International Journal of Biometeorology

, Volume 57, Issue 4, pp 579–588

Recurring weather extremes alter the flowering phenology of two common temperate shrubs

Authors

    • Disturbance EcologyUniversity of Bayreuth
  • J. Kreyling
    • BiogeographyUniversity of Bayreuth
  • E. Gellesch
    • Disturbance EcologyUniversity of Bayreuth
  • C. Beierkuhnlein
    • BiogeographyUniversity of Bayreuth
  • A. Jentsch
    • Disturbance EcologyUniversity of Bayreuth
Original Paper

DOI: 10.1007/s00484-012-0585-z

Cite this article as:
Nagy, L., Kreyling, J., Gellesch, E. et al. Int J Biometeorol (2013) 57: 579. doi:10.1007/s00484-012-0585-z
  • 363 Views

Abstract

The aim of this study is to explore the effects of heavy rain and drought on the flowering phenology of two shrub species Genista tinctoria and Calluna vulgaris. We conducted a field experiment over five consecutive years in Central Europe, applying annually recurring extreme drought and heavy rain events on constructed shrubland communities and recorded the flowering status. Further, we correlated spring temperature and precipitation with the onset of flowering. Both species showed a response to extreme weather events: drought delayed the mid flowering date of Genista tinctoria in 3 of 5 years by about 1 month and in 1 year advanced the mid flowering date by 10 days, but did not affect the length of flowering. Mid flowering date of Calluna vulgaris was not affected by drought, but the length of flowering was extended in 2 years by 6 and 10 days. For C. vulgaris the closer the drought occurred to the time of flowering, the larger the impact on the flowering length. Heavy rainfall advanced mid flowering date and reduced the length of flowering of Genista tinctoria by about 2 months in 1 year. Mid flowering date of Calluna vulgaris was not affected by heavy rain, but the length of flowering was reduced in 1 year by 4 days. Our data suggest that extreme weather events, including alterations to the precipitation regime, induce phenological shifts of plant species of a substantial magnitude. Thus, the impacts of climate extremes on plant life cycles may be as influential as gradual warming. Particularly, the variability in the timing of precipitation events appears to have a greater influence on flowering dynamics than the magnitude of the precipitation.

Keywords

FloweringClimate changeExtreme droughtPhenological alterationDwarf shrubPrecipitation change

Introduction

In 1735, Reaumur was the first scientist to suggest that the phenological differences between years and locations can be explained by differences in temperatures (Dose and Menzel 2006). Today, temperature is widely accepted to be the most important driver of the onset of flowering (Penuelas and Filella 2001; Sparks and Menzel 2002; Cleland et al. 2007; Sparks et al. 2009). Phenological shifts are therefore regarded as a “fingerprint” of global warming (Walther et al. 2002). Since 1960s, the onset of spring has been advancing in the northern hemisphere on average by 2.5–2.8 days every decade, or 4.6 days for every 1 °C increase in temperature (Prieto et al. 2009; Memmott et al. 2007; Parmesan 2007). Likewise, the length of the average annual growing season has increased on average by 10.8 days since the early 1960s (Menzel 2000).

Evidence suggests that the frequency and magnitude of climate and weather extremes such as severe drought, heat waves, heavy rain and late frost events are increasing (IPCC 2007; O’Gorman and Schneider 2009). In the regions of Siberia, South Africa, northern Japan and the eastern part of the Mediterranean, mean precipitation has not changed noticeably during the rainy season; however, frequency of heavy precipitation has been increasing. The scientific interest in the ecological implications of extreme weather events is growing (Jentsch and Beierkuhnlein 2008). Extreme climatic events can cause the local breakdown of populations (Breshears et al. 2005), resulting in declining ecosystem functionality (Royer et al. 2011) and are likely to trigger ecosystems-level disturbance (Parmesan 2000). These disturbances may affect species composition and diversity. However, research on phenology has predominantly focused on increasing temperature so far. Yet, precipitation regimes and water availability can cause complex biotic responses and can lead to phenological shifts (Jentsch et al. 2009). Limited soil moisture may slow development during bud and fruit development stages or lengthen those stages and shorten the period when flowers are open (Galen et al. 1999). Several studies further indicate that drought delays the onset of flowering and shortens the length of flowering (Prieto et al. 2008; Jentsch et al. 2009; Gordon et al. 1999b). The duration of flowering is important for pollinators and therefore for the fitness of both plants and pollinators (Memmott 2007). Changes in flowering duration may therefore alter the plant–pollinator interactions and their life cycles. Increased precipitation in preceding months advances flowering of Globularia alypum and Erica multiflora (Prieto et al. 2008). Furthermore, development of leaf bud break of Betula pubenscens was positively influenced by precipitation as well as the development of bud break to flowering of ‘Victoria’ plums (Wielgolaski 2001). However, other studies found no correlation between rainfall and phenology (Cleland et al. 2007). The length of flowering is expected to show significant impacts from temperature or precipitation (Sherry et al. 2011).

Here, we focus on the phenological responses of Central European plants subjected to recurrent simulated precipitation extremes over five consecutive years. We address the following hypotheses: (i) Drought delays mid flowering date and reduces the length of flowering; (ii) With regard to the conflicting evidence in the literature, we further test the null hypothesis that heavy rainfall has no effect on mid flowering date and the length of flowering; and (iii) we further investigate the effect of mean monthly temperature and how the monthly precipitation sum of the months preceding the onset of flowering affect mid flowering date and the length of flowering. Here, we expect that the temperature sensitivity of flowering phenology is smaller than the effects of extreme precipitation events.

Methods and data

Methods

Experimental design

Within the EVENT-experiment (http://www.bayceer.uni-bayreuth.de/EVENT/) (Jentsch 2007), we explore effects of extreme weather events on ecosystem functions (Jentsch and Beierkuhnlein 2010). Here, we analyzed how extreme weather events can affect plant phenology. The experimental site is located at the Ecological Botanical Garden of the University of Bayreuth, Germany (49°55’19”N, 11°34’55”E. 365 m a.s.l.) with a mean annual temperature of 8.2 °C, and a mean annual precipitation of 724 mm (1971–2000). Precipitation is distributed bimodally with a major peak in June/July and a second peak in December/January. The soil was homogenized and drained by a passive drainage system at 80 cm soil depths as commonly used in agriculture. The soil constituency is 82 % sand, 13 % silt and 5 % clay with a pH of 4.5 in the upper soil layer and 6.2 in the lower soil layer (Jentsch et al. 2009).

The focus of this study was on two late-flowering species of our experiment: Genista tinctoria and Calluna vulgaris. These species were selected because they flowered after the experimental weather manipulations and because they are important species for insect pollinators. Both shrubs are nectar resources for insects. We focus on dwarf-shrubs because their phenology is mainly dependent on climatic parameters and not on management actions such as mowing or fertilization. Both species are perennial shrubs. C. vulgaris is a slow-growing evergreen shrub that is classified as a stress-tolerant competitor (Gordon et al. 1999a) and flowers between late summer and autumn. For C. vulgaris, for example, nitrogen in combination with drought increases water-use efficiency; this shows that C. vulgaris can acclimate under strong stress conditions (Gordon et al. 1999b). G. tinctoria is an indicator of high soil moisture content (Fw6) (Ellenberg 1979) and flowers between early summer and autumn. That means G. tinctoria usually grows on soil that is well supplied with water. C. vulgaris show an indifferent moisture indicator value and both species grow on lowest nitrogen substrate (N1) (Ellenberg 1979).

The study plants were obtained from a nursery as 2-year-old individuals. Each of the target species was planted together with three other species (Agrostis stolonifera, Deschampsia flexuosa, and Vaccinium myrtillus), that were selected on the grounds of their affiliation with defined functional groups (grasses, legumes and dwarf-shrubs), their life-span (perennials), their overall importance in Central European ecosystems (common key species) and the fact that the species are found naturally on substrates similar to the one used in the experiment. In early April 2005, 100 plant specimens (25 individuals per species in separate plots for the target species) per plot (2 × 2 m) were planted in a defined quantitative composition in a systematic hexagonal grid with a distance of 20 cm between individuals. The three weather manipulations (see below) were arranged in a randomized block design with five replicates, resulting in 15 plots per target species and a total of 30 plots for both target species combined.

From 2006–2010, weekly observations of the flowering status of four individuals per plot were carried out in the central square meter of each plot, in order to avoid edge effects. Individuals were considered to be ‘flowering’ when anthers were visible in at least one flower. The first day of flowering represents the date of the mean date of flowering of the four monitored individuals per plot.

Weather manipulations

The weather manipulations considered in this study consisted of extreme drought and heavy rainfall. Ambient conditions served as a control. Gumbel I distributions were fitted to the annual extremes and 100- as well as 1000-year recurrence events were calculated (Gumbel 1958). In 2006 and 2007 the intensity of the treatments was based on the local 100-year extreme event in each category derived from Gumbel I distributions from the time series of local weather stations. In 2008 and 2009 the intensity of the treatments was based on the local 1000-year extreme event. The vegetation periods (March to September) of 1961–2000 were used as a reference period (data from the German Weather Service). Drought was defined as the number of consecutive days with <1 mm daily precipitation. The resulting 100-year event for Bayreuth is a drought period of 32 days or heavy rainfall of 170 mm over 14 days. The treatments were carried out in May and June before or at the beginning of the flowering of Genista tinctoria and Calluna vulgaris [drought manipulations: days of the year (DOY) 143–175 in 2006 and 139–171 in 2007; heavy rainfall manipulations: DOY 160–174 in 2006 and 158–172 in 2007]. The corresponding 1000-year extremes are a drought event of 42 days or a heavy rainfall of 21 days and 210 mm (drought manipulations: 139–181 in 2008; 139–181 in 2009; 131–173 in 2010; heavy rainfall manipulations: DOY 162–183 in 2008; 160–181 in 2009 and 152–173 in 2010).

Drought was induced using rainout shelters that permitted the penetration of almost 90 % of photosynthetically active radiation. Undesired greenhouse effects were avoided by constructing the roof at a height of 80 cm (Jentsch et al. 2009) allowing for near-surface air exchange. After the experimental drought periods the roofs were removed. With experimental data, there might be confounding between drought and warming effects on flower phenology. Ultimately, we cannot determine whether early flowering and an extension of the flowering length observed in our drought treatment was really due to drought or to warming or to a combination of both. A roof artefact control with five replicates of the rain-out shelters was established in 2006 (Jentsch et al. 2009). Adding the same amount of water as occurred naturally in daily resolution below intact shelters during the drought manipulation did not result in any differences between the ‘control’ and ‘artefact control’ with respect to the ‘mid-flowering day’ and the ‘length of the flowering period’. The slight increase of 1.6 °C due to an installation of rainout shelters was not statistically significant and only lasted for a limited period of 32 or 42 days.

Heavy rainfall was applied using portable irrigation systems. The drop size and the rainfall intensity resembled natural heavy rainfall events by using Veejet 80 100 nozzles that are also used in erosion research (Kehl et al. 2005). A lateral surface flow was avoided by using plastic sheet pilings around treated plots down to a depth of 10 cm. The calculated amount of added water was divided into two applications per day to ensure constantly high soil water saturation and to avoid any surface runoff over the pilings. If natural precipitation occurred, the amount was subtracted from the respective dose.

Data

Soil moisture was recorded by ECH2O EC-5 Moisture Sensors (Decagon Device, USA) placed at −5 cm. Furthermore, air temperature at plant height and precipitation on site were measured.

Data analysis

The species-specific onset of flowering, the mid flowering date and the length of flowering were considered. The onset date is defined by the first day of flowering. The first day of flowering represents the mean first date of flowering of the four monitored individuals per plot. The mid flowering date was calculated as the date of the 50th percentile of the flowering curve over the entire period of flowering (Jentsch et al. 2009). The length of flowering was calculated as the difference between the days of the years for the 25th percentile and the 75th percentile of the flowering curve over time (Jentsch et al. 2009). Concerning the flowering length and onset of flowering, we initially tested different quantiles (e.g., 90 %). However, results were qualitatively very similar and we decided to use a very conservative approach to avoid outliers and to be directly comparable to the results of Jentsch et al. (2009). Furthermore, a correlation was calculated between flowering length and the time between the end of the drought manipulation and the onset of flowering over all years.

Analysis of variance (ANOVA) was applied to test for significant differences between the weather manipulations. To account for the block design, block identity was used as a covariate (factorial) in the models. Each year was analyzed separately. Before statistical analysis was applied, the conditions of normality were tested by examining the residuals vs. fitted plots and the normal qq-plots of linear models (Faraway 2005). Where treatment effects were significant pairwise comparisons were made with Tukey HSD (honest significant differences).

The response to temperature over all years was calculated using a linear least-squares regression of the onset of flowering with mean monthly temperatures for the months of January/February/March and March/April/May, as well as for the months of January to June separately. The regression with temperature was analyzed using the control plots (n = 5). Temperatures shortly before onset are known to determine onset date (Menzel et al. 2006). All statistical analyses were performed using R (R Development Core Team 2010).

Results

Climate variations between years

Temperature and precipitation varied among the 5 years observed (Table 1). The years 2006, 2009 and 2010 were characterized by low monthly mean temperatures and frost in January and February. The mean annual temperatures of all observed years exceeded the local 30 year average temperature, with 2007 being the warmest and wettest year of our five observed years. Likewise, soil moisture also varied considerably among the 5 years. In the experimental manipulation the drought manipulation reduced soil moisture by an average of 53 % over all 5 years while heavy rain elevated soil moisture by an average of 37 % during the treatments. The higher the amount of ambient precipitation during the drought manipulation is, the higher the soil moisture reduction relative to the control. Likewise, for heavy rain, the higher the amount of ambient precipitation during the manipulation period, the lower the soil moisture elevation on the heavy rain plots. However, after the manipulations, soil moisture soon recovered to the level of the control plots in all years.
Table 1

Sum of mean monthly air temperature and precipitation over the years 2006–2010 as well as the local 30-year average (1971–2000). Mean soil moisture difference of drought and heavy rain plots of the treatments for each year against the soil moisture of the control plots during the treatments

Months

2006

2007

2008

2009

2010

30-year mean

Mean temperature (°C)

Jan

−3.5

4.1

2.6

−4.3

−7.0

−0.8

Feb

−1.5

4.0

2.9

−0.3

−1.0

0.1

Mar

1.7

5.2

3.9

3.6

3.0

3.9

Apr

8.0

10.9

7.9

11.7

9.0

7.4

May

12.7

14.2

14.5

14.1

12.0

12.7

June

16.6

17.6

17.4

15.1

18.0

15.6

July

21.4

17.2

17.9

17.8

22.0

16.8

Aug

14.8

16.6

17.0

18.0

17.0

16.9

Sept

15.9

15.8

11.7

11.8

12.0

12.8

Oct

10.0

11.0

7.5

8.0

8.5

8.1

Nov

2.9

6.1

2.4

4.2

5.8

3.2

Dec

−0.4

3.0

0.5

0.5

5.0

0.7

Mean

8.2

10.5

8.8

8.4

8.7

8.1

Sum of precipitation (mm)

Jan

21.3

96.2

40.6

27.5

21.0

60.8

Feb

42.5

63.0

42.3

43.2

32.4

44.6

Mar

70.1

48.2

107.7

70.2

30.5

49.2

Apr

51.2

4.3

78.3

67.6

19.0

45.7

May

111.5

115.1

13.8

54.1

116

56.3

June

35.0

91.4

44.3

60.8

57.5

78.9

July

106.2

115.2

69.7

138

61.9

85.7

Aug

84.2

54.8

71.4

33.4

190

59.3

Sept

27.5

93.3

65.4

57.7

46.5

55.6

Oct

61.4

21.2

84.0

63.7

25.6

47.2

Nov

33.1

98.0

34.1

74.6

90.1

56.9

Dec

25.1

55.1

48.2

76.7

109.4

70.3

Sum

669.2

855.8

699.8

767.5

800.0

710.4

Soil moisture (%)

Drought

−37 %

−70 %

−59 %

−59 %

−43 %

 

Heavy rain

65 %

20 %

60 %

30 %

10 %

 

Drought impacts on the mid flowering date and the length of flowering

Drought significantly delayed the mid flowering date of Genista tinctoria in 2006, 2007 and 2009 (p = 0.004, 33 days; p = 0.02, 25 days; p = 0.003, 32 days, respectively) (Fig. 1). In contrast, in 2010 the mid flowering date was significantly advanced by 10 days (p = 0.010). The length of flowering was not affected by drought. Over the 5-year period, on average the mid flowering date was delayed by 13.3 days for G. tinctoria, while the length of flowering over the 5 years was shortened by an average of 2.2 days. The correlation between flowering length and the time between the end of the drought manipulation and the onset of flowering over all years was not significant (r2 = 0.29, p = 0.348).
https://static-content.springer.com/image/art%3A10.1007%2Fs00484-012-0585-z/MediaObjects/484_2012_585_Fig1_HTML.gif
Fig. 1

Effects of extreme weather events on mid flowering date (♦) and the length of flowering of Calluna vulgaris and Genista tincoria for the vegetation periods 2006–2010. Thin lines show standard errors for the mid flowering date in white and the flowering length in black. Significance of differences between the treatments and the control indicated as * #: p <0.05, ** ## p <0.01

The mid flowering date of Calluna vulgaris was not affected by drought in any of the five observed years. In 2008 and 2009 the length of flowering however was extended significantly (p = 0.02, 10 days; p = 0.01, 6 days). For the drought treatment over the 5-year period, there was a delay in the mid flowering date of 1.7 days for C. vulgaris, and the length of flowering over the 5 years was extended by 4.7 days on average. The closer the drought occurred to the onset of flowering the longer was the flowering length (r2 = 0.78, p = 0.047).

From 2006 to 2009 G. tinctoria flowered later each year on the drought plots, and over the entire 5 years observed the length of flowering became shorter every year.

Heavy rain impacts on the mid flowering date and the length of flowering

Heavy rainfall significantly brought forward the mid flowering date of Genista tinctoria in 2008 (p = 0.001) by 57 days (Fig. 1). There is not a significant tendency towards an earlier mid flowering date in all years. The length of flowering was shortened significantly in 2008 by 56 days (p = 0.001). Even though not all years showed a significant effect there is a tendency to a shorter length of flowering in all years. For the heavy rain treatment on average over the 5-year period, the mid flowering date of G. tinctoria was advanced by 11.8 days and the length of flowering over the 5 years was shortened by an average of 21.1 days. The mid flowering date of Calluna vulgaris was not affected by heavy rain. The length of flowering was shortened significantly in 2006 by 4 days (p = 0.050). On average over the 5-year period, the mid flowering date of C. vulgaris was advanced by 0.6 days and the length of flowering over the 5 years was shortened by an average of 1.1 days.

Effects of temperature and precipitation on the onset of flowering

The onset of flowering for Genista tinctoria showed a strong negative correlation with the temperature of the months of January (Jan) (r2 = 0.81, p < 0.001) and February (r2 = 0.68, p < 0.001) as well as with those of the winter season (Jan/Feb/Mar) (r2 = 0.76, p < 0.001), implying that the higher the temperature, the earlier the onset of flowering (Table 2). The negative correlation between temperature and the onset of flowering in Calluna vulgaris was weaker (January: r2 = 0.37, p < 0.001, February: r2 = 0.35, p = 0.002, January/February/March: r2 = 0.06, p = 0.236). Onset of flowering was also significantly and negatively correlated with May temperature for both species (Calluna vulgaris: r2 = 0.57, p < 0.001; Genista tinctoria: r2 = 0.53, p < 0.001).
Table 2

Results of the linear least-squares regression between temperature and the onset of flowering of Calluna vulgaris and Genista tinctoria for the years 2006–2010. Temperature is only given for the control because all three treated plots had the same temperatures. Significant effects are set in bold

Parameter

Time

Calluna vulgaris

Genista tinctoria

r2

p

Slope

r2

p

Slope

Temperature

January/February/March

0.2

0.001

−1.34

0.67

<0.001

−2.17

March/April/May

0.32

0.003

−4.79

0.24

0.012

−3.49

January

0.37

<0.001

−1.24

0.81

<0.001

−1.51

February

0.35

0.002

−2.25

0.7

<0.001

−2.68

March

0.26

0.008

−3.78

0.38

<0.001

−3.85

April

0.06

0.227

−1.37

0

0.888

−0.14

May

0.57

<0.001

−6.52

0.53

<0.001

−5.33

June

0.1

0.115

1.81

0.03

0.391

−1.24

Discussion

Effects of drought on flower phenology

Flower phenology of one of the dwarf shrubs, Genista tinctoria, was significantly delayed by drought. Other authors also found evidence of delayed flowering after drought treatments in similar study systems with comparable plant functional types (Jentsch et al. 2009; Prieto et al. 2008; Llorens and Penuelas 2005; Gordon et al. 1999a). Jentsch et al. 2009 confirm the delay of flowering after drought for grass species, herbs and shrubs. In an experiment by Gordon et al. (1999a) the flowering of Calluna vulgaris was delayed after a 3-week drought treatment. In a 3-year experiment drought treatment caused a delay in the flowering of Erica multiflora (Prieto et al. 2008). G. tinctoria showed strongly delayed flowering in response to drought in 3 out of 5 years (Fig. 1). This leads to the conclusion that the G. tinctoria plants on the drought plots did not start flowering until the water availability was rising again. Prieto et al. (2008) also discuss the reaction of delayed flowering and a reduction in peak intensity and came to the same conclusion that water availability after a drought is insufficient and that rain is needed to stimulate plants to start flowering. For Globularia alypum more than 10 mm of rain after a drought was necessary for anthesis to start (Prieto et al. 2008), whereas the first rewetting after a drought triggers flowering of Globularia alypum. The timing of drought seems to be more important for determining flowering than the accumulated precipitation amount (Prieto et al. 2008). We can confirm this importance of timing in our study for both species. The rainfall amount between the end of drought and the onset of flowering, however, was not significantly related to the date of flowering, implying that there is no specific rainfall amount (threshold) that was necessary for onset of flowering after a drought. As observed within other studies, the effects were species-specific, and water availability plays an important role in determining year-to-year shifts in flowering (Llorens and Penuelas 2005; Prieto et al. 2008). The mid flowering date of C. vulgaris was not affected by drought. The response of C. vulgaris may be linked to a conservative use of water, which is a strategy that allows species to maintain a higher water potential and thus more continuity. This reaction was also found for Erica multiflora after a drought treatment (Prieto et al. 2008).

Our hypothesis stating that drought reduces the length of flowering was not supported by our data. In 2008 and 2009 the length of flowering of C. vulgaris was significantly extended and there is a trend towards expansion in all years. The drought manipulation resulted in an extended length of flowering in ten grass, herb and shrub species over a period of 1 year (Jentsch et al. 2009). The drought treatment may lengthen the flowering period by imposing stress that reduces the plant’s energy, and thereby, its ability to produce an entire blossom in one go, thus increasing the time required to regenerate. In other studies, water stress increased the duration of flowering by up to 15 days in Lesquerella fendleri (Ploschuk et al. 2001), whereas drought reduced the flowering period of Globularia alypum (Prieto et al. 2008). This kind of response could also be observed in 2010 for G. tinctoria. In this particular year the plants that were treated with drought flowered for the shortest period out of all 5 years, flowering for only 24 days from the beginning to the end of July. 2010 included the warmest and driest July of all observed years, while June 2010 was also very warm (Table 1). Such high temperatures could be the factor responsible for limiting the length of flowering in this particular year. Temperature is commonly assumed to be the major driver of phenological shifts in the northern hemisphere. Studies focusing on climate warming reported a phenological shift of 2 days per decade for Central Europe (Menzel and Fabian 1999), 4.5 days per decade for the British Isles (Fitter and Fitter 2002), 1.2–2.0 days per decade for North America (Walther et al. 2002), and 1.4–2.3 days per decade for global datasets (Penuelas and Filella 2001; Parmesan and Yohe 2003). However, our results suggest that for some species, a single extreme drought event can affect the phenology of flowering to a similar or even higher degree than gradual warming. The consequences for plant species can be higher for extreme events than for mean change, because individual species are more sensitive to interannual variability and extreme events compared to mean changes in environmental and resource conditions (Zavaleta et al. 2003). Drought and warming can cause similar reactions in plants such as effects on leaf water status either due to reduced soil moisture content during a drought or due to a higher evapotranspiration during warming. Furthermore, both of these factors can affect ABA (abscisic acid) mediated signalling (Toh et al. 2008), which is a major hormonal driver of flowering dynamics. Thus, mechanistic understanding of how extreme weather events affect the onset of flowering and duration of flowering is still widely lacking. An extreme event is already of extraordinary magnitude, but extreme events with an annual return interval over five consecutive years are even more extreme. Our results indicate that no clear difference was obtained in the response when the manipulation was changed from 100-year events to 1000-year events. This implies that using other distributions with shifts in the manipulation by very few days is not crucial for obtained results. However, the different flowering patterns over the 5 years may be due to alternative mechanisms. The individual plant age changed from one year to the next so the sensitivity to drought may have changed. The sensitivity to drought of G. tinctoria seems to have changed over the experimental time, as from 2006 to 2009 (but not in 2010). G. tinctoria flowered later each year on the drought plots except for 2010, and over the whole 5 years observed the length of flowering became shorter every year. This reaction could either be an adaptation or an acclimation to drought. Roots under dry soil conditions grow slower than in well-watered conditions (Chaves et al. 2003). Plants, especially legumes, have the ability to store reserves in some organs, such as stems and roots, and to mobilize them if necessary. Soil drying can also induce a decrease in nutrients available to the plants, in particular nitrogen, which can have a strong effect on plant growth and function (McDonald and Davies 1996). In drought-exposed plants the ability to mobilize reserves is increased (Rodrigues et al. 1995). Larger plants can store more reserves and plant sensitivity can be reduced. This together with adaptation to drought might be the reason for a shorter length of flowering over the 5 years observed. Furthermore, ABA plays an important role in controlling root growth under water stress because ABA restricts ethylene production, which inhibits growth (Sharp and LeNoble 2002). With our data we were not able to confirm the results of other researchers that larger individuals flower earlier (Ollerton and Lack 1998). G. tinctoria and C. vulgaris are woody plants with small leaves and flowers. They are able to store reserves in some organs like stems and root. The small leaves and flowers of both shrubs are advantageous in avoidance of water loss. Number of flowers at one date in the years 2006 and 2007, respectively, showed no significant treatment effects (data not shown). This implies that any shortening of the flowering period also reduced the reproductive allocation.

The timing of the drought events is important. Especially for C. vulgaris the closer the drought occurs to the time of flowering, the larger the impact it is likely to cause on the flowering length. There appears to be a temporal relationship between the extreme events and the period of flowering. This is more apparent when observed separately over the years for the species. In 2008 and 2009 the timing of the drought was closest to the flowering of C. vulgaris, and in both years the length of flowering was significantly longer.

Effects of heavy rainfall on flower phenology

With the exception of G. tinctoria in 2008, heavy rain did not significantly influence the mid flowering date of either species. We can partly confirm our second hypothesis stating that heavy rain has no effect on the mid flowering date. The length of flowering was shortened for G. tinctoria in 2008 and for Calluna vulgaris in 2006. An advance in phenological development after adding water has been reported for woody species at higher latitudes (Wielgolaski 2001). An increased amount of precipitation together with an increase in the length of the rainy season by 3 weeks in spring had no consistent impact on the phenology of an annual California grassland (Cleland et al. 2006), and a double precipitation treatment without changes to the timing of rainfall had no significant effect on the flowering phenology of a perennial grassland in Oklahoma (Sherry et al. 2007). Heavy rain shortened the length of flowering of C. vulgaris in 2006. Generally, there is little known about either the effects of heavy rainfall or the increased annual precipitation on flowering phenology. The plants on heavy rain plots mostly flowered immediately after the heavy rain manipulation. In this period with little resource limitation the plants appear to be able to develop all their flowers at the same time and as a result the length of flowering was shorter. Prieto et al. (2008) also found this phenomenon of compression of flowering length after a wet spring with summer flower buds opening in a single large peak.

Effects of temperature and precipitation on the onset of flowering

As expected, the onset of flowering of the two shrub species, especially Genista tinctoria, showed a strong correlation with the temperature of the preceding winter months. The results of this experiment are well-matched to previous studies, suggesting that the higher the temperatures, the earlier the onset of flowering (Menzel and Fabian 1999; Sparks and Menzel 2002; Dose and Menzel 2006). The onset of flowering was advanced by 6.5 days for C. vulgaris and 5.3 days for G. tinctoria for temperatures that were 1 °C warmer in May. By comparison, the onset of flowering has been found to be 4.6 days earlier for every 1 ° C of temperature increase in large phenological datasets across continents since 1960 (Menzel 2000; Prieto et al. 2009; Memmott et al. 2007; Parmesan 2007). Results of a study in Germany covering entirely spring, summer and early autumn phases of plants showed a temperature response of between 2.4 and 6.7 d C−1 (Menzel 2003). Estrella et al. (2007) demonstrated the same with a response of agricultural plants in Germany to March–May temperatures of 4.3 d·°C−1 and a response of wild shrubs of 4.13 d·°C−1. Our data (Table 2) showed comparable temperature responses to the literature. In addition, our data suggest that extreme heavy rain and drought can cause a similar or higher effect on the mid flowering date than 1 °C temperature can cause on the onset of flowering. The mean manipulations effect of recurrent drought caused a delay in the mid flowering date of 13.3 days in G. tinctoria, while heavy rainfall advanced the flowering of G. tinctoria by 11.8 days.

Implications for the interaction between plants and pollinators

In addition to climate warming and extreme weather events, site-specific factors including the local landscape, the land cover type, the microclimate, water availability, vegetation patterns, plant communities as well as the genetic background of a species all have an impact on the flower phenology. Phenophases react in a species-specific manner and show different sensitivities to environmental changes (Schleip et al. 2009). Synchronization between the timing of flowering and pollination is crucial for plant regeneration and insect pollinators. A change in flowering time, caused by drought, can induce a shift in the life cycles of pollinators, the number of flowers and the seed set, which in turn affect reproductive fitness. For instance, snow melt date as a climate driver has several effects on pollinator population growth as early snow melt decreases floral resources, hence the nectar availability, which determines the fecundity (Boggs and Inouye 2012). Climate change may disrupt the overlap in seasonal timing between flowering and pollinator activity and can result in an alteration of plant–animal interactions (Harrison 2000; Wall et al. 2003). It has been observed, for example, that global warming advances onset of flowering, but that it also advances the seasonal flight activity of some pollinators by 4 days for every 1 °C (Memmott et al. 2007). Especially at higher latitudes phenological advancements driven by climate change are identified, so that migrants from lower latitude may arrive after the availability of seasonal resources (McKinney et al. 2012). Thomson (2010) acknowledged increasing pollination limitation conceivably because of asynchronization between plants and pollinators. Pollinators that do not adapt their behaviour and life cycles to extreme events and an alteration in flowering of their host plants are vulnerable to local extinction (Memmott et al. 2007). Food limitation indicates a reduction in fecundity and longevity (Murphy et al. 1983) and reduces the population densities and growth rates of pollinators. Memmott et al. (2007) predict that between 17 % and 50 % of all pollinator species will suffer a disruption of food supply if plant phenology advances as much as 1–3 weeks. Plants will be less able to reproduce if they lose pollinators. However, plants could be protected if specialized pollinators decrease and the plants are visited by generalists (Melian and Bascompte 2002; Memmott et al. 2004). If plants and pollinators respond differently to extreme events in their life cycles, this could have consequences for both plants and pollinators, especially in terms of species richness and the survival of species.

Conclusions

Extreme climatic events such as drought and heavy rain might be just as important drivers for change in the relative success of dwarf shrubs as long-term environmental changes, such as increased temperature. Extreme weather events are projected to increase in magnitude and frequency, potentially having far-reaching consequences for ecology. The timing of extreme events, especially that of drought, is very important. Drought alters the flower development of species, especially those where the onset of flowering is close to the timing of drought and can also have unforeseen consequences for the structure and the dynamics of their pollinators. There is a need to increase our understanding of the dynamics of pollinators towards extreme events.

The complexity of these phenological changes showed that precipitation variability in frequency and timing may have a greater influence on flowering than the magnitude of precipitation. These results demonstrate the need to consider any changes in the climate such as heavy rain or drought, as other important drivers of climate change that may lead to significant phenological changes.

Acknowledgments

This research project was funded by the “Bavarian Climate Programme 2020” at the joint research center “FORKAST” as well as by DFG (JE 282/6–1). We gratefully acknowledge the help of students and field workers in recording flower phenology and maintaining the plots in the EVENT-experiment. We thank the anonymous reviewers as well as David Inouye as a reviewer for their useful comments.

Copyright information

© ISB 2012