Full size image
Table 4 ANOVA Results for the Effects of Pre-exposure (“Ambient Control,” “Drought,” “Heavy Rain,” and “Regular Watering”) on Tissue Die-back and Root to Shoot Ratio of the Different Experimental Heath (H4− and H2−) and Grassland (G4− and G4+) Plant Communities from Harvest 2–8
Full size table

Aboveground Biomass of Plant Communities

Significant pre-exposure effects were found for grassland plant communities (G4− and G4+) in the aboveground biomass during the very severe drought (harvest 2–8; see Figure 3, top panel; Appendix Table S5). Grassland plant communities after pre-exposure to “heavy rain” produced higher amounts of aboveground biomass compared to “ambient control” (both G4− and G4+). The grassland plant community without a legume species (G4−) produced higher amounts of aboveground biomass also after pre-exposure to “drought” compared to “ambient control.”

Figure 3
figure 3

Aboveground biomass (g m−2) (top panel) and belowground biomass (g m−2) (bottom panel) of heath and grassland plant communities from harvest 2–8. The lower case letters represent significant differences as revealed by the post hoc test. Displayed are the means over all sampling dates and the standard errors. n. s. not significant.

Full size image

The heath plant communities (H4− and H2−) did not show a significant pre-exposure effect in the aboveground biomass when subjected to a very severe drought (see Figure 3, top panel; Appendix Table S5).

Belowground Biomass of Plant Communities

Belowground biomass was not affected by any pre-exposure in neither grassland nor heath plant communities (Figure 3, bottom panel; Appendix Table S5).

Root to Shoot Ratio of Plant Communities

The root to shoot ratio during the very severe drought manipulation of grassland plant communities (both G4− and G4+) was affected by the different pre-exposures (Table 4). The plants in the grassland plant community without a legume species (G4−) reached a higher root to shoot ratio, if in previous years exposed to “regular watering” or “ambient control” conditions than if previously exposed to annually recurrent “drought” (Figure 2, bottom panel). In the grassland plant community with a legume species (G4+), we found a higher root to shoot ratio of the plants in the “ambient control” than in the “heavy rain” pre-exposed plots (Figure 2, bottom panel). For heath plant communities (H4− and H2−), root to shoot ratio was not affected by pre-exposure during the very severe drought (Figure 2, bottom panel; Table 4).

Pre-exposure Effects on Plant Species in the Face of a Very Severe Drought

Tissue Die-Back of Plant Species

Tissue die-back revealed pre-exposure effects for some of the target species in the different plant communities and monocultures (analysis of harvest 2–8; Table 5). C. vulgaris (in H4− and H2−), V. myrtillus (in H2−), H. lanatus (in G4−), and P. lanceolata (in G4−) showed a reduced drought resistance when subjected to “regular watering” in previous years (Figure 4). However, there was no difference in tissue die-back after pre-exposure to “ambient control” conditions, “drought” or “heavy rain” for C. vulgaris (in H4− and H2−), H. lanatus (in G4−), and P. lanceolata (in G4−). Regarding the four monocultures, only the V. myrtillus monoculture showed a significant pre-exposure effect (Figure 4).

Table 5 ANOVA Results for the Effects of Pre-exposure (“Ambient Control,” “Drought,” “Heavy Rain,” and “Regular Watering”) on Tissue Die-back and Reproductive Biomass (Seeds and Flowers) of the Heath and Grassland Plant Species from Harvest 2–8
Full size table
Figure 4
figure 4

Tissue die-back (%) of plant species in heath and grassland communities from harvest 2–8. The lower case letters represent significant differences as revealed by the post hoc test. Displayed are the mean percentages over all sampling dates and the standard errors. n. a. not available, n. s. not significant, “+” marginal significant (post hoc test for regular watering versus heavy rain P = 0.055).

Full size image

Regarding the plant community composition effect, V. myrtillus growing together with C. vulgaris (H2−) was less affected by tissue die-back than V. myrtillus in the monoculture (M−) (Figure 4) comparing “ambient control” conditions with annually recurrent “drought” in each plant community composition, respectively (analysis of harvest 2–8). Likewise, the response of P. lanceolata to the very severe drought was modified by plant community composition. Growing in the plant community without a legume species (G4−), P. lanceolata exhibited no significant difference between “ambient control” and “heavy rain” pre-exposure, but tended toward a lower tissue die-back when previously exposed to “ambient control” conditions. The opposite was found in the plant community containing a legume species (G4+), where P. lanceolata showed a lower tissue die-back in the “heavy rain” compared to the “ambient control” pre-exposed plants (harvest 2–8; Figure 4).

Aboveground Biomass of Plant Species

The aboveground biomass of P. lanceolata in both grassland plant communities (G4− and G4+) showed significant pre-exposure effects when subjected to a very severe drought (analysis of harvest 2–8; see Appendix Table S6). In the absence of a legume species (G4−), P. lanceolata produced more aboveground biomass under very severe drought conditions when previously exposed to “drought” compared to “ambient control” and “regular watering.” On the other hand, in the presence of a legume species (G4+), P. lanceolata produced more aboveground biomass under very severe drought conditions when pre-exposed to “heavy rain” compared to “ambient control” and “drought” (see Appendix Figure S6; Table S6). Again, the response of P. lanceolata differed between the two plant community compositions G4− and G4+ for the parameter aboveground biomass. As already mentioned above, the case of P. lanceolata in the presence of a legume species was the only one with a significant pre-exposure effect also before the beginning of the very severe drought.

C. vulgaris, V. myrtillus, A. elatius, and H. lanatus did not reveal any pre-exposure effect in the aboveground biomass during the very severe drought (see Appendix Figure S6; Table S6).

Reproductive Biomass of Plant Species

Significant pre-exposure effects were found in the production of reproductive biomass (seeds and flowers) during the very severe drought for H. lanatus and P. lanceolata in grassland plant communities (G4− and G4+) (harvest 2–8; Table 5). H. lanatus produced more reproductive biomass (Figure 5) and had a higher ratio of reproductive to total biomass (see Appendix Figure S7; Table S6) after “heavy rain” than after “regular watering” (in G4−) as well as after “heavy rain” than after “drought” pre-exposure (in G4+). P. lanceolata produced more reproductive biomass if previously pre-exposed to “heavy rain” in the plant community containing a legume species (G4+). However, this was not the case in the plant community without a legume species (G4−), where P. lanceolata showed no differences between the “heavy rain” and the other pre-exposures (Figure 5). Thus, P. lanceolata revealed a differentiated response in reproductive biomass with regard to the plant community composition during the very severe drought. Furthermore, P. lanceolata did not exhibit a significant pre-exposure effect in the ratio of reproductive to total biomass as H. lanatus did (see Appendix Table S6).

Figure 5
figure 5

Reproductive biomass (g m−2) (seeds and flowers) of Holcus lanatus and Plantago lanceolata from harvest 2–8. The lower case letters represent significant differences as revealed by the post hoc test. Displayed are the means over all sampling dates and the standard errors. “+” marginal significant (post hoc test for regular watering versus drought P = 0.058).

Full size image

C. vulgaris, V. myrtillus, and A. elatius were not affected by previous exposure in reproductive biomass and in the ratio of reproductive to total biomass during the very severe drought (harvest 2–8; Table 5; Appendix Table S6).

Discussion

Recurrent Mild Drought Stress Triggers Ecological Stress Memory

Surprisingly, plant communities and species with recurrent mild drought stress over several years showed a higher drought resistance than plants without drought experiences for 3 years in the face of a very severe drought event. The higher drought resistance revealed itself in the form of lower tissue die-back (dead over total biomass). This effect was apparent in all plant communities being pre-exposed to “regular watering” (Figure 2, top panel). Focusing on the differences, the heath communities showed a lower relative tissue die-back than the grassland communities in the face of the very severe drought. The conservative life strategy of heath (longevity, lignification) compared to grassland (Larcher 2003) and thus their ability to resist might lead to the better performance. An ecological stress memory (Walter and others 2013) could be an explanation for the relatively better performance of the plants when pre-exposed to “ambient control” conditions, recurrent “drought,” or “heavy rain” events. Since plants pre-exposed to “regular watering” can be expected to have experienced natural drought periods before 2007 (see “Methods” section), the length of the observed ecological stress memory appears to be shorter than 4 years. Moreover, our results are strengthened by the missing pre-exposure effects (except for P. lanceolata in aboveground biomass) before the beginning of the very severe drought, implying that the observed pre-exposure effects are no simple carry-overs but rather true stress memory effects.

Our finding suggests that the events which trigger drought resistance do not have to be extreme themselves. Plants pre-exposed to “ambient control” conditions and annually recurrent “heavy rain” events experienced 26 and 25 days with volumetric soil water content dropping below the permanent wilting point over the 3 years before the very severe drought. By contrast, plants pre-exposed to “regular watering” experienced only two such days over the same period. Thus, the variability and recurrence of water stress might lead to a higher drought resistance of the plants as compared to permanently favorable conditions with respect to the very severe drought in 2011.

However, an influence of recurrent mild or extreme drought stress could not be found when considering other parameters than tissue die-back such as total aboveground biomass (Figure 3, top panel; Appendix Figures S3, S5, and S6). The growth performance of the plant communities and species with different pre-exposures was mostly similar. However, tissue die-back did suggest that plant adaptations to recurrent drought exist. In the cases considered, the way the biomass is distributed among living and dead components has apparently a greater responsiveness than the total amount of biomass itself.

The increasing root to shoot ratio and root dynamics into deeper soil layers are known strategies for dealing with drought stress (Kalapos and others 1996; Kahmen and others 2005; Newman and others 2006; Ehdaie and others 2012). Based on these findings, the increasing root to shoot ratio was suggested as a mechanistic explanation for ecological stress memory (Walter and others 2013). However, our results for root to shoot ratio did not show an increase in belowground biomass in the pre-stressed plants. Moreover, the belowground biomass of all plant communities clearly did not exhibit any differences driven by pre-exposure in previous years (Figure 3, bottom panel). Thus, no adaptation of the root system to recurrent drought events over subsequent years was observed in any plant community (Figure 2, bottom panel). Yet, our findings for belowground biomass are restricted to the uppermost 14 cm of the soil and thus cannot provide insights into root adaptations to recurrent drought in deeper soil layers. In accordance with our results, Kreyling and others (2008b) as well as Gilgen and Buchmann (2009) also found no alterations in plant belowground biomass in reaction to recurrent drought events. In our study, the weather extremes and climatic variability in years prior to the very severe drought have probably improved the drought resistance of the plant communities without affecting the root to shoot ratio. The higher root to shoot ratio of the plants pre-exposed to “regular watering” in the grassland community without a legume species (G4−) and of plants pre-exposed to “ambient control” conditions in both grassland communities (G4− and G4+) (Figure 2, bottom panel) may imply that shifts in root to shoot ratio are rather short-term responses of less well-adapted plants. Here, a direct stress reaction to the very severe drought conditions might explain the plant root performance instead of an ecological stress memory. This assumption is strengthened by root to shoot ratio results from September 2010 (79 days after the drought manipulation in 2010; root depth of 15 cm), where grassland communities (G4− and G4+) showed no significant pre-exposure effect (data not shown). However, root samples from September 2010 were collected with a different method compared to 2011. Thus, a direct comparison between the findings of 2010 and 2011 cannot be provided.

Other mechanisms for the explanation of the observed ecological stress memory include the accumulation of signaling proteins or transcription factors, which could promote a rapid response to subsequent recurrent stresses (see Conrath and others 2006; Bruce and others 2007). Moreover, epigenetic changes in terms of modification of DNA activity by methylation, histone modification or alterations in genome stability and chromatin organization due to the environmental stress are possible explanations for the plant response in our study and thus for the ecological stress memory (Madlung and Comai 2004; Bruce and others 2007; Boyko and Kovalchuk 2011). Furthermore, soil biotic legacy effects induced by preceding drought occurrences before planting were found by Meisner and others (2013). This was associated with an increase in inorganic nitrogen availability owing to preceding drought. Soil biotic legacy effects might therefore be an important aspect in understanding the mechanism of ecological stress memory considered in our study.

Influence of Community Composition on the Ecological Stress Memory of Plant Species

The ecological stress memory seems to be modified by community composition, that is, the presence of neighbors for V. myrtillus and P. lanceolata (Figures 4, 5, see Appendix Figures S5, S6). The respective plant neighborhood consisting of neighbors of the same plant species or of different plant species might play a role with regard to plant response to a subsequent very severe drought extreme. Especially, the presence of a legume species in the neighborhood influenced the response of P. lanceolata with regard to aboveground biomass (Figures S5, S6), tissue die-back (Figure 4), and reproductive biomass (Figure 5). However, we cannot dismiss a simple carry-over effect from the previous years in this case, as aboveground biomass of P. lanceolata already showed a pre-exposure effect before the beginning of the very severe drought (Figure S5). Importantly, however, no such carry-over by pre-exposure occurred in the community without a legume. The explanation is in accordance with findings in aboveground biomass within the EVENT-I experiment in previous years (2007–2010). Arfin Khan and others (2014) showed that P. lanceolata was facilitated by the legume species when exposed to “heavy rain,” but not under “drought” conditions. The missing legume effect under drought exposure seems associated with decreased N-uptake of the plant species rather than by a decline in legume N-fixation.

Plant neighbors have been shown in previous studies to affect plant performance under drought stress (for example, Novoplansky and Goldberg 2001; Kreyling and others 2008a). Competition and facilitation under varying stress situations (Callaway 1997; Callaway and Walker 1997) might be relevant factors in the progress of the ecological stress memory of different plant species. For instance, Grant and others (unpublished data) show that an extreme drought event led to contrasting plant–plant interactions. Although A. elatius was facilitated by neighboring plants under drought, the same drought event increased the competitive pressure of neighboring species on Lotus corniculatus.

However, our community compositions were fixed and not repeated with other plant species diversity levels or rigorously tested for neighborhood effects by pairwise combinations of the grassland species. Therefore, the findings might be explained by sampling effects of the used plant species (for example, the potential legume effect could also be a species identity effect by the one legume used here, L. corniculatus). Further experiments are required with other plant species combinations to confirm the generality of our results. Still our findings for V. myrtillus and potentially also for P. lanceolata partly confirm our stated expectation and are in accordance with the current knowledge about the influence of plant neighborhood and community composition under drought conditions. This finding of community composition altering the pre-exposure effects of at least some target species represents a new aspect in ecological stress memory research, where most studies up to now focused on the single plant level (Goh and others 2003; Molinier and others 2006; Whittle and others 2009; Cuk and others 2010; Walter and others 2011).

Conclusions

In summary, our findings suggest that mild drought stress over several years seems to influence plant resistance and adaptation positively against extreme drought events. In particular, the lack of drought history in “regular watering” over years reduced the drought resistance of plant species and communities. Interestingly, this study indicates that community composition and plant neighborhood might play a role in plant response and thus in alterations of this ecological stress memory. The complexity of plant–plant interactions under drought conditions has to be taken into account in further ecological stress memory research. By neglecting the competitive aspects within a plant community, experimental approaches might lead to false ecological implications concerning the ecological stress memory. Thereby, ecological stress memory findings on the single plant level have to be taken with caution. Additionally, thresholds and limited longevity of ecological stress memory effects need further research efforts. Finally, ecological stress memory seems to be a supporting tool for heath and grassland plant communities and species to adapt to a changing and more extreme climate.