Whole-system responses of experimental plant communities to climate extremes imposed in different seasons


  • Hans J. De Boeck,

    1. Research Group of Plant and Vegetation Ecology, Department of Biology, Universiteit Antwerpen (Campus Drie Eiken), Universiteitsplein 1, B-2610 Wilrijk, Belgium
    Search for more papers by this author
  • Freja E. Dreesen,

    1. Research Group of Plant and Vegetation Ecology, Department of Biology, Universiteit Antwerpen (Campus Drie Eiken), Universiteitsplein 1, B-2610 Wilrijk, Belgium
    Search for more papers by this author
  • Ivan A. Janssens,

    1. Research Group of Plant and Vegetation Ecology, Department of Biology, Universiteit Antwerpen (Campus Drie Eiken), Universiteitsplein 1, B-2610 Wilrijk, Belgium
    Search for more papers by this author
  • Ivan Nijs

    1. Research Group of Plant and Vegetation Ecology, Department of Biology, Universiteit Antwerpen (Campus Drie Eiken), Universiteitsplein 1, B-2610 Wilrijk, Belgium
    2. King Saud University, Riyadh, Saudi Arabia
    Search for more papers by this author

Author for correspondence:
Hans J. De Boeck
Tel.: +32 3 265 22 82
Email: hans.deboeck@ua.ac.be


  • Discrete climate events such as heat waves and droughts can have a disproportionate impact on ecosystems relative to the temporal scale over which they occur. Research oriented towards (extreme) events rather than (gradual) trends is therefore urgently needed.
  • Here, we imposed heat waves and droughts (50-yr return time) in a full factorial design on experimental plant communities in spring, summer or autumn. Droughts were created by removing the controlled water table (rainout shelters prevented precipitation), while heat waves were imposed with infrared heaters.
  • Measurements of whole-system CO2 exchange, growth and biomass production revealed multiple interactions between treatments and the season in which they occurred. Heat waves had only small and transient effects, with infrared imaging showing little heat stress because of transpirational cooling. If heat waves were combined with drought, negative effects observed in single factor drought treatments were exacerbated through intensified soil drying, and heat stress in summer. Plant recovery from stress differed, affecting the biomass yield.
  • In conclusion, the timing of extreme events is critical regarding their impact, and synergisms between heat waves and drought aggravate the negative effects of these extremes on plant growth and functioning.


By nature, organisms are more sensitive to abrupt than to gradual change, and extreme climate events are considered to be the main instigator of evolution (Gutschick & BassiriRad, 2003) as they ‘weed out’ genotypes that are not well enough adapted (Combes, 2008). Importantly, current climate change is increasing both the likelihood and the intensity of extreme climate events because small shifts in the climate mean give rise to pronounced changes in the tails of the probabilistic distributions (Schär et al., 2004; Sterl et al., 2008). The European summer drought and heat in 2003 illustrated the profound impact that extreme events may have. Apart from short- and longer-term effects on (semi-)natural systems (Dobbertin et al., 2005; Reichstein et al., 2007), the extreme summer saw financial losses exceeding $10 billion because of harvest losses and forest fires (Schär & Jendritzky, 2004). Jentsch et al. (2007) showed that research on extreme events is currently underrepresented in the body of climate manipulation experiments, despite its apparent importance. Here, we investigated how two types of extreme climate event (heat waves and droughts) affect experimental plant communities. Both immediate responses during and recovery after the extreme events, imposed as single factors and in combination, were quantified by means of whole-system measurements of growth, CO2 exchange and productivity. Because extreme deviations from the normal weather can occur throughout the year, triggering potentially varying plant responses (De Boeck et al., 2010), we added seasonality as a factor by imposing droughts, heat waves and both combined in spring, summer and autumn.

Low water availability caused by drought is one of the major limitations for plant growth, affecting vegetation structure, plant productivity and interactions between plants (Chaves & Oliveira, 2004). It typically arises from low precipitation inputs or high rates of water loss as a result of high atmospheric vapour pressure deficits. Plant strategies generally range from drought avoidance (e.g. by tapping into deep soil water layers or by hibernating as seeds), to drought tolerance (resistance), to allowing transient senescence followed by recovery once the stressor disappears (resilience). Most temperate herbaceous species are resilient rather than resistant (Ingram & Bartels, 1996), and should therefore be vulnerable to drought at least in the short term. In temperate grasslands, drought events have been reported to reduce productivity (Kahmen et al., 2005), lower reproductive success (Aragon et al., 2008), and ultimately alter the species composition (Grime et al., 2000). Negative effects primarily arise from diminished leaf carbon fixation as a result of stomatal closure and a concomitant or even earlier inhibition of growth (Chaves & Oliveira, 2004) as cell division and expansion are directly inhibited by water stress (Zhu, 2001). The impact of drought ultimately depends on a multitude of factors, including antecedent effects (such as the soil water status at the start of the drought), the plant age (Law et al., 2003), the intensity of the drought (which is related to its duration and the soil water potentials reached) and the diversity of the plant community (Kahmen et al., 2005; De Boeck et al., 2008a). The discussion is further complicated by the fact that there is no uniform definition of drought (Heim, 2002). Most ecological studies have considered drought as a period without significant precipitation, independent of soil water status. This is why some studies have found beneficial effects of drought, in the case where water stress (in wetlands etc.) was lifted by decreased precipitation (e.g. Schwalm et al., 2010). Precipitation deficits thus do not necessarily lead to soil drought. Unfortunately, few historical data on soil water content exist (Kreyling et al., 2008), which is why, in simulating rare events, researchers have to rely on precipitation records. Drought experiments are therefore in reality mostly precipitation deficit experiments. Extreme precipitation deficits can occur in any season, with the exact characteristics depending on the probabilistic distribution of rainfall in that period of the year. These deficits cause drought especially in times of high evapotranspiration, which is more likely in summer, when leaf area is at its peak and temperatures are high (De Boeck et al., 2010).

Although high temperatures can have direct negative effects on plants by the exceedance of temperature optima (Larcher, 2003; Niu et al., 2006), there is growing evidence that heat effects are mostly indirect. Unless precipitation and relative humidity increase under heat wave conditions, which is highly unlikely (De Boeck et al., 2010), soil water depletion will be accelerated because of high evapotranspiration associated with the increased vapour pressure deficits. Such warming-induced drought has been observed in many temperature manipulation studies (e.g. Saleska et al., 1999; Marchand et al., 2006; De Boeck et al., 2008a), and can cause growth reductions, especially in warmer ecosystems (Rustad et al., 2001). The connection between high temperatures and drought was highlighted in a study by Arnone et al. (2008), which showed that declines in tallgrass prairie productivity observed in an anomalously warm year were comparable to declines observed in response to drought in other tallgrass prairie studies (Knapp et al., 2001). Other indirect effects of high temperatures include an increased risk of pest outbreaks (Rouault et al., 2006), effects on nutrient cycling (Schmidt et al., 2002) and a possible lengthening of the growing season, especially by advancing the onset of spring (Menzel et al., 2006; Piao et al., 2008), although not for all species (Körner & Basler, 2010). The season in which a heat wave occurs probably affects its impact greatly. The occurrence of heat stress, when a certain absolute temperature is exceeded, is most likely during summer heat waves, while positive effects are possible in spring (Delpierre et al., 2009; Vesala et al., 2010) and autumn, when ambient temperatures are usually suboptimal in the temperate zone.

Evaluating how multifactor interactions (here, between heat waves and droughts) affect ecosystem functioning is critical to understanding the global change responses in the real world (Luo et al., 2008). Indeed, when interactive effects predominate over single factor effects, single factor experiments are less useful for understanding ecosystem changes. High temperatures can exacerbate droughts by increasing vapour pressure deficits and hence evapotranspiration, leading to an earlier reduction in carbon assimilation rates because of stomatal closure (Arnone et al., 2008), thereby potentially affecting productivity further. Negative interactions such as these may be most pronounced in summer, when high light and temperature stress are most common, leading to further down-regulation of photosynthesis (Chaves & Oliveira, 2004).

Based on the aforementioned scientific knowledge of plant responses to heat and/or drought, our study specifically aimed to address the following hypotheses: a heat wave may stimulate growth in spring and might also do so in autumn; effects of drought and the combined treatment (heat + drought) on plant performance will be uniformly negative and most pronounced in summer; the combination of heat and drought causes effects that are greater than the sum of single heat and drought effects. These hypotheses were tested on experimental communities consisting of common herbaceous species in Belgium in a fully factorial set-up.

Materials and Methods

Study site and experimental set-up

This study was conducted at the Drie Eiken Campus of the University of Antwerp (Belgium, 51°09′N, 04°24′E) in 2009. The local climate is characterized by mild winters and cool summers, with an average annual air temperature of 9.6°C and 776 mm of rainfall, equally distributed throughout the year. Experimental plant communities containing three common herbaceous species (Plantago lanceolata L., Rumex acetosella L. and Trifolium repens L.) were established in the previous year (2008) to minimize soil disturbance carryover effects and to avoid working with fledgling communities. We opted to use experimental plant communities because these have the advantage that differences between communities (in soil structure, nutrient availability, composition, and plant numbers) are minimized, resulting in improved comparability between treatments and higher statistical power, as unwanted variation is lower. The three species used were selected because they represent common short-statured perennial forbs in central Europe that also have leaves large enough to allow straightforward leaf-cuvette gas exchange measurements. We opted for a species mixture, as the response of one individual species (in monocultures) is more indicative of species-specific characteristics (e.g. drought tolerant or drought sensitive) than of general responses. Each of the communities consisted of 10 plants (per three mesocosms, a different species was allocated four individuals to ensure equal representation) arranged in a hexagonal design with only interspecific neighbours (cf. De Boeck et al., 2008a), in PVC tubes of 20 cm diameter and 40 cm depth. Plants were treated regularly to avoid fungal infection and insect damage, and we weeded manually throughout the experiment. We clipped the vegetation on 19–20 June and 26–30 October. The soil used in the mesocosm experiment was a sandy soil (96% sand) with an initial carbon content of 1.3%, 19 mg nitrate-nitrogen (N) and 1.1 mg ammonium-N kg−1 dry soil, 13 mg phosphorus (P) kg−1 dry soil and a pH of 7.6. The water content was 0.15 m3 m−3 at field capacity (pF 2.5) and 0.037 m3 m−3 at wilting point (pF 4.2), determined by a soil laboratory (Bodemkundige Dienst, Leuven, Belgium). Commercially available pot plant fertilizer (Substral Universal, Scotts Benelux, Sint-Niklaas, Belgium) (in the recommended doses) was added at the beginning of the experiment; application was halted before the treatments started. The mesocosms were placed in watertight wooden boxes (135 × 135 cm), fully separated into two halves, each containing 18 mesocosms. A set of two boxes thus contained 18 mesocosms subjected to drought, 18 subjected to heat, 18 subjected to both, and 18 controls. The boxes, six in total, were dug into the soil to provide more realistic soil temperatures. Permanent rainout shelters of height 200–240 cm (angular roofs), sticking out 1 m at each side of the boxes and covered with highly transparent polycarbonate, prevented rainfall from reaching the communities. Light attenuation was c. 5–15%, depending on the solar angle. Because of their stature, the shelters had excellent air flow and both build-up of heat (+0.2°C on average) and changes in relative humidity (+3% on average) were modest. A water table was created in the wooden boxes so the plant communities had ample water (see Fig. 1). At the start of a drought period, the water table was removed in the drought and drought + heat wave treatments. The mesocosms were relocated twice between boxes (in June and August) to avoid location-specific effects.

Figure 1.

 Course of soil water content (averages from four communities per treatment) measured with 30-cm-long moisture sensors. Data are shown for control, heat wave, drought and heat wave + drought mesocosms exposed to extreme events in (a) spring, (b) summer or (c) autumn. DOY, day of the year; day 1 is 1 January 2009.

Extreme events with a return period of c. 50 yr were simulated. Such extremes will probably become commonplace events in the next few decades (IPCC, 2007). The drought + heat wave treatment combined the two individual events, which each had a 50-yr return time, as required for a fully factorial experiment. The estimated probability of occurrence for such a combined event would be significantly less than 2500 yr, however, as precipitation-free periods and heat waves are not independent and co-occur c. 25% of the time in Western Europe (H. J. De Boeck, unpublished data). This would imply a return time of c. 200 yr for the combined extreme event in our experiment. Note that the 2003 summer heat and drought in Europe had an estimated statistical return time of several thousands of years, which is projected to drop to merely 2 yr by the end of the century (Schär et al., 2004). For droughts, the length of the precipitation-free period was derived from the precipitation records of the Royal Meteorological Institute in Ukkel near Brussels (c. 50 km from the experimental site), dating back 130 yr. A day with precipitation < 1 mm was considered as a precipitation-free day (amounts < 1 mm are generally intercepted by the canopy and thus do not replenish soil water reserves). We derived drought lengths for every month by analysing the data per month (with overlap to the previous and the following month because of the length of droughts) as the probabilistic distribution of precipitation differed throughout the year. The lengths of the drought periods, determined by interpolation of return time data, were: 26 d (spring treatment, 17 April–13 May), 25 d (summer treatment, 6–31 July), and 31 d (autumn treatment, 8 September–9 October). Heat waves can be characterized by two parameters: length and maximum temperatures. We opted to fix the length to 10 d (a realistic length for the region; see De Boeck et al., 2010) and determine the maximum temperatures corresponding with heat waves of that length. The air temperature record from the Royal Meteorological Institute of Belgium goes back to 1833, and was analysed similarly to the precipitation record. The average daily maximum air temperatures for 10-d heat waves with a 50-yr return time thus determined were 25.6°C for the spring treatment, 29.8°C for the summer treatment and 21.9°C for the autumn treatment (this is c. 7°C above normal maximum temperatures in all cases). As heat waves are usually preceded by a dry period (De Boeck et al., 2010), we imposed the heating treatment during the last 10 d of the drought. For example, for the summer treatment this would mean that the drought treatment lasted from 6 to 31 July, the heating-only treatment from 21 to 31 July and the combined treatment from 6 to 31 July, with heating on top of drought from 21 to 31 July. Six infrared heaters of 1500 W each were suspended c. 120 cm above one box (containing both the heating and the combination treatments) to increase the canopy temperature directly, and the air temperature indirectly. Above the paired box containing controls and communities subjected to drought, a dummy construction was placed with empty lamp enclosures (cf. Van Peer et al., 2001). Several studies have shown the advantages of warming vegetation with infrared heaters over other methods (Kimball et al., 2008; Aronson & McNulty, 2009). Our own recent analysis of meteorological conditions during heat waves (De Boeck et al., 2010) found that the natural depression of relative humidity during heat waves renders the side effect of infrared heaters of drying the air (Kimball, 2005) an additional benefit of using this technique. The power output of the heaters was adjusted to match the target maximum temperatures (already described), and the output was lowered during the night to reflect the lower nighttime vs daytime increases in temperatures during natural heat waves (De Boeck et al., 2010). The set-up performed acceptably, with average maximum temperatures during heat waves deviating −0.5°C, +3.2°C and +0.5°C from the target temperature in spring, summer and autumn, respectively. Meteorological data indicated that weather was fairly average during the imposed treatments: ambient mean air temperatures were 1.4, 0.0 and 0.6°C higher than average during the spring, summer and autumn treatments, respectively, without exceptionally high or low temperatures.


To establish the immediate response of the plant communities to the imposed treatments, we measured ecosystem CO2 fluxes during nine periods (later referred to as periods 1–9), each lasting 2 or 3 d. These periods were 1 wk before, at the end of, and 1 wk after the end of each of the three (spring, summer and autumn) extreme event simulations. We used one transparent, 40-cm-high polymethyl pentene cuvette that could be tightly fitted to the plant containers, coupled to an EGM-4 infrared gas analyzer (PP Systems, Hitchin, UK). Two fans inside the cuvette mixed the air during measurements. Each 90-s measurement of net ecosystem exchange of CO2 (NEE) was followed by another 90-s measurement of total ecosystem respiration (TER) on the same community by darkening the cuvette using an opaque cloth, thereby preventing photosynthesis (cf. De Boeck et al., 2007). Gross photosynthesis (Agross) could then be calculated as Agross = NEE − TER. Because of the partial inhibition of leaf respiration in the light (Atkin et al., 2000), Agross may be slightly overestimated. Note that we are concerned with relative differences between treatments, however, not absolute values. Measurements were performed on the same six communities per treatment and repeated once per period, resulting in 12 data points for each of the nine treatments and 36 data points for the controls (because of the greater number of control mesocosms; see above), in each of the nine periods. Measurements were generally made between 09:00 and 17:00 h, and performed criss-cross between treatments, to ensure a wide range of light intensities and temperatures for all treatments. Photo-synthetic photon flux density (PPFD) was measured with a quantum sensor (JYP 1000; SDEC, Tauxigny, France) inside the cuvette. Thermocouples shielded from direct radiation were used to measure air temperature at the canopy level, and noncontact semiconductors were used for measuring canopy surface temperatures at half-hourly intervals. The soil water content, recorded hourly, was quantified with Campbell CS616 sensors (Campbell Scientific Ltd, Loughborough, UK) installed in four communities per treatment and in 12 control communities (i.e. a total of 48 sensors). These Time Domain Reflectometry (TDR) probes were 30 cm long and thus covered most of our soil profile. The values were corrected using an in situ calibration. Aboveground biomass (living biomass was separated from dead biomass) was determined by cutting all plant communities twice a year (in June and October) c. 4 cm above the soil surface, drying the plant parts for at least 2 d at 70°C and weighing them. Vegetation height was determined at seven dates (later referred to as readings 1–7) as the average of six height measurements (taken as the height at which the highest plant part touched a ruler) within a community. Finally, surface temperature measurements were made with a TH9260 infrared camera (NEC Avio Ltd, Tokyo, Japan), which has a resolution of 640 × 480 pixels. Pictures were taken using the ‘multi-focus’ setting, which renders a composite image of 32 photographs, thus dampening erratic wind speed effects. The images were taken from a fixed position under a standardized angle (c. 45°) using a tripod. Per treatment, four images were taken at different times on each measurement day. Measurements were made on the day before the start of a heat wave, and the last three readings during the heat wave (i.e. the last 10 d of an extreme event). Average surface temperatures of the canopy were subsequently derived with irMotion software (Atus GmbH, Hamburg, Germany). To prevent so-called observer effects from skewing results (De Boeck et al., 2008b), the number of measurements was kept to a minimum, and no destructive sampling was carried out apart from the harvests.

Statistical analysis

We used spss 15.0 (SPSS Science, Woking, UK) for general linear model (GLM) univariate analysis of variance on aboveground biomass, with the treatment (control, drought, heat wave and combined) and the season (spring, summer and autumn) as fixed factors. The Games–Howell post hoc test was used to separate multiple means. The significance threshold was 0.05. To enable comparison between treatments, gross photosynthesis was plotted against photosynthetic photon flux density separately for each treatment and period, using the general function

image(Eqn 1)

(QE, the quantum efficiency; Amax, the gross photosynthesis at a photosynthetic photon flux density (PPFD) of 1500 μmol m−2 s−1.) The range of light intensities during measurements confirmed that light curves were well captured (comparable curves can be found in De Boeck et al., 2007). These graphs were statistically compared, per period and per seasonal treatment group, using GraphPad Prism 5 software (GraphPad Software Inc., La Jolla, CA, USA). With an extra sums-of-squares F-test, the null hypothesis was tested that a predefined Eqn 1 with identical parameter values (here for Amax and QE) fits both sets of data. If the null hypothesis is rejected, both data sets are best defined by different parameter values, calculated by iteration. We lowered the significance threshold to 0.01 as multiple (six) comparisons were made per period and per seasonal treatment group, increasing the probability of type I errors (cf. the Bonferroni adjustment would be 0.0083, but is widely regarded as overly conservative). Total ecosystem respiration was not plotted with temperature (respiration correlates primarily with temperature) as the temperature range within a period was too narrow. Because flux measurements were taken throughout the day for all treatments, and a comparison between treatments revealed no significant differences in temperature excluding those during the extreme events, we compare the TER data sets directly with each other for a general idea of trends. The significance level was lowered to 0.01 as multiple comparisons were made within a period and per seasonal treatment. An analogue analysis was applied to the data for vegetation height. Data were transformed for normality when necessary.


Soil water status

From the data recorded with the Campbell sensors, it is clear that plants in both the control and the heat wave treatment never suffered from drought, as the soil water content (SWC) was consistently near or above field capacity (0.15 m3 m−3) (Fig. 1). Wilting point (determined to be at 0.0365 m3 m−3) was exceeded during both summer and autumn drought plus heat, and was approached (SWC < 0.04 m3 m−3) in the spring combination treatment and during the summer drought. The graphs in Fig. 1 clearly show that the speed of soil drying was faster when both precipitation was halted and temperatures were increased. We opted not to display soil drought in terms of the water potential, as the range between pF 3 (−0.1 MPa) and pF 4.2 (−1.5 MPa) was only 0.0026 m3 m−3, which is so small compared with the accuracy of the SWC sensors that water potentials would generate a lot of noise at low values.

CO2 fluxes

In the following sections, all comparisons are significant unless stated otherwise. The only differences in Agross curves found between treatments in response to extreme events in spring were a slight positive effect of a heat wave in period 2 (at the end of the extreme event) and a small negative effect of the combined treatment in period 3 (the week after the extreme event). The maximum photosynthesis, one of the two parameters defining the Agross curves, is shown in Fig. 2. No clear significant effects were recorded for TER in the spring treatment (Fig. 3b). Extreme events in summer led to marked effects on gross photosynthesis (Fig. 2c): heat waves lowered Agross only at the end of the extreme event (period 5), but photosynthesis recovered to control levels immediately afterwards (period 6–9); drought reduced Agross to a greater extent than heat waves, and recovery was only observed after c. 50 d (period 7); the combined effect of a heat wave and a drought was so dramatic that Agross decreased more markedly than in other treatments (to near zero), and never recovered back to control levels (Fig. 2c). Effects of the summer extreme on TER were similar, but heat wave effects were never significant, and drought effects were sustained throughout the rest of the growing season. The adverse effect of combined extremes on TER was consistently larger than that of the other treatments (Fig. 3c). The impact of autumn extremes was only significant for drought after the extreme event (period 9) and for the combined treatment both at the end of the extreme event and afterwards (periods 8–9; Fig. 2d). Again, the combined heat wave–drought effects were more negative than those of other treatments. Trends in TER were comparable, the only difference being that the effect of drought was significant in both periods 8 and 9 (Fig. 3d).

Figure 2.

 Course of community photosynthesis at saturating (1500 μmol m−2 s−1) light conditions (Amax) derived from photosynthetic photon flux density–gross photosynthesis relationships (cf. De Boeck et al., 2007). (a) Absolute values under control conditions (± SE), derived from 18 control communities, sampled twice per measurement period. Differences between treatment and control values (ΔAmax) were analysed for six communities sampled twice per measurement period in response to a drought (squares), a heat wave (circles) and the combination of a heat wave and a drought (diamonds) imposed in spring (b), summer (c) or autumn (d). DOY, day of the year; day 1 is 1 January 2009. For clarity, only averages are depicted. Letters, arranged in the same vertical order as the points they refer to, indicate significant differences (a, identical to controls) derived from analysis of the entire gross photosynthesis curves. The timing of the extreme event is indicated with a box, and the time of mowing (4 cm height) is indicated by arrows.

Figure 3.

 Course of total ecosystem respiration (TER). (a) Absolute values under control conditions (± SE), derived from 18 control communities, sampled twice per measurement period. Differences between treatment and control values (ΔTER) were analysed for six communities sampled twice per measurement period in response to a drought (squares), a heat wave (circles) and the combination of a heat wave and a drought (diamonds) imposed in spring (b), summer (c) or autumn (d). DOY, day of the year; day 1 is 1 January 2009. For clarity, only averages are depicted. Letters, arranged in the same vertical order as the points they refer to, indicate significant differences (a, identical to controls). The timing of the extreme event is indicated with a box, and the time of mowing (4 cm height) is indicated by arrows.

Growth and biomass

We considered vegetation height as a proxy for growth and biomass, albeit an imperfect one. The linear regression between vegetation height and biomass, determined from the two harvests presented here and an additional harvest on identical but unused ‘test’ mesocosms, showed that vegetation height explained 30–60% of the variation in biomass (R2 between 0.3 and 0.6). The observed height differences, measured on seven days, corresponded fairly well with the CO2 fluxes. A slightly positive but transient effect of the heat wave (reading 2) and a negative effect of the combined treatment (reading 3) were found for the spring extreme event (Fig. 4b). Growth was negatively affected by drought (reading 5 only) but more so by drought plus heat (readings 5–7) in response to the summer extreme event (Fig. 4c). Only the combined treatment affected the plant height significantly after the autumn extreme event (Fig. 4d). While the vegetation height tended to decline at the end of the season (Fig. 4a), associated with senescence, this was not true for those communities that had been subjected to drought or combined heat and drought in the summer (Fig. 4c).

Figure 4.

 Course of vegetation height (see text for definition). (a) Absolute values under control conditions (± SE), derived from 18 control communities. Differences between treatment and control values (Δ vegetation height) were analysed for six communities in response to a drought (squares), a heat wave (circles) and the combination of a heat wave and a drought (diamonds) imposed in spring (b), summer (c) or autumn (d). DOY, day of the year; day 1 is 1 January 2009. For clarity, only averages are depicted. Letters, arranged in the same vertical order as the points they refer to, indicate significant differences (a, identical to controls). The timing of the extreme event is indicated with a box, and the time of mowing (4 cm height) is indicated by arrows.

Two biomass harvests were performed, but as the first was in June, it is only relevant for the spring treatments (there were no significant differences between controls and summer and autumn treatments in June). The June data show only a significant, negative effect of the combination of a heat wave and a drought, with 48% less living biomass harvested at that time (Table 1; Supporting Information Table S1). The slight positive effect of heat seen in both carbon flux data and vegetation height was thus not translated into significant biomass gains. The negative effect of heat + drought was no longer apparent in the October harvest data, implying that the productivity had recovered, which is in agreement with height and flux data. Note the large standard error, which reflects large differences in recovery between communities. A summer drought lowered living aboveground production by 23%, while the autumn drought reduced living biomass by 30%. Combined heat and drought lowered living biomass by 66 and 65% in response to summer and autumn extreme events, respectively. It is important to note that this production includes recovery effects for the summer treatments, but substantially less for the autumn treatments (biomass was harvested 2 wk after the end of the autumn extreme event). When we looked at the dead (standing) biomass, the only significances found were for the summer drought and summer combined extreme event treatments, with drops of 39 and 62%, respectively, compared with controls. Total production (biomass from both harvests + necromass) was only significantly affected in summer when two extremes were combined (−40%), while trends of production losses to summer drought (−18%) and combined heat and drought in spring and autumn (−20 and −17%, respectively) were drowned out by variation.

Table 1.   Aboveground biomass (dry weight) harvested in June and October in 18 communities per treatment
HarvestSeasonal treatmentControl (g m−2)Drought (g m−2)Heat (g m−2)Drought + heat (g m−2)
  1. Necromass was separated from living biomass exclusively in the October harvest. Significant differences (Games–Howell post hoc test) with controls at the *, < 0.05 or ***, < 0.001 level are indicated in bold.


Canopy temperature

We plotted the differences between the treatments and the controls (Fig. 5), as temperatures vary between days, making a straightforward comparison difficult. Single factor drought only raised leaf temperatures during the summer extreme event, and these increases were slight. Single factor heat waves obviously increased temperatures, but vegetation temperatures were generally below 30°C, and no consistent trends in time were observed (which was expected). The combination of a heat wave and a drought led to an increase in leaf temperatures above that of single factor heat waves, and this increase was largest at the end of the extreme event and during the summer extremes. Absolute temperatures in the combined treatment rose to c. 32°C in spring and autumn, but to 45°C in summer.

Figure 5.

 Surface temperature of the vegetation, measured with an infrared camera. Averages of four images per period per treatment relative to control conditions (to filter out daily fluctuating temperatures) are shown. Treatments were drought (squares), heat wave (circles) and the combination of heat wave and drought (diamonds). Measurements were made on the day before the start of a heat wave, and the last three readings during the heat wave (i.e. the last 10 d of an extreme event). DOY, day of the year; day 1 is 1 January 2009. For clarity, only averages are depicted.


Experimental studies that have varied the timing of extreme events are rare and have mostly altered precipitation patterns (e.g. Bates et al., 2006). Studies on heat waves have so far encompassed only summer extremes. This ignores the fact that an extreme is relative to the mean, and as climate change-induced shifts in precipitation and temperature are not confined to summer (IPCC, 2007), the frequency of extreme events will increase in all seasons. However, the plant response will probably differ depending on the timing of the extreme (De Boeck et al., 2010). Jentsch et al. (2007) therefore advised that the timing of extremes should be varied, but this is the first study to explicitly do so for heat waves and droughts.

In this study, single factor heat waves had little effect on plant growth, with only transient positive effects in spring and an equally short-lived negative effect in summer, as evidenced by changes in photosynthesis and vegetation height. Our a priori expectation was that the increase in absolute temperatures to values closer to the metabolic optimum in both spring and autumn in an unusually warm period could stimulate productivity. The fact that we only observed such a growth increase in spring is probably associated with the decreased sensitivity of photosynthesis to temperature in the autumn (Piao et al., 2008; Vesala et al., 2010). The transient nature of spring growth increases may have been a result of the fact that warming only advanced the exponential growth phase of plants (Menzel et al., 2006), without affecting maximum growth rates or growth duration. The absence of marked and lasting effects of a summer heat wave can be attributed to the fact that leaf temperatures did not increase to damaging levels (causing, for example, chlorophyll degradation) as the abundance of soil water allowed plants to continuously cool their leaves via transpiration. This is in agreement with earlier claims that heat effects work mostly indirectly, mainly through drought (Reichstein et al., 2007). The possibility that we are underestimating the effects of pest outbreaks (cf. Rouault et al., 2006) cannot be denied, as we only heated the climate locally. Such larger scale effects would be hard to simulate in manipulation experiments. Any positive effects of increased nutrient mineralization under warming (Rustad et al., 2001) were insignificant. It should be noted that heat waves are usually accompanied by important reductions in precipitation (De Boeck et al., 2010), and that their occurrence without soil water deficits may be restricted to wet ecosystems and spring (when water stocks are usually plentiful).

In a review, Chaves & Oliveira (2004) stated that the timing of stress episodes is pivotal to determining the effects produced by drought. The variability of drought effects with season we found seems to support this claim. One might assume that this could be attributed simply to different soil drought intensities, as several studies have found fairly straightforward relationships between productivity and soil water status during droughts (Gilgen & Buchmann, 2009; Prieto et al., 2009). Our results indicate that phenological factors can confound this relationship, as drought impacts on growth and biomass production were least apparent in spring, despite conspicuously lower soil water contents than in autumn. The rate of soil drying was 30% greater in spring than in autumn, which seems to correspond with a vapour pressure deficit that was 25% higher in the spring compared with the autumn treatment period. It is possible that higher plant water demand in spring could have contributed to faster soil drying, although the lower leaf area of the plants in spring seems to be in contradiction to this possibility. The difference in drought-related growth reductions between spring and autumn was probably not caused by acute drought stress, as infrared imaging did not indicate substantial stomatal closure in either spring or autumn. Rather, the longer length of the precipitation-free period in autumn compared with spring could have extended suboptimal conditions. It is also possible that the water potential within the leaves, which does not necessarily reflect that in the soil, differed between seasons. However, we have no data to confirm this. As anticipated, drought had the strongest effects during summer. In this season, evaporative demand is at its maximum (high vapour pressure deficits), and the majority of plants in temperate regions have peak leaf area. Soil water contents neared the wilting point, which led to significant declines in photosynthesis. A typical plant response to low water potential is closing of the stomata to prevent excessive transpiration, which inevitably leads to declining carbon uptake (Chaves & Oliveira, 2004). We observed the stomatal response indirectly, through the moderate increases in leaf temperature, which strongly suggest a significant decline in stomatal conductance resulting in decreased transpirational cooling. It is important to note that, at the whole-plant level, total carbon uptake is further reduced because of the concomitant or even earlier inhibition of growth as cell division and expansion have been shown to be directly inhibited by water stress (Zhu, 2001; Chaves & Oliveira, 2004). However, flux measurements show a fairly swift recovery after summer drought. Although on first inspection counterintuitive, the lower end-of-season amounts of standing dead biomass in the summer treatment compared with other seasonal treatments may reflect high plant mortality during summer (also for the combined heat + drought treatment) coupled with fairly vigorous regrowth. Dead plants would have largely been transformed to litter by the end of October, while the new shoots did not encounter further stress, hence explaining the lower fraction of standing dead biomass in these two treatments.

As we predicted, negative effects of the combination of a heat wave with a drought period were larger than the sum of the single factor effects. Prieto et al. (2009) speculated that, because they found a similar relationship between biomass and soil moisture under ambient conditions, drought, and warming, soil moisture alone would be a good predictor of biomass accumulation. We argue that the interplay between heat and drought works in two ways. Heat leads to increased vapour pressure deficits and hence increased evapotranspiration, speeding up the process of soil drought. This then leads to stomatal closure, which unavoidably increases leaf temperatures, potentially leading to heat stress. This last effect was only found during summer extremes, when absolute temperatures were higher, leading to leaf temperatures regularly exceeding 40°C in the combination treatment. The combination of excessive temperatures together with soil water contents near or below wilting point caused extreme stress, effectively halting CO2 exchange and causing senescence of almost all aboveground parts (visual observations). In a heat wave experiment in the Arctic, Marchand et al. (2006) reported that senescence of leaves was related to both drought-induced desiccation and heat-related effects on the photosynthetic apparatus. Widespread tissue die-back is consistent with recent studies demonstrating a tight regulation of leaf life span to optimize the plant carbon balance (Oikawa et al., 2008; Reich et al., 2009). Shedding of leaves could also improve whole-plant water relations by increasing the proportion of water-transporting tissues relative to leaves (Chaves et al., 2009). Whereas a recovery to pretreatment photosynthesis levels was observed in the drought-only treatment, and whereas an earlier experiment with a mild temperature increase and summer drought showed strong resilience of grasslands (Zavalloni et al., 2008), plant communities did not recover fully from the combination of a summer heat wave and drought. This was reflected in the sustained reduction in Agross and an end-of-season production that amounted to only a third of that found in the controls. A similar production loss was recorded in response to autumn heat and drought, although here, little recovery was possible in the short time after the end of the treatment. The fact that the total (yearly) production was lowered twice as much in communities exposed to heat and drought in summer compared with those exposed to extremes in autumn suggests that the impact of an autumn heat wave and drought, although significant, was less detrimental in the short term than the effect of summer events. Longer term effects, such as the impact of entering winter when potentially suffering from stress exhaustion (after autumn extremes), will be reported in a later study after data from the 2010 growing season have been compiled. Thabeet et al. (2009) showed that these after-effects can be even greater than the immediate impact. Plants in the spring combination treatment suffered less drought (in terms of intensity and duration) than those in both the summer and autumn treatments and experienced limited or no heat stress. Although there was a decrease in growth, this lasted only weeks and during the second half of the growing season productivity was restored to the same level as in controls. Similar to the single factor effects of drought, plants in spring may, at least in the short term, have been less responsive to drought and heat combined because of the duration of suboptimal growing conditions and a potentially differing leaf water status.

In a global analysis of eddy covariance-based CO2 flux estimates, Schwalm et al. (2010) found that photosynthesis was on average 50% more responsive to drought than respiration. Our observations show that, in absolute numbers, photosynthesis indeed responded more strongly to drought (with or without heat) than respiration, but as a proportion, the declines were comparable. Photosynthesis seemed to recover more rapidly than respiration. A number of factors could have dampened respiration for a prolonged period, such as, for example, the shedding of inefficient (i.e. lower carbon influx vs efflux ratio) leaves during drought (Sala et al., 2010) and the younger average age of leaves in the recovery phase. Belowground, damage to the microbial communities and/or changes in their structure may have lowered heterotrophic respiration (Schimel et al., 2007; Muhr & Borken, 2009). As we did not separate respiration into above- and belowground compartments, we cannot convincingly attribute the absence of strong recovery of respiration after drought to either compartment. It is possible that soil respiration will increase in the next year, when undecomposed organic material is broken down, as, for example, observed in the study of Arnone et al. (2008).

In conclusion, in this first experimental study on the seasonal effects of heat waves and droughts, we uncovered several interactive effects. As seasonality perturbs, for example, the relationship between soil water status and community response, it is clear that the usage of simple, annual correlations is problematic (cf. Vesala et al., 2010). Heat waves as a single factor did not generate a significant response, which can be attributed to the fact that heat stress does not develop when plants have sufficient water at their disposal. In spring, a heat wave seemed to advance the onset of the exponential growth phase somewhat, without altering the yearly biomass production. Droughts generated temporary negative effects on carbon uptake and growth, but adverse impacts increased disproportionately when accompanied by a heat wave, as this intensified soil drought. The strongest effects were observed in summer, a time of already high vapour pressure deficits and fully expanded canopies, with heat stress developing when leaves were no longer able to cool themselves because of soil drying. In contrast to the resilience exhibited after summer extremes, plant communities showed little recovery after autumn extremes, as the growing season was coming to an end. This may lead to significant after-effects in winter and the following year. These findings should be compared with observations in natural systems on a larger spatial and temporal scale, as effects such as pest outbreaks (Rouault et al., 2006), long-term vegetation shifts (Mueller et al., 2005) and pollinator impacts (Carroll et al., 2001) cannot be properly assessed in small-scale experimental studies such as our own.


H.J. De Boeck is a post-doctoral research associate of the Fund for Scientific Research–Flanders. We thank colleagues at the PLECO for help with relocating and harvesting the communities, F. Kockelbergh for technical assistance and four anonymous reviewers for helpful comments.